Return to tm.texi CVS log

Up to [Apple XNU] / GNUtools / cc

Annotation of GNUtools/cc/tm.texi, revision 1.1

1.1     ! root        1: @c Copyright (C) 1988, 1989, 1992, 1993 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: @node Target Macros
        !             6: @chapter Target Description Macros
        !             7: @cindex machine description macros
        !             8: @cindex target description macros
        !             9: @cindex macros, target description
        !            10: @cindex @file{tm.h} macros
        !            11: 
        !            12: In addition to the file @file{@var{machine}.md}, a machine description
        !            13: includes a C header file conventionally given the name
        !            14: @file{@var{machine}.h}.  This header file defines numerous macros
        !            15: that convey the information about the target machine that does not fit
        !            16: into the scheme of the @file{.md} file.  The file @file{tm.h} should be
        !            17: a link to @file{@var{machine}.h}.  The header file @file{config.h}
        !            18: includes @file{tm.h} and most compiler source files include
        !            19: @file{config.h}.
        !            20: 
        !            21: @menu
        !            22: * Driver::              Controlling how the driver runs the compilation passes.
        !            23: * Run-time Target::     Defining @samp{-m} options like @samp{-m68000} and @samp{-m68020}.
        !            24: * Storage Layout::      Defining sizes and alignments of data.
        !            25: * Type Layout::         Defining sizes and properties of basic user data types.
        !            26: * Registers::           Naming and describing the hardware registers.
        !            27: * Register Classes::    Defining the classes of hardware registers.
        !            28: * Stack and Calling::   Defining which way the stack grows and by how much.
        !            29: * Varargs::            Defining the varargs macros.
        !            30: * Trampolines::         Code set up at run time to enter a nested function.
        !            31: * Library Calls::       Controlling how library routines are implicitly called.
        !            32: * Addressing Modes::    Defining addressing modes valid for memory operands.
        !            33: * Condition Code::      Defining how insns update the condition code.
        !            34: * Costs::               Defining relative costs of different operations.
        !            35: * Sections::            Dividing storage into text, data, and other sections.
        !            36: * PIC::                        Macros for position independent code.
        !            37: * Assembler Format::    Defining how to write insns and pseudo-ops to output.
        !            38: * Debugging Info::      Defining the format of debugging output.
        !            39: * Cross-compilation::   Handling floating point for cross-compilers.
        !            40: * Misc::                Everything else.
        !            41: @end menu
        !            42: 
        !            43: @node Driver
        !            44: @section Controlling the Compilation Driver, @file{gcc}
        !            45: @cindex driver
        !            46: @cindex controlling the compilation driver
        !            47: 
        !            48: @table @code
        !            49: @findex SWITCH_TAKES_ARG
        !            50: @item SWITCH_TAKES_ARG (@var{char})
        !            51: A C expression which determines whether the option @samp{-@var{char}}
        !            52: takes arguments.  The value should be the number of arguments that
        !            53: option takes--zero, for many options.
        !            54: 
        !            55: By default, this macro is defined to handle the standard options
        !            56: properly.  You need not define it unless you wish to add additional
        !            57: options which take arguments.
        !            58: 
        !            59: @findex WORD_SWITCH_TAKES_ARG
        !            60: @item WORD_SWITCH_TAKES_ARG (@var{name})
        !            61: A C expression which determines whether the option @samp{-@var{name}}
        !            62: takes arguments.  The value should be the number of arguments that
        !            63: option takes--zero, for many options.  This macro rather than
        !            64: @code{SWITCH_TAKES_ARG} is used for multi-character option names.
        !            65: 
        !            66: By default, this macro is defined as
        !            67: @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
        !            68: properly.  You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
        !            69: wish to add additional options which take arguments.  Any redefinition
        !            70: should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
        !            71: additional options.
        !            72: 
        !            73: @findex SWITCHES_NEED_SPACES
        !            74: @item SWITCHES_NEED_SPACES
        !            75: A string-valued C expression which is nonempty if the linker needs a
        !            76: space between the @samp{-L} or @samp{-o} option and its argument.
        !            77: 
        !            78: If this macro is not defined, the default value is 0.
        !            79: 
        !            80: @findex CPP_SPEC
        !            81: @item CPP_SPEC
        !            82: A C string constant that tells the GNU CC driver program options to
        !            83: pass to CPP.  It can also specify how to translate options you
        !            84: give to GNU CC into options for GNU CC to pass to the CPP.
        !            85: 
        !            86: Do not define this macro if it does not need to do anything.
        !            87: 
        !            88: @findex NO_BUILTIN_SIZE_TYPE
        !            89: @item NO_BUILTIN_SIZE_TYPE
        !            90: If this macro is defined, the preprocessor will not define the builtin macro
        !            91: @code{__SIZE_TYPE__}.  The macro @code{__SIZE_TYPE__} must then be defined
        !            92: by @code{CPP_SPEC} instead.
        !            93: 
        !            94: This should be defined if @code{SIZE_TYPE} depends on target dependent flags
        !            95: which are not accessible to the preprocessor.  Otherwise, it should not
        !            96: be defined.
        !            97: 
        !            98: @findex NO_BUILTIN_PTRDIFF_TYPE
        !            99: @item NO_BUILTIN_PTRDIFF_TYPE
        !           100: If this macro is defined, the preprocessor will not define the builtin macro
        !           101: @code{__PTRDIFF_TYPE__}.  The macro @code{__PTRDIFF_TYPE__} must then be
        !           102: defined by @code{CPP_SPEC} instead.
        !           103: 
        !           104: This should be defined if @code{PTRDIFF_TYPE} depends on target dependent flags
        !           105: which are not accessible to the preprocessor.  Otherwise, it should not
        !           106: be defined.
        !           107: 
        !           108: @findex SIGNED_CHAR_SPEC
        !           109: @item SIGNED_CHAR_SPEC
        !           110: A C string constant that tells the GNU CC driver program options to
        !           111: pass to CPP.  By default, this macro is defined to pass the option
        !           112: @samp{-D__CHAR_UNSIGNED__} to CPP if @code{char} will be treated as
        !           113: @code{unsigned char} by @code{cc1}.
        !           114: 
        !           115: Do not define this macro unless you need to override the default
        !           116: definition.
        !           117: 
        !           118: @findex CC1_SPEC
        !           119: @item CC1_SPEC
        !           120: A C string constant that tells the GNU CC driver program options to
        !           121: pass to @code{cc1}.  It can also specify how to translate options you
        !           122: give to GNU CC into options for GNU CC to pass to the @code{cc1}.
        !           123: 
        !           124: Do not define this macro if it does not need to do anything.
        !           125: 
        !           126: @findex CC1PLUS_SPEC
        !           127: @item CC1PLUS_SPEC
        !           128: A C string constant that tells the GNU CC driver program options to
        !           129: pass to @code{cc1plus}.  It can also specify how to translate options you
        !           130: give to GNU CC into options for GNU CC to pass to the @code{cc1plus}.
        !           131: 
        !           132: Do not define this macro if it does not need to do anything.
        !           133: 
        !           134: @findex ASM_SPEC
        !           135: @item ASM_SPEC
        !           136: A C string constant that tells the GNU CC driver program options to
        !           137: pass to the assembler.  It can also specify how to translate options
        !           138: you give to GNU CC into options for GNU CC to pass to the assembler.
        !           139: See the file @file{sun3.h} for an example of this.
        !           140: 
        !           141: Do not define this macro if it does not need to do anything.
        !           142: 
        !           143: @findex ASM_FINAL_SPEC
        !           144: @item ASM_FINAL_SPEC
        !           145: A C string constant that tells the GNU CC driver program how to
        !           146: run any programs which cleanup after the normal assembler.
        !           147: Normally, this is not needed.  See the file @file{mips.h} for
        !           148: an example of this.
        !           149: 
        !           150: Do not define this macro if it does not need to do anything.
        !           151: 
        !           152: @findex LINK_SPEC
        !           153: @item LINK_SPEC
        !           154: A C string constant that tells the GNU CC driver program options to
        !           155: pass to the linker.  It can also specify how to translate options you
        !           156: give to GNU CC into options for GNU CC to pass to the linker.
        !           157: 
        !           158: Do not define this macro if it does not need to do anything.
        !           159: 
        !           160: @findex LIB_SPEC
        !           161: @item LIB_SPEC
        !           162: Another C string constant used much like @code{LINK_SPEC}.  The difference
        !           163: between the two is that @code{LIB_SPEC} is used at the end of the
        !           164: command given to the linker.
        !           165: 
        !           166: If this macro is not defined, a default is provided that
        !           167: loads the standard C library from the usual place.  See @file{gcc.c}.
        !           168: 
        !           169: @findex STARTFILE_SPEC
        !           170: @item STARTFILE_SPEC
        !           171: Another C string constant used much like @code{LINK_SPEC}.  The
        !           172: difference between the two is that @code{STARTFILE_SPEC} is used at
        !           173: the very beginning of the command given to the linker.
        !           174: 
        !           175: If this macro is not defined, a default is provided that loads the
        !           176: standard C startup file from the usual place.  See @file{gcc.c}.
        !           177: 
        !           178: @findex ENDFILE_SPEC
        !           179: @item ENDFILE_SPEC
        !           180: Another C string constant used much like @code{LINK_SPEC}.  The
        !           181: difference between the two is that @code{ENDFILE_SPEC} is used at
        !           182: the very end of the command given to the linker.
        !           183: 
        !           184: Do not define this macro if it does not need to do anything.
        !           185: 
        !           186: @findex LINK_LIBGCC_SPECIAL
        !           187: @item LINK_LIBGCC_SPECIAL
        !           188: Define this macro meaning that @code{gcc} should find the library
        !           189: @file{libgcc.a} by hand, rather than passing the argument @samp{-lgcc}
        !           190: to tell the linker to do the search; also, @code{gcc} should not
        !           191: generate @samp{-L} options to pass to the linker (as it normally does).
        !           192: 
        !           193: @findex LINK_LIBGCC_SPECIAL_1
        !           194: @item LINK_LIBGCC_SPECIAL_1
        !           195: Define this macro meaning that @code{gcc} should find the
        !           196: library @file{libgcc.a} by hand, rather than passing the argument
        !           197: @samp{-lgcc} to tell the linker to do the search.
        !           198: 
        !           199: @findex RELATIVE_PREFIX_NOT_LINKDIR
        !           200: @item RELATIVE_PREFIX_NOT_LINKDIR
        !           201: Define this macro to tell @code{gcc} that it should only translate
        !           202: a @samp{-B} prefix into a @samp{-L} linker option if the prefix
        !           203: indicates an absolute file name.
        !           204: 
        !           205: @findex STANDARD_EXEC_PREFIX
        !           206: @item STANDARD_EXEC_PREFIX
        !           207: Define this macro as a C string constant if you wish to override the
        !           208: standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
        !           209: try when searching for the executable files of the compiler.
        !           210: 
        !           211: @findex MD_EXEC_PREFIX
        !           212: @item MD_EXEC_PREFIX
        !           213: If defined, this macro is an additional prefix to try after
        !           214: @code{STANDARD_EXEC_PREFIX}.  @code{MD_EXEC_PREFIX} is not searched
        !           215: when the @samp{-b} option is used, or the compiler is built as a cross
        !           216: compiler.
        !           217: 
        !           218: @findex STANDARD_STARTFILE_PREFIX
        !           219: @item STANDARD_STARTFILE_PREFIX
        !           220: Define this macro as a C string constant if you wish to override the
        !           221: standard choice of @file{/usr/local/lib/} as the default prefix to
        !           222: try when searching for startup files such as @file{crt0.o}.
        !           223: 
        !           224: @findex MD_STARTFILE_PREFIX
        !           225: @item MD_STARTFILE_PREFIX
        !           226: If defined, this macro supplies an additional prefix to try after the
        !           227: standard prefixes.  @code{MD_EXEC_PREFIX} is not searched when the
        !           228: @samp{-b} option is used, or when the compiler is built as a cross
        !           229: compiler.
        !           230: 
        !           231: @findex MD_STARTFILE_PREFIX_1
        !           232: @item MD_STARTFILE_PREFIX_1
        !           233: If defined, this macro supplies yet another prefix to try after the
        !           234: standard prefixes.  It is not searched when the @samp{-b} option is
        !           235: used, or when the compiler is built as a cross compiler.
        !           236: 
        !           237: @findex LOCAL_INCLUDE_DIR
        !           238: @item LOCAL_INCLUDE_DIR
        !           239: Define this macro as a C string constant if you wish to override the
        !           240: standard choice of @file{/usr/local/include} as the default prefix to
        !           241: try when searching for local header files.  @code{LOCAL_INCLUDE_DIR}
        !           242: comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
        !           243: 
        !           244: Cross compilers do not use this macro and do not search either
        !           245: @file{/usr/local/include} or its replacement.
        !           246: 
        !           247: @findex SYSTEM_INCLUDE_DIR
        !           248: @item SYSTEM_INCLUDE_DIR
        !           249: Define this macro as a C string constant if you wish to specify a
        !           250: system-specific directory to search for header files before the standard
        !           251: directory.  @code{SYSTEM_INCLUDE_DIR} comes before
        !           252: @code{STANDARD_INCLUDE_DIR} in the search order.
        !           253: 
        !           254: Cross compilers do not use this macro and do not search the directory
        !           255: specified.
        !           256: 
        !           257: @findex STANDARD_INCLUDE_DIR
        !           258: @item STANDARD_INCLUDE_DIR
        !           259: Define this macro as a C string constant if you wish to override the
        !           260: standard choice of @file{/usr/include} as the default prefix to
        !           261: try when searching for header files.
        !           262: 
        !           263: Cross compilers do not use this macro and do not search either
        !           264: @file{/usr/include} or its replacement.
        !           265: 
        !           266: @findex INCLUDE_DEFAULTS
        !           267: @item INCLUDE_DEFAULTS
        !           268: Define this macro if you wish to override the entire default search path
        !           269: for include files.  The default search path includes
        !           270: @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
        !           271: @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
        !           272: @code{STANDARD_INCLUDE_DIR}.  In addition, @code{GPLUSPLUS_INCLUDE_DIR}
        !           273: and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
        !           274: and specify private search areas for GCC.  The directory
        !           275: @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
        !           276: 
        !           277: The definition should be an initializer for an array of structures.
        !           278: Each array element should have two elements: the directory name (a
        !           279: string constant) and a flag for C++-only directories.  Mark the end of
        !           280: the array with a null element.  For example, here is the definition used
        !           281: for VMS:
        !           282: 
        !           283: @example
        !           284: #define INCLUDE_DEFAULTS \
        !           285: @{                                       \
        !           286:   @{ "GNU_GXX_INCLUDE:", 1@},             \
        !           287:   @{ "GNU_CC_INCLUDE:", 0@},              \
        !           288:   @{ "SYS$SYSROOT:[SYSLIB.]", 0@},        \
        !           289:   @{ ".", 0@},                            \
        !           290:   @{ 0, 0@}                               \
        !           291: @}
        !           292: @end example
        !           293: @end table
        !           294: 
        !           295: Here is the order of prefixes tried for exec files:
        !           296: 
        !           297: @enumerate
        !           298: @item
        !           299: Any prefixes specified by the user with @samp{-B}.
        !           300: 
        !           301: @item
        !           302: The environment variable @code{GCC_EXEC_PREFIX}, if any.
        !           303: 
        !           304: @item
        !           305: The directories specified by the environment variable @code{COMPILER_PATH}.
        !           306: 
        !           307: @item
        !           308: The macro @code{STANDARD_EXEC_PREFIX}.
        !           309: 
        !           310: @item
        !           311: @file{/usr/lib/gcc/}.
        !           312: 
        !           313: @item
        !           314: The macro @code{MD_EXEC_PREFIX}, if any.
        !           315: @end enumerate
        !           316: 
        !           317: Here is the order of prefixes tried for startfiles:
        !           318: 
        !           319: @enumerate
        !           320: @item
        !           321: Any prefixes specified by the user with @samp{-B}.
        !           322: 
        !           323: @item
        !           324: The environment variable @code{GCC_EXEC_PREFIX}, if any.
        !           325: 
        !           326: @item
        !           327: The directories specified by the environment variable @code{LIBRARY_PATH}.
        !           328: 
        !           329: @item
        !           330: The macro @code{STANDARD_EXEC_PREFIX}.
        !           331: 
        !           332: @item
        !           333: @file{/usr/lib/gcc/}.
        !           334: 
        !           335: @item
        !           336: The macro @code{MD_EXEC_PREFIX}, if any.
        !           337: 
        !           338: @item
        !           339: The macro @code{MD_STARTFILE_PREFIX}, if any.
        !           340: 
        !           341: @item
        !           342: The macro @code{STANDARD_STARTFILE_PREFIX}.
        !           343: 
        !           344: @item
        !           345: @file{/lib/}.
        !           346: 
        !           347: @item
        !           348: @file{/usr/lib/}.
        !           349: @end enumerate
        !           350: 
        !           351: @node Run-time Target
        !           352: @section Run-time Target Specification
        !           353: @cindex run-time target specification
        !           354: @cindex predefined macros
        !           355: @cindex target specifications
        !           356: 
        !           357: @table @code
        !           358: @findex CPP_PREDEFINES
        !           359: @item CPP_PREDEFINES
        !           360: Define this to be a string constant containing @samp{-D} options to
        !           361: define the predefined macros that identify this machine and system.
        !           362: These macros will be predefined unless the @samp{-ansi} option is
        !           363: specified.
        !           364: 
        !           365: In addition, a parallel set of macros are predefined, whose names are
        !           366: made by appending @samp{__} at the beginning and at the end.  These
        !           367: @samp{__} macros are permitted by the ANSI standard, so they are
        !           368: predefined regardless of whether @samp{-ansi} is specified.
        !           369: 
        !           370: For example, on the Sun, one can use the following value:
        !           371: 
        !           372: @smallexample
        !           373: "-Dmc68000 -Dsun -Dunix"
        !           374: @end smallexample
        !           375: 
        !           376: The result is to define the macros @code{__mc68000__}, @code{__sun__}
        !           377: and @code{__unix__} unconditionally, and the macros @code{mc68000},
        !           378: @code{sun} and @code{unix} provided @samp{-ansi} is not specified.
        !           379: 
        !           380: @findex STDC_VALUE
        !           381: @item STDC_VALUE
        !           382: Define the value to be assigned to the built-in macro @code{__STDC__}.
        !           383: The default is the value @samp{1}.
        !           384: 
        !           385: @findex extern int target_flags
        !           386: @item extern int target_flags;
        !           387: This declaration should be present.
        !           388: 
        !           389: @cindex optional hardware or system features
        !           390: @cindex features, optional, in system conventions
        !           391: @item TARGET_@dots{}
        !           392: This series of macros is to allow compiler command arguments to
        !           393: enable or disable the use of optional features of the target machine.
        !           394: For example, one machine description serves both the 68000 and
        !           395: the 68020; a command argument tells the compiler whether it should
        !           396: use 68020-only instructions or not.  This command argument works
        !           397: by means of a macro @code{TARGET_68020} that tests a bit in
        !           398: @code{target_flags}.
        !           399: 
        !           400: Define a macro @code{TARGET_@var{featurename}} for each such option.
        !           401: Its definition should test a bit in @code{target_flags}; for example:
        !           402: 
        !           403: @smallexample
        !           404: #define TARGET_68020 (target_flags & 1)
        !           405: @end smallexample
        !           406: 
        !           407: One place where these macros are used is in the condition-expressions
        !           408: of instruction patterns.  Note how @code{TARGET_68020} appears
        !           409: frequently in the 68000 machine description file, @file{m68k.md}.
        !           410: Another place they are used is in the definitions of the other
        !           411: macros in the @file{@var{machine}.h} file.
        !           412: 
        !           413: @findex TARGET_SWITCHES
        !           414: @item TARGET_SWITCHES
        !           415: This macro defines names of command options to set and clear
        !           416: bits in @code{target_flags}.  Its definition is an initializer
        !           417: with a subgrouping for each command option.
        !           418: 
        !           419: Each subgrouping contains a string constant, that defines the option
        !           420: name, and a number, which contains the bits to set in
        !           421: @code{target_flags}.  A negative number says to clear bits instead;
        !           422: the negative of the number is which bits to clear.  The actual option
        !           423: name is made by appending @samp{-m} to the specified name.
        !           424: 
        !           425: One of the subgroupings should have a null string.  The number in
        !           426: this grouping is the default value for @code{target_flags}.  Any
        !           427: target options act starting with that value.
        !           428: 
        !           429: Here is an example which defines @samp{-m68000} and @samp{-m68020}
        !           430: with opposite meanings, and picks the latter as the default:
        !           431: 
        !           432: @smallexample
        !           433: #define TARGET_SWITCHES \
        !           434:   @{ @{ "68020", 1@},      \
        !           435:     @{ "68000", -1@},     \
        !           436:     @{ "", 1@}@}
        !           437: @end smallexample
        !           438: 
        !           439: @findex TARGET_OPTIONS
        !           440: @item TARGET_OPTIONS
        !           441: This macro is similar to @code{TARGET_SWITCHES} but defines names of command
        !           442: options that have values.  Its definition is an initializer with a
        !           443: subgrouping for each command option. 
        !           444: 
        !           445: Each subgrouping contains a string constant, that defines the fixed part
        !           446: of the option name, and the address of a variable.  The variable, type
        !           447: @code{char *}, is set to the variable part of the given option if the fixed
        !           448: part matches.  The actual option name is made by appending @samp{-m} to the
        !           449: specified name. 
        !           450: 
        !           451: Here is an example which defines @samp{-mshort-data-@var{number}}.  If the
        !           452: given option is @samp{-mshort-data-512}, the variable @code{m88k_short_data}
        !           453: will be set to the string @code{"512"}. 
        !           454: 
        !           455: @smallexample
        !           456: extern char *m88k_short_data;
        !           457: #define TARGET_OPTIONS \
        !           458:  @{ @{ "short-data-", &m88k_short_data @} @}
        !           459: @end smallexample
        !           460: 
        !           461: @findex TARGET_VERSION
        !           462: @item TARGET_VERSION
        !           463: This macro is a C statement to print on @code{stderr} a string
        !           464: describing the particular machine description choice.  Every machine
        !           465: description should define @code{TARGET_VERSION}.  For example:
        !           466: 
        !           467: @smallexample
        !           468: #ifdef MOTOROLA
        !           469: #define TARGET_VERSION \
        !           470:   fprintf (stderr, " (68k, Motorola syntax)");
        !           471: #else
        !           472: #define TARGET_VERSION \
        !           473:   fprintf (stderr, " (68k, MIT syntax)");
        !           474: #endif
        !           475: @end smallexample
        !           476: 
        !           477: @findex OVERRIDE_OPTIONS
        !           478: @item OVERRIDE_OPTIONS
        !           479: Sometimes certain combinations of command options do not make sense on
        !           480: a particular target machine.  You can define a macro
        !           481: @code{OVERRIDE_OPTIONS} to take account of this.  This macro, if
        !           482: defined, is executed once just after all the command options have been
        !           483: parsed.
        !           484: 
        !           485: Don't use this macro to turn on various extra optimizations for
        !           486: @samp{-O}.  That is what @code{OPTIMIZATION_OPTIONS} is for.
        !           487: 
        !           488: @findex OPTIMIZATION_OPTIONS
        !           489: @item OPTIMIZATION_OPTIONS (@var{level})
        !           490: Some machines may desire to change what optimizations are performed for
        !           491: various optimization levels.   This macro, if defined, is executed once
        !           492: just after the optimization level is determined and before the remainder
        !           493: of the command options have been parsed.  Values set in this macro are
        !           494: used as the default values for the other command line options.
        !           495: 
        !           496: @var{level} is the optimization level specified; 2 if -O2 is specified,
        !           497: 1 if -O is specified, and 0 if neither is specified.
        !           498: 
        !           499: @strong{Do not examine @code{write_symbols} in this macro!}
        !           500: The debugging options are not supposed to alter the generated code.
        !           501: @end table
        !           502: 
        !           503: @node Storage Layout
        !           504: @section Storage Layout
        !           505: @cindex storage layout
        !           506: 
        !           507: Note that the definitions of the macros in this table which are sizes or
        !           508: alignments measured in bits do not need to be constant.  They can be C
        !           509: expressions that refer to static variables, such as the @code{target_flags}.
        !           510: @xref{Run-time Target}.
        !           511: 
        !           512: @table @code
        !           513: @findex BITS_BIG_ENDIAN
        !           514: @item BITS_BIG_ENDIAN
        !           515: Define this macro to be the value 1 if the most significant bit in a
        !           516: byte has the lowest number; otherwise define it to be the value zero.
        !           517: This means that bit-field instructions count from the most significant
        !           518: bit.  If the machine has no bit-field instructions, then this must still
        !           519: be defined, but it doesn't matter which value it is defined to.
        !           520: 
        !           521: This macro does not affect the way structure fields are packed into
        !           522: bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
        !           523: 
        !           524: @findex BYTES_BIG_ENDIAN
        !           525: @item BYTES_BIG_ENDIAN
        !           526: Define this macro to be 1 if the most significant byte in a word has the
        !           527: lowest number.
        !           528: 
        !           529: @findex WORDS_BIG_ENDIAN
        !           530: @item WORDS_BIG_ENDIAN
        !           531: Define this macro to be 1 if, in a multiword object, the most
        !           532: significant word has the lowest number.  This applies to both memory
        !           533: locations and registers; GNU CC fundamentally assumes that the order of
        !           534: words in memory is the same as the order in registers.
        !           535: 
        !           536: @findex FLOAT_WORDS_BIG_ENDIAN
        !           537: @item FLOAT_WORDS_BIG_ENDIAN
        !           538: Define this macro to be 1 if @code{DFmode}, @code{XFmode} or
        !           539: @code{TFmode} floating point numbers are stored in memory with the word
        !           540: containing the sign bit at the lowest address; otherwise define it to be
        !           541: 0.
        !           542: 
        !           543: You need not define this macro if the ordering is the same as for
        !           544: multi-word integers.
        !           545: 
        !           546: @findex BITS_PER_UNIT
        !           547: @item BITS_PER_UNIT
        !           548: Define this macro to be the number of bits in an addressable storage
        !           549: unit (byte); normally 8.
        !           550: 
        !           551: @findex BITS_PER_WORD
        !           552: @item BITS_PER_WORD
        !           553: Number of bits in a word; normally 32.
        !           554: 
        !           555: @findex MAX_BITS_PER_WORD
        !           556: @item MAX_BITS_PER_WORD
        !           557: Maximum number of bits in a word.  If this is undefined, the default is
        !           558: @code{BITS_PER_WORD}.  Otherwise, it is the constant value that is the
        !           559: largest value that @code{BITS_PER_WORD} can have at run-time.
        !           560: 
        !           561: @findex UNITS_PER_WORD
        !           562: @item UNITS_PER_WORD
        !           563: Number of storage units in a word; normally 4.
        !           564: 
        !           565: @findex MAX_UNITS_PER_WORD
        !           566: @item MAX_UNITS_PER_WORD
        !           567: Maximum number of units in a word.  If this is undefined, the default is
        !           568: @code{UNITS_PER_WORD}.  Otherwise, it is the constant value that is the
        !           569: largest value that @code{UNITS_PER_WORD} can have at run-time.
        !           570: 
        !           571: @findex POINTER_SIZE
        !           572: @item POINTER_SIZE
        !           573: Width of a pointer, in bits.
        !           574: 
        !           575: @findex PROMOTE_MODE
        !           576: @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
        !           577: A macro to update @var{m} and @var{unsignedp} when an object whose type
        !           578: is @var{type} and which has the specified mode and signedness is to be
        !           579: stored in a register.  This macro is only called when @var{type} is a
        !           580: scalar type.
        !           581: 
        !           582: On most RISC machines, which only have operations that operate on a full
        !           583: register, define this macro to set @var{m} to @code{word_mode} if
        !           584: @var{m} is an integer mode narrower than @code{BITS_PER_WORD}.  In most
        !           585: cases, only integer modes should be widened because wider-precision
        !           586: floating-point operations are usually more expensive than their narrower
        !           587: counterparts.
        !           588: 
        !           589: For most machines, the macro definition does not change @var{unsignedp}.
        !           590: However, some machines, have instructions that preferentially handle
        !           591: either signed or unsigned quantities of certain modes.  For example, on
        !           592: the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
        !           593: sign-extend the result to 64 bits.  On such machines, set
        !           594: @var{unsignedp} according to which kind of extension is more efficient.
        !           595: 
        !           596: Do not define this macro if it would never modify @var{m}.
        !           597: 
        !           598: @findex PROMOTE_FUNCTION_ARGS
        !           599: @item PROMOTE_FUNCTION_ARGS
        !           600: Define this macro if the promotion described by @code{PROMOTE_MODE}
        !           601: should also be done for outgoing function arguments.  
        !           602: 
        !           603: @findex PROMOTE_FUNCTION_RETURN
        !           604: @item PROMOTE_FUNCTION_RETURN
        !           605: Define this macro if the promotion described by @code{PROMOTE_MODE}
        !           606: should also be done for the return value of functions.
        !           607: 
        !           608: If this macro is defined, @code{FUNCTION_VALUE} must perform the same
        !           609: promotions done by @code{PROMOTE_MODE}.
        !           610: 
        !           611: @findex PARM_BOUNDARY
        !           612: @item PARM_BOUNDARY
        !           613: Normal alignment required for function parameters on the stack, in
        !           614: bits.  All stack parameters receive at least this much alignment
        !           615: regardless of data type.  On most machines, this is the same as the
        !           616: size of an integer.
        !           617: 
        !           618: @findex STACK_BOUNDARY
        !           619: @item STACK_BOUNDARY
        !           620: Define this macro if you wish to preserve a certain alignment for
        !           621: the stack pointer.  The definition is a C expression
        !           622: for the desired alignment (measured in bits).
        !           623: 
        !           624: @cindex @code{PUSH_ROUNDING}, interaction with @code{STACK_BOUNDARY}
        !           625: If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
        !           626: to the specified boundary.  If @code{PUSH_ROUNDING} is defined and specifies a
        !           627: less strict alignment than @code{STACK_BOUNDARY}, the stack may be
        !           628: momentarily unaligned while pushing arguments.
        !           629: 
        !           630: @findex FUNCTION_BOUNDARY
        !           631: @item FUNCTION_BOUNDARY
        !           632: Alignment required for a function entry point, in bits.
        !           633: 
        !           634: @findex BIGGEST_ALIGNMENT
        !           635: @item BIGGEST_ALIGNMENT
        !           636: Biggest alignment that any data type can require on this machine, in bits.
        !           637: 
        !           638: @findex BIGGEST_FIELD_ALIGNMENT
        !           639: @item BIGGEST_FIELD_ALIGNMENT
        !           640: Biggest alignment that any structure field can require on this machine,
        !           641: in bits.  If defined, this overrides @code{BIGGEST_ALIGNMENT} for
        !           642: structure fields only.
        !           643: 
        !           644: @findex MAX_OFILE_ALIGNMENT
        !           645: @item MAX_OFILE_ALIGNMENT
        !           646: Biggest alignment supported by the object file format of this machine.
        !           647: Use this macro to limit the alignment which can be specified using the
        !           648: @code{__attribute__ ((aligned (@var{n})))} construct.  If not defined,
        !           649: the default value is @code{BIGGEST_ALIGNMENT}.
        !           650: 
        !           651: @findex DATA_ALIGNMENT
        !           652: @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
        !           653: If defined, a C expression to compute the alignment for a static
        !           654: variable.  @var{type} is the data type, and @var{basic-align} is the
        !           655: alignment that the object would ordinarily have.  The value of this
        !           656: macro is used instead of that alignment to align the object.
        !           657: 
        !           658: If this macro is not defined, then @var{basic-align} is used.
        !           659: 
        !           660: @findex strcpy
        !           661: One use of this macro is to increase alignment of medium-size data to
        !           662: make it all fit in fewer cache lines.  Another is to cause character
        !           663: arrays to be word-aligned so that @code{strcpy} calls that copy
        !           664: constants to character arrays can be done inline.
        !           665: 
        !           666: @findex CONSTANT_ALIGNMENT
        !           667: @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
        !           668: If defined, a C expression to compute the alignment given to a constant
        !           669: that is being placed in memory.  @var{constant} is the constant and
        !           670: @var{basic-align} is the alignment that the object would ordinarily
        !           671: have.  The value of this macro is used instead of that alignment to
        !           672: align the object.
        !           673: 
        !           674: If this macro is not defined, then @var{basic-align} is used.
        !           675: 
        !           676: The typical use of this macro is to increase alignment for string
        !           677: constants to be word aligned so that @code{strcpy} calls that copy
        !           678: constants can be done inline.
        !           679: 
        !           680: @findex EMPTY_FIELD_BOUNDARY
        !           681: @item EMPTY_FIELD_BOUNDARY
        !           682: Alignment in bits to be given to a structure bit field that follows an
        !           683: empty field such as @code{int : 0;}.
        !           684: 
        !           685: Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
        !           686: that results from an empty field.
        !           687: 
        !           688: @findex STRUCTURE_SIZE_BOUNDARY
        !           689: @item STRUCTURE_SIZE_BOUNDARY
        !           690: Number of bits which any structure or union's size must be a multiple of.
        !           691: Each structure or union's size is rounded up to a multiple of this.
        !           692: 
        !           693: If you do not define this macro, the default is the same as
        !           694: @code{BITS_PER_UNIT}.
        !           695: 
        !           696: @findex STRICT_ALIGNMENT
        !           697: @item STRICT_ALIGNMENT
        !           698: Define this macro to be the value 1 if instructions will fail to work
        !           699: if given data not on the nominal alignment.  If instructions will merely
        !           700: go slower in that case, define this macro as 0.
        !           701: 
        !           702: @findex PCC_BITFIELD_TYPE_MATTERS
        !           703: @item PCC_BITFIELD_TYPE_MATTERS
        !           704: Define this if you wish to imitate the way many other C compilers handle
        !           705: alignment of bitfields and the structures that contain them.
        !           706: 
        !           707: The behavior is that the type written for a bitfield (@code{int},
        !           708: @code{short}, or other integer type) imposes an alignment for the
        !           709: entire structure, as if the structure really did contain an ordinary
        !           710: field of that type.  In addition, the bitfield is placed within the
        !           711: structure so that it would fit within such a field, not crossing a
        !           712: boundary for it.
        !           713: 
        !           714: Thus, on most machines, a bitfield whose type is written as @code{int}
        !           715: would not cross a four-byte boundary, and would force four-byte
        !           716: alignment for the whole structure.  (The alignment used may not be four
        !           717: bytes; it is controlled by the other alignment parameters.)
        !           718: 
        !           719: If the macro is defined, its definition should be a C expression;
        !           720: a nonzero value for the expression enables this behavior.
        !           721: 
        !           722: Note that if this macro is not defined, or its value is zero, some
        !           723: bitfields may cross more than one alignment boundary.  The compiler can
        !           724: support such references if there are @samp{insv}, @samp{extv}, and
        !           725: @samp{extzv} insns that can directly reference memory.
        !           726: 
        !           727: The other known way of making bitfields work is to define
        !           728: @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
        !           729: Then every structure can be accessed with fullwords.
        !           730: 
        !           731: Unless the machine has bitfield instructions or you define
        !           732: @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
        !           733: @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
        !           734: 
        !           735: If your aim is to make GNU CC use the same conventions for laying out
        !           736: bitfields as are used by another compiler, here is how to investigate
        !           737: what the other compiler does.  Compile and run this program:
        !           738: 
        !           739: @example
        !           740: struct foo1
        !           741: @{
        !           742:   char x;
        !           743:   char :0;
        !           744:   char y;
        !           745: @};
        !           746: 
        !           747: struct foo2
        !           748: @{
        !           749:   char x;
        !           750:   int :0;
        !           751:   char y;
        !           752: @};
        !           753: 
        !           754: main ()
        !           755: @{
        !           756:   printf ("Size of foo1 is %d\n",
        !           757:           sizeof (struct foo1));
        !           758:   printf ("Size of foo2 is %d\n",
        !           759:           sizeof (struct foo2));
        !           760:   exit (0);
        !           761: @}
        !           762: @end example
        !           763: 
        !           764: If this prints 2 and 5, then the compiler's behavior is what you would
        !           765: get from @code{PCC_BITFIELD_TYPE_MATTERS}.
        !           766: 
        !           767: @findex BITFIELD_NBYTES_LIMITED
        !           768: @item BITFIELD_NBYTES_LIMITED
        !           769: Like PCC_BITFIELD_TYPE_MATTERS except that its effect is limited to
        !           770: aligning a bitfield within the structure.
        !           771: 
        !           772: @findex ROUND_TYPE_SIZE
        !           773: @item ROUND_TYPE_SIZE (@var{struct}, @var{size}, @var{align})
        !           774: Define this macro as an expression for the overall size of a structure 
        !           775: (given by @var{struct} as a tree node) when the size computed from the
        !           776: fields is @var{size} and the alignment is @var{align}.
        !           777: 
        !           778: The default is to round @var{size} up to a multiple of @var{align}.
        !           779: 
        !           780: @findex ROUND_TYPE_ALIGN
        !           781: @item ROUND_TYPE_ALIGN (@var{struct}, @var{computed}, @var{specified})
        !           782: Define this macro as an expression for the alignment of a structure 
        !           783: (given by @var{struct} as a tree node) if the alignment computed in the
        !           784: usual way is @var{computed} and the alignment explicitly specified was
        !           785: @var{specified}.
        !           786: 
        !           787: The default is to use @var{specified} if it is larger; otherwise, use
        !           788: the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
        !           789: 
        !           790: @findex MAX_FIXED_MODE_SIZE
        !           791: @item MAX_FIXED_MODE_SIZE
        !           792: An integer expression for the size in bits of the largest integer
        !           793: machine mode that should actually be used.  All integer machine modes of
        !           794: this size or smaller can be used for structures and unions with the
        !           795: appropriate sizes.  If this macro is undefined, @code{GET_MODE_BITSIZE
        !           796: (DImode)} is assumed.
        !           797: 
        !           798: @findex CHECK_FLOAT_VALUE
        !           799: @item CHECK_FLOAT_VALUE (@var{mode}, @var{value})
        !           800: A C statement to validate the value @var{value} (of type
        !           801: @code{double}) for mode @var{mode}.  This means that you check whether
        !           802: @var{value} fits within the possible range of values for mode
        !           803: @var{mode} on this target machine.  The mode @var{mode} is always
        !           804: @code{SFmode} or @code{DFmode}.
        !           805: 
        !           806: @findex error
        !           807: If @var{value} is not valid, you should call @code{error} to print an
        !           808: error message and then assign some valid value to @var{value}.
        !           809: Allowing an invalid value to go through the compiler can produce
        !           810: incorrect assembler code which may even cause Unix assemblers to
        !           811: crash.
        !           812: 
        !           813: This macro need not be defined if there is no work for it to do.
        !           814: 
        !           815: @findex TARGET_FLOAT_FORMAT
        !           816: @item TARGET_FLOAT_FORMAT
        !           817: A code distinguishing the floating point format of the target machine.
        !           818: There are three defined values:
        !           819: 
        !           820: @table @code
        !           821: @findex IEEE_FLOAT_FORMAT
        !           822: @item IEEE_FLOAT_FORMAT
        !           823: This code indicates IEEE floating point.  It is the default; there is no
        !           824: need to define this macro when the format is IEEE.
        !           825: 
        !           826: @findex VAX_FLOAT_FORMAT
        !           827: @item VAX_FLOAT_FORMAT
        !           828: This code indicates the peculiar format used on the Vax.
        !           829: 
        !           830: @findex UNKNOWN_FLOAT_FORMAT
        !           831: @item UNKNOWN_FLOAT_FORMAT
        !           832: This code indicates any other format.
        !           833: @end table
        !           834: 
        !           835: The value of this macro is compared with @code{HOST_FLOAT_FORMAT}
        !           836: (@pxref{Config}) to determine whether the target machine has the same
        !           837: format as the host machine.  If any other formats are actually in use on
        !           838: supported machines, new codes should be defined for them.
        !           839: 
        !           840: The ordering of the component words of floating point values stored in
        !           841: memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN} for the target
        !           842: machine and @code{HOST_FLOAT_WORDS_BIG_ENDIAN} for the host.
        !           843: @end table
        !           844: 
        !           845: @node Type Layout
        !           846: @section Layout of Source Language Data Types
        !           847: 
        !           848: These macros define the sizes and other characteristics of the standard
        !           849: basic data types used in programs being compiled.  Unlike the macros in
        !           850: the previous section, these apply to specific features of C and related
        !           851: languages, rather than to fundamental aspects of storage layout.
        !           852: 
        !           853: @table @code
        !           854: @findex INT_TYPE_SIZE
        !           855: @item INT_TYPE_SIZE
        !           856: A C expression for the size in bits of the type @code{int} on the
        !           857: target machine.  If you don't define this, the default is one word.
        !           858: 
        !           859: @findex MAX_INT_TYPE_SIZE
        !           860: @item MAX_INT_TYPE_SIZE
        !           861: Maximum number for the size in bits of the type @code{int} on the target
        !           862: machine.  If this is undefined, the default is @code{INT_TYPE_SIZE}.
        !           863: Otherwise, it is the constant value that is the largest value that
        !           864: @code{INT_TYPE_SIZE} can have at run-time.  This is used in @code{cpp}.
        !           865: 
        !           866: @findex SHORT_TYPE_SIZE
        !           867: @item SHORT_TYPE_SIZE
        !           868: A C expression for the size in bits of the type @code{short} on the
        !           869: target machine.  If you don't define this, the default is half a word.
        !           870: (If this would be less than one storage unit, it is rounded up to one
        !           871: unit.)
        !           872: 
        !           873: @findex LONG_TYPE_SIZE
        !           874: @item LONG_TYPE_SIZE
        !           875: A C expression for the size in bits of the type @code{long} on the
        !           876: target machine.  If you don't define this, the default is one word.
        !           877: 
        !           878: @findex MAX_LONG_TYPE_SIZE
        !           879: @item MAX_LONG_TYPE_SIZE
        !           880: Maximum number for the size in bits of the type @code{long} on the
        !           881: target machine.  If this is undefined, the default is
        !           882: @code{LONG_TYPE_SIZE}.  Otherwise, it is the constant value that is the
        !           883: largest value that @code{LONG_TYPE_SIZE} can have at run-time.  This is
        !           884: used in @code{cpp}.
        !           885: 
        !           886: @findex LONG_LONG_TYPE_SIZE
        !           887: @item LONG_LONG_TYPE_SIZE
        !           888: A C expression for the size in bits of the type @code{long long} on the
        !           889: target machine.  If you don't define this, the default is two
        !           890: words.
        !           891: 
        !           892: @findex CHAR_TYPE_SIZE
        !           893: @item CHAR_TYPE_SIZE
        !           894: A C expression for the size in bits of the type @code{char} on the
        !           895: target machine.  If you don't define this, the default is one quarter
        !           896: of a word.  (If this would be less than one storage unit, it is rounded up
        !           897: to one unit.)
        !           898: 
        !           899: @findex MAX_CHAR_TYPE_SIZE
        !           900: @item MAX_CHAR_TYPE_SIZE
        !           901: Maximum number for the size in bits of the type @code{char} on the
        !           902: target machine.  If this is undefined, the default is
        !           903: @code{CHAR_TYPE_SIZE}.  Otherwise, it is the constant value that is the
        !           904: largest value that @code{CHAR_TYPE_SIZE} can have at run-time.  This is
        !           905: used in @code{cpp}.
        !           906: 
        !           907: @findex FLOAT_TYPE_SIZE
        !           908: @item FLOAT_TYPE_SIZE
        !           909: A C expression for the size in bits of the type @code{float} on the
        !           910: target machine.  If you don't define this, the default is one word.
        !           911: 
        !           912: @findex DOUBLE_TYPE_SIZE
        !           913: @item DOUBLE_TYPE_SIZE
        !           914: A C expression for the size in bits of the type @code{double} on the
        !           915: target machine.  If you don't define this, the default is two
        !           916: words.
        !           917: 
        !           918: @findex LONG_DOUBLE_TYPE_SIZE
        !           919: @item LONG_DOUBLE_TYPE_SIZE
        !           920: A C expression for the size in bits of the type @code{long double} on
        !           921: the target machine.  If you don't define this, the default is two
        !           922: words.
        !           923: 
        !           924: @findex DEFAULT_SIGNED_CHAR
        !           925: @item DEFAULT_SIGNED_CHAR
        !           926: An expression whose value is 1 or 0, according to whether the type
        !           927: @code{char} should be signed or unsigned by default.  The user can
        !           928: always override this default with the options @samp{-fsigned-char}
        !           929: and @samp{-funsigned-char}.
        !           930: 
        !           931: @findex DEFAULT_SHORT_ENUMS
        !           932: @item DEFAULT_SHORT_ENUMS
        !           933: A C expression to determine whether to give an @code{enum} type 
        !           934: only as many bytes as it takes to represent the range of possible values
        !           935: of that type.  A nonzero value means to do that; a zero value means all
        !           936: @code{enum} types should be allocated like @code{int}.
        !           937: 
        !           938: If you don't define the macro, the default is 0.
        !           939: 
        !           940: @findex SIZE_TYPE
        !           941: @item SIZE_TYPE
        !           942: A C expression for a string describing the name of the data type to use
        !           943: for size values.  The typedef name @code{size_t} is defined using the
        !           944: contents of the string.
        !           945: 
        !           946: The string can contain more than one keyword.  If so, separate them with
        !           947: spaces, and write first any length keyword, then @code{unsigned} if
        !           948: appropriate, and finally @code{int}.  The string must exactly match one
        !           949: of the data type names defined in the function
        !           950: @code{init_decl_processing} in the file @file{c-decl.c}.  You may not
        !           951: omit @code{int} or change the order---that would cause the compiler to
        !           952: crash on startup.
        !           953: 
        !           954: If you don't define this macro, the default is @code{"long unsigned
        !           955: int"}.
        !           956: 
        !           957: @findex PTRDIFF_TYPE
        !           958: @item PTRDIFF_TYPE
        !           959: A C expression for a string describing the name of the data type to use
        !           960: for the result of subtracting two pointers.  The typedef name
        !           961: @code{ptrdiff_t} is defined using the contents of the string.  See
        !           962: @code{SIZE_TYPE} above for more information.
        !           963: 
        !           964: If you don't define this macro, the default is @code{"long int"}.
        !           965: 
        !           966: @findex WCHAR_TYPE
        !           967: @item WCHAR_TYPE
        !           968: A C expression for a string describing the name of the data type to use
        !           969: for wide characters.  The typedef name @code{wchar_t} is defined using
        !           970: the contents of the string.  See @code{SIZE_TYPE} above for more
        !           971: information.
        !           972: 
        !           973: If you don't define this macro, the default is @code{"int"}.
        !           974: 
        !           975: @findex WCHAR_TYPE_SIZE
        !           976: @item WCHAR_TYPE_SIZE
        !           977: A C expression for the size in bits of the data type for wide
        !           978: characters.  This is used in @code{cpp}, which cannot make use of
        !           979: @code{WCHAR_TYPE}.
        !           980: 
        !           981: @findex MAX_WCHAR_TYPE_SIZE
        !           982: @item MAX_WCHAR_TYPE_SIZE
        !           983: Maximum number for the size in bits of the data type for wide
        !           984: characters.  If this is undefined, the default is
        !           985: @code{WCHAR_TYPE_SIZE}.  Otherwise, it is the constant value that is the
        !           986: largest value that @code{WCHAR_TYPE_SIZE} can have at run-time.  This is
        !           987: used in @code{cpp}.
        !           988: 
        !           989: @findex OBJC_INT_SELECTORS
        !           990: @item OBJC_INT_SELECTORS
        !           991: Define this macro if the type of Objective C selectors should be
        !           992: @code{int}.
        !           993: 
        !           994: If this macro is not defined, then selectors should have the type
        !           995: @code{struct objc_selector *}.
        !           996: 
        !           997: @findex OBJC_SELECTORS_WITHOUT_LABELS
        !           998: @item OBJC_SELECTORS_WITHOUT_LABELS
        !           999: Define this macro if the compiler can group all the selectors together
        !          1000: into a vector and use just one label at the beginning of the vector.
        !          1001: Otherwise, the compiler must give each selector its own assembler
        !          1002: label.
        !          1003: 
        !          1004: On certain machines, it is important to have a separate label for each
        !          1005: selector because this enables the linker to eliminate duplicate selectors.
        !          1006: 
        !          1007: @findex TARGET_BELL
        !          1008: @item TARGET_BELL
        !          1009: A C constant expression for the integer value for escape sequence
        !          1010: @samp{\a}.
        !          1011: 
        !          1012: @findex TARGET_TAB
        !          1013: @findex TARGET_BS
        !          1014: @findex TARGET_NEWLINE
        !          1015: @item TARGET_BS
        !          1016: @itemx TARGET_TAB
        !          1017: @itemx TARGET_NEWLINE
        !          1018: C constant expressions for the integer values for escape sequences
        !          1019: @samp{\b}, @samp{\t} and @samp{\n}.
        !          1020: 
        !          1021: @findex TARGET_VT
        !          1022: @findex TARGET_FF
        !          1023: @findex TARGET_CR
        !          1024: @item TARGET_VT
        !          1025: @itemx TARGET_FF
        !          1026: @itemx TARGET_CR
        !          1027: C constant expressions for the integer values for escape sequences
        !          1028: @samp{\v}, @samp{\f} and @samp{\r}.
        !          1029: @end table
        !          1030: 
        !          1031: @node Registers
        !          1032: @section Register Usage
        !          1033: @cindex register usage
        !          1034: 
        !          1035: This section explains how to describe what registers the target machine
        !          1036: has, and how (in general) they can be used.
        !          1037: 
        !          1038: The description of which registers a specific instruction can use is
        !          1039: done with register classes; see @ref{Register Classes}.  For information
        !          1040: on using registers to access a stack frame, see @ref{Frame Registers}.
        !          1041: For passing values in registers, see @ref{Register Arguments}.
        !          1042: For returning values in registers, see @ref{Scalar Return}.
        !          1043: 
        !          1044: @menu
        !          1045: * Register Basics::            Number and kinds of registers.
        !          1046: * Allocation Order::           Order in which registers are allocated.
        !          1047: * Values in Registers::                What kinds of values each reg can hold.
        !          1048: * Leaf Functions::             Renumbering registers for leaf functions.
        !          1049: * Stack Registers::            Handling a register stack such as 80387.
        !          1050: * Obsolete Register Macros::   Macros formerly used for the 80387.
        !          1051: @end menu
        !          1052: 
        !          1053: @node Register Basics
        !          1054: @subsection Basic Characteristics of Registers
        !          1055: 
        !          1056: @table @code
        !          1057: @findex FIRST_PSEUDO_REGISTER
        !          1058: @item FIRST_PSEUDO_REGISTER
        !          1059: Number of hardware registers known to the compiler.  They receive
        !          1060: numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
        !          1061: pseudo register's number really is assigned the number
        !          1062: @code{FIRST_PSEUDO_REGISTER}.
        !          1063: 
        !          1064: @item FIXED_REGISTERS
        !          1065: @findex FIXED_REGISTERS
        !          1066: @cindex fixed register
        !          1067: An initializer that says which registers are used for fixed purposes
        !          1068: all throughout the compiled code and are therefore not available for
        !          1069: general allocation.  These would include the stack pointer, the frame
        !          1070: pointer (except on machines where that can be used as a general
        !          1071: register when no frame pointer is needed), the program counter on
        !          1072: machines where that is considered one of the addressable registers,
        !          1073: and any other numbered register with a standard use.
        !          1074: 
        !          1075: This information is expressed as a sequence of numbers, separated by
        !          1076: commas and surrounded by braces.  The @var{n}th number is 1 if
        !          1077: register @var{n} is fixed, 0 otherwise.
        !          1078: 
        !          1079: The table initialized from this macro, and the table initialized by
        !          1080: the following one, may be overridden at run time either automatically,
        !          1081: by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
        !          1082: the user with the command options @samp{-ffixed-@var{reg}},
        !          1083: @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}}.
        !          1084: 
        !          1085: @findex CALL_USED_REGISTERS
        !          1086: @item CALL_USED_REGISTERS
        !          1087: @cindex call-used register
        !          1088: @cindex call-clobbered register
        !          1089: @cindex call-saved register
        !          1090: Like @code{FIXED_REGISTERS} but has 1 for each register that is
        !          1091: clobbered (in general) by function calls as well as for fixed
        !          1092: registers.  This macro therefore identifies the registers that are not
        !          1093: available for general allocation of values that must live across
        !          1094: function calls.
        !          1095: 
        !          1096: If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
        !          1097: automatically saves it on function entry and restores it on function
        !          1098: exit, if the register is used within the function.
        !          1099: 
        !          1100: @findex CONDITIONAL_REGISTER_USAGE
        !          1101: @findex fixed_regs
        !          1102: @findex call_used_regs
        !          1103: @item CONDITIONAL_REGISTER_USAGE
        !          1104: Zero or more C statements that may conditionally modify two variables
        !          1105: @code{fixed_regs} and @code{call_used_regs} (both of type @code{char
        !          1106: []}) after they have been initialized from the two preceding macros.
        !          1107: 
        !          1108: This is necessary in case the fixed or call-clobbered registers depend
        !          1109: on target flags.
        !          1110: 
        !          1111: You need not define this macro if it has no work to do.
        !          1112: 
        !          1113: @cindex disabling certain registers
        !          1114: @cindex controlling register usage 
        !          1115: If the usage of an entire class of registers depends on the target
        !          1116: flags, you may indicate this to GCC by using this macro to modify
        !          1117: @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
        !          1118: registers in the classes which should not be used by GCC.  Also define
        !          1119: the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
        !          1120: is called with a letter for a class that shouldn't be used.
        !          1121: 
        !          1122: (However, if this class is not included in @code{GENERAL_REGS} and all
        !          1123: of the insn patterns whose constraints permit this class are
        !          1124: controlled by target switches, then GCC will automatically avoid using
        !          1125: these registers when the target switches are opposed to them.)
        !          1126: 
        !          1127: @findex NON_SAVING_SETJMP
        !          1128: @item NON_SAVING_SETJMP
        !          1129: If this macro is defined and has a nonzero value, it means that
        !          1130: @code{setjmp} and related functions fail to save the registers, or that
        !          1131: @code{longjmp} fails to restore them.  To compensate, the compiler
        !          1132: avoids putting variables in registers in functions that use
        !          1133: @code{setjmp}.
        !          1134: 
        !          1135: @findex INCOMING_REGNO
        !          1136: @item INCOMING_REGNO (@var{out})
        !          1137: Define this macro if the target machine has register windows.  This C
        !          1138: expression returns the register number as seen by the called function
        !          1139: corresponding to the register number @var{out} as seen by the calling
        !          1140: function.  Return @var{out} if register number @var{out} is not an
        !          1141: outbound register.
        !          1142: 
        !          1143: @findex OUTGOING_REGNO
        !          1144: @item OUTGOING_REGNO (@var{in})
        !          1145: Define this macro if the target machine has register windows.  This C
        !          1146: expression returns the register number as seen by the calling function
        !          1147: corresponding to the register number @var{in} as seen by the called
        !          1148: function.  Return @var{in} if register number @var{in} is not an inbound
        !          1149: register.
        !          1150: 
        !          1151: @ignore
        !          1152: @findex PC_REGNUM
        !          1153: @item PC_REGNUM
        !          1154: If the program counter has a register number, define this as that
        !          1155: register number.  Otherwise, do not define it.
        !          1156: @end ignore
        !          1157: @end table
        !          1158: 
        !          1159: @node Allocation Order
        !          1160: @subsection Order of Allocation of Registers
        !          1161: @cindex order of register allocation
        !          1162: @cindex register allocation order
        !          1163: 
        !          1164: @table @code
        !          1165: @findex REG_ALLOC_ORDER
        !          1166: @item REG_ALLOC_ORDER
        !          1167: If defined, an initializer for a vector of integers, containing the
        !          1168: numbers of hard registers in the order in which GNU CC should prefer
        !          1169: to use them (from most preferred to least).
        !          1170: 
        !          1171: If this macro is not defined, registers are used lowest numbered first
        !          1172: (all else being equal).
        !          1173: 
        !          1174: One use of this macro is on machines where the highest numbered
        !          1175: registers must always be saved and the save-multiple-registers
        !          1176: instruction supports only sequences of consecutive registers.  On such
        !          1177: machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
        !          1178: the highest numbered allocatable register first.
        !          1179: 
        !          1180: @findex ORDER_REGS_FOR_LOCAL_ALLOC
        !          1181: @item ORDER_REGS_FOR_LOCAL_ALLOC
        !          1182: A C statement (sans semicolon) to choose the order in which to allocate
        !          1183: hard registers for pseudo-registers local to a basic block.
        !          1184: 
        !          1185: Store the desired register order in the array @code{reg_alloc_order}.
        !          1186: Element 0 should be the register to allocate first; element 1, the next
        !          1187: register; and so on.
        !          1188: 
        !          1189: The macro body should not assume anything about the contents of
        !          1190: @code{reg_alloc_order} before execution of the macro.
        !          1191: 
        !          1192: On most machines, it is not necessary to define this macro.
        !          1193: @end table
        !          1194: 
        !          1195: @node Values in Registers
        !          1196: @subsection How Values Fit in Registers
        !          1197: 
        !          1198: This section discusses the macros that describe which kinds of values
        !          1199: (specifically, which machine modes) each register can hold, and how many
        !          1200: consecutive registers are needed for a given mode.
        !          1201: 
        !          1202: @table @code
        !          1203: @findex HARD_REGNO_NREGS
        !          1204: @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
        !          1205: A C expression for the number of consecutive hard registers, starting
        !          1206: at register number @var{regno}, required to hold a value of mode
        !          1207: @var{mode}.
        !          1208: 
        !          1209: On a machine where all registers are exactly one word, a suitable
        !          1210: definition of this macro is
        !          1211: 
        !          1212: @smallexample
        !          1213: #define HARD_REGNO_NREGS(REGNO, MODE)            \
        !          1214:    ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1)  \
        !          1215:     / UNITS_PER_WORD))
        !          1216: @end smallexample
        !          1217: 
        !          1218: @findex HARD_REGNO_MODE_OK
        !          1219: @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
        !          1220: A C expression that is nonzero if it is permissible to store a value
        !          1221: of mode @var{mode} in hard register number @var{regno} (or in several
        !          1222: registers starting with that one).  For a machine where all registers
        !          1223: are equivalent, a suitable definition is
        !          1224: 
        !          1225: @smallexample
        !          1226: #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
        !          1227: @end smallexample
        !          1228: 
        !          1229: It is not necessary for this macro to check for the numbers of fixed
        !          1230: registers, because the allocation mechanism considers them to be always
        !          1231: occupied.
        !          1232: 
        !          1233: @cindex register pairs
        !          1234: On some machines, double-precision values must be kept in even/odd
        !          1235: register pairs.  The way to implement that is to define this macro
        !          1236: to reject odd register numbers for such modes.
        !          1237: 
        !          1238: @ignore
        !          1239: @c I think this is not true now
        !          1240: GNU CC assumes that it can always move values between registers and
        !          1241: (suitably addressed) memory locations.  If it is impossible to move a
        !          1242: value of a certain mode between memory and certain registers, then
        !          1243: @code{HARD_REGNO_MODE_OK} must not allow this mode in those registers.
        !          1244: @end ignore
        !          1245: 
        !          1246: The minimum requirement for a mode to be OK in a register is that the
        !          1247: @samp{mov@var{mode}} instruction pattern support moves between the
        !          1248: register and any other hard register for which the mode is OK; and that
        !          1249: moving a value into the register and back out not alter it.
        !          1250: 
        !          1251: Since the same instruction used to move @code{SImode} will work for all
        !          1252: narrower integer modes, it is not necessary on any machine for
        !          1253: @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
        !          1254: you define patterns @samp{movhi}, etc., to take advantage of this.  This
        !          1255: is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
        !          1256: and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
        !          1257: to be tieable.
        !          1258: 
        !          1259: Many machines have special registers for floating point arithmetic.
        !          1260: Often people assume that floating point machine modes are allowed only
        !          1261: in floating point registers.  This is not true.  Any registers that
        !          1262: can hold integers can safely @emph{hold} a floating point machine
        !          1263: mode, whether or not floating arithmetic can be done on it in those
        !          1264: registers.  Integer move instructions can be used to move the values.
        !          1265: 
        !          1266: On some machines, though, the converse is true: fixed-point machine
        !          1267: modes may not go in floating registers.  This is true if the floating
        !          1268: registers normalize any value stored in them, because storing a
        !          1269: non-floating value there would garble it.  In this case,
        !          1270: @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
        !          1271: floating registers.  But if the floating registers do not automatically
        !          1272: normalize, if you can store any bit pattern in one and retrieve it
        !          1273: unchanged without a trap, then any machine mode may go in a floating
        !          1274: register, so you can define this macro to say so.
        !          1275: 
        !          1276: On some machines, such as the Sparc and the Mips, we get better code
        !          1277: by defining @code{HARD_REGNO_MODE_OK} to forbid integers in floating
        !          1278: registers, even though the hardware is capable of handling them.  This
        !          1279: is because transferring values between floating registers and general
        !          1280: registers is so slow that it is better to keep the integer in memory.
        !          1281: 
        !          1282: The primary significance of special floating registers is rather that
        !          1283: they are the registers acceptable in floating point arithmetic
        !          1284: instructions.  However, this is of no concern to
        !          1285: @code{HARD_REGNO_MODE_OK}.  You handle it by writing the proper
        !          1286: constraints for those instructions.
        !          1287: 
        !          1288: On some machines, the floating registers are especially slow to access,
        !          1289: so that it is better to store a value in a stack frame than in such a
        !          1290: register if floating point arithmetic is not being done.  As long as the
        !          1291: floating registers are not in class @code{GENERAL_REGS}, they will not
        !          1292: be used unless some pattern's constraint asks for one.
        !          1293: 
        !          1294: @findex MODES_TIEABLE_P
        !          1295: @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
        !          1296: A C expression that is nonzero if it is desirable to choose register
        !          1297: allocation so as to avoid move instructions between a value of mode
        !          1298: @var{mode1} and a value of mode @var{mode2}.
        !          1299: 
        !          1300: If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
        !          1301: @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are ever different
        !          1302: for any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1},
        !          1303: @var{mode2})} must be zero.
        !          1304: @end table
        !          1305: 
        !          1306: @node Leaf Functions
        !          1307: @subsection Handling Leaf Functions
        !          1308: 
        !          1309: @cindex leaf functions
        !          1310: @cindex functions, leaf
        !          1311: On some machines, a leaf function (i.e., one which makes no calls) can run
        !          1312: more efficiently if it does not make its own register window.  Often this
        !          1313: means it is required to receive its arguments in the registers where they
        !          1314: are passed by the caller, instead of the registers where they would
        !          1315: normally arrive.
        !          1316: 
        !          1317: The special treatment for leaf functions generally applies only when
        !          1318: other conditions are met; for example, often they may use only those
        !          1319: registers for its own variables and temporaries.  We use the term ``leaf
        !          1320: function'' to mean a function that is suitable for this special
        !          1321: handling, so that functions with no calls are not necessarily ``leaf
        !          1322: functions''.
        !          1323: 
        !          1324: GNU CC assigns register numbers before it knows whether the function is
        !          1325: suitable for leaf function treatment.  So it needs to renumber the
        !          1326: registers in order to output a leaf function.  The following macros
        !          1327: accomplish this.
        !          1328: 
        !          1329: @table @code
        !          1330: @findex LEAF_REGISTERS
        !          1331: @item LEAF_REGISTERS
        !          1332: A C initializer for a vector, indexed by hard register number, which
        !          1333: contains 1 for a register that is allowable in a candidate for leaf
        !          1334: function treatment.
        !          1335: 
        !          1336: If leaf function treatment involves renumbering the registers, then the
        !          1337: registers marked here should be the ones before renumbering---those that
        !          1338: GNU CC would ordinarily allocate.  The registers which will actually be
        !          1339: used in the assembler code, after renumbering, should not be marked with 1
        !          1340: in this vector.
        !          1341: 
        !          1342: Define this macro only if the target machine offers a way to optimize
        !          1343: the treatment of leaf functions.
        !          1344: 
        !          1345: @findex LEAF_REG_REMAP
        !          1346: @item LEAF_REG_REMAP (@var{regno})
        !          1347: A C expression whose value is the register number to which @var{regno}
        !          1348: should be renumbered, when a function is treated as a leaf function.
        !          1349: 
        !          1350: If @var{regno} is a register number which should not appear in a leaf
        !          1351: function before renumbering, then the expression should yield -1, which
        !          1352: will cause the compiler to abort.
        !          1353: 
        !          1354: Define this macro only if the target machine offers a way to optimize the
        !          1355: treatment of leaf functions, and registers need to be renumbered to do
        !          1356: this.
        !          1357: 
        !          1358: @findex REG_LEAF_ALLOC_ORDER
        !          1359: @item REG_LEAF_ALLOC_ORDER
        !          1360: If defined, an initializer for a vector of integers, containing the
        !          1361: numbers of hard registers in the order in which the GNU CC should prefer
        !          1362: to use them (from most preferred to least) in a leaf function.  If this
        !          1363: macro is not defined, REG_ALLOC_ORDER is used for both non-leaf and
        !          1364: leaf-functions.
        !          1365: @c *important*.. is the above all-caps phrase supposed tobe in code
        !          1366: @c font? --mew 2feb93
        !          1367: @end table
        !          1368: 
        !          1369: @findex leaf_function
        !          1370: Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
        !          1371: treat leaf functions specially.  It can test the C variable
        !          1372: @code{leaf_function} which is nonzero for leaf functions.  (The variable
        !          1373: @code{leaf_function} is defined only if @code{LEAF_REGISTERS} is
        !          1374: defined.)
        !          1375: @c changed this to fix overfull.  ALSO:  why the "it" at the beginning
        !          1376: @c of the next paragraph?!  --mew 2feb93 
        !          1377: 
        !          1378: @node Stack Registers
        !          1379: @subsection Registers That Form a Stack
        !          1380: 
        !          1381: There are special features to handle computers where some of the
        !          1382: ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
        !          1383: Stack registers are normally written by pushing onto the stack, and are
        !          1384: numbered relative to the top of the stack.
        !          1385: 
        !          1386: Currently, GNU CC can only handle one group of stack-like registers, and
        !          1387: they must be consecutively numbered.
        !          1388: 
        !          1389: @table @code
        !          1390: @findex STACK_REGS
        !          1391: @item STACK_REGS
        !          1392: Define this if the machine has any stack-like registers.
        !          1393: 
        !          1394: @findex FIRST_STACK_REG
        !          1395: @item FIRST_STACK_REG
        !          1396: The number of the first stack-like register.  This one is the top
        !          1397: of the stack.
        !          1398: 
        !          1399: @findex LAST_STACK_REG
        !          1400: @item LAST_STACK_REG
        !          1401: The number of the last stack-like register.  This one is the bottom of
        !          1402: the stack.
        !          1403: @end table
        !          1404: 
        !          1405: @node Obsolete Register Macros
        !          1406: @subsection Obsolete Macros for Controlling Register Usage
        !          1407: 
        !          1408: These features do not work very well.  They exist because they used to
        !          1409: be required to generate correct code for the 80387 coprocessor of the
        !          1410: 80386.  They are no longer used by that machine description and may be
        !          1411: removed in a later version of the compiler.  Don't use them!
        !          1412: 
        !          1413: @table @code
        !          1414: @findex OVERLAPPING_REGNO_P 
        !          1415: @item OVERLAPPING_REGNO_P (@var{regno})
        !          1416: If defined, this is a C expression whose value is nonzero if hard
        !          1417: register number @var{regno} is an overlapping register.  This means a
        !          1418: hard register which overlaps a hard register with a different number.
        !          1419: (Such overlap is undesirable, but occasionally it allows a machine to
        !          1420: be supported which otherwise could not be.)  This macro must return
        !          1421: nonzero for @emph{all} the registers which overlap each other.  GNU CC
        !          1422: can use an overlapping register only in certain limited ways.  It can
        !          1423: be used for allocation within a basic block, and may be spilled for
        !          1424: reloading; that is all.
        !          1425: 
        !          1426: If this macro is not defined, it means that none of the hard registers
        !          1427: overlap each other.  This is the usual situation.
        !          1428: 
        !          1429: @findex INSN_CLOBBERS_REGNO_P
        !          1430: @item INSN_CLOBBERS_REGNO_P (@var{insn}, @var{regno})
        !          1431: If defined, this is a C expression whose value should be nonzero if
        !          1432: the insn @var{insn} has the effect of mysteriously clobbering the
        !          1433: contents of hard register number @var{regno}.  By ``mysterious'' we
        !          1434: mean that the insn's RTL expression doesn't describe such an effect.
        !          1435: 
        !          1436: If this macro is not defined, it means that no insn clobbers registers
        !          1437: mysteriously.  This is the usual situation; all else being equal,
        !          1438: it is best for the RTL expression to show all the activity.
        !          1439: 
        !          1440: @cindex death notes
        !          1441: @findex PRESERVE_DEATH_INFO_REGNO_P
        !          1442: @item PRESERVE_DEATH_INFO_REGNO_P (@var{regno})
        !          1443: If defined, this is a C expression whose value is nonzero if accurate
        !          1444: @code{REG_DEAD} notes are needed for hard register number @var{regno}
        !          1445: at the time of outputting the assembler code.  When this is so, a few
        !          1446: optimizations that take place after register allocation and could
        !          1447: invalidate the death notes are not done when this register is
        !          1448: involved.
        !          1449: 
        !          1450: You would arrange to preserve death info for a register when some of the
        !          1451: code in the machine description which is executed to write the assembler
        !          1452: code looks at the death notes.  This is necessary only when the actual
        !          1453: hardware feature which GNU CC thinks of as a register is not actually a
        !          1454: register of the usual sort.  (It might, for example, be a hardware
        !          1455: stack.)
        !          1456: 
        !          1457: If this macro is not defined, it means that no death notes need to be
        !          1458: preserved.  This is the usual situation.
        !          1459: @end table
        !          1460: 
        !          1461: @node Register Classes
        !          1462: @section Register Classes
        !          1463: @cindex register class definitions
        !          1464: @cindex class definitions, register
        !          1465: 
        !          1466: On many machines, the numbered registers are not all equivalent.
        !          1467: For example, certain registers may not be allowed for indexed addressing;
        !          1468: certain registers may not be allowed in some instructions.  These machine
        !          1469: restrictions are described to the compiler using @dfn{register classes}.
        !          1470: 
        !          1471: You define a number of register classes, giving each one a name and saying
        !          1472: which of the registers belong to it.  Then you can specify register classes
        !          1473: that are allowed as operands to particular instruction patterns.
        !          1474: 
        !          1475: @findex ALL_REGS
        !          1476: @findex NO_REGS
        !          1477: In general, each register will belong to several classes.  In fact, one
        !          1478: class must be named @code{ALL_REGS} and contain all the registers.  Another
        !          1479: class must be named @code{NO_REGS} and contain no registers.  Often the
        !          1480: union of two classes will be another class; however, this is not required.
        !          1481: 
        !          1482: @findex GENERAL_REGS
        !          1483: One of the classes must be named @code{GENERAL_REGS}.  There is nothing
        !          1484: terribly special about the name, but the operand constraint letters
        !          1485: @samp{r} and @samp{g} specify this class.  If @code{GENERAL_REGS} is
        !          1486: the same as @code{ALL_REGS}, just define it as a macro which expands
        !          1487: to @code{ALL_REGS}.
        !          1488: 
        !          1489: Order the classes so that if class @var{x} is contained in class @var{y}
        !          1490: then @var{x} has a lower class number than @var{y}.
        !          1491: 
        !          1492: The way classes other than @code{GENERAL_REGS} are specified in operand
        !          1493: constraints is through machine-dependent operand constraint letters.
        !          1494: You can define such letters to correspond to various classes, then use
        !          1495: them in operand constraints.
        !          1496: 
        !          1497: You should define a class for the union of two classes whenever some
        !          1498: instruction allows both classes.  For example, if an instruction allows
        !          1499: either a floating point (coprocessor) register or a general register for a
        !          1500: certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
        !          1501: which includes both of them.  Otherwise you will get suboptimal code.
        !          1502: 
        !          1503: You must also specify certain redundant information about the register
        !          1504: classes: for each class, which classes contain it and which ones are
        !          1505: contained in it; for each pair of classes, the largest class contained
        !          1506: in their union.
        !          1507: 
        !          1508: When a value occupying several consecutive registers is expected in a
        !          1509: certain class, all the registers used must belong to that class.
        !          1510: Therefore, register classes cannot be used to enforce a requirement for
        !          1511: a register pair to start with an even-numbered register.  The way to
        !          1512: specify this requirement is with @code{HARD_REGNO_MODE_OK}.
        !          1513: 
        !          1514: Register classes used for input-operands of bitwise-and or shift
        !          1515: instructions have a special requirement: each such class must have, for
        !          1516: each fixed-point machine mode, a subclass whose registers can transfer that
        !          1517: mode to or from memory.  For example, on some machines, the operations for
        !          1518: single-byte values (@code{QImode}) are limited to certain registers.  When
        !          1519: this is so, each register class that is used in a bitwise-and or shift
        !          1520: instruction must have a subclass consisting of registers from which
        !          1521: single-byte values can be loaded or stored.  This is so that
        !          1522: @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
        !          1523: 
        !          1524: @table @code
        !          1525: @findex enum reg_class
        !          1526: @item enum reg_class
        !          1527: An enumeral type that must be defined with all the register class names
        !          1528: as enumeral values.  @code{NO_REGS} must be first.  @code{ALL_REGS}
        !          1529: must be the last register class, followed by one more enumeral value,
        !          1530: @code{LIM_REG_CLASSES}, which is not a register class but rather
        !          1531: tells how many classes there are.
        !          1532: 
        !          1533: Each register class has a number, which is the value of casting
        !          1534: the class name to type @code{int}.  The number serves as an index
        !          1535: in many of the tables described below.
        !          1536: 
        !          1537: @findex N_REG_CLASSES
        !          1538: @item N_REG_CLASSES
        !          1539: The number of distinct register classes, defined as follows:
        !          1540: 
        !          1541: @example
        !          1542: #define N_REG_CLASSES (int) LIM_REG_CLASSES
        !          1543: @end example
        !          1544: 
        !          1545: @findex REG_CLASS_NAMES
        !          1546: @item REG_CLASS_NAMES
        !          1547: An initializer containing the names of the register classes as C string
        !          1548: constants.  These names are used in writing some of the debugging dumps.
        !          1549: 
        !          1550: @findex REG_CLASS_CONTENTS
        !          1551: @item REG_CLASS_CONTENTS
        !          1552: An initializer containing the contents of the register classes, as integers
        !          1553: which are bit masks.  The @var{n}th integer specifies the contents of class
        !          1554: @var{n}.  The way the integer @var{mask} is interpreted is that
        !          1555: register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
        !          1556: 
        !          1557: When the machine has more than 32 registers, an integer does not suffice.
        !          1558: Then the integers are replaced by sub-initializers, braced groupings containing
        !          1559: several integers.  Each sub-initializer must be suitable as an initializer
        !          1560: for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
        !          1561: 
        !          1562: @findex REGNO_REG_CLASS 
        !          1563: @item REGNO_REG_CLASS (@var{regno})
        !          1564: A C expression whose value is a register class containing hard register
        !          1565: @var{regno}.  In general there is more than one such class; choose a class
        !          1566: which is @dfn{minimal}, meaning that no smaller class also contains the
        !          1567: register.
        !          1568: 
        !          1569: @findex BASE_REG_CLASS
        !          1570: @item BASE_REG_CLASS
        !          1571: A macro whose definition is the name of the class to which a valid
        !          1572: base register must belong.  A base register is one used in an address
        !          1573: which is the register value plus a displacement.
        !          1574: 
        !          1575: @findex INDEX_REG_CLASS
        !          1576: @item INDEX_REG_CLASS
        !          1577: A macro whose definition is the name of the class to which a valid
        !          1578: index register must belong.  An index register is one used in an
        !          1579: address where its value is either multiplied by a scale factor or
        !          1580: added to another register (as well as added to a displacement).
        !          1581: 
        !          1582: @findex REG_CLASS_FROM_LETTER
        !          1583: @item REG_CLASS_FROM_LETTER (@var{char})
        !          1584: A C expression which defines the machine-dependent operand constraint
        !          1585: letters for register classes.  If @var{char} is such a letter, the
        !          1586: value should be the register class corresponding to it.  Otherwise,
        !          1587: the value should be @code{NO_REGS}.  The register letter @samp{r},
        !          1588: corresponding to class @code{GENERAL_REGS}, will not be passed
        !          1589: to this macro; you do not need to handle it.
        !          1590: 
        !          1591: @findex REGNO_OK_FOR_BASE_P
        !          1592: @item REGNO_OK_FOR_BASE_P (@var{num})
        !          1593: A C expression which is nonzero if register number @var{num} is
        !          1594: suitable for use as a base register in operand addresses.  It may be
        !          1595: either a suitable hard register or a pseudo register that has been
        !          1596: allocated such a hard register.
        !          1597: 
        !          1598: @findex REGNO_OK_FOR_INDEX_P
        !          1599: @item REGNO_OK_FOR_INDEX_P (@var{num})
        !          1600: A C expression which is nonzero if register number @var{num} is
        !          1601: suitable for use as an index register in operand addresses.  It may be
        !          1602: either a suitable hard register or a pseudo register that has been
        !          1603: allocated such a hard register.
        !          1604: 
        !          1605: The difference between an index register and a base register is that
        !          1606: the index register may be scaled.  If an address involves the sum of
        !          1607: two registers, neither one of them scaled, then either one may be
        !          1608: labeled the ``base'' and the other the ``index''; but whichever
        !          1609: labeling is used must fit the machine's constraints of which registers
        !          1610: may serve in each capacity.  The compiler will try both labelings,
        !          1611: looking for one that is valid, and will reload one or both registers
        !          1612: only if neither labeling works.
        !          1613: 
        !          1614: @findex PREFERRED_RELOAD_CLASS
        !          1615: @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
        !          1616: A C expression that places additional restrictions on the register class
        !          1617: to use when it is necessary to copy value @var{x} into a register in class
        !          1618: @var{class}.  The value is a register class; perhaps @var{class}, or perhaps
        !          1619: another, smaller class.  On many machines, the following definition is
        !          1620: safe: 
        !          1621: 
        !          1622: @example
        !          1623: #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
        !          1624: @end example
        !          1625: 
        !          1626: Sometimes returning a more restrictive class makes better code.  For
        !          1627: example, on the 68000, when @var{x} is an integer constant that is in range
        !          1628: for a @samp{moveq} instruction, the value of this macro is always
        !          1629: @code{DATA_REGS} as long as @var{class} includes the data registers.
        !          1630: Requiring a data register guarantees that a @samp{moveq} will be used.
        !          1631: 
        !          1632: If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
        !          1633: you can force @var{x} into a memory constant.  This is useful on
        !          1634: certain machines where immediate floating values cannot be loaded into
        !          1635: certain kinds of registers.
        !          1636: 
        !          1637: @findex PREFERRED_OUTPUT_RELOAD_CLASS
        !          1638: @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
        !          1639: Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
        !          1640: input reloads.  If you don't define this macro, the default is to use
        !          1641: @var{class}, unchanged.
        !          1642: 
        !          1643: @findex LIMIT_RELOAD_CLASS
        !          1644: @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
        !          1645: A C expression that places additional restrictions on the register class
        !          1646: to use when it is necessary to be able to hold a value of mode
        !          1647: @var{mode} in a reload register for which class @var{class} would
        !          1648: ordinarily be used.
        !          1649: 
        !          1650: Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
        !          1651: there are certain modes that simply can't go in certain reload classes.
        !          1652: 
        !          1653: The value is a register class; perhaps @var{class}, or perhaps another,
        !          1654: smaller class.
        !          1655: 
        !          1656: Don't define this macro unless the target machine has limitations which
        !          1657: require the macro to do something nontrivial.
        !          1658: 
        !          1659: @findex SECONDARY_RELOAD_CLASS
        !          1660: @findex SECONDARY_INPUT_RELOAD_CLASS
        !          1661: @findex SECONDARY_OUTPUT_RELOAD_CLASS
        !          1662: @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
        !          1663: @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
        !          1664: @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
        !          1665: Many machines have some registers that cannot be copied directly to or
        !          1666: from memory or even from other types of registers.  An example is the
        !          1667: @samp{MQ} register, which on most machines, can only be copied to or
        !          1668: from general registers, but not memory.  Some machines allow copying all
        !          1669: registers to and from memory, but require a scratch register for stores
        !          1670: to some memory locations (e.g., those with symbolic address on the RT,
        !          1671: and those with certain symbolic address on the Sparc when compiling
        !          1672: PIC).  In some cases, both an intermediate and a scratch register are
        !          1673: required.
        !          1674: 
        !          1675: You should define these macros to indicate to the reload phase that it may
        !          1676: need to allocate at least one register for a reload in addition to the
        !          1677: register to contain the data.  Specifically, if copying @var{x} to a
        !          1678: register @var{class} in @var{mode} requires an intermediate register,
        !          1679: you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
        !          1680: largest register class all of whose registers can be used as
        !          1681: intermediate registers or scratch registers.
        !          1682: 
        !          1683: If copying a register @var{class} in @var{mode} to @var{x} requires an
        !          1684: intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
        !          1685: should be defined to return the largest register class required.  If the
        !          1686: requirements for input and output reloads are the same, the macro
        !          1687: @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
        !          1688: macros identically.
        !          1689: 
        !          1690: The values returned by these macros are often @code{GENERAL_REGS}.
        !          1691: Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
        !          1692: can be directly copied to or from a register of @var{class} in
        !          1693: @var{mode} without requiring a scratch register.  Do not define this
        !          1694: macro if it would always return @code{NO_REGS}.
        !          1695: 
        !          1696: If a scratch register is required (either with or without an
        !          1697: intermediate register), you should define patterns for
        !          1698: @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
        !          1699: (@pxref{Standard Names}.  These patterns, which will normally be
        !          1700: implemented with a @code{define_expand}, should be similar to the
        !          1701: @samp{mov@var{m}} patterns, except that operand 2 is the scratch
        !          1702: register. 
        !          1703: 
        !          1704: Define constraints for the reload register and scratch register that
        !          1705: contain a single register class.  If the original reload register (whose
        !          1706: class is @var{class}) can meet the constraint given in the pattern, the
        !          1707: value returned by these macros is used for the class of the scratch
        !          1708: register.  Otherwise, two additional reload registers are required.
        !          1709: Their classes are obtained from the constraints in the insn pattern.
        !          1710: 
        !          1711: @var{x} might be a pseudo-register or a @code{subreg} of a
        !          1712: pseudo-register, which could either be in a hard register or in memory.
        !          1713: Use @code{true_regnum} to find out; it will return -1 if the pseudo is
        !          1714: in memory and the hard register number if it is in a register.
        !          1715: 
        !          1716: These macros should not be used in the case where a particular class of
        !          1717: registers can only be copied to memory and not to another class of
        !          1718: registers.  In that case, secondary reload registers are not needed and
        !          1719: would not be helpful.  Instead, a stack location must be used to perform
        !          1720: the copy and the @code{mov@var{m}} pattern should use memory as a
        !          1721: intermediate storage.  This case often occurs between floating-point and
        !          1722: general registers.
        !          1723: 
        !          1724: @findex SECONDARY_MEMORY_NEEDED
        !          1725: @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
        !          1726: Certain machines have the property that some registers cannot be copied
        !          1727: to some other registers without using memory.  Define this macro on
        !          1728: those machines to be a C expression that is non-zero if objects of mode
        !          1729: @var{m} in registers of @var{class1} can only be copied to registers of
        !          1730: class @var{class2} by storing a register of @var{class1} into memory
        !          1731: and loading that memory location into a register of @var{class2}.
        !          1732: 
        !          1733: Do not define this macro if its value would always be zero. 
        !          1734: 
        !          1735: @findex SECONDARY_MEMORY_NEEDED_RTX
        !          1736: @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
        !          1737: Normally, when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
        !          1738: will allocate a stack slot when a memory location for a register copy
        !          1739: is needed.  If this macro is defined, the compiler instead uses the
        !          1740: memory location defined by this macro.
        !          1741: 
        !          1742: @findex SMALL_REGISTER_CLASSES
        !          1743: @item SMALL_REGISTER_CLASSES
        !          1744: Normally the compiler will avoid choosing spill registers from registers
        !          1745: that have been explicitly mentioned in the rtl (these registers are
        !          1746: normally those used to pass parameters and return values).  However,
        !          1747: some machines have so few registers of certain classes that there would
        !          1748: not be enough registers to use as spill registers if this were done.
        !          1749: 
        !          1750: You should define @code{SMALL_REGISTER_CLASSES} on those machines.  When
        !          1751: it is defined, the compiler allows registers explicitly used in the rtl
        !          1752: to be used as spill registers but prevents the compiler from extending
        !          1753: the lifetime of these registers.
        !          1754: 
        !          1755: Defining this macro is always safe, but unnecessarily defining this macro
        !          1756: will reduce the amount of optimizations that can be performed in some
        !          1757: cases.  If this macro is not defined but needs to be, the compiler will
        !          1758: run out of reload registers and print a fatal error message.
        !          1759: 
        !          1760: For most machines, this macro should not be defined.
        !          1761: 
        !          1762: @findex CLASS_LIKELY_SPILLED_P
        !          1763: @item CLASS_LIKELY_SPILLED_P (@var{class})
        !          1764: A C expression whose value is nonzero if pseudos that have been assigned
        !          1765: to registers of class @var{class} would likely be spilled because
        !          1766: registers of @var{class} are needed for spill registers.
        !          1767: 
        !          1768: The default value of this macro returns 1 if @var{class} has exactly one
        !          1769: register and zero otherwise.  On most machines, this default should be
        !          1770: used.  Only define this macro to some other expression if pseudo
        !          1771: allocated by @file{local-alloc.c} end up in memory because their hard
        !          1772: registers were needed for spill regisers.  If this macro returns nonzero
        !          1773: for those classes, those pseudos will only be allocated by
        !          1774: @file{global.c}, which knows how to reallocate the pseudo to another
        !          1775: register.  If there would not be another register available for
        !          1776: reallocation, you should not change the definition of this macro since
        !          1777: the only effect of such a definition would be to slow down register
        !          1778: allocation.
        !          1779: 
        !          1780: @findex CLASS_MAX_NREGS
        !          1781: @item CLASS_MAX_NREGS (@var{class}, @var{mode})
        !          1782: A C expression for the maximum number of consecutive registers
        !          1783: of class @var{class} needed to hold a value of mode @var{mode}.
        !          1784: 
        !          1785: This is closely related to the macro @code{HARD_REGNO_NREGS}.
        !          1786: In fact, the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
        !          1787: should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno}, @var{mode})}
        !          1788: for all @var{regno} values in the class @var{class}.
        !          1789: 
        !          1790: This macro helps control the handling of multiple-word values
        !          1791: in the reload pass.
        !          1792: @end table
        !          1793: 
        !          1794: Three other special macros describe which operands fit which constraint
        !          1795: letters.
        !          1796: 
        !          1797: @table @code
        !          1798: @findex CONST_OK_FOR_LETTER_P
        !          1799: @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
        !          1800: A C expression that defines the machine-dependent operand constraint letters
        !          1801: that specify particular ranges of integer values.  If @var{c} is one
        !          1802: of those letters, the expression should check that @var{value}, an integer,
        !          1803: is in the appropriate range and return 1 if so, 0 otherwise.  If @var{c} is
        !          1804: not one of those letters, the value should be 0 regardless of @var{value}.
        !          1805: 
        !          1806: @findex CONST_DOUBLE_OK_FOR_LETTER_P
        !          1807: @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
        !          1808: A C expression that defines the machine-dependent operand constraint
        !          1809: letters that specify particular ranges of @code{const_double} values.
        !          1810: 
        !          1811: If @var{c} is one of those letters, the expression should check that
        !          1812: @var{value}, an RTX of code @code{const_double}, is in the appropriate
        !          1813: range and return 1 if so, 0 otherwise.  If @var{c} is not one of those
        !          1814: letters, the value should be 0 regardless of @var{value}.
        !          1815: 
        !          1816: @code{const_double} is used for all floating-point constants and for
        !          1817: @code{DImode} fixed-point constants.  A given letter can accept either
        !          1818: or both kinds of values.  It can use @code{GET_MODE} to distinguish
        !          1819: between these kinds.
        !          1820: 
        !          1821: @findex EXTRA_CONSTRAINT
        !          1822: @item EXTRA_CONSTRAINT (@var{value}, @var{c})
        !          1823: A C expression that defines the optional machine-dependent constraint
        !          1824: letters that can be used to segregate specific types of operands,
        !          1825: usually memory references, for the target machine.  Normally this macro
        !          1826: will not be defined.  If it is required for a particular target machine,
        !          1827: it should return 1 if @var{value} corresponds to the operand type
        !          1828: represented by the constraint letter @var{c}.  If @var{c} is not defined
        !          1829: as an extra constraint, the value returned should be 0 regardless of
        !          1830: @var{value}.
        !          1831: 
        !          1832: For example, on the ROMP, load instructions cannot have their output in r0 if
        !          1833: the memory reference contains a symbolic address.  Constraint letter
        !          1834: @samp{Q} is defined as representing a memory address that does
        !          1835: @emph{not} contain a symbolic address.  An alternative is specified with
        !          1836: a @samp{Q} constraint on the input and @samp{r} on the output.  The next
        !          1837: alternative specifies @samp{m} on the input and a register class that
        !          1838: does not include r0 on the output.
        !          1839: @end table
        !          1840: 
        !          1841: @node Stack and Calling
        !          1842: @section Stack Layout and Calling Conventions
        !          1843: @cindex calling conventions
        !          1844: 
        !          1845: @menu
        !          1846: * Frame Layout::
        !          1847: * Frame Registers::
        !          1848: * Elimination::                        
        !          1849: * Stack Arguments::
        !          1850: * Register Arguments::
        !          1851: * Scalar Return::
        !          1852: * Aggregate Return::
        !          1853: * Caller Saves::
        !          1854: * Function Entry::
        !          1855: * Profiling::
        !          1856: @end menu
        !          1857: 
        !          1858: @node Frame Layout
        !          1859: @subsection Basic Stack Layout
        !          1860: @cindex stack frame layout
        !          1861: @cindex frame layout
        !          1862: 
        !          1863: @table @code
        !          1864: @findex STACK_GROWS_DOWNWARD
        !          1865: @item STACK_GROWS_DOWNWARD
        !          1866: Define this macro if pushing a word onto the stack moves the stack
        !          1867: pointer to a smaller address.
        !          1868: 
        !          1869: When we say, ``define this macro if @dots{},'' it means that the
        !          1870: compiler checks this macro only with @code{#ifdef} so the precise
        !          1871: definition used does not matter.
        !          1872: 
        !          1873: @findex FRAME_GROWS_DOWNWARD
        !          1874: @item FRAME_GROWS_DOWNWARD
        !          1875: Define this macro if the addresses of local variable slots are at negative
        !          1876: offsets from the frame pointer.
        !          1877: 
        !          1878: @findex ARGS_GROW_DOWNWARD
        !          1879: @item ARGS_GROW_DOWNWARD
        !          1880: Define this macro if successive arguments to a function occupy decreasing
        !          1881: addresses on the stack.
        !          1882: 
        !          1883: @findex STARTING_FRAME_OFFSET
        !          1884: @item STARTING_FRAME_OFFSET
        !          1885: Offset from the frame pointer to the first local variable slot to be allocated.
        !          1886: 
        !          1887: If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
        !          1888: subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
        !          1889: Otherwise, it is found by adding the length of the first slot to the
        !          1890: value @code{STARTING_FRAME_OFFSET}.
        !          1891: @c i'm not sure if the above is still correct.. had to change it to get
        !          1892: @c rid of an overfull.  --mew 2feb93
        !          1893: 
        !          1894: @findex STACK_POINTER_OFFSET
        !          1895: @item STACK_POINTER_OFFSET
        !          1896: Offset from the stack pointer register to the first location at which
        !          1897: outgoing arguments are placed.  If not specified, the default value of
        !          1898: zero is used.  This is the proper value for most machines.
        !          1899: 
        !          1900: If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
        !          1901: the first location at which outgoing arguments are placed.
        !          1902: 
        !          1903: @findex FIRST_PARM_OFFSET
        !          1904: @item FIRST_PARM_OFFSET (@var{fundecl})
        !          1905: Offset from the argument pointer register to the first argument's
        !          1906: address.  On some machines it may depend on the data type of the
        !          1907: function. 
        !          1908: 
        !          1909: If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
        !          1910: the first argument's address.
        !          1911: 
        !          1912: @findex STACK_DYNAMIC_OFFSET
        !          1913: @item STACK_DYNAMIC_OFFSET (@var{fundecl})
        !          1914: Offset from the stack pointer register to an item dynamically allocated
        !          1915: on the stack, e.g., by @code{alloca}.
        !          1916: 
        !          1917: The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
        !          1918: length of the outgoing arguments.  The default is correct for most
        !          1919: machines.  See @file{function.c} for details.
        !          1920: 
        !          1921: @findex DYNAMIC_CHAIN_ADDRESS
        !          1922: @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
        !          1923: A C expression whose value is RTL representing the address in a stack
        !          1924: frame where the pointer to the caller's frame is stored.  Assume that
        !          1925: @var{frameaddr} is an RTL expression for the address of the stack frame
        !          1926: itself.
        !          1927: 
        !          1928: If you don't define this macro, the default is to return the value
        !          1929: of @var{frameaddr}---that is, the stack frame address is also the
        !          1930: address of the stack word that points to the previous frame.
        !          1931: 
        !          1932: @findex SETUP_FRAME_ADDRESSES
        !          1933: @item SERTUP_FRAME_ADDRESSES ()
        !          1934: If defined, a C expression that produces the machine-specific code to
        !          1935: setup the stack so that arbitrary frames can be accessed.  For example,
        !          1936: on the Sparc, we must flush all of the register windows to the stack
        !          1937: before we can access arbitrary stack frames.
        !          1938: This macro will seldom need to be defined.
        !          1939: 
        !          1940: @findex RETURN_ADDR_RTX
        !          1941: @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
        !          1942: A C expression whose value is RTL representing the value of the return
        !          1943: address for the frame @var{count} steps up from the current frame.
        !          1944: @var{frameaddr} is the frame pointer of the @var{count} frame, or
        !          1945: the frame pointer of the @var{count} @minus{} 1 frame if
        !          1946: @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
        !          1947: 
        !          1948: @findex RETURN_ADDR_IN_PREVIOUS_FRAME
        !          1949: @item RETURN_ADDR_IN_PREVIOUS_FRAME
        !          1950: Define this if the return address of a particular stack frame is accessed
        !          1951: from the frame pointer of the previous stack frame.
        !          1952: @end table
        !          1953: 
        !          1954: @need 2000
        !          1955: @node Frame Registers
        !          1956: @subsection Registers That Address the Stack Frame 
        !          1957: 
        !          1958: @table @code
        !          1959: @findex STACK_POINTER_REGNUM
        !          1960: @item STACK_POINTER_REGNUM
        !          1961: The register number of the stack pointer register, which must also be a
        !          1962: fixed register according to @code{FIXED_REGISTERS}.  On most machines,
        !          1963: the hardware determines which register this is.
        !          1964: 
        !          1965: @findex FRAME_POINTER_REGNUM
        !          1966: @item FRAME_POINTER_REGNUM
        !          1967: The register number of the frame pointer register, which is used to
        !          1968: access automatic variables in the stack frame.  On some machines, the
        !          1969: hardware determines which register this is.  On other machines, you can
        !          1970: choose any register you wish for this purpose.
        !          1971: 
        !          1972: @findex HARD_FRAME_POINTER_REGNUM
        !          1973: @item HARD_FRAME_POINTER_REGNUM
        !          1974: On some machines the offset between the frame pointer and starting
        !          1975: offset of the automatic variables is not known until after register
        !          1976: allocation has been done (for example, because the saved registers are
        !          1977: between these two locations).  On those machines,
        !          1978: @code{FRAME_POINTER_REGNUM} as a special, fixed register to be used
        !          1979: internally until the offset is known, and define
        !          1980: @code{HARD_FRAME_POINTER_REGNUM} to be the hard register used for the
        !          1981: frame pointer.
        !          1982: 
        !          1983: You should define this macro only in the very rare circumstances when it
        !          1984: is not possible to calculate the offset between the frame pointer and
        !          1985: the automatic variables until after register allocation has been
        !          1986: completed.  When this macro is defined, you must also indicate in your
        !          1987: definition of @code{ELIMINABLE_REGS} how to eliminate
        !          1988: @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
        !          1989: or @code{STACK_POINTER_REGNUM}.
        !          1990: 
        !          1991: Do not define this macro if it would be the same as
        !          1992: @code{FRAME_POINTER_REGNUM}.
        !          1993: 
        !          1994: @findex ARG_POINTER_REGNUM
        !          1995: @item ARG_POINTER_REGNUM
        !          1996: The register number of the arg pointer register, which is used to access
        !          1997: the function's argument list.  On some machines, this is the same as the
        !          1998: frame pointer register.  On some machines, the hardware determines which
        !          1999: register this is.  On other machines, you can choose any register you
        !          2000: wish for this purpose.  If this is not the same register as the frame
        !          2001: pointer register, then you must mark it as a fixed register according to
        !          2002: @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
        !          2003: (@pxref{Elimination}).
        !          2004: 
        !          2005: @findex STATIC_CHAIN_REGNUM
        !          2006: @findex STATIC_CHAIN_INCOMING_REGNUM
        !          2007: @item STATIC_CHAIN_REGNUM
        !          2008: @itemx STATIC_CHAIN_INCOMING_REGNUM
        !          2009: Register numbers used for passing a function's static chain pointer.  If
        !          2010: register windows are used, the register number as seen by the called
        !          2011: function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
        !          2012: number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}.  If
        !          2013: these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
        !          2014: not be defined.@refill
        !          2015: 
        !          2016: The static chain register need not be a fixed register.
        !          2017: 
        !          2018: If the static chain is passed in memory, these macros should not be
        !          2019: defined; instead, the next two macros should be defined.
        !          2020: 
        !          2021: @findex STATIC_CHAIN
        !          2022: @findex STATIC_CHAIN_INCOMING
        !          2023: @item STATIC_CHAIN
        !          2024: @itemx STATIC_CHAIN_INCOMING
        !          2025: If the static chain is passed in memory, these macros provide rtx giving
        !          2026: @code{mem} expressions that denote where they are stored.
        !          2027: @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
        !          2028: as seen by the calling and called functions, respectively.  Often the former
        !          2029: will be at an offset from the stack pointer and the latter at an offset from
        !          2030: the frame pointer.@refill
        !          2031: 
        !          2032: @findex stack_pointer_rtx
        !          2033: @findex frame_pointer_rtx
        !          2034: @findex arg_pointer_rtx
        !          2035: The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
        !          2036: @code{arg_pointer_rtx} will have been initialized prior to the use of these
        !          2037: macros and should be used to refer to those items.
        !          2038: 
        !          2039: If the static chain is passed in a register, the two previous macros should
        !          2040: be defined instead.
        !          2041: @end table
        !          2042: 
        !          2043: @node Elimination
        !          2044: @subsection Eliminating Frame Pointer and Arg Pointer
        !          2045: 
        !          2046: @table @code
        !          2047: @findex FRAME_POINTER_REQUIRED
        !          2048: @item FRAME_POINTER_REQUIRED
        !          2049: A C expression which is nonzero if a function must have and use a frame
        !          2050: pointer.  This expression is evaluated  in the reload pass.  If its value is
        !          2051: nonzero the function will have a frame pointer.
        !          2052: 
        !          2053: The expression can in principle examine the current function and decide
        !          2054: according to the facts, but on most machines the constant 0 or the
        !          2055: constant 1 suffices.  Use 0 when the machine allows code to be generated
        !          2056: with no frame pointer, and doing so saves some time or space.  Use 1
        !          2057: when there is no possible advantage to avoiding a frame pointer.
        !          2058: 
        !          2059: In certain cases, the compiler does not know how to produce valid code
        !          2060: without a frame pointer.  The compiler recognizes those cases and
        !          2061: automatically gives the function a frame pointer regardless of what
        !          2062: @code{FRAME_POINTER_REQUIRED} says.  You don't need to worry about
        !          2063: them.@refill
        !          2064: 
        !          2065: In a function that does not require a frame pointer, the frame pointer
        !          2066: register can be allocated for ordinary usage, unless you mark it as a
        !          2067: fixed register.  See @code{FIXED_REGISTERS} for more information.
        !          2068: 
        !          2069: This macro is ignored and you do not need to define it if the function
        !          2070: @code{ELIMINABLE_REGS} is defined.
        !          2071: 
        !          2072: @findex INITIAL_FRAME_POINTER_OFFSET
        !          2073: @findex get_frame_size
        !          2074: @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
        !          2075: A C statement to store in the variable @var{depth-var} the difference
        !          2076: between the frame pointer and the stack pointer values immediately after
        !          2077: the function prologue.  The value would be computed from information
        !          2078: such as the result of @code{get_frame_size ()} and the tables of
        !          2079: registers @code{regs_ever_live} and @code{call_used_regs}.
        !          2080: 
        !          2081: If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
        !          2082: need not be defined.  Otherwise, it must be defined even if
        !          2083: @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
        !          2084: case, you may set @var{depth-var} to anything.
        !          2085: 
        !          2086: @findex ELIMINABLE_REGS
        !          2087: @item ELIMINABLE_REGS
        !          2088: If defined, this macro specifies a table of register pairs used to
        !          2089: eliminate unneeded registers that point into the stack frame.  If it is not
        !          2090: defined, the only elimination attempted by the compiler is to replace
        !          2091: references to the frame pointer with references to the stack pointer.
        !          2092: 
        !          2093: The definition of this macro is a list of structure initializations, each
        !          2094: of which specifies an original and replacement register.
        !          2095: 
        !          2096: On some machines, the position of the argument pointer is not known until
        !          2097: the compilation is completed.  In such a case, a separate hard register
        !          2098: must be used for the argument pointer.  This register can be eliminated by
        !          2099: replacing it with either the frame pointer or the argument pointer,
        !          2100: depending on whether or not the frame pointer has been eliminated.
        !          2101: 
        !          2102: In this case, you might specify:
        !          2103: @example
        !          2104: #define ELIMINABLE_REGS  \
        !          2105: @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
        !          2106:  @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
        !          2107:  @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
        !          2108: @end example
        !          2109: 
        !          2110: Note that the elimination of the argument pointer with the stack pointer is
        !          2111: specified first since that is the preferred elimination.
        !          2112: 
        !          2113: @findex CAN_ELIMINATE
        !          2114: @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
        !          2115: A C expression that returns non-zero if the compiler is allowed to try
        !          2116: to replace register number @var{from-reg} with register number
        !          2117: @var{to-reg}.  This macro need only be defined if @code{ELIMINABLE_REGS}
        !          2118: is defined, and will usually be the constant 1, since most of the cases
        !          2119: preventing register elimination are things that the compiler already
        !          2120: knows about.
        !          2121: 
        !          2122: @findex INITIAL_ELIMINATION_OFFSET
        !          2123: @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
        !          2124: This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}.  It
        !          2125: specifies the initial difference between the specified pair of
        !          2126: registers.  This macro must be defined if @code{ELIMINABLE_REGS} is
        !          2127: defined.
        !          2128: 
        !          2129: @findex LONGJMP_RESTORE_FROM_STACK
        !          2130: @item LONGJMP_RESTORE_FROM_STACK
        !          2131: Define this macro if the @code{longjmp} function restores registers from
        !          2132: the stack frames, rather than from those saved specifically by
        !          2133: @code{setjmp}.  Certain quantities must not be kept in registers across
        !          2134: a call to @code{setjmp} on such machines.
        !          2135: @end table
        !          2136: 
        !          2137: @node Stack Arguments
        !          2138: @subsection Passing Function Arguments on the Stack
        !          2139: @cindex arguments on stack
        !          2140: @cindex stack arguments
        !          2141: 
        !          2142: The macros in this section control how arguments are passed
        !          2143: on the stack.  See the following section for other macros that
        !          2144: control passing certain arguments in registers.
        !          2145: 
        !          2146: @table @code
        !          2147: @findex PROMOTE_PROTOTYPES
        !          2148: @item PROMOTE_PROTOTYPES
        !          2149: Define this macro if an argument declared in a prototype as an
        !          2150: integral type smaller than @code{int} should actually be passed as an
        !          2151: @code{int}.  In addition to avoiding errors in certain cases of
        !          2152: mismatch, it also makes for better code on certain machines.
        !          2153: 
        !          2154: @findex PUSH_ROUNDING
        !          2155: @item PUSH_ROUNDING (@var{npushed})
        !          2156: A C expression that is the number of bytes actually pushed onto the
        !          2157: stack when an instruction attempts to push @var{npushed} bytes.
        !          2158: 
        !          2159: If the target machine does not have a push instruction, do not define
        !          2160: this macro.  That directs GNU CC to use an alternate strategy: to
        !          2161: allocate the entire argument block and then store the arguments into
        !          2162: it.
        !          2163: 
        !          2164: On some machines, the definition
        !          2165: 
        !          2166: @example
        !          2167: #define PUSH_ROUNDING(BYTES) (BYTES)
        !          2168: @end example
        !          2169: 
        !          2170: @noindent
        !          2171: will suffice.  But on other machines, instructions that appear
        !          2172: to push one byte actually push two bytes in an attempt to maintain
        !          2173: alignment.  Then the definition should be
        !          2174: 
        !          2175: @example
        !          2176: #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
        !          2177: @end example
        !          2178: 
        !          2179: @findex ACCUMULATE_OUTGOING_ARGS
        !          2180: @findex current_function_outgoing_args_size
        !          2181: @item ACCUMULATE_OUTGOING_ARGS
        !          2182: If defined, the maximum amount of space required for outgoing arguments
        !          2183: will be computed and placed into the variable
        !          2184: @code{current_function_outgoing_args_size}.  No space will be pushed
        !          2185: onto the stack for each call; instead, the function prologue should
        !          2186: increase the stack frame size by this amount.
        !          2187: 
        !          2188: Defining both @code{PUSH_ROUNDING} and @code{ACCUMULATE_OUTGOING_ARGS}
        !          2189: is not proper.
        !          2190: 
        !          2191: @findex REG_PARM_STACK_SPACE
        !          2192: @item REG_PARM_STACK_SPACE (@var{fndecl})
        !          2193: Define this macro if functions should assume that stack space has been
        !          2194: allocated for arguments even when their values are passed in
        !          2195: registers.
        !          2196: 
        !          2197: The value of this macro is the size, in bytes, of the area reserved for
        !          2198: arguments passed in registers for the function represented by @var{fndecl}.
        !          2199: 
        !          2200: This space can be allocated by the caller, or be a part of the
        !          2201: machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
        !          2202: which.
        !          2203: @c above is overfull.  not sure what to do.  --mew 5feb93  did
        !          2204: @c something, not sure if it looks good.  --mew 10feb93
        !          2205: 
        !          2206: @findex MAYBE_REG_PARM_STACK_SPACE
        !          2207: @findex FINAL_REG_PARM_STACK_SPACE
        !          2208: @item MAYBE_REG_PARM_STACK_SPACE
        !          2209: @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
        !          2210: Define these macros in addition to the one above if functions might
        !          2211: allocate stack space for arguments even when their values are passed
        !          2212: in registers.  These should be used when the stack space allocated
        !          2213: for arguments in registers is not a simple constant independent of the
        !          2214: function declaration.
        !          2215: 
        !          2216: The value of the first macro is the size, in bytes, of the area that
        !          2217: we should initially assume would be reserved for arguments passed in registers.
        !          2218: 
        !          2219: The value of the second macro is the actual size, in bytes, of the area
        !          2220: that will be reserved for arguments passed in registers.  This takes two
        !          2221: arguments: an integer representing the number of bytes of fixed sized
        !          2222: arguments on the stack, and a tree representing the number of bytes of
        !          2223: variable sized arguments on the stack.
        !          2224: 
        !          2225: When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
        !          2226: called for libcall functions, the current function, or for a function
        !          2227: being called when it is known that such stack space must be allocated.
        !          2228: In each case this value can be easily computed.
        !          2229: 
        !          2230: When deciding whether a called function needs such stack space, and how
        !          2231: much space to reserve, GNU CC uses these two macros instead of
        !          2232: @code{REG_PARM_STACK_SPACE}.
        !          2233: 
        !          2234: @findex OUTGOING_REG_PARM_STACK_SPACE
        !          2235: @item OUTGOING_REG_PARM_STACK_SPACE
        !          2236: Define this if it is the responsibility of the caller to allocate the area
        !          2237: reserved for arguments passed in registers.
        !          2238: 
        !          2239: If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
        !          2240: whether the space for these arguments counts in the value of
        !          2241: @code{current_function_outgoing_args_size}.
        !          2242: 
        !          2243: @findex STACK_PARMS_IN_REG_PARM_AREA
        !          2244: @item STACK_PARMS_IN_REG_PARM_AREA
        !          2245: Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
        !          2246: stack parameters don't skip the area specified by it.
        !          2247: @c i changed this, makes more sens and it should have taken care of the
        !          2248: @c overfull.. not as specific, tho.  --mew 5feb93
        !          2249: 
        !          2250: Normally, when a parameter is not passed in registers, it is placed on the
        !          2251: stack beyond the @code{REG_PARM_STACK_SPACE} area.  Defining this macro
        !          2252: suppresses this behavior and causes the parameter to be passed on the
        !          2253: stack in its natural location.
        !          2254: 
        !          2255: @findex RETURN_POPS_ARGS
        !          2256: @item RETURN_POPS_ARGS (@var{funtype}, @var{stack-size})
        !          2257: A C expression that should indicate the number of bytes of its own
        !          2258: arguments that a function pops on returning, or 0 if the
        !          2259: function pops no arguments and the caller must therefore pop them all
        !          2260: after the function returns.
        !          2261: 
        !          2262: @var{funtype} is a C variable whose value is a tree node that
        !          2263: describes the function in question.  Normally it is a node of type
        !          2264: @code{FUNCTION_TYPE} that describes the data type of the function.
        !          2265: From this it is possible to obtain the data types of the value and
        !          2266: arguments (if known).
        !          2267: 
        !          2268: When a call to a library function is being considered, @var{funtype}
        !          2269: will contain an identifier node for the library function.  Thus, if
        !          2270: you need to distinguish among various library functions, you can do so
        !          2271: by their names.  Note that ``library function'' in this context means
        !          2272: a function used to perform arithmetic, whose name is known specially
        !          2273: in the compiler and was not mentioned in the C code being compiled.
        !          2274: 
        !          2275: @var{stack-size} is the number of bytes of arguments passed on the
        !          2276: stack.  If a variable number of bytes is passed, it is zero, and
        !          2277: argument popping will always be the responsibility of the calling function.
        !          2278: 
        !          2279: On the Vax, all functions always pop their arguments, so the definition
        !          2280: of this macro is @var{stack-size}.  On the 68000, using the standard
        !          2281: calling convention, no functions pop their arguments, so the value of
        !          2282: the macro is always 0 in this case.  But an alternative calling
        !          2283: convention is available in which functions that take a fixed number of
        !          2284: arguments pop them but other functions (such as @code{printf}) pop
        !          2285: nothing (the caller pops all).  When this convention is in use,
        !          2286: @var{funtype} is examined to determine whether a function takes a fixed
        !          2287: number of arguments.
        !          2288: @end table
        !          2289: 
        !          2290: @node Register Arguments
        !          2291: @subsection Passing Arguments in Registers
        !          2292: @cindex arguments in registers
        !          2293: @cindex registers arguments
        !          2294: 
        !          2295: This section describes the macros which let you control how various
        !          2296: types of arguments are passed in registers or how they are arranged in
        !          2297: the stack.
        !          2298: 
        !          2299: @table @code
        !          2300: @findex FUNCTION_ARG
        !          2301: @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
        !          2302: A C expression that controls whether a function argument is passed
        !          2303: in a register, and which register.
        !          2304: 
        !          2305: The arguments are @var{cum}, which summarizes all the previous
        !          2306: arguments; @var{mode}, the machine mode of the argument; @var{type},
        !          2307: the data type of the argument as a tree node or 0 if that is not known
        !          2308: (which happens for C support library functions); and @var{named},
        !          2309: which is 1 for an ordinary argument and 0 for nameless arguments that
        !          2310: correspond to @samp{@dots{}} in the called function's prototype.
        !          2311: 
        !          2312: The value of the expression should either be a @code{reg} RTX for the
        !          2313: hard register in which to pass the argument, or zero to pass the
        !          2314: argument on the stack.
        !          2315: 
        !          2316: For machines like the Vax and 68000, where normally all arguments are
        !          2317: pushed, zero suffices as a definition.
        !          2318: 
        !          2319: @cindex @file{stdarg.h} and register arguments
        !          2320: The usual way to make the ANSI library @file{stdarg.h} work on a machine
        !          2321: where some arguments are usually passed in registers, is to cause
        !          2322: nameless arguments to be passed on the stack instead.  This is done
        !          2323: by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
        !          2324: 
        !          2325: @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
        !          2326: @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
        !          2327: You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
        !          2328: in the definition of this macro to determine if this argument is of a
        !          2329: type that must be passed in the stack.  If @code{REG_PARM_STACK_SPACE}
        !          2330: is not defined and @code{FUNCTION_ARG} returns non-zero for such an
        !          2331: argument, the compiler will abort.  If @code{REG_PARM_STACK_SPACE} is
        !          2332: defined, the argument will be computed in the stack and then loaded into
        !          2333: a register.
        !          2334: 
        !          2335: @findex FUNCTION_INCOMING_ARG
        !          2336: @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
        !          2337: Define this macro if the target machine has ``register windows'', so
        !          2338: that the register in which a function sees an arguments is not
        !          2339: necessarily the same as the one in which the caller passed the
        !          2340: argument.
        !          2341: 
        !          2342: For such machines, @code{FUNCTION_ARG} computes the register in which
        !          2343: the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
        !          2344: be defined in a similar fashion to tell the function being called
        !          2345: where the arguments will arrive.
        !          2346: 
        !          2347: If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
        !          2348: serves both purposes.@refill
        !          2349: 
        !          2350: @findex FUNCTION_ARG_PARTIAL_NREGS
        !          2351: @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
        !          2352: A C expression for the number of words, at the beginning of an
        !          2353: argument, must be put in registers.  The value must be zero for
        !          2354: arguments that are passed entirely in registers or that are entirely
        !          2355: pushed on the stack.
        !          2356: 
        !          2357: On some machines, certain arguments must be passed partially in
        !          2358: registers and partially in memory.  On these machines, typically the
        !          2359: first @var{n} words of arguments are passed in registers, and the rest
        !          2360: on the stack.  If a multi-word argument (a @code{double} or a
        !          2361: structure) crosses that boundary, its first few words must be passed
        !          2362: in registers and the rest must be pushed.  This macro tells the
        !          2363: compiler when this occurs, and how many of the words should go in
        !          2364: registers.
        !          2365: 
        !          2366: @code{FUNCTION_ARG} for these arguments should return the first
        !          2367: register to be used by the caller for this argument; likewise
        !          2368: @code{FUNCTION_INCOMING_ARG}, for the called function.
        !          2369: 
        !          2370: @findex FUNCTION_ARG_PASS_BY_REFERENCE
        !          2371: @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
        !          2372: A C expression that indicates when an argument must be passed by reference.
        !          2373: If nonzero for an argument, a copy of that argument is made in memory and a
        !          2374: pointer to the argument is passed instead of the argument itself.
        !          2375: The pointer is passed in whatever way is appropriate for passing a pointer
        !          2376: to that type.
        !          2377: 
        !          2378: On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
        !          2379: definition of this macro might be
        !          2380: @smallexample
        !          2381: #define FUNCTION_ARG_PASS_BY_REFERENCE\
        !          2382: (CUM, MODE, TYPE, NAMED)  \
        !          2383:   MUST_PASS_IN_STACK (MODE, TYPE)
        !          2384: @end smallexample
        !          2385: @c this is *still* too long.  --mew 5feb93 
        !          2386: 
        !          2387: @findex FUNCTION_ARG_CALLEE_COPIES
        !          2388: @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
        !          2389: If defined, a C expression that indicates when it is the called function's
        !          2390: responsibility to make a copy of arguments passed by invisible reference.
        !          2391: Normally, the caller makes a copy and passes the address of the copy to the
        !          2392: routine being called.  When FUNCTION_ARG_CALLEE_COPIES is defined and is
        !          2393: nonzero, the caller does not make a copy.  Instead, it passes a pointer to the
        !          2394: ``live'' value.  The called function must not modify this value.  If it can be
        !          2395: determined that the value won't be modified, it need not make a copy;
        !          2396: otherwise a copy must be made.
        !          2397: 
        !          2398: @findex CUMULATIVE_ARGS
        !          2399: @item CUMULATIVE_ARGS
        !          2400: A C type for declaring a variable that is used as the first argument of
        !          2401: @code{FUNCTION_ARG} and other related values.  For some target machines,
        !          2402: the type @code{int} suffices and can hold the number of bytes of
        !          2403: argument so far.
        !          2404: 
        !          2405: There is no need to record in @code{CUMULATIVE_ARGS} anything about the
        !          2406: arguments that have been passed on the stack.  The compiler has other
        !          2407: variables to keep track of that.  For target machines on which all
        !          2408: arguments are passed on the stack, there is no need to store anything in
        !          2409: @code{CUMULATIVE_ARGS}; however, the data structure must exist and
        !          2410: should not be empty, so use @code{int}.
        !          2411: 
        !          2412: @findex INIT_CUMULATIVE_ARGS
        !          2413: @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname})
        !          2414: A C statement (sans semicolon) for initializing the variable @var{cum}
        !          2415: for the state at the beginning of the argument list.  The variable has
        !          2416: type @code{CUMULATIVE_ARGS}.  The value of @var{fntype} is the tree node
        !          2417: for the data type of the function which will receive the args, or 0
        !          2418: if the args are to a compiler support library function.
        !          2419: 
        !          2420: When processing a call to a compiler support library function,
        !          2421: @var{libname} identifies which one.  It is a @code{symbol_ref} rtx which
        !          2422: contains the name of the function, as a string.  @var{libname} is 0 when
        !          2423: an ordinary C function call is being processed.  Thus, each time this
        !          2424: macro is called, either @var{libname} or @var{fntype} is nonzero, but
        !          2425: never both of them at once.
        !          2426: 
        !          2427: @findex INIT_CUMULATIVE_INCOMING_ARGS
        !          2428: @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
        !          2429: Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
        !          2430: finding the arguments for the function being compiled.  If this macro is
        !          2431: undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
        !          2432: 
        !          2433: The value passed for @var{libname} is always 0, since library routines
        !          2434: with special calling conventions are never compiled with GNU CC.  The
        !          2435: argument @var{libname} exists for symmetry with
        !          2436: @code{INIT_CUMULATIVE_ARGS}.
        !          2437: @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
        !          2438: @c --mew 5feb93   i switched the order of the sentences.  --mew 10feb93
        !          2439: 
        !          2440: @findex FUNCTION_ARG_ADVANCE
        !          2441: @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
        !          2442: A C statement (sans semicolon) to update the summarizer variable
        !          2443: @var{cum} to advance past an argument in the argument list.  The
        !          2444: values @var{mode}, @var{type} and @var{named} describe that argument.
        !          2445: Once this is done, the variable @var{cum} is suitable for analyzing
        !          2446: the @emph{following} argument with @code{FUNCTION_ARG}, etc.@refill
        !          2447: 
        !          2448: This macro need not do anything if the argument in question was passed
        !          2449: on the stack.  The compiler knows how to track the amount of stack space
        !          2450: used for arguments without any special help.
        !          2451: 
        !          2452: @findex FUNCTION_ARG_PADDING
        !          2453: @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
        !          2454: If defined, a C expression which determines whether, and in which direction,
        !          2455: to pad out an argument with extra space.  The value should be of type
        !          2456: @code{enum direction}: either @code{upward} to pad above the argument,
        !          2457: @code{downward} to pad below, or @code{none} to inhibit padding.
        !          2458: 
        !          2459: The @emph{amount} of padding is always just enough to reach the next
        !          2460: multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
        !          2461: it.
        !          2462: 
        !          2463: This macro has a default definition which is right for most systems.
        !          2464: For little-endian machines, the default is to pad upward.  For
        !          2465: big-endian machines, the default is to pad downward for an argument of
        !          2466: constant size shorter than an @code{int}, and upward otherwise.
        !          2467: 
        !          2468: @findex FUNCTION_ARG_BOUNDARY
        !          2469: @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
        !          2470: If defined, a C expression that gives the alignment boundary, in bits,
        !          2471: of an argument with the specified mode and type.  If it is not defined, 
        !          2472: @code{PARM_BOUNDARY} is used for all arguments.
        !          2473: 
        !          2474: @findex FUNCTION_ARG_REGNO_P
        !          2475: @item FUNCTION_ARG_REGNO_P (@var{regno})
        !          2476: A C expression that is nonzero if @var{regno} is the number of a hard
        !          2477: register in which function arguments are sometimes passed.  This does
        !          2478: @emph{not} include implicit arguments such as the static chain and
        !          2479: the structure-value address.  On many machines, no registers can be
        !          2480: used for this purpose since all function arguments are pushed on the
        !          2481: stack.
        !          2482: @end table
        !          2483: 
        !          2484: @node Scalar Return
        !          2485: @subsection How Scalar Function Values Are Returned
        !          2486: @cindex return values in registers
        !          2487: @cindex values, returned by functions
        !          2488: @cindex scalars, returned as values
        !          2489: 
        !          2490: This section discusses the macros that control returning scalars as
        !          2491: values---values that can fit in registers.
        !          2492: 
        !          2493: @table @code
        !          2494: @findex TRADITIONAL_RETURN_FLOAT
        !          2495: @item TRADITIONAL_RETURN_FLOAT
        !          2496: Define this macro if @samp{-traditional} should not cause functions 
        !          2497: declared to return @code{float} to convert the value to @code{double}.
        !          2498: 
        !          2499: @findex FUNCTION_VALUE
        !          2500: @item FUNCTION_VALUE (@var{valtype}, @var{func})
        !          2501: A C expression to create an RTX representing the place where a
        !          2502: function returns a value of data type @var{valtype}.  @var{valtype} is
        !          2503: a tree node representing a data type.  Write @code{TYPE_MODE
        !          2504: (@var{valtype})} to get the machine mode used to represent that type.
        !          2505: On many machines, only the mode is relevant.  (Actually, on most
        !          2506: machines, scalar values are returned in the same place regardless of
        !          2507: mode).@refill
        !          2508: 
        !          2509: If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
        !          2510: promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
        !          2511: scalar type.
        !          2512: 
        !          2513: If the precise function being called is known, @var{func} is a tree
        !          2514: node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
        !          2515: pointer.  This makes it possible to use a different value-returning
        !          2516: convention for specific functions when all their calls are
        !          2517: known.@refill
        !          2518: 
        !          2519: @code{FUNCTION_VALUE} is not used for return vales with aggregate data
        !          2520: types, because these are returned in another way.  See
        !          2521: @code{STRUCT_VALUE_REGNUM} and related macros, below.
        !          2522: 
        !          2523: @findex FUNCTION_OUTGOING_VALUE
        !          2524: @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
        !          2525: Define this macro if the target machine has ``register windows''
        !          2526: so that the register in which a function returns its value is not
        !          2527: the same as the one in which the caller sees the value.
        !          2528: 
        !          2529: For such machines, @code{FUNCTION_VALUE} computes the register in which
        !          2530: the caller will see the value.  @code{FUNCTION_OUTGOING_VALUE} should be
        !          2531: defined in a similar fashion to tell the function where to put the
        !          2532: value.@refill
        !          2533: 
        !          2534: If @code{FUNCTION_OUTGOING_VALUE} is not defined,
        !          2535: @code{FUNCTION_VALUE} serves both purposes.@refill
        !          2536: 
        !          2537: @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
        !          2538: aggregate data types, because these are returned in another way.  See
        !          2539: @code{STRUCT_VALUE_REGNUM} and related macros, below.
        !          2540: 
        !          2541: @findex LIBCALL_VALUE
        !          2542: @item LIBCALL_VALUE (@var{mode})
        !          2543: A C expression to create an RTX representing the place where a library
        !          2544: function returns a value of mode @var{mode}.  If the precise function
        !          2545: being called is known, @var{func} is a tree node
        !          2546: (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
        !          2547: pointer.  This makes it possible to use a different value-returning
        !          2548: convention for specific functions when all their calls are
        !          2549: known.@refill
        !          2550: 
        !          2551: Note that ``library function'' in this context means a compiler
        !          2552: support routine, used to perform arithmetic, whose name is known
        !          2553: specially by the compiler and was not mentioned in the C code being
        !          2554: compiled.
        !          2555: 
        !          2556: The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
        !          2557: data types, because none of the library functions returns such types.
        !          2558: 
        !          2559: @findex FUNCTION_VALUE_REGNO_P
        !          2560: @item FUNCTION_VALUE_REGNO_P (@var{regno})
        !          2561: A C expression that is nonzero if @var{regno} is the number of a hard
        !          2562: register in which the values of called function may come back.
        !          2563: 
        !          2564: A register whose use for returning values is limited to serving as the
        !          2565: second of a pair (for a value of type @code{double}, say) need not be
        !          2566: recognized by this macro.  So for most machines, this definition
        !          2567: suffices:
        !          2568: 
        !          2569: @example
        !          2570: #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
        !          2571: @end example
        !          2572: 
        !          2573: If the machine has register windows, so that the caller and the called
        !          2574: function use different registers for the return value, this macro
        !          2575: should recognize only the caller's register numbers.
        !          2576: 
        !          2577: @findex APPLY_RESULT_SIZE
        !          2578: @item APPLY_RESULT_SIZE
        !          2579: Define this macro if @samp{untyped_call} and @samp{untyped_return}
        !          2580: need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
        !          2581: saving and restoring an arbitrary return value.
        !          2582: @end table
        !          2583: 
        !          2584: @node Aggregate Return
        !          2585: @subsection How Large Values Are Returned
        !          2586: @cindex aggregates as return values
        !          2587: @cindex large return values
        !          2588: @cindex returning aggregate values
        !          2589: @cindex structure value address
        !          2590: 
        !          2591: When a function value's mode is @code{BLKmode} (and in some other
        !          2592: cases), the value is not returned according to @code{FUNCTION_VALUE}
        !          2593: (@pxref{Scalar Return}).  Instead, the caller passes the address of a
        !          2594: block of memory in which the value should be stored.  This address
        !          2595: is called the @dfn{structure value address}.
        !          2596: 
        !          2597: This section describes how to control returning structure values in
        !          2598: memory.
        !          2599: 
        !          2600: @table @code
        !          2601: @findex RETURN_IN_MEMORY
        !          2602: @item RETURN_IN_MEMORY (@var{type})
        !          2603: A C expression which can inhibit the returning of certain function
        !          2604: values in registers, based on the type of value.  A nonzero value says
        !          2605: to return the function value in memory, just as large structures are
        !          2606: always returned.  Here @var{type} will be a C expression of type
        !          2607: @code{tree}, representing the data type of the value.
        !          2608: 
        !          2609: Note that values of mode @code{BLKmode} must be explicitly handled
        !          2610: by this macro.  Also, the option @samp{-fpcc-struct-return}
        !          2611: takes effect regardless of this macro.  On most systems, it is
        !          2612: possible to leave the macro undefined; this causes a default
        !          2613: definition to be used, whose value is the constant 1 for @code{BLKmode}
        !          2614: values, and 0 otherwise.
        !          2615: 
        !          2616: Do not use this macro to indicate that structures and unions should always
        !          2617: be returned in memory.  You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
        !          2618: to indicate this.
        !          2619: 
        !          2620: @findex DEFAULT_PCC_STRUCT_RETURN
        !          2621: @item DEFAULT_PCC_STRUCT_RETURN
        !          2622: Define this macro to be 1 if all structure and union return values must be
        !          2623: in memory.  Since this results in slower code, this should be defined
        !          2624: only if needed for compatibility with other compilers or with an ABI.
        !          2625: If you define this macro to be 0, then the conventions used for structure
        !          2626: and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
        !          2627: 
        !          2628: If not defined, this defaults to the value 1.
        !          2629: 
        !          2630: @findex STRUCT_VALUE_REGNUM
        !          2631: @item STRUCT_VALUE_REGNUM
        !          2632: If the structure value address is passed in a register, then
        !          2633: @code{STRUCT_VALUE_REGNUM} should be the number of that register.
        !          2634: 
        !          2635: @findex STRUCT_VALUE
        !          2636: @item STRUCT_VALUE
        !          2637: If the structure value address is not passed in a register, define
        !          2638: @code{STRUCT_VALUE} as an expression returning an RTX for the place
        !          2639: where the address is passed.  If it returns 0, the address is passed as
        !          2640: an ``invisible'' first argument.
        !          2641: 
        !          2642: @findex STRUCT_VALUE_INCOMING_REGNUM
        !          2643: @item STRUCT_VALUE_INCOMING_REGNUM
        !          2644: On some architectures the place where the structure value address
        !          2645: is found by the called function is not the same place that the
        !          2646: caller put it.  This can be due to register windows, or it could
        !          2647: be because the function prologue moves it to a different place.
        !          2648: 
        !          2649: If the incoming location of the structure value address is in a
        !          2650: register, define this macro as the register number.
        !          2651: 
        !          2652: @findex STRUCT_VALUE_INCOMING
        !          2653: @item STRUCT_VALUE_INCOMING
        !          2654: If the incoming location is not a register, then you should define
        !          2655: @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
        !          2656: called function should find the value.  If it should find the value on
        !          2657: the stack, define this to create a @code{mem} which refers to the frame
        !          2658: pointer.  A definition of 0 means that the address is passed as an
        !          2659: ``invisible'' first argument.
        !          2660: 
        !          2661: @findex PCC_STATIC_STRUCT_RETURN
        !          2662: @item PCC_STATIC_STRUCT_RETURN
        !          2663: Define this macro if the usual system convention on the target machine
        !          2664: for returning structures and unions is for the called function to return
        !          2665: the address of a static variable containing the value.
        !          2666: 
        !          2667: Do not define this if the usual system convention is for the caller to
        !          2668: pass an address to the subroutine.
        !          2669: 
        !          2670: This macro has effect in @samp{-fpcc-struct-return} mode, but it does
        !          2671: nothing when you use @samp{-freg-struct-return} mode.
        !          2672: @end table
        !          2673: 
        !          2674: @node Caller Saves
        !          2675: @subsection Caller-Saves Register Allocation
        !          2676: 
        !          2677: If you enable it, GNU CC can save registers around function calls.  This
        !          2678: makes it possible to use call-clobbered registers to hold variables that
        !          2679: must live across calls.
        !          2680: 
        !          2681: @table @code
        !          2682: @findex DEFAULT_CALLER_SAVES
        !          2683: @item DEFAULT_CALLER_SAVES
        !          2684: Define this macro if function calls on the target machine do not preserve
        !          2685: any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
        !          2686: for all registers.  This macro enables @samp{-fcaller-saves} by default.
        !          2687: Eventually that option will be enabled by default on all machines and both
        !          2688: the option and this macro will be eliminated.
        !          2689: 
        !          2690: @findex CALLER_SAVE_PROFITABLE
        !          2691: @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
        !          2692: A C expression to determine whether it is worthwhile to consider placing
        !          2693: a pseudo-register in a call-clobbered hard register and saving and
        !          2694: restoring it around each function call.  The expression should be 1 when
        !          2695: this is worth doing, and 0 otherwise.
        !          2696: 
        !          2697: If you don't define this macro, a default is used which is good on most
        !          2698: machines: @code{4 * @var{calls} < @var{refs}}.
        !          2699: @end table
        !          2700: 
        !          2701: @node Function Entry
        !          2702: @subsection Function Entry and Exit
        !          2703: @cindex function entry and exit
        !          2704: @cindex prologue
        !          2705: @cindex epilogue
        !          2706: 
        !          2707: This section describes the macros that output function entry
        !          2708: (@dfn{prologue}) and exit (@dfn{epilogue}) code.
        !          2709: 
        !          2710: @table @code
        !          2711: @findex FUNCTION_PROLOGUE
        !          2712: @item FUNCTION_PROLOGUE (@var{file}, @var{size})
        !          2713: A C compound statement that outputs the assembler code for entry to a
        !          2714: function.  The prologue is responsible for setting up the stack frame,
        !          2715: initializing the frame pointer register, saving registers that must be
        !          2716: saved, and allocating @var{size} additional bytes of storage for the
        !          2717: local variables.  @var{size} is an integer.  @var{file} is a stdio
        !          2718: stream to which the assembler code should be output.
        !          2719: 
        !          2720: The label for the beginning of the function need not be output by this
        !          2721: macro.  That has already been done when the macro is run.
        !          2722: 
        !          2723: @findex regs_ever_live
        !          2724: To determine which registers to save, the macro can refer to the array
        !          2725: @code{regs_ever_live}: element @var{r} is nonzero if hard register
        !          2726: @var{r} is used anywhere within the function.  This implies the function
        !          2727: prologue should save register @var{r}, provided it is not one of the
        !          2728: call-used registers.  (@code{FUNCTION_EPILOGUE} must likewise use
        !          2729: @code{regs_ever_live}.)
        !          2730: 
        !          2731: On machines that have ``register windows'', the function entry code does
        !          2732: not save on the stack the registers that are in the windows, even if
        !          2733: they are supposed to be preserved by function calls; instead it takes
        !          2734: appropriate steps to ``push'' the register stack, if any non-call-used
        !          2735: registers are used in the function.
        !          2736: 
        !          2737: @findex frame_pointer_needed
        !          2738: On machines where functions may or may not have frame-pointers, the
        !          2739: function entry code must vary accordingly; it must set up the frame
        !          2740: pointer if one is wanted, and not otherwise.  To determine whether a
        !          2741: frame pointer is in wanted, the macro can refer to the variable
        !          2742: @code{frame_pointer_needed}.  The variable's value will be 1 at run
        !          2743: time in a function that needs a frame pointer.  @xref{Elimination}.
        !          2744: 
        !          2745: The function entry code is responsible for allocating any stack space
        !          2746: required for the function.  This stack space consists of the regions
        !          2747: listed below.  In most cases, these regions are allocated in the
        !          2748: order listed, with the last listed region closest to the top of the
        !          2749: stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
        !          2750: the highest address if it is not defined).  You can use a different order
        !          2751: for a machine if doing so is more convenient or required for
        !          2752: compatibility reasons.  Except in cases where required by standard
        !          2753: or by a debugger, there is no reason why the stack layout used by GCC
        !          2754: need agree with that used by other compilers for a machine.
        !          2755: 
        !          2756: @itemize @bullet
        !          2757: @item
        !          2758: @findex current_function_pretend_args_size
        !          2759: A region of @code{current_function_pretend_args_size} bytes of
        !          2760: uninitialized space just underneath the first argument arriving on the
        !          2761: stack.  (This may not be at the very start of the allocated stack region
        !          2762: if the calling sequence has pushed anything else since pushing the stack
        !          2763: arguments.  But usually, on such machines, nothing else has been pushed
        !          2764: yet, because the function prologue itself does all the pushing.)  This
        !          2765: region is used on machines where an argument may be passed partly in
        !          2766: registers and partly in memory, and, in some cases to support the
        !          2767: features in @file{varargs.h} and @file{stdargs.h}.
        !          2768: 
        !          2769: @item
        !          2770: An area of memory used to save certain registers used by the function.
        !          2771: The size of this area, which may also include space for such things as
        !          2772: the return address and pointers to previous stack frames, is
        !          2773: machine-specific and usually depends on which registers have been used
        !          2774: in the function.  Machines with register windows often do not require
        !          2775: a save area.
        !          2776: 
        !          2777: @item
        !          2778: A region of at least @var{size} bytes, possibly rounded up to an allocation
        !          2779: boundary, to contain the local variables of the function.  On some machines,
        !          2780: this region and the save area may occur in the opposite order, with the
        !          2781: save area closer to the top of the stack.
        !          2782: 
        !          2783: @item
        !          2784: @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
        !          2785: Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
        !          2786: @code{current_function_outgoing_args_size} bytes to be used for outgoing
        !          2787: argument lists of the function.  @xref{Stack Arguments}.
        !          2788: @end itemize
        !          2789: 
        !          2790: Normally, it is necessary for the macros @code{FUNCTION_PROLOGUE} and
        !          2791: @code{FUNCTION_EPILOGUE} to treat leaf functions specially.  The C
        !          2792: variable @code{leaf_function} is nonzero for such a function.
        !          2793: 
        !          2794: @findex EXIT_IGNORE_STACK
        !          2795: @item EXIT_IGNORE_STACK
        !          2796: Define this macro as a C expression that is nonzero if the return
        !          2797: instruction or the function epilogue ignores the value of the stack
        !          2798: pointer; in other words, if it is safe to delete an instruction to
        !          2799: adjust the stack pointer before a return from the function.
        !          2800: 
        !          2801: Note that this macro's value is relevant only for functions for which
        !          2802: frame pointers are maintained.  It is never safe to delete a final
        !          2803: stack adjustment in a function that has no frame pointer, and the
        !          2804: compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
        !          2805: 
        !          2806: @findex FUNCTION_EPILOGUE
        !          2807: @item FUNCTION_EPILOGUE (@var{file}, @var{size})
        !          2808: A C compound statement that outputs the assembler code for exit from a
        !          2809: function.  The epilogue is responsible for restoring the saved
        !          2810: registers and stack pointer to their values when the function was
        !          2811: called, and returning control to the caller.  This macro takes the
        !          2812: same arguments as the macro @code{FUNCTION_PROLOGUE}, and the
        !          2813: registers to restore are determined from @code{regs_ever_live} and
        !          2814: @code{CALL_USED_REGISTERS} in the same way.
        !          2815: 
        !          2816: On some machines, there is a single instruction that does all the work
        !          2817: of returning from the function.  On these machines, give that
        !          2818: instruction the name @samp{return} and do not define the macro
        !          2819: @code{FUNCTION_EPILOGUE} at all.
        !          2820: 
        !          2821: Do not define a pattern named @samp{return} if you want the
        !          2822: @code{FUNCTION_EPILOGUE} to be used.  If you want the target switches
        !          2823: to control whether return instructions or epilogues are used, define a
        !          2824: @samp{return} pattern with a validity condition that tests the target
        !          2825: switches appropriately.  If the @samp{return} pattern's validity
        !          2826: condition is false, epilogues will be used.
        !          2827: 
        !          2828: On machines where functions may or may not have frame-pointers, the
        !          2829: function exit code must vary accordingly.  Sometimes the code for these
        !          2830: two cases is completely different.  To determine whether a frame pointer
        !          2831: is wanted, the macro can refer to the variable
        !          2832: @code{frame_pointer_needed}.  The variable's value will be 1 when compiling
        !          2833: a function that needs a frame pointer.
        !          2834: 
        !          2835: Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
        !          2836: treat leaf functions specially.  The C variable @code{leaf_function} is
        !          2837: nonzero for such a function.  @xref{Leaf Functions}.
        !          2838: 
        !          2839: On some machines, some functions pop their arguments on exit while
        !          2840: others leave that for the caller to do.  For example, the 68020 when
        !          2841: given @samp{-mrtd} pops arguments in functions that take a fixed
        !          2842: number of arguments.
        !          2843: 
        !          2844: @findex current_function_pops_args
        !          2845: Your definition of the macro @code{RETURN_POPS_ARGS} decides which
        !          2846: functions pop their own arguments.  @code{FUNCTION_EPILOGUE} needs to
        !          2847: know what was decided.  The variable that is called
        !          2848: @code{current_function_pops_args} is the number of bytes of its
        !          2849: arguments that a function should pop.  @xref{Scalar Return}.
        !          2850: @c what is the "its arguments" in the above sentence referring to, pray
        !          2851: @c tell?  --mew 5feb93
        !          2852: 
        !          2853: @findex DELAY_SLOTS_FOR_EPILOGUE
        !          2854: @item DELAY_SLOTS_FOR_EPILOGUE
        !          2855: Define this macro if the function epilogue contains delay slots to which
        !          2856: instructions from the rest of the function can be ``moved''.  The
        !          2857: definition should be a C expression whose value is an integer
        !          2858: representing the number of delay slots there.
        !          2859: 
        !          2860: @findex ELIGIBLE_FOR_EPILOGUE_DELAY
        !          2861: @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
        !          2862: A C expression that returns 1 if @var{insn} can be placed in delay
        !          2863: slot number @var{n} of the epilogue.
        !          2864: 
        !          2865: The argument @var{n} is an integer which identifies the delay slot now
        !          2866: being considered (since different slots may have different rules of
        !          2867: eligibility).  It is never negative and is always less than the number
        !          2868: of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
        !          2869: If you reject a particular insn for a given delay slot, in principle, it
        !          2870: may be reconsidered for a subsequent delay slot.  Also, other insns may
        !          2871: (at least in principle) be considered for the so far unfilled delay
        !          2872: slot.
        !          2873: 
        !          2874: @findex current_function_epilogue_delay_list
        !          2875: @findex final_scan_insn
        !          2876: The insns accepted to fill the epilogue delay slots are put in an RTL
        !          2877: list made with @code{insn_list} objects, stored in the variable
        !          2878: @code{current_function_epilogue_delay_list}.  The insn for the first
        !          2879: delay slot comes first in the list.  Your definition of the macro
        !          2880: @code{FUNCTION_EPILOGUE} should fill the delay slots by outputting the
        !          2881: insns in this list, usually by calling @code{final_scan_insn}.
        !          2882: 
        !          2883: You need not define this macro if you did not define
        !          2884: @code{DELAY_SLOTS_FOR_EPILOGUE}.
        !          2885: @end table
        !          2886: 
        !          2887: @node Profiling
        !          2888: @subsection Generating Code for Profiling
        !          2889: @cindex profiling, code generation
        !          2890: 
        !          2891: These macros will help you generate code for profiling.
        !          2892: 
        !          2893: @table @code
        !          2894: @findex FUNCTION_PROFILER 
        !          2895: @item FUNCTION_PROFILER (@var{file}, @var{labelno})
        !          2896: A C statement or compound statement to output to @var{file} some
        !          2897: assembler code to call the profiling subroutine @code{mcount}.
        !          2898: Before calling, the assembler code must load the address of a
        !          2899: counter variable into a register where @code{mcount} expects to
        !          2900: find the address.  The name of this variable is @samp{LP} followed
        !          2901: by the number @var{labelno}, so you would generate the name using
        !          2902: @samp{LP%d} in a @code{fprintf}.
        !          2903: 
        !          2904: @findex mcount
        !          2905: The details of how the address should be passed to @code{mcount} are
        !          2906: determined by your operating system environment, not by GNU CC.  To
        !          2907: figure them out, compile a small program for profiling using the
        !          2908: system's installed C compiler and look at the assembler code that
        !          2909: results.
        !          2910: 
        !          2911: @findex PROFILE_BEFORE_PROLOGUE
        !          2912: @item PROFILE_BEFORE_PROLOGUE
        !          2913: Define this macro if the code for function profiling should come before
        !          2914: the function prologue.  Normally, the profiling code comes after.
        !          2915: 
        !          2916: @findex FUNCTION_BLOCK_PROFILER
        !          2917: @findex __bb_init_func
        !          2918: @item FUNCTION_BLOCK_PROFILER (@var{file}, @var{labelno})
        !          2919: A C statement or compound statement to output to @var{file} some
        !          2920: assembler code to initialize basic-block profiling for the current
        !          2921: object module.  This code should call the subroutine
        !          2922: @code{__bb_init_func} once per object module, passing it as its sole
        !          2923: argument the address of a block allocated in the object module.
        !          2924: 
        !          2925: The name of the block is a local symbol made with this statement:
        !          2926: 
        !          2927: @example
        !          2928: ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
        !          2929: @end example
        !          2930: 
        !          2931: Of course, since you are writing the definition of
        !          2932: @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
        !          2933: can take a short cut in the definition of this macro and use the name
        !          2934: that you know will result.
        !          2935: 
        !          2936: The first word of this block is a flag which will be nonzero if the
        !          2937: object module has already been initialized.  So test this word first,
        !          2938: and do not call @code{__bb_init_func} if the flag is nonzero.
        !          2939: 
        !          2940: @findex BLOCK_PROFILER
        !          2941: @item BLOCK_PROFILER (@var{file}, @var{blockno})
        !          2942: A C statement or compound statement to increment the count associated
        !          2943: with the basic block number @var{blockno}.  Basic blocks are numbered
        !          2944: separately from zero within each compilation.  The count associated
        !          2945: with block number @var{blockno} is at index @var{blockno} in a vector
        !          2946: of words; the name of this array is a local symbol made with this
        !          2947: statement:
        !          2948: 
        !          2949: @example
        !          2950: ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 2);
        !          2951: @end example
        !          2952: 
        !          2953: @c This paragraph is the same as one a few paragraphs up.
        !          2954: @c That is not an error.
        !          2955: Of course, since you are writing the definition of
        !          2956: @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
        !          2957: can take a short cut in the definition of this macro and use the name
        !          2958: that you know will result.
        !          2959: 
        !          2960: @findex BLOCK_PROFILER_CODE
        !          2961: @item BLOCK_PROFILER_CODE
        !          2962: A C function or functions which are needed in the library to
        !          2963: support block profiling.
        !          2964: @end table
        !          2965: 
        !          2966: @node Varargs
        !          2967: @section Implementing the Varargs Macros
        !          2968: @cindex varargs implementation
        !          2969: 
        !          2970: GNU CC comes with an implementation of @file{varargs.h} and
        !          2971: @file{stdarg.h} that work without change on machines that pass arguments
        !          2972: on the stack.  Other machines require their own implementations of
        !          2973: varargs, and the two machine independent header files must have
        !          2974: conditionals to include it.
        !          2975: 
        !          2976: ANSI @file{stdarg.h} differs from traditional @file{varargs.h} mainly in
        !          2977: the calling convention for @code{va_start}.  The traditional
        !          2978: implementation takes just one argument, which is the variable in which
        !          2979: to store the argument pointer.  The ANSI implementation of
        !          2980: @code{va_start} takes an additional second argument.  The user is
        !          2981: supposed to write the last named argument of the function here.
        !          2982: 
        !          2983: However, @code{va_start} should not use this argument.  The way to find
        !          2984: the end of the named arguments is with the built-in functions described
        !          2985: below.
        !          2986: 
        !          2987: @table @code
        !          2988: @findex __builtin_saveregs
        !          2989: @item __builtin_saveregs ()
        !          2990: Use this built-in function to save the argument registers in memory so
        !          2991: that the varargs mechanism can access them.  Both ANSI and traditional
        !          2992: versions of @code{va_start} must use @code{__builtin_saveregs}, unless
        !          2993: you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
        !          2994: 
        !          2995: On some machines, @code{__builtin_saveregs} is open-coded under the
        !          2996: control of the macro @code{EXPAND_BUILTIN_SAVEREGS}.  On other machines,
        !          2997: it calls a routine written in assembler language, found in
        !          2998: @file{libgcc2.c}.
        !          2999: 
        !          3000: Code generated for the call to @code{__builtin_saveregs} appears at the
        !          3001: beginning of the function, as opposed to where the call to
        !          3002: @code{__builtin_saveregs} is written, regardless of what the code is.
        !          3003: This is because the registers must be saved before the function starts
        !          3004: to use them for its own purposes.
        !          3005: @c i rewrote the first sentence above to fix an overfull hbox. --mew
        !          3006: @c 10feb93 
        !          3007: 
        !          3008: @findex __builtin_args_info
        !          3009: @item __builtin_args_info (@var{category})
        !          3010: Use this built-in function to find the first anonymous arguments in
        !          3011: registers.
        !          3012: 
        !          3013: In general, a machine may have several categories of registers used for
        !          3014: arguments, each for a particular category of data types.  (For example,
        !          3015: on some machines, floating-point registers are used for floating-point
        !          3016: arguments while other arguments are passed in the general registers.)
        !          3017: To make non-varargs functions use the proper calling convention, you
        !          3018: have defined the @code{CUMULATIVE_ARGS} data type to record how many
        !          3019: registers in each category have been used so far
        !          3020: 
        !          3021: @code{__builtin_args_info} accesses the same data structure of type
        !          3022: @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
        !          3023: with it, with @var{category} specifying which word to access.  Thus, the
        !          3024: value indicates the first unused register in a given category.
        !          3025: 
        !          3026: Normally, you would use @code{__builtin_args_info} in the implementation
        !          3027: of @code{va_start}, accessing each category just once and storing the
        !          3028: value in the @code{va_list} object.  This is because @code{va_list} will
        !          3029: have to update the values, and there is no way to alter the
        !          3030: values accessed by @code{__builtin_args_info}.
        !          3031: 
        !          3032: @findex __builtin_next_arg
        !          3033: @item __builtin_next_arg ()
        !          3034: This is the equivalent of @code{__builtin_args_info}, for stack
        !          3035: arguments.  It returns the address of the first anonymous stack
        !          3036: argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
        !          3037: returns the address of the location above the first anonymous stack
        !          3038: argument. Use it in @code{va_start} to initialize the pointer for
        !          3039: fetching arguments from the stack. 
        !          3040: 
        !          3041: @findex __builtin_classify_type
        !          3042: @item __builtin_classify_type (@var{object})
        !          3043: Since each machine has its own conventions for which data types are
        !          3044: passed in which kind of register, your implementation of @code{va_arg}
        !          3045: has to embody these conventions.  The easiest way to categorize the
        !          3046: specified data type is to use @code{__builtin_classify_type} together
        !          3047: with @code{sizeof} and @code{__alignof__}.
        !          3048: 
        !          3049: @code{__builtin_classify_type} ignores the value of @var{object},
        !          3050: considering only its data type.  It returns an integer describing what
        !          3051: kind of type that is---integer, floating, pointer, structure, and so on.
        !          3052: 
        !          3053: The file @file{typeclass.h} defines an enumeration that you can use to
        !          3054: interpret the values of @code{__builtin_classify_type}.
        !          3055: @end table
        !          3056: 
        !          3057: These machine description macros help implement varargs: 
        !          3058: 
        !          3059: @table @code
        !          3060: @findex EXPAND_BUILTIN_SAVEREGS
        !          3061: @item EXPAND_BUILTIN_SAVEREGS (@var{args})
        !          3062: If defined, is a C expression that produces the machine-specific code
        !          3063: for a call to @code{__builtin_saveregs}.  This code will be moved to the
        !          3064: very beginning of the function, before any parameter access are made.
        !          3065: The return value of this function should be an RTX that contains the
        !          3066: value to use as the return of @code{__builtin_saveregs}.
        !          3067: 
        !          3068: The argument @var{args} is a @code{tree_list} containing the arguments
        !          3069: that were passed to @code{__builtin_saveregs}.
        !          3070: 
        !          3071: If this macro is not defined, the compiler will output an ordinary
        !          3072: call to the library function @samp{__builtin_saveregs}.
        !          3073: 
        !          3074: @c !!! a bug in texinfo; how to make the entry on the @item line allow
        !          3075: @c more than one line of text... help...  --mew 10feb93
        !          3076: @findex SETUP_INCOMING_VARARGS
        !          3077: @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type},
        !          3078: @var{pretend_args_size}, @var{second_time})  
        !          3079: This macro offers an alternative to using @code{__builtin_saveregs} and
        !          3080: defining the macro @code{EXPAND_BUILTIN_SAVEREGS}.  Use it to store the
        !          3081: anonymous register arguments into the stack so that all the arguments
        !          3082: appear to have been passed consecutively on the stack.  Once this is
        !          3083: done, you can use the standard implementation of varargs that works for
        !          3084: machines that pass all their arguments on the stack.
        !          3085: 
        !          3086: The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
        !          3087: structure, containing the values that obtain after processing of the
        !          3088: named arguments.  The arguments @var{mode} and @var{type} describe the
        !          3089: last named argument---its machine mode and its data type as a tree node.
        !          3090: 
        !          3091: The macro implementation should do two things: first, push onto the
        !          3092: stack all the argument registers @emph{not} used for the named
        !          3093: arguments, and second, store the size of the data thus pushed into the
        !          3094: @code{int}-valued variable whose name is supplied as the argument
        !          3095: @var{pretend_args_size}.  The value that you store here will serve as
        !          3096: additional offset for setting up the stack frame.
        !          3097: 
        !          3098: Because you must generate code to push the anonymous arguments at
        !          3099: compile time without knowing their data types,
        !          3100: @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
        !          3101: a single category of argument register and use it uniformly for all data
        !          3102: types.
        !          3103: 
        !          3104: If the argument @var{second_time} is nonzero, it means that the
        !          3105: arguments of the function are being analyzed for the second time.  This
        !          3106: happens for an inline function, which is not actually compiled until the
        !          3107: end of the source file.  The macro @code{SETUP_INCOMING_VARARGS} should
        !          3108: not generate any instructions in this case.
        !          3109: @end table
        !          3110: 
        !          3111: @node Trampolines
        !          3112: @section Trampolines for Nested Functions
        !          3113: @cindex trampolines for nested functions
        !          3114: @cindex nested functions, trampolines for
        !          3115: 
        !          3116: A @dfn{trampoline} is a small piece of code that is created at run time
        !          3117: when the address of a nested function is taken.  It normally resides on
        !          3118: the stack, in the stack frame of the containing function.  These macros
        !          3119: tell GNU CC how to generate code to allocate and initialize a
        !          3120: trampoline.
        !          3121: 
        !          3122: The instructions in the trampoline must do two things: load a constant
        !          3123: address into the static chain register, and jump to the real address of
        !          3124: the nested function.  On CISC machines such as the m68k, this requires
        !          3125: two instructions, a move immediate and a jump.  Then the two addresses
        !          3126: exist in the trampoline as word-long immediate operands.  On RISC
        !          3127: machines, it is often necessary to load each address into a register in
        !          3128: two parts.  Then pieces of each address form separate immediate
        !          3129: operands.
        !          3130: 
        !          3131: The code generated to initialize the trampoline must store the variable
        !          3132: parts---the static chain value and the function address---into the
        !          3133: immediate operands of the instructions.  On a CISC machine, this is
        !          3134: simply a matter of copying each address to a memory reference at the
        !          3135: proper offset from the start of the trampoline.  On a RISC machine, it
        !          3136: may be necessary to take out pieces of the address and store them
        !          3137: separately.
        !          3138: 
        !          3139: @table @code
        !          3140: @findex TRAMPOLINE_TEMPLATE
        !          3141: @item TRAMPOLINE_TEMPLATE (@var{file})
        !          3142: A C statement to output, on the stream @var{file}, assembler code for a
        !          3143: block of data that contains the constant parts of a trampoline.  This
        !          3144: code should not include a label---the label is taken care of
        !          3145: automatically.
        !          3146: 
        !          3147: @findex TRAMPOLINE_SECTION
        !          3148: @item TRAMPOLINE_SECTION
        !          3149: The name of a subroutine to switch to the section in which the
        !          3150: trampoline template is to be placed (@pxref{Sections}).  The default is
        !          3151: a value of @samp{readonly_data_section}, which places the trampoline in
        !          3152: the section containing read-only data. 
        !          3153: 
        !          3154: @findex TRAMPOLINE_SIZE
        !          3155: @item TRAMPOLINE_SIZE
        !          3156: A C expression for the size in bytes of the trampoline, as an integer.
        !          3157: 
        !          3158: @findex TRAMPOLINE_ALIGNMENT
        !          3159: @item TRAMPOLINE_ALIGNMENT
        !          3160: Alignment required for trampolines, in bits.
        !          3161: 
        !          3162: If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
        !          3163: is used for aligning trampolines.
        !          3164: 
        !          3165: @findex INITIALIZE_TRAMPOLINE
        !          3166: @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
        !          3167: A C statement to initialize the variable parts of a trampoline.
        !          3168: @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
        !          3169: an RTX for the address of the nested function; @var{static_chain} is an
        !          3170: RTX for the static chain value that should be passed to the function
        !          3171: when it is called.
        !          3172: 
        !          3173: @findex ALLOCATE_TRAMPOLINE
        !          3174: @item ALLOCATE_TRAMPOLINE (@var{fp})
        !          3175: A C expression to allocate run-time space for a trampoline.  The
        !          3176: expression value should be an RTX representing a memory reference to the
        !          3177: space for the trampoline.
        !          3178: 
        !          3179: @cindex @code{FUNCTION_EPILOGUE} and trampolines
        !          3180: @cindex @code{FUNCTION_PROLOGUE} and trampolines
        !          3181: If this macro is not defined, by default the trampoline is allocated as
        !          3182: a stack slot.  This default is right for most machines.  The exceptions
        !          3183: are machines where it is impossible to execute instructions in the stack
        !          3184: area.  On such machines, you may have to implement a separate stack,
        !          3185: using this macro in conjunction with @code{FUNCTION_PROLOGUE} and
        !          3186: @code{FUNCTION_EPILOGUE}.
        !          3187: 
        !          3188: @var{fp} points to a data structure, a @code{struct function}, which
        !          3189: describes the compilation status of the immediate containing function of
        !          3190: the function which the trampoline is for.  Normally (when
        !          3191: @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
        !          3192: trampoline is in the stack frame of this containing function.  Other
        !          3193: allocation strategies probably must do something analogous with this
        !          3194: information.
        !          3195: @end table
        !          3196: 
        !          3197: Implementing trampolines is difficult on many machines because they have
        !          3198: separate instruction and data caches.  Writing into a stack location
        !          3199: fails to clear the memory in the instruction cache, so when the program
        !          3200: jumps to that location, it executes the old contents.
        !          3201: 
        !          3202: Here are two possible solutions.  One is to clear the relevant parts of
        !          3203: the instruction cache whenever a trampoline is set up.  The other is to
        !          3204: make all trampolines identical, by having them jump to a standard
        !          3205: subroutine.  The former technique makes trampoline execution faster; the
        !          3206: latter makes initialization faster.
        !          3207: 
        !          3208: To clear the instruction cache when a trampoline is initialized, define
        !          3209: the following macros which describe the shape of the cache.
        !          3210: 
        !          3211: @table @code
        !          3212: @findex INSN_CACHE_SIZE
        !          3213: @item INSN_CACHE_SIZE
        !          3214: The total size in bytes of the cache.
        !          3215: 
        !          3216: @findex INSN_CACHE_LINE_WIDTH
        !          3217: @item INSN_CACHE_LINE_WIDTH
        !          3218: The length in bytes of each cache line.  The cache is divided into cache
        !          3219: lines which are disjoint slots, each holding a contiguous chunk of data
        !          3220: fetched from memory.  Each time data is brought into the cache, an
        !          3221: entire line is read at once.  The data loaded into a cache line is 
        !          3222: always aligned on a boundary equal to the line size.
        !          3223: 
        !          3224: @findex INSN_CACHE_DEPTH
        !          3225: @item INSN_CACHE_DEPTH
        !          3226: The number of alternative cache lines that can hold any particular memory
        !          3227: location.
        !          3228: @end table
        !          3229: 
        !          3230: Alternatively, if the machine has system calls or instructions to clear
        !          3231: the instruction cache directly, you can define the following macro.
        !          3232: 
        !          3233: @table @code
        !          3234: @findex CLEAR_INSN_CACHE (BEG, END)
        !          3235: @item
        !          3236: If defined, expands to a C expression clearing the @emph{instruction
        !          3237: cache} in the specified interval.  If it is not defined, and the macro
        !          3238: INSN_CACHE_SIZE is defined, some generic code is generated to clear the
        !          3239: cache.  The definition of this macro would typically be a series of
        !          3240: @code{asm} statements.  Both @var{BEG} and @var{END} are both pointer 
        !          3241: expressions. 
        !          3242: @end table
        !          3243: 
        !          3244: To use a standard subroutine, define the following macro.  In addition,
        !          3245: you must make sure that the instructions in a trampoline fill an entire
        !          3246: cache line with identical instructions, or else ensure that the
        !          3247: beginning of the trampoline code is always aligned at the same point in
        !          3248: its cache line.  Look in @file{m68k.h} as a guide.
        !          3249: 
        !          3250: @table @code
        !          3251: @findex TRANSFER_FROM_TRAMPOLINE
        !          3252: @item TRANSFER_FROM_TRAMPOLINE
        !          3253: Define this macro if trampolines need a special subroutine to do their
        !          3254: work.  The macro should expand to a series of @code{asm} statements
        !          3255: which will be compiled with GNU CC.  They go in a library function named
        !          3256: @code{__transfer_from_trampoline}.
        !          3257: 
        !          3258: If you need to avoid executing the ordinary prologue code of a compiled
        !          3259: C function when you jump to the subroutine, you can do so by placing a
        !          3260: special label of your own in the assembler code.  Use one @code{asm}
        !          3261: statement to generate an assembler label, and another to make the label
        !          3262: global.  Then trampolines can use that label to jump directly to your
        !          3263: special assembler code.
        !          3264: @end table
        !          3265: 
        !          3266: @node Library Calls
        !          3267: @section Implicit Calls to Library Routines
        !          3268: @cindex library subroutine names
        !          3269: @cindex @file{libgcc.a}
        !          3270: 
        !          3271: @table @code
        !          3272: @findex MULSI3_LIBCALL
        !          3273: @item MULSI3_LIBCALL
        !          3274: A C string constant giving the name of the function to call for
        !          3275: multiplication of one signed full-word by another.  If you do not
        !          3276: define this macro, the default name is used, which is @code{__mulsi3},
        !          3277: a function defined in @file{libgcc.a}.
        !          3278: 
        !          3279: @findex DIVSI3_LIBCALL
        !          3280: @item DIVSI3_LIBCALL
        !          3281: A C string constant giving the name of the function to call for
        !          3282: division of one signed full-word by another.  If you do not define
        !          3283: this macro, the default name is used, which is @code{__divsi3}, a
        !          3284: function defined in @file{libgcc.a}.
        !          3285: 
        !          3286: @findex UDIVSI3_LIBCALL
        !          3287: @item UDIVSI3_LIBCALL
        !          3288: A C string constant giving the name of the function to call for
        !          3289: division of one unsigned full-word by another.  If you do not define
        !          3290: this macro, the default name is used, which is @code{__udivsi3}, a
        !          3291: function defined in @file{libgcc.a}.
        !          3292: 
        !          3293: @findex MODSI3_LIBCALL
        !          3294: @item MODSI3_LIBCALL
        !          3295: A C string constant giving the name of the function to call for the
        !          3296: remainder in division of one signed full-word by another.  If you do
        !          3297: not define this macro, the default name is used, which is
        !          3298: @code{__modsi3}, a function defined in @file{libgcc.a}.
        !          3299: 
        !          3300: @findex UMODSI3_LIBCALL
        !          3301: @item UMODSI3_LIBCALL
        !          3302: A C string constant giving the name of the function to call for the
        !          3303: remainder in division of one unsigned full-word by another.  If you do
        !          3304: not define this macro, the default name is used, which is
        !          3305: @code{__umodsi3}, a function defined in @file{libgcc.a}.
        !          3306: 
        !          3307: @findex MULDI3_LIBCALL
        !          3308: @item MULDI3_LIBCALL
        !          3309: A C string constant giving the name of the function to call for
        !          3310: multiplication of one signed double-word by another.  If you do not
        !          3311: define this macro, the default name is used, which is @code{__muldi3},
        !          3312: a function defined in @file{libgcc.a}.
        !          3313: 
        !          3314: @findex DIVDI3_LIBCALL
        !          3315: @item DIVDI3_LIBCALL
        !          3316: A C string constant giving the name of the function to call for
        !          3317: division of one signed double-word by another.  If you do not define
        !          3318: this macro, the default name is used, which is @code{__divdi3}, a
        !          3319: function defined in @file{libgcc.a}.
        !          3320: 
        !          3321: @findex UDIVDI3_LIBCALL
        !          3322: @item UDIVDI3_LIBCALL
        !          3323: A C string constant giving the name of the function to call for
        !          3324: division of one unsigned full-word by another.  If you do not define
        !          3325: this macro, the default name is used, which is @code{__udivdi3}, a
        !          3326: function defined in @file{libgcc.a}.
        !          3327: 
        !          3328: @findex MODDI3_LIBCALL
        !          3329: @item MODDI3_LIBCALL
        !          3330: A C string constant giving the name of the function to call for the
        !          3331: remainder in division of one signed double-word by another.  If you do
        !          3332: not define this macro, the default name is used, which is
        !          3333: @code{__moddi3}, a function defined in @file{libgcc.a}.
        !          3334: 
        !          3335: @findex UMODDI3_LIBCALL
        !          3336: @item UMODDI3_LIBCALL
        !          3337: A C string constant giving the name of the function to call for the
        !          3338: remainder in division of one unsigned full-word by another.  If you do
        !          3339: not define this macro, the default name is used, which is
        !          3340: @code{__umoddi3}, a function defined in @file{libgcc.a}.
        !          3341: 
        !          3342: @findex TARGET_EDOM
        !          3343: @cindex @code{EDOM}, implicit usage
        !          3344: @item TARGET_EDOM
        !          3345: The value of @code{EDOM} on the target machine, as a C integer constant
        !          3346: expression.  If you don't define this macro, GNU CC does not attempt to
        !          3347: deposit the value of @code{EDOM} into @code{errno} directly.  Look in
        !          3348: @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
        !          3349: system.
        !          3350: 
        !          3351: If you do not define @code{TARGET_EDOM}, then compiled code reports
        !          3352: domain errors by calling the library function and letting it report the
        !          3353: error.  If mathematical functions on your system use @code{matherr} when
        !          3354: there is an error, then you should leave @code{TARGET_EDOM} undefined so
        !          3355: that @code{matherr} is used normally.
        !          3356: 
        !          3357: @findex GEN_ERRNO_RTX
        !          3358: @cindex @code{errno}, implicit usage
        !          3359: @item GEN_ERRNO_RTX
        !          3360: Define this macro as a C expression to create an rtl expression that
        !          3361: refers to the global ``variable'' @code{errno}.  (On certain systems,
        !          3362: @code{errno} may not actually be a variable.)  If you don't define this
        !          3363: macro, a reasonable default is used.
        !          3364: 
        !          3365: @findex TARGET_MEM_FUNCTIONS
        !          3366: @cindex @code{bcopy}, implicit usage
        !          3367: @cindex @code{memcpy}, implicit usage
        !          3368: @cindex @code{bzero}, implicit usage
        !          3369: @cindex @code{memset}, implicit usage
        !          3370: @item TARGET_MEM_FUNCTIONS
        !          3371: Define this macro if GNU CC should generate calls to the System V
        !          3372: (and ANSI C) library functions @code{memcpy} and @code{memset}
        !          3373: rather than the BSD functions @code{bcopy} and @code{bzero}.
        !          3374: 
        !          3375: @findex LIBGCC_NEEDS_DOUBLE
        !          3376: @item LIBGCC_NEEDS_DOUBLE
        !          3377: Define this macro if only @code{float} arguments cannot be passed to
        !          3378: library routines (so they must be converted to @code{double}).  This
        !          3379: macro affects both how library calls are generated and how the library
        !          3380: routines in @file{libgcc1.c} accept their arguments.  It is useful on
        !          3381: machines where floating and fixed point arguments are passed
        !          3382: differently, such as the i860.
        !          3383: 
        !          3384: @findex FLOAT_ARG_TYPE
        !          3385: @item FLOAT_ARG_TYPE
        !          3386: Define this macro to override the type used by the library routines to
        !          3387: pick up arguments of type @code{float}.  (By default, they use a union
        !          3388: of @code{float} and @code{int}.)
        !          3389: 
        !          3390: The obvious choice would be @code{float}---but that won't work with
        !          3391: traditional C compilers that expect all arguments declared as @code{float}
        !          3392: to arrive as @code{double}.  To avoid this conversion, the library routines
        !          3393: ask for the value as some other type and then treat it as a @code{float}.
        !          3394: 
        !          3395: On some systems, no other type will work for this.  For these systems,
        !          3396: you must use @code{LIBGCC_NEEDS_DOUBLE} instead, to force conversion of
        !          3397: the values @code{double} before they are passed.
        !          3398: 
        !          3399: @findex FLOATIFY
        !          3400: @item FLOATIFY (@var{passed-value})
        !          3401: Define this macro to override the way library routines redesignate a
        !          3402: @code{float} argument as a @code{float} instead of the type it was
        !          3403: passed as.  The default is an expression which takes the @code{float}
        !          3404: field of the union.
        !          3405: 
        !          3406: @findex FLOAT_VALUE_TYPE
        !          3407: @item FLOAT_VALUE_TYPE
        !          3408: Define this macro to override the type used by the library routines to
        !          3409: return values that ought to have type @code{float}.  (By default, they
        !          3410: use @code{int}.)
        !          3411: 
        !          3412: The obvious choice would be @code{float}---but that won't work with
        !          3413: traditional C compilers gratuitously convert values declared as
        !          3414: @code{float} into @code{double}.
        !          3415: 
        !          3416: @findex INTIFY
        !          3417: @item INTIFY (@var{float-value})
        !          3418: Define this macro to override the way the value of a
        !          3419: @code{float}-returning library routine should be packaged in order to
        !          3420: return it.  These functions are actually declared to return type 
        !          3421: @code{FLOAT_VALUE_TYPE} (normally @code{int}).
        !          3422: 
        !          3423: These values can't be returned as type @code{float} because traditional
        !          3424: C compilers would gratuitously convert the value to a @code{double}.
        !          3425: 
        !          3426: A local variable named @code{intify} is always available when the macro
        !          3427: @code{INTIFY} is used.  It is a union of a @code{float} field named
        !          3428: @code{f} and a field named @code{i} whose type is
        !          3429: @code{FLOAT_VALUE_TYPE} or @code{int}.
        !          3430: 
        !          3431: If you don't define this macro, the default definition works by copying
        !          3432: the value through that union.
        !          3433: 
        !          3434: @findex nongcc_SI_type
        !          3435: @item nongcc_SI_type
        !          3436: Define this macro as the name of the data type corresponding to
        !          3437: @code{SImode} in the system's own C compiler.
        !          3438: 
        !          3439: You need not define this macro if that type is @code{long int}, as it usually
        !          3440: is.
        !          3441: 
        !          3442: @findex nongcc_word_type
        !          3443: @item nongcc_word_type
        !          3444: Define this macro as the name of the data type corresponding to the 
        !          3445: word_mode in the system's own C compiler.
        !          3446: 
        !          3447: You need not define this macro if that type is @code{long int}, as it usually
        !          3448: is.
        !          3449: 
        !          3450: @findex perform_@dots{}
        !          3451: @item perform_@dots{}
        !          3452: Define these macros to supply explicit C statements to carry out various
        !          3453: arithmetic operations on types @code{float} and @code{double} in the
        !          3454: library routines in @file{libgcc1.c}.  See that file for a full list
        !          3455: of these macros and their arguments.
        !          3456: 
        !          3457: On most machines, you don't need to define any of these macros, because
        !          3458: the C compiler that comes with the system takes care of doing them.
        !          3459: 
        !          3460: @findex NEXT_OBJC_RUNTIME
        !          3461: @item NEXT_OBJC_RUNTIME
        !          3462: Define this macro to generate code for Objective C message sending using
        !          3463: the calling convention of the NeXT system.  This calling convention
        !          3464: involves passing the object, the selector and the method arguments all
        !          3465: at once to the method-lookup library function.
        !          3466: 
        !          3467: The default calling convention passes just the object and the selector
        !          3468: to the lookup function, which returns a pointer to the method.
        !          3469: @end table
        !          3470: 
        !          3471: @node Addressing Modes
        !          3472: @section Addressing Modes
        !          3473: @cindex addressing modes
        !          3474: 
        !          3475: @table @code
        !          3476: @findex HAVE_POST_INCREMENT
        !          3477: @item HAVE_POST_INCREMENT
        !          3478: Define this macro if the machine supports post-increment addressing.
        !          3479: 
        !          3480: @findex HAVE_PRE_INCREMENT
        !          3481: @findex HAVE_POST_DECREMENT
        !          3482: @findex HAVE_PRE_DECREMENT
        !          3483: @item HAVE_PRE_INCREMENT
        !          3484: @itemx HAVE_POST_DECREMENT
        !          3485: @itemx HAVE_PRE_DECREMENT
        !          3486: Similar for other kinds of addressing.
        !          3487: 
        !          3488: @findex CONSTANT_ADDRESS_P
        !          3489: @item CONSTANT_ADDRESS_P (@var{x})
        !          3490: A C expression that is 1 if the RTX @var{x} is a constant which
        !          3491: is a valid address.  On most machines, this can be defined as
        !          3492: @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
        !          3493: in which constant addresses are supported.
        !          3494: 
        !          3495: @findex CONSTANT_P
        !          3496: @code{CONSTANT_P} accepts integer-values expressions whose values are
        !          3497: not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
        !          3498: @code{high} expressions and @code{const} arithmetic expressions, in
        !          3499: addition to @code{const_int} and @code{const_double} expressions.
        !          3500: 
        !          3501: @findex MAX_REGS_PER_ADDRESS
        !          3502: @item MAX_REGS_PER_ADDRESS
        !          3503: A number, the maximum number of registers that can appear in a valid
        !          3504: memory address.  Note that it is up to you to specify a value equal to
        !          3505: the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
        !          3506: accept.
        !          3507: 
        !          3508: @findex GO_IF_LEGITIMATE_ADDRESS
        !          3509: @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
        !          3510: A C compound statement with a conditional @code{goto @var{label};}
        !          3511: executed if @var{x} (an RTX) is a legitimate memory address on the
        !          3512: target machine for a memory operand of mode @var{mode}.
        !          3513: 
        !          3514: It usually pays to define several simpler macros to serve as
        !          3515: subroutines for this one.  Otherwise it may be too complicated to
        !          3516: understand.
        !          3517: 
        !          3518: This macro must exist in two variants: a strict variant and a
        !          3519: non-strict one.  The strict variant is used in the reload pass.  It
        !          3520: must be defined so that any pseudo-register that has not been
        !          3521: allocated a hard register is considered a memory reference.  In
        !          3522: contexts where some kind of register is required, a pseudo-register
        !          3523: with no hard register must be rejected.
        !          3524: 
        !          3525: The non-strict variant is used in other passes.  It must be defined to
        !          3526: accept all pseudo-registers in every context where some kind of
        !          3527: register is required.
        !          3528: 
        !          3529: @findex REG_OK_STRICT
        !          3530: Compiler source files that want to use the strict variant of this
        !          3531: macro define the macro @code{REG_OK_STRICT}.  You should use an
        !          3532: @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
        !          3533: in that case and the non-strict variant otherwise.
        !          3534: 
        !          3535: Subroutines to check for acceptable registers for various purposes (one
        !          3536: for base registers, one for index registers, and so on) are typically
        !          3537: among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
        !          3538: Then only these subroutine macros need have two variants; the higher
        !          3539: levels of macros may be the same whether strict or not.@refill
        !          3540: 
        !          3541: Normally, constant addresses which are the sum of a @code{symbol_ref}
        !          3542: and an integer are stored inside a @code{const} RTX to mark them as
        !          3543: constant.  Therefore, there is no need to recognize such sums
        !          3544: specifically as legitimate addresses.  Normally you would simply
        !          3545: recognize any @code{const} as legitimate.
        !          3546: 
        !          3547: Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
        !          3548: sums that are not marked with  @code{const}.  It assumes that a naked
        !          3549: @code{plus} indicates indexing.  If so, then you @emph{must} reject such
        !          3550: naked constant sums as illegitimate addresses, so that none of them will
        !          3551: be given to @code{PRINT_OPERAND_ADDRESS}.
        !          3552: 
        !          3553: @cindex @code{ENCODE_SECTION_INFO} and address validation
        !          3554: On some machines, whether a symbolic address is legitimate depends on
        !          3555: the section that the address refers to.  On these machines, define the
        !          3556: macro @code{ENCODE_SECTION_INFO} to store the information into the
        !          3557: @code{symbol_ref}, and then check for it here.  When you see a
        !          3558: @code{const}, you will have to look inside it to find the
        !          3559: @code{symbol_ref} in order to determine the section.  @xref{Assembler
        !          3560: Format}.
        !          3561: 
        !          3562: @findex saveable_obstack
        !          3563: The best way to modify the name string is by adding text to the
        !          3564: beginning, with suitable punctuation to prevent any ambiguity.  Allocate
        !          3565: the new name in @code{saveable_obstack}.  You will have to modify
        !          3566: @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
        !          3567: output the name accordingly, and define @code{STRIP_NAME_ENCODING} to
        !          3568: access the original name string.
        !          3569: 
        !          3570: You can check the information stored here into the @code{symbol_ref} in
        !          3571: the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
        !          3572: @code{PRINT_OPERAND_ADDRESS}.
        !          3573: 
        !          3574: @findex REG_OK_FOR_BASE_P
        !          3575: @item REG_OK_FOR_BASE_P (@var{x})
        !          3576: A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
        !          3577: RTX) is valid for use as a base register.  For hard registers, it
        !          3578: should always accept those which the hardware permits and reject the
        !          3579: others.  Whether the macro accepts or rejects pseudo registers must be
        !          3580: controlled by @code{REG_OK_STRICT} as described above.  This usually
        !          3581: requires two variant definitions, of which @code{REG_OK_STRICT}
        !          3582: controls the one actually used.
        !          3583: 
        !          3584: @findex REG_OK_FOR_INDEX_P
        !          3585: @item REG_OK_FOR_INDEX_P (@var{x})
        !          3586: A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
        !          3587: RTX) is valid for use as an index register.
        !          3588: 
        !          3589: The difference between an index register and a base register is that
        !          3590: the index register may be scaled.  If an address involves the sum of
        !          3591: two registers, neither one of them scaled, then either one may be
        !          3592: labeled the ``base'' and the other the ``index''; but whichever
        !          3593: labeling is used must fit the machine's constraints of which registers
        !          3594: may serve in each capacity.  The compiler will try both labelings,
        !          3595: looking for one that is valid, and will reload one or both registers
        !          3596: only if neither labeling works.
        !          3597: 
        !          3598: @findex LEGITIMIZE_ADDRESS
        !          3599: @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
        !          3600: A C compound statement that attempts to replace @var{x} with a valid
        !          3601: memory address for an operand of mode @var{mode}.  @var{win} will be a
        !          3602: C statement label elsewhere in the code; the macro definition may use
        !          3603: 
        !          3604: @example
        !          3605: GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
        !          3606: @end example
        !          3607: 
        !          3608: @noindent
        !          3609: to avoid further processing if the address has become legitimate.
        !          3610: 
        !          3611: @findex break_out_memory_refs
        !          3612: @var{x} will always be the result of a call to @code{break_out_memory_refs},
        !          3613: and @var{oldx} will be the operand that was given to that function to produce
        !          3614: @var{x}.
        !          3615: 
        !          3616: The code generated by this macro should not alter the substructure of
        !          3617: @var{x}.  If it transforms @var{x} into a more legitimate form, it
        !          3618: should assign @var{x} (which will always be a C variable) a new value.
        !          3619: 
        !          3620: It is not necessary for this macro to come up with a legitimate
        !          3621: address.  The compiler has standard ways of doing so in all cases.  In
        !          3622: fact, it is safe for this macro to do nothing.  But often a
        !          3623: machine-dependent strategy can generate better code.
        !          3624: 
        !          3625: @findex GO_IF_MODE_DEPENDENT_ADDRESS
        !          3626: @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
        !          3627: A C statement or compound statement with a conditional @code{goto
        !          3628: @var{label};} executed if memory address @var{x} (an RTX) can have
        !          3629: different meanings depending on the machine mode of the memory
        !          3630: reference it is used for or if the address is valid for some modes
        !          3631: but not others.
        !          3632: 
        !          3633: Autoincrement and autodecrement addresses typically have mode-dependent
        !          3634: effects because the amount of the increment or decrement is the size
        !          3635: of the operand being addressed.  Some machines have other mode-dependent
        !          3636: addresses.  Many RISC machines have no mode-dependent addresses.
        !          3637: 
        !          3638: You may assume that @var{addr} is a valid address for the machine.
        !          3639: 
        !          3640: @findex LEGITIMATE_CONSTANT_P
        !          3641: @item LEGITIMATE_CONSTANT_P (@var{x})
        !          3642: A C expression that is nonzero if @var{x} is a legitimate constant for
        !          3643: an immediate operand on the target machine.  You can assume that
        !          3644: @var{x} satisfies @code{CONSTANT_P}, so you need not check this.  In fact,
        !          3645: @samp{1} is a suitable definition for this macro on machines where
        !          3646: anything @code{CONSTANT_P} is valid.@refill
        !          3647: @end table
        !          3648: 
        !          3649: @node Condition Code
        !          3650: @section Condition Code Status
        !          3651: @cindex condition code status
        !          3652: 
        !          3653: @findex cc_status
        !          3654: The file @file{conditions.h} defines a variable @code{cc_status} to
        !          3655: describe how the condition code was computed (in case the interpretation of
        !          3656: the condition code depends on the instruction that it was set by).  This
        !          3657: variable contains the RTL expressions on which the condition code is
        !          3658: currently based, and several standard flags.
        !          3659: 
        !          3660: Sometimes additional machine-specific flags must be defined in the machine
        !          3661: description header file.  It can also add additional machine-specific
        !          3662: information by defining @code{CC_STATUS_MDEP}.
        !          3663: 
        !          3664: @table @code
        !          3665: @findex CC_STATUS_MDEP
        !          3666: @item CC_STATUS_MDEP
        !          3667: C code for a data type which is used for declaring the @code{mdep}
        !          3668: component of @code{cc_status}.  It defaults to @code{int}.
        !          3669: 
        !          3670: This macro is not used on machines that do not use @code{cc0}.
        !          3671: 
        !          3672: @findex CC_STATUS_MDEP_INIT
        !          3673: @item CC_STATUS_MDEP_INIT
        !          3674: A C expression to initialize the @code{mdep} field to ``empty''.
        !          3675: The default definition does nothing, since most machines don't use
        !          3676: the field anyway.  If you want to use the field, you should probably
        !          3677: define this macro to initialize it.
        !          3678: 
        !          3679: This macro is not used on machines that do not use @code{cc0}.
        !          3680: 
        !          3681: @findex NOTICE_UPDATE_CC
        !          3682: @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
        !          3683: A C compound statement to set the components of @code{cc_status}
        !          3684: appropriately for an insn @var{insn} whose body is @var{exp}.  It is
        !          3685: this macro's responsibility to recognize insns that set the condition
        !          3686: code as a byproduct of other activity as well as those that explicitly
        !          3687: set @code{(cc0)}.
        !          3688: 
        !          3689: This macro is not used on machines that do not use @code{cc0}.
        !          3690: 
        !          3691: If there are insns that do not set the condition code but do alter
        !          3692: other machine registers, this macro must check to see whether they
        !          3693: invalidate the expressions that the condition code is recorded as
        !          3694: reflecting.  For example, on the 68000, insns that store in address
        !          3695: registers do not set the condition code, which means that usually
        !          3696: @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
        !          3697: insns.  But suppose that the previous insn set the condition code
        !          3698: based on location @samp{a4@@(102)} and the current insn stores a new
        !          3699: value in @samp{a4}.  Although the condition code is not changed by
        !          3700: this, it will no longer be true that it reflects the contents of
        !          3701: @samp{a4@@(102)}.  Therefore, @code{NOTICE_UPDATE_CC} must alter
        !          3702: @code{cc_status} in this case to say that nothing is known about the
        !          3703: condition code value.
        !          3704: 
        !          3705: The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
        !          3706: with the results of peephole optimization: insns whose patterns are
        !          3707: @code{parallel} RTXs containing various @code{reg}, @code{mem} or
        !          3708: constants which are just the operands.  The RTL structure of these
        !          3709: insns is not sufficient to indicate what the insns actually do.  What
        !          3710: @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
        !          3711: @code{CC_STATUS_INIT}.
        !          3712: 
        !          3713: A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
        !          3714: that looks at an attribute (@pxref{Insn Attributes}) named, for example,
        !          3715: @samp{cc}.  This avoids having detailed information about patterns in
        !          3716: two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
        !          3717: 
        !          3718: @findex EXTRA_CC_MODES
        !          3719: @item EXTRA_CC_MODES
        !          3720: A list of names to be used for additional modes for condition code
        !          3721: values in registers (@pxref{Jump Patterns}).  These names are added
        !          3722: to @code{enum machine_mode} and all have class @code{MODE_CC}.  By
        !          3723: convention, they should start with @samp{CC} and end with @samp{mode}.
        !          3724: 
        !          3725: You should only define this macro if your machine does not use @code{cc0}
        !          3726: and only if additional modes are required.
        !          3727: 
        !          3728: @findex EXTRA_CC_NAMES
        !          3729: @item EXTRA_CC_NAMES
        !          3730: A list of C strings giving the names for the modes listed in
        !          3731: @code{EXTRA_CC_MODES}.  For example, the Sparc defines this macro and
        !          3732: @code{EXTRA_CC_MODES} as
        !          3733: 
        !          3734: @smallexample
        !          3735: #define EXTRA_CC_MODES CC_NOOVmode, CCFPmode
        !          3736: #define EXTRA_CC_NAMES "CC_NOOV", "CCFP"
        !          3737: @end smallexample
        !          3738: 
        !          3739: This macro is not required if @code{EXTRA_CC_MODES} is not defined.
        !          3740: 
        !          3741: @findex SELECT_CC_MODE
        !          3742: @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
        !          3743: Returns a mode from class @code{MODE_CC} to be used when comparison
        !          3744: operation code @var{op} is applied to rtx @var{x} and @var{y}.  For
        !          3745: example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
        !          3746: @pxref{Jump Patterns} for a description of the reason for this
        !          3747: definition)
        !          3748: 
        !          3749: @smallexample
        !          3750: #define SELECT_CC_MODE(OP,X,Y) \
        !          3751:   (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT          \
        !          3752:    ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode)    \
        !          3753:    : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS    \
        !          3754:        || GET_CODE (X) == NEG) \
        !          3755:       ? CC_NOOVmode : CCmode))
        !          3756: @end smallexample
        !          3757: 
        !          3758: This macro is not required if @code{EXTRA_CC_MODES} is not defined.
        !          3759: @end table
        !          3760: 
        !          3761: @node Costs
        !          3762: @section Describing Relative Costs of Operations
        !          3763: @cindex costs of instructions
        !          3764: @cindex relative costs
        !          3765: @cindex speed of instructions
        !          3766: 
        !          3767: These macros let you describe the relative speed of various operations
        !          3768: on the target machine.
        !          3769: 
        !          3770: @table @code
        !          3771: @findex CONST_COSTS 
        !          3772: @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
        !          3773: A part of a C @code{switch} statement that describes the relative costs
        !          3774: of constant RTL expressions.  It must contain @code{case} labels for
        !          3775: expression codes @code{const_int}, @code{const}, @code{symbol_ref},
        !          3776: @code{label_ref} and @code{const_double}.  Each case must ultimately
        !          3777: reach a @code{return} statement to return the relative cost of the use
        !          3778: of that kind of constant value in an expression.  The cost may depend on
        !          3779: the precise value of the constant, which is available for examination in
        !          3780: @var{x}, and the rtx code of the expression in which it is contained,
        !          3781: found in @var{outer_code}.
        !          3782: 
        !          3783: @var{code} is the expression code---redundant, since it can be
        !          3784: obtained with @code{GET_CODE (@var{x})}.
        !          3785: 
        !          3786: @findex RTX_COSTS 
        !          3787: @findex COSTS_N_INSNS
        !          3788: @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
        !          3789: Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
        !          3790: This can be used, for example, to indicate how costly a multiply
        !          3791: instruction is.  In writing this macro, you can use the construct
        !          3792: @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
        !          3793: instructions.  @var{outer_code} is the code of the expression in which
        !          3794: @var{x} is contained.
        !          3795: 
        !          3796: This macro is optional; do not define it if the default cost assumptions
        !          3797: are adequate for the target machine.
        !          3798: 
        !          3799: @findex ADDRESS_COST
        !          3800: @item ADDRESS_COST (@var{address})
        !          3801: An expression giving the cost of an addressing mode that contains
        !          3802: @var{address}.  If not defined, the cost is computed from
        !          3803: the @var{address} expression and the @code{CONST_COSTS} values.
        !          3804: 
        !          3805: For most CISC machines, the default cost is a good approximation of the
        !          3806: true cost of the addressing mode.  However, on RISC machines, all
        !          3807: instructions normally have the same length and execution time.  Hence
        !          3808: all addresses will have equal costs.
        !          3809: 
        !          3810: In cases where more than one form of an address is known, the form with
        !          3811: the lowest cost will be used.  If multiple forms have the same, lowest,
        !          3812: cost, the one that is the most complex will be used.
        !          3813: 
        !          3814: For example, suppose an address that is equal to the sum of a register
        !          3815: and a constant is used twice in the same basic block.  When this macro
        !          3816: is not defined, the address will be computed in a register and memory
        !          3817: references will be indirect through that register.  On machines where
        !          3818: the cost of the addressing mode containing the sum is no higher than
        !          3819: that of a simple indirect reference, this will produce an additional
        !          3820: instruction and possibly require an additional register.  Proper
        !          3821: specification of this macro eliminates this overhead for such machines.
        !          3822: 
        !          3823: Similar use of this macro is made in strength reduction of loops.
        !          3824: 
        !          3825: @var{address} need not be valid as an address.  In such a case, the cost
        !          3826: is not relevant and can be any value; invalid addresses need not be
        !          3827: assigned a different cost.
        !          3828: 
        !          3829: On machines where an address involving more than one register is as
        !          3830: cheap as an address computation involving only one register, defining
        !          3831: @code{ADDRESS_COST} to reflect this can cause two registers to be live
        !          3832: over a region of code where only one would have been if
        !          3833: @code{ADDRESS_COST} were not defined in that manner.  This effect should
        !          3834: be considered in the definition of this macro.  Equivalent costs should
        !          3835: probably only be given to addresses with different numbers of registers
        !          3836: on machines with lots of registers.
        !          3837: 
        !          3838: This macro will normally either not be defined or be defined as a
        !          3839: constant.
        !          3840: 
        !          3841: @findex REGISTER_MOVE_COST
        !          3842: @item REGISTER_MOVE_COST (@var{from}, @var{to})
        !          3843: A C expression for the cost of moving data from a register in class
        !          3844: @var{from} to one in class @var{to}.  The classes are expressed using
        !          3845: the enumeration values such as @code{GENERAL_REGS}.  A value of 4 is the
        !          3846: default; other values are interpreted relative to that.
        !          3847: 
        !          3848: It is not required that the cost always equal 2 when @var{from} is the
        !          3849: same as @var{to}; on some machines it is expensive to move between
        !          3850: registers if they are not general registers.
        !          3851: 
        !          3852: If reload sees an insn consisting of a single @code{set} between two
        !          3853: hard registers, and if @code{REGISTER_MOVE_COST} applied to their
        !          3854: classes returns a value of 2, reload does not check to ensure that the
        !          3855: constraints of the insn are met.  Setting a cost of other than 2 will
        !          3856: allow reload to verify that the constraints are met.  You should do this
        !          3857: if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
        !          3858: 
        !          3859: @findex MEMORY_MOVE_COST
        !          3860: @item MEMORY_MOVE_COST (@var{m})
        !          3861: A C expression for the cost of moving data of mode @var{m} between a
        !          3862: register and memory.  A value of 2 is the default; this cost is relative
        !          3863: to those in @code{REGISTER_MOVE_COST}.
        !          3864: 
        !          3865: If moving between registers and memory is more expensive than between
        !          3866: two registers, you should define this macro to express the relative cost.
        !          3867: 
        !          3868: @findex BRANCH_COST
        !          3869: @item BRANCH_COST
        !          3870: A C expression for the cost of a branch instruction.  A value of 1 is
        !          3871: the default; other values are interpreted relative to that.
        !          3872: @end table
        !          3873: 
        !          3874: Here are additional macros which do not specify precise relative costs,
        !          3875: but only that certain actions are more expensive than GNU CC would
        !          3876: ordinarily expect.
        !          3877: 
        !          3878: @table @code
        !          3879: @findex SLOW_BYTE_ACCESS
        !          3880: @item SLOW_BYTE_ACCESS
        !          3881: Define this macro as a C expression which is nonzero if accessing less
        !          3882: than a word of memory (i.e. a @code{char} or a @code{short}) is no
        !          3883: faster than accessing a word of memory, i.e., if such access
        !          3884: require more than one instruction or if there is no difference in cost
        !          3885: between byte and (aligned) word loads.
        !          3886: 
        !          3887: When this macro is not defined, the compiler will access a field by
        !          3888: finding the smallest containing object; when it is defined, a fullword
        !          3889: load will be used if alignment permits.  Unless bytes accesses are
        !          3890: faster than word accesses, using word accesses is preferable since it
        !          3891: may eliminate subsequent memory access if subsequent accesses occur to
        !          3892: other fields in the same word of the structure, but to different bytes.
        !          3893: 
        !          3894: @findex SLOW_ZERO_EXTEND
        !          3895: @item SLOW_ZERO_EXTEND
        !          3896: Define this macro if zero-extension (of a @code{char} or @code{short}
        !          3897: to an @code{int}) can be done faster if the destination is a register
        !          3898: that is known to be zero.
        !          3899: 
        !          3900: If you define this macro, you must have instruction patterns that
        !          3901: recognize RTL structures like this:
        !          3902: 
        !          3903: @smallexample
        !          3904: (set (strict_low_part (subreg:QI (reg:SI @dots{}) 0)) @dots{})
        !          3905: @end smallexample
        !          3906: 
        !          3907: @noindent
        !          3908: and likewise for @code{HImode}.
        !          3909: 
        !          3910: @findex SLOW_UNALIGNED_ACCESS
        !          3911: @item SLOW_UNALIGNED_ACCESS
        !          3912: Define this macro to be the value 1 if unaligned accesses have a cost
        !          3913: many times greater than aligned accesses, for example if they are
        !          3914: emulated in a trap handler.
        !          3915: 
        !          3916: When this macro is non-zero, the compiler will act as if
        !          3917: @code{STRICT_ALIGNMENT} were non-zero when generating code for block
        !          3918: moves.  This can cause significantly more instructions to be produced.
        !          3919: Therefore, do not set this macro non-zero if unaligned accesses only add a
        !          3920: cycle or two to the time for a memory access.
        !          3921: 
        !          3922: If the value of this macro is always zero, it need not be defined.
        !          3923: 
        !          3924: @findex DONT_REDUCE_ADDR
        !          3925: @item DONT_REDUCE_ADDR
        !          3926: Define this macro to inhibit strength reduction of memory addresses.
        !          3927: (On some machines, such strength reduction seems to do harm rather
        !          3928: than good.)
        !          3929: 
        !          3930: @findex MOVE_RATIO
        !          3931: @item MOVE_RATIO
        !          3932: The number of scalar move insns which should be generated instead of a
        !          3933: string move insn or a library call.  Increasing the value will always
        !          3934: make code faster, but eventually incurs high cost in increased code size.
        !          3935: 
        !          3936: If you don't define this, a reasonable default is used.
        !          3937: 
        !          3938: @findex NO_FUNCTION_CSE
        !          3939: @item NO_FUNCTION_CSE
        !          3940: Define this macro if it is as good or better to call a constant
        !          3941: function address than to call an address kept in a register.
        !          3942: 
        !          3943: @findex NO_RECURSIVE_FUNCTION_CSE
        !          3944: @item NO_RECURSIVE_FUNCTION_CSE
        !          3945: Define this macro if it is as good or better for a function to call
        !          3946: itself with an explicit address than to call an address kept in a
        !          3947: register.
        !          3948: 
        !          3949: @findex ADJUST_COST
        !          3950: @item ADJUST_COST (@var{insn}, @var{link}, @var{dep_insn}, @var{cost})
        !          3951: A C statement (sans semicolon) to update the integer variable @var{cost}
        !          3952: based on the relationship between @var{insn} that is dependent on
        !          3953: @var{dep_insn} through the dependence @var{link}.  The default is to
        !          3954: make no adjustment to @var{cost}.  This can be used for example to
        !          3955: specify to the scheduler that an output- or anti-dependence does not
        !          3956: incur the same cost as a data-dependence.
        !          3957: @end table
        !          3958: 
        !          3959: @node Sections
        !          3960: @section Dividing the Output into Sections (Texts, Data, @dots{})
        !          3961: @c the above section title is WAY too long.  maybe cut the part between
        !          3962: @c the (...)?  --mew 10feb93
        !          3963: 
        !          3964: An object file is divided into sections containing different types of
        !          3965: data.  In the most common case, there are three sections: the @dfn{text
        !          3966: section}, which holds instructions and read-only data; the @dfn{data
        !          3967: section}, which holds initialized writable data; and the @dfn{bss
        !          3968: section}, which holds uninitialized data.  Some systems have other kinds
        !          3969: of sections.
        !          3970: 
        !          3971: The compiler must tell the assembler when to switch sections.  These
        !          3972: macros control what commands to output to tell the assembler this.  You
        !          3973: can also define additional sections.
        !          3974: 
        !          3975: @table @code
        !          3976: @findex TEXT_SECTION_ASM_OP
        !          3977: @item TEXT_SECTION_ASM_OP
        !          3978: A C expression whose value is a string containing the assembler
        !          3979: operation that should precede instructions and read-only data.  Normally
        !          3980: @code{".text"} is right.
        !          3981: 
        !          3982: @findex DATA_SECTION_ASM_OP
        !          3983: @item DATA_SECTION_ASM_OP
        !          3984: A C expression whose value is a string containing the assembler
        !          3985: operation to identify the following data as writable initialized data.
        !          3986: Normally @code{".data"} is right.
        !          3987: 
        !          3988: @findex SHARED_SECTION_ASM_OP
        !          3989: @item SHARED_SECTION_ASM_OP
        !          3990: if defined, a C expression whose value is a string containing the
        !          3991: assembler operation to identify the following data as shared data.  If
        !          3992: not defined, @code{DATA_SECTION_ASM_OP} will be used.
        !          3993: 
        !          3994: @findex INIT_SECTION_ASM_OP
        !          3995: @item INIT_SECTION_ASM_OP
        !          3996: if defined, a C expression whose value is a string containing the
        !          3997: assembler operation to identify the following data as initialization
        !          3998: code.  If not defined, GNU CC will assume such a section does not
        !          3999: exist.
        !          4000: 
        !          4001: @findex EXTRA_SECTIONS
        !          4002: @findex in_text
        !          4003: @findex in_data
        !          4004: @item EXTRA_SECTIONS
        !          4005: A list of names for sections other than the standard two, which are
        !          4006: @code{in_text} and @code{in_data}.  You need not define this macro
        !          4007: on a system with no other sections (that GCC needs to use).
        !          4008: 
        !          4009: @findex EXTRA_SECTION_FUNCTIONS
        !          4010: @findex text_section
        !          4011: @findex data_section
        !          4012: @item EXTRA_SECTION_FUNCTIONS
        !          4013: One or more functions to be defined in @file{varasm.c}.  These
        !          4014: functions should do jobs analogous to those of @code{text_section} and
        !          4015: @code{data_section}, for your additional sections.  Do not define this
        !          4016: macro if you do not define @code{EXTRA_SECTIONS}.
        !          4017: 
        !          4018: @findex READONLY_DATA_SECTION
        !          4019: @item READONLY_DATA_SECTION
        !          4020: On most machines, read-only variables, constants, and jump tables are
        !          4021: placed in the text section.  If this is not the case on your machine,
        !          4022: this macro should be defined to be the name of a function (either
        !          4023: @code{data_section} or a function defined in @code{EXTRA_SECTIONS}) that
        !          4024: switches to the section to be used for read-only items.
        !          4025: 
        !          4026: If these items should be placed in the text section, this macro should
        !          4027: not be defined.
        !          4028: 
        !          4029: @findex SELECT_SECTION
        !          4030: @item SELECT_SECTION (@var{exp}, @var{reloc})
        !          4031: A C statement or statements to switch to the appropriate section for
        !          4032: output of @var{exp}.  You can assume that @var{exp} is either a
        !          4033: @code{VAR_DECL} node or a constant of some sort.  @var{reloc}
        !          4034: indicates whether the initial value of @var{exp} requires link-time
        !          4035: relocations.  Select the section by calling @code{text_section} or one
        !          4036: of the alternatives for other sections.
        !          4037: 
        !          4038: Do not define this macro if you put all read-only variables and
        !          4039: constants in the read-only data section (usually the text section).
        !          4040: 
        !          4041: @findex SELECT_RTX_SECTION
        !          4042: @item SELECT_RTX_SECTION (@var{mode}, @var{rtx})
        !          4043: A C statement or statements to switch to the appropriate section for
        !          4044: output of @var{rtx} in mode @var{mode}.  You can assume that @var{rtx}
        !          4045: is some kind of constant in RTL.  The argument @var{mode} is redundant
        !          4046: except in the case of a @code{const_int} rtx.  Select the section by
        !          4047: calling @code{text_section} or one of the alternatives for other
        !          4048: sections.
        !          4049: 
        !          4050: Do not define this macro if you put all constants in the read-only
        !          4051: data section.
        !          4052: 
        !          4053: @findex JUMP_TABLES_IN_TEXT_SECTION
        !          4054: @item JUMP_TABLES_IN_TEXT_SECTION
        !          4055: Define this macro if jump tables (for @code{tablejump} insns) should be
        !          4056: output in the text section, along with the assembler instructions.
        !          4057: Otherwise, the readonly data section is used.
        !          4058: 
        !          4059: This macro is irrelevant if there is no separate readonly data section.
        !          4060: 
        !          4061: @findex ENCODE_SECTION_INFO
        !          4062: @item ENCODE_SECTION_INFO (@var{decl})
        !          4063: Define this macro if references to a symbol must be treated differently
        !          4064: depending on something about the variable or function named by the
        !          4065: symbol (such as what section it is in).
        !          4066: 
        !          4067: The macro definition, if any, is executed immediately after the rtl for
        !          4068: @var{decl} has been created and stored in @code{DECL_RTL (@var{decl})}.
        !          4069: The value of the rtl will be a @code{mem} whose address is a
        !          4070: @code{symbol_ref}.
        !          4071: 
        !          4072: @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO}
        !          4073: The usual thing for this macro to do is to record a flag in the
        !          4074: @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
        !          4075: modified name string in the @code{symbol_ref} (if one bit is not enough
        !          4076: information).
        !          4077: 
        !          4078: @findex STRIP_NAME_ENCODING
        !          4079: @item STRIP_NAME_ENCODING (@var{var}, @var{sym_name})
        !          4080: Decode @var{sym_name} and store the real name part in @var{var}, sans
        !          4081: the characters that encode section info.  Define this macro if
        !          4082: @code{ENCODE_SECTION_INFO} alters the symbol's name string.
        !          4083: @end table
        !          4084: 
        !          4085: @node PIC
        !          4086: @section Position Independent Code
        !          4087: @cindex position independent code
        !          4088: @cindex PIC
        !          4089: 
        !          4090: This section describes macros that help implement generation of position
        !          4091: independent code.  Simply defining these macros is not enough to
        !          4092: generate valid PIC; you must also add support to the macros
        !          4093: @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
        !          4094: well as @code{LEGITIMIZE_ADDRESS}.  You must modify the definition of
        !          4095: @samp{movsi} to do something appropriate when the source operand
        !          4096: contains a symbolic address.  You may also need to alter the handling of
        !          4097: switch statements so that they use relative addresses.
        !          4098: @c i rearranged the order of the macros above to try to force one of
        !          4099: @c them to the next line, to eliminate an overfull hbox. --mew 10feb93 
        !          4100: 
        !          4101: @table @code
        !          4102: @findex PIC_OFFSET_TABLE_REGNUM
        !          4103: @item PIC_OFFSET_TABLE_REGNUM
        !          4104: The register number of the register used to address a table of static
        !          4105: data addresses in memory.  In some cases this register is defined by a
        !          4106: processor's ``application binary interface'' (ABI).  When this macro
        !          4107: is defined, RTL is generated for this register once, as with the stack
        !          4108: pointer and frame pointer registers.  If this macro is not defined, it
        !          4109: is up to the machine-dependent files to allocate such a register (if
        !          4110: necessary).
        !          4111: 
        !          4112: @findex FINALIZE_PIC
        !          4113: @item FINALIZE_PIC
        !          4114: By generating position-independent code, when two different programs (A
        !          4115: and B) share a common library (libC.a), the text of the library can be
        !          4116: shared whether or not the library is linked at the same address for both
        !          4117: programs.  In some of these environments, position-independent code
        !          4118: requires not only the use of different addressing modes, but also
        !          4119: special code to enable the use of these addressing modes.
        !          4120: 
        !          4121: The @code{FINALIZE_PIC} macro serves as a hook to emit these special
        !          4122: codes once the function is being compiled into assembly code, but not
        !          4123: before.  (It is not done before, because in the case of compiling an
        !          4124: inline function, it would lead to multiple PIC prologues being
        !          4125: included in functions which used inline functions and were compiled to
        !          4126: assembly language.)
        !          4127: 
        !          4128: @findex LEGITIMATE_PIC_OPERAND_P
        !          4129: @item LEGITIMATE_PIC_OPERAND_P (@var{x})
        !          4130: A C expression that is nonzero if @var{x} is a legitimate immediate
        !          4131: operand on the target machine when generating position independent code.
        !          4132: You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
        !          4133: check this.  You can also assume @var{flag_pic} is true, so you need not
        !          4134: check it either.  You need not define this macro if all constants 
        !          4135: (including @code{SYMBOL_REF}) can be immediate operands when generating 
        !          4136: position independent code.
        !          4137: @end table
        !          4138: 
        !          4139: @node Assembler Format
        !          4140: @section Defining the Output Assembler Language
        !          4141: 
        !          4142: This section describes macros whose principal purpose is to describe how
        !          4143: to write instructions in assembler language--rather than what the
        !          4144: instructions do.
        !          4145: 
        !          4146: @menu
        !          4147: * File Framework::       Structural information for the assembler file.
        !          4148: * Data Output::          Output of constants (numbers, strings, addresses).
        !          4149: * Uninitialized Data::   Output of uninitialized variables.
        !          4150: * Label Output::         Output and generation of labels.
        !          4151: * Initialization::       General principles of initialization
        !          4152:                           and termination routines.
        !          4153: * Macros for Initialization::
        !          4154:                         Specific macros that control the handling of 
        !          4155:                           initialization and termination routines.
        !          4156: * Instruction Output::   Output of actual instructions.
        !          4157: * Dispatch Tables::      Output of jump tables.
        !          4158: * Alignment Output::     Pseudo ops for alignment and skipping data.
        !          4159: @end menu
        !          4160: 
        !          4161: @node File Framework
        !          4162: @subsection The Overall Framework of an Assembler File 
        !          4163: @cindex assembler format
        !          4164: @cindex output of assembler code
        !          4165: 
        !          4166: @table @code
        !          4167: @findex ASM_FILE_START
        !          4168: @item ASM_FILE_START (@var{stream})
        !          4169: A C expression which outputs to the stdio stream @var{stream}
        !          4170: some appropriate text to go at the start of an assembler file.
        !          4171: 
        !          4172: Normally this macro is defined to output a line containing
        !          4173: @samp{#NO_APP}, which is a comment that has no effect on most
        !          4174: assemblers but tells the GNU assembler that it can save time by not
        !          4175: checking for certain assembler constructs.
        !          4176: 
        !          4177: On systems that use SDB, it is necessary to output certain commands;
        !          4178: see @file{attasm.h}.
        !          4179: 
        !          4180: @findex ASM_FILE_END
        !          4181: @item ASM_FILE_END (@var{stream})
        !          4182: A C expression which outputs to the stdio stream @var{stream}
        !          4183: some appropriate text to go at the end of an assembler file.
        !          4184: 
        !          4185: If this macro is not defined, the default is to output nothing
        !          4186: special at the end of the file.  Most systems don't require any
        !          4187: definition.
        !          4188: 
        !          4189: On systems that use SDB, it is necessary to output certain commands;
        !          4190: see @file{attasm.h}.
        !          4191: 
        !          4192: @findex ASM_IDENTIFY_GCC
        !          4193: @item ASM_IDENTIFY_GCC (@var{file})
        !          4194: A C statement to output assembler commands which will identify
        !          4195: the object file as having been compiled with GNU CC (or another
        !          4196: GNU compiler).
        !          4197: 
        !          4198: If you don't define this macro, the string @samp{gcc_compiled.:}
        !          4199: is output.  This string is calculated to define a symbol which,
        !          4200: on BSD systems, will never be defined for any other reason.
        !          4201: GDB checks for the presence of this symbol when reading the
        !          4202: symbol table of an executable.
        !          4203: 
        !          4204: On non-BSD systems, you must arrange communication with GDB in
        !          4205: some other fashion.  If GDB is not used on your system, you can
        !          4206: define this macro with an empty body.
        !          4207: 
        !          4208: @findex ASM_COMMENT_START
        !          4209: @item ASM_COMMENT_START
        !          4210: A C string constant describing how to begin a comment in the target
        !          4211: assembler language.  The compiler assumes that the comment will end at
        !          4212: the end of the line.
        !          4213: 
        !          4214: @findex ASM_APP_ON
        !          4215: @item ASM_APP_ON
        !          4216: A C string constant for text to be output before each @code{asm}
        !          4217: statement or group of consecutive ones.  Normally this is
        !          4218: @code{"#APP"}, which is a comment that has no effect on most
        !          4219: assemblers but tells the GNU assembler that it must check the lines
        !          4220: that follow for all valid assembler constructs.
        !          4221: 
        !          4222: @findex ASM_APP_OFF
        !          4223: @item ASM_APP_OFF
        !          4224: A C string constant for text to be output after each @code{asm}
        !          4225: statement or group of consecutive ones.  Normally this is
        !          4226: @code{"#NO_APP"}, which tells the GNU assembler to resume making the
        !          4227: time-saving assumptions that are valid for ordinary compiler output.
        !          4228: 
        !          4229: @findex ASM_OUTPUT_SOURCE_FILENAME
        !          4230: @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
        !          4231: A C statement to output COFF information or DWARF debugging information
        !          4232: which indicates that filename @var{name} is the current source file to
        !          4233: the stdio stream @var{stream}.
        !          4234: 
        !          4235: This macro need not be defined if the standard form of output
        !          4236: for the file format in use is appropriate.
        !          4237: 
        !          4238: @findex ASM_OUTPUT_SOURCE_LINE
        !          4239: @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
        !          4240: A C statement to output DBX or SDB debugging information before code
        !          4241: for line number @var{line} of the current source file to the
        !          4242: stdio stream @var{stream}.
        !          4243: 
        !          4244: This macro need not be defined if the standard form of debugging
        !          4245: information for the debugger in use is appropriate.
        !          4246: 
        !          4247: @findex ASM_OUTPUT_IDENT
        !          4248: @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
        !          4249: A C statement to output something to the assembler file to handle a
        !          4250: @samp{#ident} directive containing the text @var{string}.  If this
        !          4251: macro is not defined, nothing is output for a @samp{#ident} directive.
        !          4252: 
        !          4253: @findex OBJC_PROLOGUE
        !          4254: @item OBJC_PROLOGUE
        !          4255: A C statement to output any assembler statements which are required to
        !          4256: precede any Objective C object definitions or message sending.  The
        !          4257: statement is executed only when compiling an Objective C program.
        !          4258: @end table
        !          4259: 
        !          4260: @need 2000
        !          4261: @node Data Output
        !          4262: @subsection Output of Data
        !          4263: 
        !          4264: @table @code
        !          4265: @findex ASM_OUTPUT_LONG_DOUBLE
        !          4266: @findex ASM_OUTPUT_DOUBLE
        !          4267: @findex ASM_OUTPUT_FLOAT
        !          4268: @item ASM_OUTPUT_LONG_DOUBLE (@var{stream}, @var{value})
        !          4269: @itemx ASM_OUTPUT_DOUBLE (@var{stream}, @var{value})
        !          4270: @itemx ASM_OUTPUT_FLOAT (@var{stream}, @var{value})
        !          4271: A C statement to output to the stdio stream @var{stream} an assembler
        !          4272: instruction to assemble a floating-point constant of @code{TFmode},
        !          4273: @code{DFmode} or @code{SFmode}, respectively, whose value is
        !          4274: @var{value}.  @var{value} will be a C expression of type
        !          4275: @code{REAL_VALUE_TYPE}.  Macros such as
        !          4276: @code{REAL_VALUE_TO_TARGET_DOUBLE} are useful for writing these
        !          4277: definitions.
        !          4278: 
        !          4279: @findex ASM_OUTPUT_QUADRUPLE_INT
        !          4280: @findex ASM_OUTPUT_DOUBLE_INT
        !          4281: @findex ASM_OUTPUT_INT
        !          4282: @findex ASM_OUTPUT_SHORT
        !          4283: @findex ASM_OUTPUT_CHAR
        !          4284: @findex output_addr_const
        !          4285: @item ASM_OUTPUT_QUADRUPLE_INT (@var{stream}, @var{exp})
        !          4286: @itemx ASM_OUTPUT_DOUBLE_INT (@var{stream}, @var{exp})
        !          4287: @itemx ASM_OUTPUT_INT (@var{stream}, @var{exp})
        !          4288: @itemx ASM_OUTPUT_SHORT (@var{stream}, @var{exp})
        !          4289: @itemx ASM_OUTPUT_CHAR (@var{stream}, @var{exp})
        !          4290: A C statement to output to the stdio stream @var{stream} an assembler
        !          4291: instruction to assemble an integer of 16, 8, 4, 2 or 1 bytes,
        !          4292: respectively, whose value is @var{value}.  The argument @var{exp} will
        !          4293: be an RTL expression which represents a constant value.  Use
        !          4294: @samp{output_addr_const (@var{stream}, @var{exp})} to output this value
        !          4295: as an assembler expression.@refill
        !          4296: 
        !          4297: For sizes larger than @code{UNITS_PER_WORD}, if the action of a macro
        !          4298: would be identical to repeatedly calling the macro corresponding to
        !          4299: a size of @code{UNITS_PER_WORD}, once for each word, you need not define
        !          4300: the macro.
        !          4301: 
        !          4302: @findex ASM_OUTPUT_BYTE
        !          4303: @item ASM_OUTPUT_BYTE (@var{stream}, @var{value})
        !          4304: A C statement to output to the stdio stream @var{stream} an assembler
        !          4305: instruction to assemble a single byte containing the number @var{value}.
        !          4306: 
        !          4307: @findex ASM_BYTE_OP
        !          4308: @item ASM_BYTE_OP
        !          4309: A C string constant giving the pseudo-op to use for a sequence of
        !          4310: single-byte constants.  If this macro is not defined, the default is
        !          4311: @code{"byte"}.
        !          4312: 
        !          4313: @findex ASM_OUTPUT_ASCII
        !          4314: @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
        !          4315: A C statement to output to the stdio stream @var{stream} an assembler
        !          4316: instruction to assemble a string constant containing the @var{len}
        !          4317: bytes at @var{ptr}.  @var{ptr} will be a C expression of type
        !          4318: @code{char *} and @var{len} a C expression of type @code{int}.
        !          4319: 
        !          4320: If the assembler has a @code{.ascii} pseudo-op as found in the
        !          4321: Berkeley Unix assembler, do not define the macro
        !          4322: @code{ASM_OUTPUT_ASCII}.
        !          4323: 
        !          4324: @findex ASM_OUTPUT_POOL_PROLOGUE
        !          4325: @item ASM_OUTPUT_POOL_PROLOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
        !          4326: A C statement to output assembler commands to define the start of the
        !          4327: constant pool for a function.  @var{funname} is a string giving
        !          4328: the name of the function.  Should the return type of the function
        !          4329: be required, it can be obtained via @var{fundecl}.  @var{size}
        !          4330: is the size, in bytes, of the constant pool that will be written
        !          4331: immediately after this call.
        !          4332: 
        !          4333: If no constant-pool prefix is required, the usual case, this macro need
        !          4334: not be defined.
        !          4335: 
        !          4336: @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
        !          4337: @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
        !          4338: A C statement (with or without semicolon) to output a constant in the
        !          4339: constant pool, if it needs special treatment.  (This macro need not do
        !          4340: anything for RTL expressions that can be output normally.)
        !          4341: 
        !          4342: The argument @var{file} is the standard I/O stream to output the
        !          4343: assembler code on.  @var{x} is the RTL expression for the constant to
        !          4344: output, and @var{mode} is the machine mode (in case @var{x} is a
        !          4345: @samp{const_int}).  @var{align} is the required alignment for the value
        !          4346: @var{x}; you should output an assembler directive to force this much
        !          4347: alignment.
        !          4348: 
        !          4349: The argument @var{labelno} is a number to use in an internal label for
        !          4350: the address of this pool entry.  The definition of this macro is
        !          4351: responsible for outputting the label definition at the proper place.
        !          4352: Here is how to do this:
        !          4353: 
        !          4354: @example
        !          4355: ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
        !          4356: @end example
        !          4357: 
        !          4358: When you output a pool entry specially, you should end with a
        !          4359: @code{goto} to the label @var{jumpto}.  This will prevent the same pool
        !          4360: entry from being output a second time in the usual manner.
        !          4361: 
        !          4362: You need not define this macro if it would do nothing.
        !          4363: 
        !          4364: @findex ASM_OPEN_PAREN
        !          4365: @findex ASM_CLOSE_PAREN
        !          4366: @item ASM_OPEN_PAREN
        !          4367: @itemx ASM_CLOSE_PAREN
        !          4368: These macros are defined as C string constant, describing the syntax
        !          4369: in the assembler for grouping arithmetic expressions.  The following
        !          4370: definitions are correct for most assemblers:
        !          4371: 
        !          4372: @example
        !          4373: #define ASM_OPEN_PAREN "("
        !          4374: #define ASM_CLOSE_PAREN ")"
        !          4375: @end example
        !          4376: @end table
        !          4377: 
        !          4378:   These macros are provided by @file{real.h} for writing the definitions
        !          4379: of @code{ASM_OUTPUT_DOUBLE} and the like:
        !          4380: 
        !          4381: @table @code
        !          4382: @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
        !          4383: @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
        !          4384: @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
        !          4385: @findex REAL_VALUE_TO_TARGET_SINGLE
        !          4386: @findex REAL_VALUE_TO_TARGET_DOUBLE
        !          4387: @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
        !          4388: These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
        !          4389: floating point representation, and store its bit pattern in the array of
        !          4390: @code{long int} whose address is @var{l}.  The number of elements in the
        !          4391: output array is determined by the size of the desired target floating
        !          4392: point data type: 32 bits of it go in each @code{long int} array
        !          4393: element.  Each array element holds 32 bits of the result, even if
        !          4394: @code{long int} is wider than 32 bits on the host machine.
        !          4395: 
        !          4396: The array element values are designed so that you can print them out
        !          4397: using @code{fprintf} in the order they should appear in the target
        !          4398: machine's memory.
        !          4399: 
        !          4400: @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
        !          4401: @findex REAL_VALUE_TO_DECIMAL
        !          4402: This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
        !          4403: decimal number and stores it as a string into @var{string}.
        !          4404: You must pass, as @var{string}, the address of a long enough block
        !          4405: of space to hold the result.
        !          4406: 
        !          4407: The argument @var{format} is a @code{printf}-specification that serves
        !          4408: as a suggestion for how to format the output string.
        !          4409: @end table
        !          4410: 
        !          4411: @node Uninitialized Data
        !          4412: @subsection Output of Uninitialized Variables
        !          4413: 
        !          4414: Each of the macros in this section is used to do the whole job of
        !          4415: outputting a single uninitialized variable.
        !          4416: 
        !          4417: @table @code
        !          4418: @findex ASM_OUTPUT_COMMON
        !          4419: @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
        !          4420: A C statement (sans semicolon) to output to the stdio stream
        !          4421: @var{stream} the assembler definition of a common-label named
        !          4422: @var{name} whose size is @var{size} bytes.  The variable @var{rounded}
        !          4423: is the size rounded up to whatever alignment the caller wants.
        !          4424: 
        !          4425: Use the expression @code{assemble_name (@var{stream}, @var{name})} to
        !          4426: output the name itself; before and after that, output the additional
        !          4427: assembler syntax for defining the name, and a newline.
        !          4428: 
        !          4429: This macro controls how the assembler definitions of uninitialized
        !          4430: global variables are output.
        !          4431: 
        !          4432: @findex ASM_OUTPUT_ALIGNED_COMMON
        !          4433: @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
        !          4434: Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
        !          4435: separate, explicit argument.  If you define this macro, it is used in
        !          4436: place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
        !          4437: handling the required alignment of the variable.
        !          4438: 
        !          4439: @findex ASM_OUTPUT_SHARED_COMMON
        !          4440: @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
        !          4441: If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
        !          4442: is used when @var{name} is shared.  If not defined, @code{ASM_OUTPUT_COMMON}
        !          4443: will be used.
        !          4444: 
        !          4445: @findex ASM_OUTPUT_LOCAL
        !          4446: @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
        !          4447: A C statement (sans semicolon) to output to the stdio stream
        !          4448: @var{stream} the assembler definition of a local-common-label named
        !          4449: @var{name} whose size is @var{size} bytes.  The variable @var{rounded}
        !          4450: is the size rounded up to whatever alignment the caller wants.
        !          4451: 
        !          4452: Use the expression @code{assemble_name (@var{stream}, @var{name})} to
        !          4453: output the name itself; before and after that, output the additional
        !          4454: assembler syntax for defining the name, and a newline.
        !          4455: 
        !          4456: This macro controls how the assembler definitions of uninitialized
        !          4457: static variables are output.
        !          4458: 
        !          4459: @findex ASM_OUTPUT_ALIGNED_LOCAL
        !          4460: @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
        !          4461: Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
        !          4462: separate, explicit argument.  If you define this macro, it is used in
        !          4463: place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
        !          4464: handling the required alignment of the variable.
        !          4465: 
        !          4466: @findex ASM_OUTPUT_SHARED_LOCAL
        !          4467: @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
        !          4468: If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
        !          4469: is used when @var{name} is shared.  If not defined, @code{ASM_OUTPUT_LOCAL}
        !          4470: will be used.
        !          4471: @end table
        !          4472: 
        !          4473: @node Label Output
        !          4474: @subsection Output and Generation of Labels
        !          4475: 
        !          4476: @table @code
        !          4477: @findex ASM_OUTPUT_LABEL
        !          4478: @findex assemble_name
        !          4479: @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
        !          4480: A C statement (sans semicolon) to output to the stdio stream
        !          4481: @var{stream} the assembler definition of a label named @var{name}.
        !          4482: Use the expression @code{assemble_name (@var{stream}, @var{name})} to
        !          4483: output the name itself; before and after that, output the additional
        !          4484: assembler syntax for defining the name, and a newline.
        !          4485: 
        !          4486: @findex ASM_DECLARE_FUNCTION_NAME
        !          4487: @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
        !          4488: A C statement (sans semicolon) to output to the stdio stream
        !          4489: @var{stream} any text necessary for declaring the name @var{name} of a
        !          4490: function which is being defined.  This macro is responsible for
        !          4491: outputting the label definition (perhaps using
        !          4492: @code{ASM_OUTPUT_LABEL}).  The argument @var{decl} is the
        !          4493: @code{FUNCTION_DECL} tree node representing the function.
        !          4494: 
        !          4495: If this macro is not defined, then the function name is defined in the
        !          4496: usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
        !          4497: 
        !          4498: @findex ASM_DECLARE_FUNCTION_SIZE
        !          4499: @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
        !          4500: A C statement (sans semicolon) to output to the stdio stream
        !          4501: @var{stream} any text necessary for declaring the size of a function
        !          4502: which is being defined.  The argument @var{name} is the name of the
        !          4503: function.  The argument @var{decl} is the @code{FUNCTION_DECL} tree node
        !          4504: representing the function.
        !          4505: 
        !          4506: If this macro is not defined, then the function size is not defined.
        !          4507: 
        !          4508: @findex ASM_DECLARE_OBJECT_NAME
        !          4509: @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
        !          4510: A C statement (sans semicolon) to output to the stdio stream
        !          4511: @var{stream} any text necessary for declaring the name @var{name} of an
        !          4512: initialized variable which is being defined.  This macro must output the
        !          4513: label definition (perhaps using @code{ASM_OUTPUT_LABEL}).  The argument
        !          4514: @var{decl} is the @code{VAR_DECL} tree node representing the variable.
        !          4515: 
        !          4516: If this macro is not defined, then the variable name is defined in the
        !          4517: usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
        !          4518: 
        !          4519: @findex  ASM_FINISH_DECLARE_OBJECT
        !          4520: @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
        !          4521: A C statement (sans semicolon) to finish up declaring a variable name
        !          4522: once the compiler has processed its initializer fully and thus has had a
        !          4523: chance to determine the size of an array when controlled by an
        !          4524: initializer.  This is used on systems where it's necessary to declare
        !          4525: something about the size of the object.
        !          4526: 
        !          4527: If you don't define this macro, that is equivalent to defining it to do
        !          4528: nothing.
        !          4529: 
        !          4530: @findex ASM_GLOBALIZE_LABEL
        !          4531: @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
        !          4532: A C statement (sans semicolon) to output to the stdio stream
        !          4533: @var{stream} some commands that will make the label @var{name} global;
        !          4534: that is, available for reference from other files.  Use the expression
        !          4535: @code{assemble_name (@var{stream}, @var{name})} to output the name
        !          4536: itself; before and after that, output the additional assembler syntax
        !          4537: for making that name global, and a newline.
        !          4538: 
        !          4539: @findex ASM_OUTPUT_EXTERNAL
        !          4540: @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
        !          4541: A C statement (sans semicolon) to output to the stdio stream
        !          4542: @var{stream} any text necessary for declaring the name of an external
        !          4543: symbol named @var{name} which is referenced in this compilation but
        !          4544: not defined.  The value of @var{decl} is the tree node for the
        !          4545: declaration.
        !          4546: 
        !          4547: This macro need not be defined if it does not need to output anything.
        !          4548: The GNU assembler and most Unix assemblers don't require anything.
        !          4549: 
        !          4550: @findex ASM_OUTPUT_EXTERNAL_LIBCALL
        !          4551: @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
        !          4552: A C statement (sans semicolon) to output on @var{stream} an assembler
        !          4553: pseudo-op to declare a library function name external.  The name of the
        !          4554: library function is given by @var{symref}, which has type @code{rtx} and
        !          4555: is a @code{symbol_ref}.
        !          4556: 
        !          4557: This macro need not be defined if it does not need to output anything.
        !          4558: The GNU assembler and most Unix assemblers don't require anything.
        !          4559: 
        !          4560: @findex ASM_OUTPUT_LABELREF
        !          4561: @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
        !          4562: A C statement (sans semicolon) to output to the stdio stream
        !          4563: @var{stream} a reference in assembler syntax to a label named
        !          4564: @var{name}.  This should add @samp{_} to the front of the name, if that
        !          4565: is customary on your operating system, as it is in most Berkeley Unix
        !          4566: systems.  This macro is used in @code{assemble_name}.
        !          4567: 
        !          4568: @ignore @c Seems not to exist anymore.
        !          4569: @findex ASM_OUTPUT_LABELREF_AS_INT
        !          4570: @item ASM_OUTPUT_LABELREF_AS_INT (@var{file}, @var{label})
        !          4571: Define this macro for systems that use the program @code{collect2}.
        !          4572: The definition should be a C statement to output a word containing
        !          4573: a reference to the label @var{label}.
        !          4574: @end ignore
        !          4575: 
        !          4576: @findex ASM_OUTPUT_INTERNAL_LABEL
        !          4577: @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
        !          4578: A C statement to output to the stdio stream @var{stream} a label whose
        !          4579: name is made from the string @var{prefix} and the number @var{num}.
        !          4580: 
        !          4581: It is absolutely essential that these labels be distinct from the labels
        !          4582: used for user-level functions and variables.  Otherwise, certain programs
        !          4583: will have name conflicts with internal labels.
        !          4584: 
        !          4585: It is desirable to exclude internal labels from the symbol table of the
        !          4586: object file.  Most assemblers have a naming convention for labels that
        !          4587: should be excluded; on many systems, the letter @samp{L} at the
        !          4588: beginning of a label has this effect.  You should find out what
        !          4589: convention your system uses, and follow it.
        !          4590: 
        !          4591: The usual definition of this macro is as follows:
        !          4592: 
        !          4593: @example
        !          4594: fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
        !          4595: @end example
        !          4596: 
        !          4597: @findex ASM_GENERATE_INTERNAL_LABEL
        !          4598: @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
        !          4599: A C statement to store into the string @var{string} a label whose name
        !          4600: is made from the string @var{prefix} and the number @var{num}.
        !          4601: 
        !          4602: This string, when output subsequently by @code{assemble_name}, should
        !          4603: produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
        !          4604: with the same @var{prefix} and @var{num}.
        !          4605: 
        !          4606: If the string begins with @samp{*}, then @code{assemble_name} will
        !          4607: output the rest of the string unchanged.  It is often convenient for
        !          4608: @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way.  If the
        !          4609: string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
        !          4610: to output the string, and may change it.  (Of course,
        !          4611: @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
        !          4612: you should know what it does on your machine.)
        !          4613: 
        !          4614: @findex ASM_FORMAT_PRIVATE_NAME
        !          4615: @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
        !          4616: A C expression to assign to @var{outvar} (which is a variable of type
        !          4617: @code{char *}) a newly allocated string made from the string
        !          4618: @var{name} and the number @var{number}, with some suitable punctuation
        !          4619: added.  Use @code{alloca} to get space for the string.
        !          4620: 
        !          4621: The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
        !          4622: produce an assembler label for an internal static variable whose name is
        !          4623: @var{name}.  Therefore, the string must be such as to result in valid
        !          4624: assembler code.  The argument @var{number} is different each time this
        !          4625: macro is executed; it prevents conflicts between similarly-named
        !          4626: internal static variables in different scopes.
        !          4627: 
        !          4628: Ideally this string should not be a valid C identifier, to prevent any
        !          4629: conflict with the user's own symbols.  Most assemblers allow periods
        !          4630: or percent signs in assembler symbols; putting at least one of these
        !          4631: between the name and the number will suffice.
        !          4632: 
        !          4633: @findex OBJC_GEN_METHOD_LABEL
        !          4634: @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
        !          4635: Define this macro to override the default assembler names used for
        !          4636: Objective C methods.
        !          4637: 
        !          4638: The default name is a unique method number followed by the name of the
        !          4639: class (e.g.@: @samp{_1_Foo}).  For methods in categories, the name of
        !          4640: the category is also included in the assembler name (e.g.@:
        !          4641: @samp{_1_Foo_Bar}).
        !          4642: 
        !          4643: These names are safe on most systems, but make debugging difficult since
        !          4644: the method's selector is not present in the name.  Therefore, particular
        !          4645: systems define other ways of computing names.
        !          4646: 
        !          4647: @var{buf} is an expression of type @code{char *} which gives you a
        !          4648: buffer in which to store the name; its length is as long as
        !          4649: @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
        !          4650: 50 characters extra.
        !          4651: 
        !          4652: The argument @var{is_inst} specifies whether the method is an instance
        !          4653: method or a class method; @var{class_name} is the name of the class;
        !          4654: @var{cat_name} is the name of the category (or NULL if the method is not
        !          4655: in a category); and @var{sel_name} is the name of the selector.
        !          4656: 
        !          4657: On systems where the assembler can handle quoted names, you can use this
        !          4658: macro to provide more human-readable names.
        !          4659: @end table
        !          4660: 
        !          4661: @node Initialization
        !          4662: @subsection How Initialization Functions Are Handled
        !          4663: @cindex initialization routines
        !          4664: @cindex termination routines
        !          4665: @cindex constructors, output of
        !          4666: @cindex destructors, output of
        !          4667: 
        !          4668: The compiled code for certain languages includes @dfn{constructors}
        !          4669: (also called @dfn{initialization routines})---functions to initialize
        !          4670: data in the program when the program is started.  These functions need
        !          4671: to be called before the program is ``started''---that is to say, before
        !          4672: @code{main} is called.
        !          4673: 
        !          4674: Compiling some languages generates @dfn{destructors} (also called
        !          4675: @dfn{termination routines}) that should be called when the program
        !          4676: terminates.
        !          4677: 
        !          4678: To make the initialization and termination functions work, the compiler
        !          4679: must output something in the assembler code to cause those functions to
        !          4680: be called at the appropriate time.  When you port the compiler to a new
        !          4681: system, you need to specify how to do this.
        !          4682: 
        !          4683: There are two major ways that GCC currently supports the execution of
        !          4684: initialization and termination functions.  Each way has two variants.
        !          4685: Much of the structure is common to all four variations.
        !          4686: 
        !          4687: @findex __CTOR_LIST__
        !          4688: @findex __DTOR_LIST__
        !          4689: The linker must build two lists of these functions---a list of
        !          4690: initialization functions, called @code{__CTOR_LIST__}, and a list of
        !          4691: termination functions, called @code{__DTOR_LIST__}.
        !          4692: 
        !          4693: Each list always begins with an ignored function pointer (which may hold
        !          4694: 0, @minus{}1, or a count of the function pointers after it, depending on
        !          4695: the environment).  This is followed by a series of zero or more function
        !          4696: pointers to constructors (or destructors), followed by a function
        !          4697: pointer containing zero.
        !          4698: 
        !          4699: Depending on the operating system and its executable file format, either
        !          4700: @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
        !          4701: time and exit time.  Constructors are called in forward order of the
        !          4702: list; destructors in reverse order.
        !          4703: 
        !          4704: The best way to handle static constructors works only for object file
        !          4705: formats which provide arbitrarily-named sections.  A section is set
        !          4706: aside for a list of constructors, and another for a list of destructors.
        !          4707: Traditionally these are called @samp{.ctors} and @samp{.dtors}.  Each
        !          4708: object file that defines an initialization function also puts a word in
        !          4709: the constructor section to point to that function.  The linker
        !          4710: accumulates all these words into one contiguous @samp{.ctors} section.
        !          4711: Termination functions are handled similarly.
        !          4712: 
        !          4713: To use this method, you need appropriate definitions of the macros
        !          4714: @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}.  Usually
        !          4715: you can get them by including @file{svr4.h}.
        !          4716: 
        !          4717: When arbitrary sections are available, there are two variants, depending
        !          4718: upon how the code in @file{crtstuff.c} is called.  On systems that
        !          4719: support an @dfn{init} section which is executed at program startup,
        !          4720: parts of @file{crtstuff.c} are compiled into that section.  The
        !          4721: program is linked by the @code{gcc} driver like this:
        !          4722: 
        !          4723: @example
        !          4724: ld -o @var{output_file} crtbegin.o @dots{} crtend.o -lgcc
        !          4725: @end example
        !          4726: 
        !          4727: The head of a function (@code{__do_global_ctors}) appears in the init
        !          4728: section of @file{crtbegin.o}; the remainder of the function appears in
        !          4729: the init section of @file{crtend.o}.  The linker will pull these two
        !          4730: parts of the section together, making a whole function.  If any of the
        !          4731: user's object files linked into the middle of it contribute code, then that
        !          4732: code will be executed as part of the body of @code{__do_global_ctors}.
        !          4733: 
        !          4734: To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
        !          4735: macro properly.
        !          4736: 
        !          4737: If no init section is available, do not define
        !          4738: @code{INIT_SECTION_ASM_OP}.  Then @code{__do_global_ctors} is built into
        !          4739: the text section like all other functions, and resides in
        !          4740: @file{libgcc.a}.  When GCC compiles any function called @code{main}, it
        !          4741: inserts a procedure call to @code{__main} as the first executable code
        !          4742: after the function prologue.  The @code{__main} function, also defined
        !          4743: in @file{libgcc2.c}, simply calls @file{__do_global_ctors}.
        !          4744: 
        !          4745: In file formats that don't support arbitrary sections, there are again
        !          4746: two variants.  In the simplest variant, the GNU linker (GNU @code{ld})
        !          4747: and an `a.out' format must be used.  In this case,
        !          4748: @code{ASM_OUTPUT_CONSTRUCTOR} is defined to produce a @code{.stabs}
        !          4749: entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
        !          4750: and with the address of the void function containing the initialization
        !          4751: code as its value.  The GNU linker recognizes this as a request to add
        !          4752: the value to a ``set''; the values are accumulated, and are eventually
        !          4753: placed in the executable as a vector in the format described above, with
        !          4754: a leading (ignored) count and a trailing zero element.
        !          4755: @code{ASM_OUTPUT_DESTRUCTOR} is handled similarly.  Since no init
        !          4756: section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
        !          4757: the compilation of @code{main} to call @code{__main} as above, starting
        !          4758: the initialization process.
        !          4759: 
        !          4760: The last variant uses neither arbitrary sections nor the GNU linker.
        !          4761: This is preferable when you want to do dynamic linking and when using
        !          4762: file formats which the GNU linker does not support, such as `ECOFF'.  In
        !          4763: this case, @code{ASM_OUTPUT_CONSTRUCTOR} does not produce an
        !          4764: @code{N_SETT} symbol; initialization and termination functions are
        !          4765: recognized simply by their names.  This requires an extra program in the
        !          4766: linkage step, called @code{collect2}.  This program pretends to be the
        !          4767: linker, for use with GNU CC; it does its job by running the ordinary
        !          4768: linker, but also arranges to include the vectors of initialization and
        !          4769: termination functions.  These functions are called via @code{__main} as
        !          4770: described above.
        !          4771: 
        !          4772: Choosing among these configuration options has been simplified by a set
        !          4773: of operating-system-dependent files in the @file{config} subdirectory.
        !          4774: These files define all of the relevant parameters.  Usually it is
        !          4775: sufficient to include one into your specific machine-dependent
        !          4776: configuration file.  These files are:
        !          4777: 
        !          4778: @table @file
        !          4779: @item aoutos.h
        !          4780: For operating systems using the `a.out' format.
        !          4781: 
        !          4782: @item next.h
        !          4783: For operating systems using the `MachO' format.
        !          4784: 
        !          4785: @item svr3.h
        !          4786: For System V Release 3 and similar systems using `COFF' format.
        !          4787: 
        !          4788: @item svr4.h
        !          4789: For System V Release 4 and similar systems using `ELF' format.
        !          4790: 
        !          4791: @item vms.h
        !          4792: For the VMS operating system.
        !          4793: @end table
        !          4794: 
        !          4795: @ifinfo
        !          4796: The following section describes the specific macros that control and
        !          4797: customize the handling of initialization and termination functions.
        !          4798: @end ifinfo
        !          4799: 
        !          4800: @node Macros for Initialization
        !          4801: @subsection Macros Controlling Initialization Routines
        !          4802: 
        !          4803: Here are the macros that control how the compiler handles initialization
        !          4804: and termination functions:
        !          4805: 
        !          4806: @table @code
        !          4807: @findex INIT_SECTION_ASM_OP
        !          4808: @item INIT_SECTION_ASM_OP
        !          4809: If defined, a C string constant for the assembler operation to identify
        !          4810: the following data as initialization code.  If not defined, GNU CC will
        !          4811: assume such a section does not exist.  When you are using special
        !          4812: sections for initialization and termination functions, this macro also
        !          4813: controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to run the
        !          4814: initialization functions.
        !          4815: 
        !          4816: @item ASM_OUTPUT_CONSTRUCTOR (@var{stream}, @var{name})
        !          4817: @findex ASM_OUTPUT_CONSTRUCTOR
        !          4818: Define this macro as a C statement to output on the stream @var{stream}
        !          4819: the assembler code to arrange to call the function named @var{name} at
        !          4820: initialization time.
        !          4821: 
        !          4822: Assume that @var{name} is the name of a C function generated
        !          4823: automatically by the compiler.  This function takes no arguments.  Use
        !          4824: the function @code{assemble_name} to output the name @var{name}; this
        !          4825: performs any system-specific syntactic transformations such as adding an
        !          4826: underscore.
        !          4827: 
        !          4828: If you don't define this macro, nothing special is output to arrange to
        !          4829: call the function.  This is correct when the function will be called in
        !          4830: some other manner---for example, by means of the @code{collect2} program,
        !          4831: which looks through the symbol table to find these functions by their
        !          4832: names.
        !          4833: 
        !          4834: @item ASM_OUTPUT_DESTRUCTOR (@var{stream}, @var{name})
        !          4835: @findex ASM_OUTPUT_DESTRUCTOR
        !          4836: This is like @code{ASM_OUTPUT_CONSTRUCTOR} but used for termination
        !          4837: functions rather than initialization functions.
        !          4838: @end table
        !          4839: 
        !          4840: If your system uses @code{collect2} as the means of processing
        !          4841: constructors, then that program normally uses @code{nm} to scan an
        !          4842: object file for constructor functions to be called.  On certain kinds of
        !          4843: systems, you can define these macros to make @code{collect2} work faster
        !          4844: (and, in some cases, make it work at all):
        !          4845: 
        !          4846: @table @code
        !          4847: @findex OBJECT_FORMAT_COFF
        !          4848: @item OBJECT_FORMAT_COFF
        !          4849: Define this macro if the system uses COFF (Common Object File Format)
        !          4850: object files, so that @code{collect2} can assume this format and scan
        !          4851: object files directly for dynamic constructor/destructor functions.
        !          4852: 
        !          4853: @findex OBJECT_FORMAT_ROSE
        !          4854: @item OBJECT_FORMAT_ROSE
        !          4855: Define this macro if the system uses ROSE format object files, so that
        !          4856: @code{collect2} can assume this format and scan object files directly
        !          4857: for dynamic constructor/destructor functions.
        !          4858: @end table
        !          4859: 
        !          4860: These macros are effective only in a native compiler; @code{collect2} as
        !          4861: part of a cross compiler always uses @code{nm}.
        !          4862: 
        !          4863: @table @code
        !          4864: @findex REAL_NM_FILE_NAME
        !          4865: @item REAL_NM_FILE_NAME
        !          4866: Define this macro as a C string constant containing the file name to use
        !          4867: to execute @code{nm}.  The default is to search the path normally for
        !          4868: @code{nm}.
        !          4869: @end table
        !          4870: 
        !          4871: @node Instruction Output
        !          4872: @subsection Output of Assembler Instructions
        !          4873: 
        !          4874: @table @code
        !          4875: @findex REGISTER_NAMES
        !          4876: @item REGISTER_NAMES
        !          4877: A C initializer containing the assembler's names for the machine
        !          4878: registers, each one as a C string constant.  This is what translates
        !          4879: register numbers in the compiler into assembler language.
        !          4880: 
        !          4881: @findex ADDITIONAL_REGISTER_NAMES
        !          4882: @item ADDITIONAL_REGISTER_NAMES
        !          4883: If defined, a C initializer for an array of structures containing a name
        !          4884: and a register number.  This macro defines additional names for hard
        !          4885: registers, thus allowing the @code{asm} option in declarations to refer
        !          4886: to registers using alternate names.
        !          4887: 
        !          4888: @findex ASM_OUTPUT_OPCODE
        !          4889: @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
        !          4890: Define this macro if you are using an unusual assembler that
        !          4891: requires different names for the machine instructions.
        !          4892: 
        !          4893: The definition is a C statement or statements which output an
        !          4894: assembler instruction opcode to the stdio stream @var{stream}.  The
        !          4895: macro-operand @var{ptr} is a variable of type @code{char *} which
        !          4896: points to the opcode name in its ``internal'' form---the form that is
        !          4897: written in the machine description.  The definition should output the
        !          4898: opcode name to @var{stream}, performing any translation you desire, and
        !          4899: increment the variable @var{ptr} to point at the end of the opcode
        !          4900: so that it will not be output twice.
        !          4901: 
        !          4902: In fact, your macro definition may process less than the entire opcode
        !          4903: name, or more than the opcode name; but if you want to process text
        !          4904: that includes @samp{%}-sequences to substitute operands, you must take
        !          4905: care of the substitution yourself.  Just be sure to increment
        !          4906: @var{ptr} over whatever text should not be output normally.
        !          4907: 
        !          4908: @findex recog_operand
        !          4909: If you need to look at the operand values, they can be found as the
        !          4910: elements of @code{recog_operand}.
        !          4911: 
        !          4912: If the macro definition does nothing, the instruction is output
        !          4913: in the usual way.
        !          4914: 
        !          4915: @findex FINAL_PRESCAN_INSN
        !          4916: @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
        !          4917: If defined, a C statement to be executed just prior to the output of
        !          4918: assembler code for @var{insn}, to modify the extracted operands so
        !          4919: they will be output differently.
        !          4920: 
        !          4921: Here the argument @var{opvec} is the vector containing the operands
        !          4922: extracted from @var{insn}, and @var{noperands} is the number of
        !          4923: elements of the vector which contain meaningful data for this insn.
        !          4924: The contents of this vector are what will be used to convert the insn
        !          4925: template into assembler code, so you can change the assembler output
        !          4926: by changing the contents of the vector.
        !          4927: 
        !          4928: This macro is useful when various assembler syntaxes share a single
        !          4929: file of instruction patterns; by defining this macro differently, you
        !          4930: can cause a large class of instructions to be output differently (such
        !          4931: as with rearranged operands).  Naturally, variations in assembler
        !          4932: syntax affecting individual insn patterns ought to be handled by
        !          4933: writing conditional output routines in those patterns.
        !          4934: 
        !          4935: If this macro is not defined, it is equivalent to a null statement.
        !          4936: 
        !          4937: @findex PRINT_OPERAND
        !          4938: @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
        !          4939: A C compound statement to output to stdio stream @var{stream} the
        !          4940: assembler syntax for an instruction operand @var{x}.  @var{x} is an
        !          4941: RTL expression.
        !          4942: 
        !          4943: @var{code} is a value that can be used to specify one of several ways
        !          4944: of printing the operand.  It is used when identical operands must be
        !          4945: printed differently depending on the context.  @var{code} comes from
        !          4946: the @samp{%} specification that was used to request printing of the
        !          4947: operand.  If the specification was just @samp{%@var{digit}} then
        !          4948: @var{code} is 0; if the specification was @samp{%@var{ltr}
        !          4949: @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
        !          4950: 
        !          4951: @findex reg_names
        !          4952: If @var{x} is a register, this macro should print the register's name.
        !          4953: The names can be found in an array @code{reg_names} whose type is
        !          4954: @code{char *[]}.  @code{reg_names} is initialized from
        !          4955: @code{REGISTER_NAMES}.
        !          4956: 
        !          4957: When the machine description has a specification @samp{%@var{punct}}
        !          4958: (a @samp{%} followed by a punctuation character), this macro is called
        !          4959: with a null pointer for @var{x} and the punctuation character for
        !          4960: @var{code}.
        !          4961: 
        !          4962: @findex PRINT_OPERAND_PUNCT_VALID_P
        !          4963: @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
        !          4964: A C expression which evaluates to true if @var{code} is a valid
        !          4965: punctuation character for use in the @code{PRINT_OPERAND} macro.  If
        !          4966: @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
        !          4967: punctuation characters (except for the standard one, @samp{%}) are used
        !          4968: in this way.
        !          4969: 
        !          4970: @findex PRINT_OPERAND_ADDRESS
        !          4971: @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
        !          4972: A C compound statement to output to stdio stream @var{stream} the
        !          4973: assembler syntax for an instruction operand that is a memory reference
        !          4974: whose address is @var{x}.  @var{x} is an RTL expression.
        !          4975: 
        !          4976: @cindex @code{ENCODE_SECTION_INFO} usage
        !          4977: On some machines, the syntax for a symbolic address depends on the
        !          4978: section that the address refers to.  On these machines, define the macro
        !          4979: @code{ENCODE_SECTION_INFO} to store the information into the
        !          4980: @code{symbol_ref}, and then check for it here.  @xref{Assembler Format}.
        !          4981: 
        !          4982: @findex DBR_OUTPUT_SEQEND
        !          4983: @findex dbr_sequence_length
        !          4984: @item DBR_OUTPUT_SEQEND(@var{file})
        !          4985: A C statement, to be executed after all slot-filler instructions have
        !          4986: been output.  If necessary, call @code{dbr_sequence_length} to
        !          4987: determine the number of slots filled in a sequence (zero if not
        !          4988: currently outputting a sequence), to decide how many no-ops to output,
        !          4989: or whatever.
        !          4990: 
        !          4991: Don't define this macro if it has nothing to do, but it is helpful in
        !          4992: reading assembly output if the extent of the delay sequence is made
        !          4993: explicit (e.g. with white space).
        !          4994: 
        !          4995: @findex final_sequence
        !          4996: Note that output routines for instructions with delay slots must be
        !          4997: prepared to deal with not being output as part of a sequence (i.e.
        !          4998: when the scheduling pass is not run, or when no slot fillers could be
        !          4999: found.)  The variable @code{final_sequence} is null when not
        !          5000: processing a sequence, otherwise it contains the @code{sequence} rtx
        !          5001: being output.
        !          5002: 
        !          5003: @findex REGISTER_PREFIX
        !          5004: @findex LOCAL_LABEL_PREFIX
        !          5005: @findex USER_LABEL_PREFIX
        !          5006: @findex IMMEDIATE_PREFIX
        !          5007: @findex asm_fprintf
        !          5008: @item REGISTER_PREFIX
        !          5009: @itemx LOCAL_LABEL_PREFIX
        !          5010: @itemx USER_LABEL_PREFIX
        !          5011: @itemx IMMEDIATE_PREFIX
        !          5012: If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
        !          5013: @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
        !          5014: @file{final.c}).  These are useful when a single @file{md} file must
        !          5015: support multiple assembler formats.  In that case, the various @file{tm.h}
        !          5016: files can define these macros differently.
        !          5017: 
        !          5018: @findex ASSEMBLER_DIALECT
        !          5019: @item ASSEMBLER_DIALECT
        !          5020: If your target supports multiple dialects of assembler language (such as
        !          5021: different opcodes), define this macro as a C expression that gives the
        !          5022: numeric index of the assembler langauge dialect to use, with zero as the
        !          5023: first variant.
        !          5024: 
        !          5025: If this macro is defined, you may use
        !          5026: @samp{@{option0|option1|option2@dots{}@}} constructs in the output
        !          5027: templates of patterns (@pxref{Output Template}) or in the first argument
        !          5028: of @code{asm_fprintf}.  This construct outputs @samp{option0},
        !          5029: @samp{option1} or @samp{option2}, etc., if the value of
        !          5030: @code{ASSEMBLER_DIALECT} is zero, one or two, etc.  Any special
        !          5031: characters within these strings retain their usual meaning.
        !          5032: 
        !          5033: If you do not define this macro, the characters @samp{@{}, @samp{|} and
        !          5034: @samp{@}} do not have any special meaning when used in templates or
        !          5035: operands to @code{asm_fprintf}.
        !          5036: 
        !          5037: Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
        !          5038: @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
        !          5039: the variations in assemble language syntax with that mechanism.  Define
        !          5040: @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
        !          5041: if the syntax variant are larger and involve such things as different
        !          5042: opcodes or operand order.
        !          5043: 
        !          5044: @findex ASM_OUTPUT_REG_PUSH
        !          5045: @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
        !          5046: A C expression to output to @var{stream} some assembler code
        !          5047: which will push hard register number @var{regno} onto the stack.
        !          5048: The code need not be optimal, since this macro is used only when
        !          5049: profiling.
        !          5050: 
        !          5051: @findex ASM_OUTPUT_REG_POP
        !          5052: @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
        !          5053: A C expression to output to @var{stream} some assembler code
        !          5054: which will pop hard register number @var{regno} off of the stack.
        !          5055: The code need not be optimal, since this macro is used only when
        !          5056: profiling.
        !          5057: @end table
        !          5058: 
        !          5059: @node Dispatch Tables
        !          5060: @subsection Output of Dispatch Tables
        !          5061: 
        !          5062: @table @code
        !          5063: @cindex dispatch table
        !          5064: @findex ASM_OUTPUT_ADDR_DIFF_ELT
        !          5065: @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{value}, @var{rel})
        !          5066: This macro should be provided on machines where the addresses
        !          5067: in a dispatch table are relative to the table's own address.
        !          5068: 
        !          5069: The definition should be a C statement to output to the stdio stream
        !          5070: @var{stream} an assembler pseudo-instruction to generate a difference
        !          5071: between two labels.  @var{value} and @var{rel} are the numbers of two
        !          5072: internal labels.  The definitions of these labels are output using
        !          5073: @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
        !          5074: way here.  For example,
        !          5075: 
        !          5076: @example
        !          5077: fprintf (@var{stream}, "\t.word L%d-L%d\n",
        !          5078:          @var{value}, @var{rel})
        !          5079: @end example
        !          5080: 
        !          5081: @findex ASM_OUTPUT_ADDR_VEC_ELT
        !          5082: @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
        !          5083: This macro should be provided on machines where the addresses
        !          5084: in a dispatch table are absolute.
        !          5085: 
        !          5086: The definition should be a C statement to output to the stdio stream
        !          5087: @var{stream} an assembler pseudo-instruction to generate a reference to
        !          5088: a label.  @var{value} is the number of an internal label whose
        !          5089: definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
        !          5090: For example,
        !          5091: 
        !          5092: @example
        !          5093: fprintf (@var{stream}, "\t.word L%d\n", @var{value})
        !          5094: @end example
        !          5095: 
        !          5096: @findex ASM_OUTPUT_CASE_LABEL
        !          5097: @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
        !          5098: Define this if the label before a jump-table needs to be output
        !          5099: specially.  The first three arguments are the same as for
        !          5100: @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
        !          5101: jump-table which follows (a @code{jump_insn} containing an
        !          5102: @code{addr_vec} or @code{addr_diff_vec}).
        !          5103: 
        !          5104: This feature is used on system V to output a @code{swbeg} statement
        !          5105: for the table.
        !          5106: 
        !          5107: If this macro is not defined, these labels are output with
        !          5108: @code{ASM_OUTPUT_INTERNAL_LABEL}.
        !          5109: 
        !          5110: @findex ASM_OUTPUT_CASE_END
        !          5111: @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
        !          5112: Define this if something special must be output at the end of a
        !          5113: jump-table.  The definition should be a C statement to be executed
        !          5114: after the assembler code for the table is written.  It should write
        !          5115: the appropriate code to stdio stream @var{stream}.  The argument
        !          5116: @var{table} is the jump-table insn, and @var{num} is the label-number
        !          5117: of the preceding label.
        !          5118: 
        !          5119: If this macro is not defined, nothing special is output at the end of
        !          5120: the jump-table.
        !          5121: @end table
        !          5122: 
        !          5123: @node Alignment Output
        !          5124: @subsection Assembler Commands for Alignment
        !          5125: 
        !          5126: @table @code
        !          5127: @findex ASM_OUTPUT_ALIGN_CODE
        !          5128: @item ASM_OUTPUT_ALIGN_CODE (@var{file})
        !          5129: A C expression to output text to align the location counter in the way
        !          5130: that is desirable at a point in the code that is reached only by
        !          5131: jumping.
        !          5132: 
        !          5133: This macro need not be defined if you don't want any special alignment
        !          5134: to be done at such a time.  Most machine descriptions do not currently
        !          5135: define the macro.
        !          5136: 
        !          5137: @findex ASM_OUTPUT_LOOP_ALIGN
        !          5138: @item ASM_OUTPUT_LOOP_ALIGN (@var{file})
        !          5139: A C expression to output text to align the location counter in the way
        !          5140: that is desirable at the beginning of a loop.
        !          5141: 
        !          5142: This macro need not be defined if you don't want any special alignment
        !          5143: to be done at such a time.  Most machine descriptions do not currently
        !          5144: define the macro.
        !          5145: 
        !          5146: @findex ASM_OUTPUT_SKIP
        !          5147: @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
        !          5148: A C statement to output to the stdio stream @var{stream} an assembler
        !          5149: instruction to advance the location counter by @var{nbytes} bytes.
        !          5150: Those bytes should be zero when loaded.  @var{nbytes} will be a C
        !          5151: expression of type @code{int}.
        !          5152: 
        !          5153: @findex ASM_NO_SKIP_IN_TEXT
        !          5154: @item ASM_NO_SKIP_IN_TEXT
        !          5155: Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
        !          5156: text section because it fails put zeros in the bytes that are skipped.
        !          5157: This is true on many Unix systems, where the pseudo--op to skip bytes
        !          5158: produces no-op instructions rather than zeros when used in the text
        !          5159: section.
        !          5160: 
        !          5161: @findex ASM_OUTPUT_ALIGN
        !          5162: @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
        !          5163: A C statement to output to the stdio stream @var{stream} an assembler
        !          5164: command to advance the location counter to a multiple of 2 to the
        !          5165: @var{power} bytes.  @var{power} will be a C expression of type @code{int}.
        !          5166: @end table
        !          5167: 
        !          5168: @need 3000
        !          5169: @node Debugging Info
        !          5170: @section Controlling Debugging Information Format
        !          5171: @c this should have some text in here....  --mew 10feb93
        !          5172: 
        !          5173: @menu
        !          5174: * All Debuggers::      Macros that affect all debugging formats uniformly.
        !          5175: * DBX Options::        Macros enabling specific options in DBX format.
        !          5176: * DBX Hooks::          Hook macros for varying DBX format.
        !          5177: * File Names and DBX:: Macros controlling output of file names in DBX format.
        !          5178: * SDB and DWARF::      Macros for SDB (COFF) and DWARF formats.
        !          5179: @end menu
        !          5180: 
        !          5181: @node All Debuggers
        !          5182: @subsection Macros Affecting All Debugging Formats
        !          5183: 
        !          5184: @table @code
        !          5185: @findex DBX_REGISTER_NUMBER
        !          5186: @item DBX_REGISTER_NUMBER (@var{regno})
        !          5187: A C expression that returns the DBX register number for the compiler
        !          5188: register number @var{regno}.  In simple cases, the value of this
        !          5189: expression may be @var{regno} itself.  But sometimes there are some
        !          5190: registers that the compiler knows about and DBX does not, or vice
        !          5191: versa.  In such cases, some register may need to have one number in
        !          5192: the compiler and another for DBX.
        !          5193: 
        !          5194: If two registers have consecutive numbers inside GNU CC, and they can be
        !          5195: used as a pair to hold a multiword value, then they @emph{must} have
        !          5196: consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
        !          5197: Otherwise, debuggers will be unable to access such a pair, because they
        !          5198: expect register pairs to be consecutive in their own numbering scheme.
        !          5199: 
        !          5200: If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
        !          5201: does not preserve register pairs, then what you must do instead is
        !          5202: redefine the actual register numbering scheme.
        !          5203: 
        !          5204: @findex DEBUGGER_AUTO_OFFSET
        !          5205: @item DEBUGGER_AUTO_OFFSET (@var{x})
        !          5206: A C expression that returns the integer offset value for an automatic
        !          5207: variable having address @var{x} (an RTL expression).  The default
        !          5208: computation assumes that @var{x} is based on the frame-pointer and
        !          5209: gives the offset from the frame-pointer.  This is required for targets
        !          5210: that produce debugging output for DBX or COFF-style debugging output
        !          5211: for SDB and allow the frame-pointer to be eliminated when the
        !          5212: @samp{-g} options is used.
        !          5213: 
        !          5214: @findex DEBUGGER_ARG_OFFSET
        !          5215: @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
        !          5216: A C expression that returns the integer offset value for an argument
        !          5217: having address @var{x} (an RTL expression).  The nominal offset is
        !          5218: @var{offset}.
        !          5219: 
        !          5220: @findex PREFERRED_DEBUGGING_TYPE
        !          5221: @item PREFERRED_DEBUGGING_TYPE
        !          5222: A C expression that returns the type of debugging output GNU CC produces
        !          5223: when the user specifies @samp{-g} or @samp{-ggdb}.  Define this if you
        !          5224: have arranged for GNU CC to support more than one format of debugging
        !          5225: output.  Currently, the allowable values are @code{DBX_DEBUG},
        !          5226: @code{SDB_DEBUG}, @code{DWARF_DEBUG}, and @code{XCOFF_DEBUG}.
        !          5227: 
        !          5228: The value of this macro only affects the default debugging output; the
        !          5229: user can always get a specific type of output by using @samp{-gstabs},
        !          5230: @samp{-gcoff}, @samp{-gdwarf}, or @samp{-gxcoff}.
        !          5231: @end table
        !          5232: 
        !          5233: @node DBX Options
        !          5234: @subsection Specific Options for DBX Output
        !          5235: 
        !          5236: @table @code
        !          5237: @findex DBX_DEBUGGING_INFO
        !          5238: @item DBX_DEBUGGING_INFO
        !          5239: Define this macro if GNU CC should produce debugging output for DBX
        !          5240: in response to the @samp{-g} option.
        !          5241: 
        !          5242: @findex XCOFF_DEBUGGING_INFO
        !          5243: @item XCOFF_DEBUGGING_INFO
        !          5244: Define this macro if GNU CC should produce XCOFF format debugging output
        !          5245: in response to the @samp{-g} option.  This is a variant of DBX format.
        !          5246: 
        !          5247: @findex DEFAULT_GDB_EXTENSIONS
        !          5248: @item DEFAULT_GDB_EXTENSIONS
        !          5249: Define this macro to control whether GNU CC should by default generate
        !          5250: GDB's extended version of DBX debugging information (assuming DBX-format
        !          5251: debugging information is enabled at all).  If you don't define the
        !          5252: macro, the default is 1: always generate the extended information
        !          5253: if there is any occasion to.
        !          5254: 
        !          5255: @findex DEBUG_SYMS_TEXT
        !          5256: @item DEBUG_SYMS_TEXT
        !          5257: Define this macro if all @code{.stabs} commands should be output while
        !          5258: in the text section.
        !          5259: 
        !          5260: @findex ASM_STABS_OP
        !          5261: @item ASM_STABS_OP
        !          5262: A C string constant naming the assembler pseudo op to use instead of
        !          5263: @code{.stabs} to define an ordinary debugging symbol.  If you don't
        !          5264: define this macro, @code{.stabs} is used.  This macro applies only to
        !          5265: DBX debugging information format.
        !          5266: 
        !          5267: @findex ASM_STABD_OP
        !          5268: @item ASM_STABD_OP
        !          5269: A C string constant naming the assembler pseudo op to use instead of
        !          5270: @code{.stabd} to define a debugging symbol whose value is the current
        !          5271: location.  If you don't define this macro, @code{.stabd} is used.
        !          5272: This macro applies only to DBX debugging information format.
        !          5273: 
        !          5274: @findex ASM_STABN_OP
        !          5275: @item ASM_STABN_OP
        !          5276: A C string constant naming the assembler pseudo op to use instead of
        !          5277: @code{.stabn} to define a debugging symbol with no name.  If you don't
        !          5278: define this macro, @code{.stabn} is used.  This macro applies only to
        !          5279: DBX debugging information format.
        !          5280: 
        !          5281: @findex DBX_NO_XREFS
        !          5282: @item DBX_NO_XREFS
        !          5283: Define this macro if DBX on your system does not support the construct
        !          5284: @samp{xs@var{tagname}}.  On some systems, this construct is used to
        !          5285: describe a forward reference to a structure named @var{tagname}.
        !          5286: On other systems, this construct is not supported at all.
        !          5287: 
        !          5288: @findex DBX_CONTIN_LENGTH
        !          5289: @item DBX_CONTIN_LENGTH
        !          5290: A symbol name in DBX-format debugging information is normally
        !          5291: continued (split into two separate @code{.stabs} directives) when it
        !          5292: exceeds a certain length (by default, 80 characters).  On some
        !          5293: operating systems, DBX requires this splitting; on others, splitting
        !          5294: must not be done.  You can inhibit splitting by defining this macro
        !          5295: with the value zero.  You can override the default splitting-length by
        !          5296: defining this macro as an expression for the length you desire.
        !          5297: 
        !          5298: @findex DBX_CONTIN_CHAR
        !          5299: @item DBX_CONTIN_CHAR
        !          5300: Normally continuation is indicated by adding a @samp{\} character to
        !          5301: the end of a @code{.stabs} string when a continuation follows.  To use
        !          5302: a different character instead, define this macro as a character
        !          5303: constant for the character you want to use.  Do not define this macro
        !          5304: if backslash is correct for your system.
        !          5305: 
        !          5306: @findex DBX_STATIC_STAB_DATA_SECTION
        !          5307: @item DBX_STATIC_STAB_DATA_SECTION
        !          5308: Define this macro if it is necessary to go to the data section before
        !          5309: outputting the @samp{.stabs} pseudo-op for a non-global static
        !          5310: variable.
        !          5311: 
        !          5312: @findex DBX_TYPE_DECL_STABS_CODE
        !          5313: @item DBX_TYPE_DECL_STABS_CODE
        !          5314: The value to use in the ``code'' field of the @code{.stabs} directive
        !          5315: for a typedef.  The default is @code{N_LSYM}.
        !          5316: 
        !          5317: @findex DBX_STATIC_CONST_VAR_CODE
        !          5318: @item DBX_STATIC_CONST_VAR_CODE
        !          5319: The value to use in the ``code'' field of the @code{.stabs} directive
        !          5320: for a static variable located in the text section.  DBX format does not
        !          5321: provide any ``right'' way to do this.  The default is @code{N_FUN}.
        !          5322: 
        !          5323: @findex DBX_REGPARM_STABS_CODE
        !          5324: @item DBX_REGPARM_STABS_CODE
        !          5325: The value to use in the ``code'' field of the @code{.stabs} directive
        !          5326: for a parameter passed in registers.  DBX format does not provide any
        !          5327: ``right'' way to do this.  The default is @code{N_RSYM}.
        !          5328: 
        !          5329: @findex DBX_REGPARM_STABS_LETTER
        !          5330: @item DBX_REGPARM_STABS_LETTER
        !          5331: The letter to use in DBX symbol data to identify a symbol as a parameter
        !          5332: passed in registers.  DBX format does not customarily provide any way to
        !          5333: do this.  The default is @code{'P'}.
        !          5334: 
        !          5335: @findex DBX_MEMPARM_STABS_LETTER
        !          5336: @item DBX_MEMPARM_STABS_LETTER
        !          5337: The letter to use in DBX symbol data to identify a symbol as a stack
        !          5338: parameter.  The default is @code{'p'}.
        !          5339: 
        !          5340: @findex DBX_FUNCTION_FIRST
        !          5341: @item DBX_FUNCTION_FIRST
        !          5342: Define this macro if the DBX information for a function and its
        !          5343: arguments should precede the assembler code for the function.  Normally,
        !          5344: in DBX format, the debugging information entirely follows the assembler
        !          5345: code.
        !          5346: 
        !          5347: @findex DBX_LBRAC_FIRST
        !          5348: @item DBX_LBRAC_FIRST
        !          5349: Define this macro if the @code{N_LBRAC} symbol for a block should
        !          5350: precede the debugging information for variables and functions defined in
        !          5351: that block.  Normally, in DBX format, the @code{N_LBRAC} symbol comes
        !          5352: first.
        !          5353: @end table
        !          5354: 
        !          5355: @node DBX Hooks
        !          5356: @subsection Open-Ended Hooks for DBX Format
        !          5357: 
        !          5358: @table @code
        !          5359: @findex DBX_OUTPUT_LBRAC
        !          5360: @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
        !          5361: Define this macro to say how to output to @var{stream} the debugging
        !          5362: information for the start of a scope level for variable names.  The
        !          5363: argument @var{name} is the name of an assembler symbol (for use with
        !          5364: @code{assemble_name}) whose value is the address where the scope begins.
        !          5365: 
        !          5366: @findex DBX_OUTPUT_RBRAC
        !          5367: @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
        !          5368: Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
        !          5369: 
        !          5370: @findex DBX_OUTPUT_ENUM
        !          5371: @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
        !          5372: Define this macro if the target machine requires special handling to
        !          5373: output an enumeration type.  The definition should be a C statement
        !          5374: (sans semicolon) to output the appropriate information to @var{stream}
        !          5375: for the type @var{type}.
        !          5376: 
        !          5377: @findex DBX_OUTPUT_FUNCTION_END
        !          5378: @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
        !          5379: Define this macro if the target machine requires special output at the
        !          5380: end of the debugging information for a function.  The definition should
        !          5381: be a C statement (sans semicolon) to output the appropriate information
        !          5382: to @var{stream}.  @var{function} is the @code{FUNCTION_DECL} node for
        !          5383: the function.
        !          5384: 
        !          5385: @findex DBX_OUTPUT_STANDARD_TYPES
        !          5386: @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
        !          5387: Define this macro if you need to control the order of output of the
        !          5388: standard data types at the beginning of compilation.  The argument
        !          5389: @var{syms} is a @code{tree} which is a chain of all the predefined
        !          5390: global symbols, including names of data types.
        !          5391: 
        !          5392: Normally, DBX output starts with definitions of the types for integers
        !          5393: and characters, followed by all the other predefined types of the
        !          5394: particular language in no particular order.
        !          5395: 
        !          5396: On some machines, it is necessary to output different particular types
        !          5397: first.  To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
        !          5398: those symbols in the necessary order.  Any predefined types that you
        !          5399: don't explicitly output will be output afterward in no particular order.
        !          5400: 
        !          5401: Be careful not to define this macro so that it works only for C.  There
        !          5402: are no global variables to access most of the built-in types, because
        !          5403: another language may have another set of types.  The way to output a
        !          5404: particular type is to look through @var{syms} to see if you can find it.
        !          5405: Here is an example:
        !          5406: 
        !          5407: @smallexample
        !          5408: @{
        !          5409:   tree decl;
        !          5410:   for (decl = syms; decl; decl = TREE_CHAIN (decl))
        !          5411:     if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
        !          5412:                  "long int"))
        !          5413:       dbxout_symbol (decl);
        !          5414:   @dots{}
        !          5415: @}
        !          5416: @end smallexample
        !          5417: 
        !          5418: @noindent
        !          5419: This does nothing if the expected type does not exist.
        !          5420: 
        !          5421: See the function @code{init_decl_processing} in @file{c-decl.c} to find
        !          5422: the names to use for all the built-in C types.
        !          5423: 
        !          5424: Here is another way of finding a particular type:
        !          5425: 
        !          5426: @c this is still overfull.  --mew 10feb93
        !          5427: @smallexample
        !          5428: @{
        !          5429:   tree decl;
        !          5430:   for (decl = syms; decl; decl = TREE_CHAIN (decl))
        !          5431:     if (TREE_CODE (decl) == TYPE_DECL
        !          5432:         && (TREE_CODE (TREE_TYPE (decl))
        !          5433:             == INTEGER_CST)
        !          5434:         && TYPE_PRECISION (TREE_TYPE (decl)) == 16
        !          5435:         && TYPE_UNSIGNED (TREE_TYPE (decl)))
        !          5436: @group
        !          5437:       /* @r{This must be @code{unsigned short}.}  */
        !          5438:       dbxout_symbol (decl);
        !          5439:   @dots{}
        !          5440: @}
        !          5441: @end group
        !          5442: @end smallexample
        !          5443: @end table
        !          5444: 
        !          5445: @node File Names and DBX
        !          5446: @subsection File Names in DBX Format
        !          5447: 
        !          5448: @table @code
        !          5449: @findex DBX_WORKING_DIRECTORY
        !          5450: @item DBX_WORKING_DIRECTORY
        !          5451: Define this if DBX wants to have the current directory recorded in each
        !          5452: object file.
        !          5453: 
        !          5454: Note that the working directory is always recorded if GDB extensions are
        !          5455: enabled.
        !          5456: 
        !          5457: @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
        !          5458: @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
        !          5459: A C statement to output DBX debugging information to the stdio stream
        !          5460: @var{stream} which indicates that file @var{name} is the main source
        !          5461: file---the file specified as the input file for compilation.
        !          5462: This macro is called only once, at the beginning of compilation.
        !          5463: 
        !          5464: This macro need not be defined if the standard form of output
        !          5465: for DBX debugging information is appropriate.
        !          5466: 
        !          5467: @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
        !          5468: @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
        !          5469: A C statement to output DBX debugging information to the stdio stream
        !          5470: @var{stream} which indicates that the current directory during
        !          5471: compilation is named @var{name}.
        !          5472: 
        !          5473: This macro need not be defined if the standard form of output
        !          5474: for DBX debugging information is appropriate.
        !          5475: 
        !          5476: @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
        !          5477: @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
        !          5478: A C statement to output DBX debugging information at the end of
        !          5479: compilation of the main source file @var{name}.
        !          5480: 
        !          5481: If you don't define this macro, nothing special is output at the end
        !          5482: of compilation, which is correct for most machines.
        !          5483: 
        !          5484: @findex DBX_OUTPUT_SOURCE_FILENAME
        !          5485: @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
        !          5486: A C statement to output DBX debugging information to the stdio stream
        !          5487: @var{stream} which indicates that file @var{name} is the current source
        !          5488: file.  This output is generated each time input shifts to a different
        !          5489: source file as a result of @samp{#include}, the end of an included file,
        !          5490: or a @samp{#line} command.
        !          5491: 
        !          5492: This macro need not be defined if the standard form of output
        !          5493: for DBX debugging information is appropriate.
        !          5494: @end table
        !          5495: 
        !          5496: @need 2000
        !          5497: @node SDB and DWARF
        !          5498: @subsection Macros for SDB and DWARF Output
        !          5499: 
        !          5500: @table @code
        !          5501: @findex SDB_DEBUGGING_INFO
        !          5502: @item SDB_DEBUGGING_INFO
        !          5503: Define this macro if GNU CC should produce COFF-style debugging output
        !          5504: for SDB in response to the @samp{-g} option.
        !          5505: 
        !          5506: @findex DWARF_DEBUGGING_INFO
        !          5507: @item DWARF_DEBUGGING_INFO
        !          5508: Define this macro if GNU CC should produce dwarf format debugging output 
        !          5509: in response to the @samp{-g} option.
        !          5510: 
        !          5511: @findex PUT_SDB_@dots{}
        !          5512: @item PUT_SDB_@dots{}
        !          5513: Define these macros to override the assembler syntax for the special
        !          5514: SDB assembler directives.  See @file{sdbout.c} for a list of these
        !          5515: macros and their arguments.  If the standard syntax is used, you need
        !          5516: not define them yourself.
        !          5517: 
        !          5518: @findex SDB_DELIM
        !          5519: @item SDB_DELIM
        !          5520: Some assemblers do not support a semicolon as a delimiter, even between
        !          5521: SDB assembler directives.  In that case, define this macro to be the
        !          5522: delimiter to use (usually @samp{\n}).  It is not necessary to define
        !          5523: a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
        !          5524: required.
        !          5525: 
        !          5526: @findex SDB_GENERATE_FAKE
        !          5527: @item SDB_GENERATE_FAKE
        !          5528: Define this macro to override the usual method of constructing a dummy
        !          5529: name for anonymous structure and union types.  See @file{sdbout.c} for
        !          5530: more information.
        !          5531: 
        !          5532: @findex SDB_ALLOW_UNKNOWN_REFERENCES
        !          5533: @item SDB_ALLOW_UNKNOWN_REFERENCES
        !          5534: Define this macro to allow references to unknown structure,
        !          5535: union, or enumeration tags to be emitted.  Standard COFF does not
        !          5536: allow handling of unknown references, MIPS ECOFF has support for
        !          5537: it.
        !          5538: 
        !          5539: @findex SDB_ALLOW_FORWARD_REFERENCES
        !          5540: @item SDB_ALLOW_FORWARD_REFERENCES
        !          5541: Define this macro to allow references to structure, union, or
        !          5542: enumeration tags that have not yet been seen to be handled.  Some
        !          5543: assemblers choke if forward tags are used, while some require it.
        !          5544: @end table
        !          5545: 
        !          5546: @node Cross-compilation
        !          5547: @section Cross Compilation and Floating Point
        !          5548: @cindex cross compilation and floating point 
        !          5549: @cindex floating point and cross compilation
        !          5550: 
        !          5551: While all modern machines use 2's complement representation for integers,
        !          5552: there are a variety of representations for floating point numbers.  This
        !          5553: means that in a cross-compiler the representation of floating point numbers
        !          5554: in the compiled program may be different from that used in the machine
        !          5555: doing the compilation.
        !          5556: 
        !          5557: @findex atof
        !          5558: Because different representation systems may offer different amounts of
        !          5559: range and precision, the cross compiler cannot safely use the host
        !          5560: machine's floating point arithmetic.  Therefore, floating point constants
        !          5561: must be represented in the target machine's format.  This means that the
        !          5562: cross compiler cannot use @code{atof} to parse a floating point constant;
        !          5563: it must have its own special routine to use instead.  Also, constant
        !          5564: folding must emulate the target machine's arithmetic (or must not be done
        !          5565: at all).
        !          5566: 
        !          5567: The macros in the following table should be defined only if you are cross
        !          5568: compiling between different floating point formats.
        !          5569: 
        !          5570: Otherwise, don't define them.  Then default definitions will be set up which
        !          5571: use @code{double} as the data type, @code{==} to test for equality, etc.
        !          5572: 
        !          5573: You don't need to worry about how many times you use an operand of any
        !          5574: of these macros.  The compiler never uses operands which have side effects.
        !          5575: 
        !          5576: @table @code
        !          5577: @findex REAL_VALUE_TYPE
        !          5578: @item REAL_VALUE_TYPE
        !          5579: A macro for the C data type to be used to hold a floating point value
        !          5580: in the target machine's format.  Typically this would be a
        !          5581: @code{struct} containing an array of @code{int}.
        !          5582: 
        !          5583: @findex REAL_VALUES_EQUAL
        !          5584: @item REAL_VALUES_EQUAL (@var{x}, @var{y})
        !          5585: A macro for a C expression which compares for equality the two values,
        !          5586: @var{x} and @var{y}, both of type @code{REAL_VALUE_TYPE}.
        !          5587: 
        !          5588: @findex REAL_VALUES_LESS
        !          5589: @item REAL_VALUES_LESS (@var{x}, @var{y})
        !          5590: A macro for a C expression which tests whether @var{x} is less than
        !          5591: @var{y}, both values being of type @code{REAL_VALUE_TYPE} and
        !          5592: interpreted as floating point numbers in the target machine's
        !          5593: representation.
        !          5594: 
        !          5595: @findex REAL_VALUE_LDEXP
        !          5596: @findex ldexp
        !          5597: @item REAL_VALUE_LDEXP (@var{x}, @var{scale})
        !          5598: A macro for a C expression which performs the standard library
        !          5599: function @code{ldexp}, but using the target machine's floating point
        !          5600: representation.  Both @var{x} and the value of the expression have
        !          5601: type @code{REAL_VALUE_TYPE}.  The second argument, @var{scale}, is an
        !          5602: integer.
        !          5603: 
        !          5604: @findex REAL_VALUE_FIX
        !          5605: @item REAL_VALUE_FIX (@var{x})
        !          5606: A macro whose definition is a C expression to convert the target-machine
        !          5607: floating point value @var{x} to a signed integer.  @var{x} has type
        !          5608: @code{REAL_VALUE_TYPE}.
        !          5609: 
        !          5610: @findex REAL_VALUE_UNSIGNED_FIX
        !          5611: @item REAL_VALUE_UNSIGNED_FIX (@var{x})
        !          5612: A macro whose definition is a C expression to convert the target-machine
        !          5613: floating point value @var{x} to an unsigned integer.  @var{x} has type
        !          5614: @code{REAL_VALUE_TYPE}.
        !          5615: 
        !          5616: @findex REAL_VALUE_RNDZINT
        !          5617: @item REAL_VALUE_RNDZINT (@var{x})
        !          5618: A macro whose definition is a C expression to round the target-machine
        !          5619: floating point value @var{x} towards zero to an integer value (but still
        !          5620: as a floating point number).  @var{x} has type @code{REAL_VALUE_TYPE},
        !          5621: and so does the value.
        !          5622: 
        !          5623: @findex REAL_VALUE_UNSIGNED_RNDZINT
        !          5624: @item REAL_VALUE_UNSIGNED_RNDZINT (@var{x})
        !          5625: A macro whose definition is a C expression to round the target-machine
        !          5626: floating point value @var{x} towards zero to an unsigned integer value
        !          5627: (but still represented as a floating point number).  @var{x} has type
        !          5628: @code{REAL_VALUE_TYPE}, and so does the value.
        !          5629: 
        !          5630: @findex REAL_VALUE_ATOF
        !          5631: @item REAL_VALUE_ATOF (@var{string}, @var{mode})
        !          5632: A macro for a C expression which converts @var{string}, an expression of
        !          5633: type @code{char *}, into a floating point number in the target machine's
        !          5634: representation for mode @var{mode}.  The value has type
        !          5635: @code{REAL_VALUE_TYPE}.
        !          5636: 
        !          5637: @findex REAL_INFINITY
        !          5638: @item REAL_INFINITY
        !          5639: Define this macro if infinity is a possible floating point value, and
        !          5640: therefore division by 0 is legitimate.
        !          5641: 
        !          5642: @findex REAL_VALUE_ISINF
        !          5643: @findex isinf
        !          5644: @item REAL_VALUE_ISINF (@var{x})
        !          5645: A macro for a C expression which determines whether @var{x}, a floating
        !          5646: point value, is infinity.  The value has type @code{int}.
        !          5647: By default, this is defined to call @code{isinf}.
        !          5648: 
        !          5649: @findex REAL_VALUE_ISNAN
        !          5650: @findex isnan
        !          5651: @item REAL_VALUE_ISNAN (@var{x})
        !          5652: A macro for a C expression which determines whether @var{x}, a floating
        !          5653: point value, is a ``nan'' (not-a-number).  The value has type
        !          5654: @code{int}.  By default, this is defined to call @code{isnan}.
        !          5655: @end table
        !          5656: 
        !          5657: @cindex constant folding and floating point
        !          5658: Define the following additional macros if you want to make floating
        !          5659: point constant folding work while cross compiling.  If you don't
        !          5660: define them, cross compilation is still possible, but constant folding
        !          5661: will not happen for floating point values.
        !          5662: 
        !          5663: @table @code
        !          5664: @findex REAL_ARITHMETIC
        !          5665: @item REAL_ARITHMETIC (@var{output}, @var{code}, @var{x}, @var{y})
        !          5666: A macro for a C statement which calculates an arithmetic operation of
        !          5667: the two floating point values @var{x} and @var{y}, both of type
        !          5668: @code{REAL_VALUE_TYPE} in the target machine's representation, to
        !          5669: produce a result of the same type and representation which is stored
        !          5670: in @var{output} (which will be a variable).
        !          5671: 
        !          5672: The operation to be performed is specified by @var{code}, a tree code
        !          5673: which will always be one of the following: @code{PLUS_EXPR},
        !          5674: @code{MINUS_EXPR}, @code{MULT_EXPR}, @code{RDIV_EXPR},
        !          5675: @code{MAX_EXPR}, @code{MIN_EXPR}.@refill
        !          5676: 
        !          5677: @cindex overflow while constant folding
        !          5678: The expansion of this macro is responsible for checking for overflow.
        !          5679: If overflow happens, the macro expansion should execute the statement
        !          5680: @code{return 0;}, which indicates the inability to perform the
        !          5681: arithmetic operation requested.
        !          5682: 
        !          5683: @findex REAL_VALUE_NEGATE
        !          5684: @item REAL_VALUE_NEGATE (@var{x})
        !          5685: A macro for a C expression which returns the negative of the floating
        !          5686: point value @var{x}.  Both @var{x} and the value of the expression
        !          5687: have type @code{REAL_VALUE_TYPE} and are in the target machine's
        !          5688: floating point representation.
        !          5689: 
        !          5690: There is no way for this macro to report overflow, since overflow
        !          5691: can't happen in the negation operation.
        !          5692: 
        !          5693: @findex REAL_VALUE_TRUNCATE
        !          5694: @item REAL_VALUE_TRUNCATE (@var{mode}, @var{x})
        !          5695: A macro for a C expression which converts the floating point value
        !          5696: @var{x} to mode @var{mode}.
        !          5697: 
        !          5698: Both @var{x} and the value of the expression are in the target machine's
        !          5699: floating point representation and have type @code{REAL_VALUE_TYPE}.
        !          5700: However, the value should have an appropriate bit pattern to be output
        !          5701: properly as a floating constant whose precision accords with mode
        !          5702: @var{mode}.
        !          5703: 
        !          5704: There is no way for this macro to report overflow.
        !          5705: 
        !          5706: @findex REAL_VALUE_TO_INT
        !          5707: @item REAL_VALUE_TO_INT (@var{low}, @var{high}, @var{x})
        !          5708: A macro for a C expression which converts a floating point value
        !          5709: @var{x} into a double-precision integer which is then stored into
        !          5710: @var{low} and @var{high}, two variables of type @var{int}.
        !          5711: 
        !          5712: @item REAL_VALUE_FROM_INT (@var{x}, @var{low}, @var{high})
        !          5713: @findex REAL_VALUE_FROM_INT
        !          5714: A macro for a C expression which converts a double-precision integer
        !          5715: found in @var{low} and @var{high}, two variables of type @var{int},
        !          5716: into a floating point value which is then stored into @var{x}.
        !          5717: @end table
        !          5718: 
        !          5719: @node Misc
        !          5720: @section Miscellaneous Parameters
        !          5721: @cindex parameters, miscellaneous
        !          5722: 
        !          5723: @table @code
        !          5724: @item PREDICATE_CODES
        !          5725: @findex PREDICATE_CODES
        !          5726: Define this if you have defined special-purpose predicates in the file
        !          5727: @file{@var{machine}.c}.  This macro is called within an initializer of an
        !          5728: array of structures.  The first field in the structure is the name of a
        !          5729: predicate and the second field is an array of rtl codes.  For each
        !          5730: predicate, list all rtl codes that can be in expressions matched by the
        !          5731: predicate.  The list should have a trailing comma.  Here is an example
        !          5732: of two entries in the list for a typical RISC machine:
        !          5733: 
        !          5734: @smallexample
        !          5735: #define PREDICATE_CODES \
        !          5736:   @{"gen_reg_rtx_operand", @{SUBREG, REG@}@},  \
        !          5737:   @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
        !          5738: @end smallexample
        !          5739: 
        !          5740: Defining this macro does not affect the generated code (however,
        !          5741: incorrect definitions that omit an rtl code that may be matched by the
        !          5742: predicate can cause the compiler to malfunction).  Instead, it allows
        !          5743: the table built by @file{genrecog} to be more compact and efficient,
        !          5744: thus speeding up the compiler.  The most important predicates to include
        !          5745: in the list specified by this macro are thoses used in the most insn
        !          5746: patterns.
        !          5747: 
        !          5748: @findex CASE_VECTOR_MODE
        !          5749: @item CASE_VECTOR_MODE
        !          5750: An alias for a machine mode name.  This is the machine mode that
        !          5751: elements of a jump-table should have.
        !          5752: 
        !          5753: @findex CASE_VECTOR_PC_RELATIVE
        !          5754: @item CASE_VECTOR_PC_RELATIVE
        !          5755: Define this macro if jump-tables should contain relative addresses.
        !          5756: 
        !          5757: @findex CASE_DROPS_THROUGH
        !          5758: @item CASE_DROPS_THROUGH
        !          5759: Define this if control falls through a @code{case} insn when the index
        !          5760: value is out of range.  This means the specified default-label is
        !          5761: actually ignored by the @code{case} insn proper.
        !          5762: 
        !          5763: @findex CASE_VALUES_THRESHOLD
        !          5764: @item CASE_VALUES_THRESHOLD
        !          5765: Define this to be the smallest number of different values for which it
        !          5766: is best to use a jump-table instead of a tree of conditional branches.
        !          5767: The default is four for machines with a @code{casesi} instruction and
        !          5768: five otherwise.  This is best for most machines.
        !          5769: 
        !          5770: @findex WORD_REGISTER_OPERATIONS
        !          5771: @item WORD_REGISTER_OPERATIONS
        !          5772: Define this macro if operations between registers with integral mode
        !          5773: smaller than a word are always performed on the entire register.
        !          5774: Most RISC machines have this property and most CISC machines do not.
        !          5775: 
        !          5776: @findex LOAD_EXTEND_OP
        !          5777: @item LOAD_EXTEND_OP (@var{mode})
        !          5778: Define this macro to be a C expression indicating when insns that read
        !          5779: memory in @var{mode}, an integral mode narrower than a word, set the
        !          5780: bits outside of @var{mode} to be either the sign-extension or the
        !          5781: zero-extension of the data read.  Return @code{SIGN_EXTEND} for values
        !          5782: of @var{mode} for which the
        !          5783: insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
        !          5784: @code{NIL} for other modes.
        !          5785: 
        !          5786: This macro is not called with @var{mode} non-integral or with a width
        !          5787: greater than or equal to @code{BITS_PER_WORD}, so you may return any
        !          5788: value in this case.  Do not define this macro if it would always return
        !          5789: @code{NIL}.  On machines where this macro is defined, you will normally
        !          5790: define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
        !          5791: 
        !          5792: @findex IMPLICIT_FIX_EXPR
        !          5793: @item IMPLICIT_FIX_EXPR
        !          5794: An alias for a tree code that should be used by default for conversion
        !          5795: of floating point values to fixed point.  Normally,
        !          5796: @code{FIX_ROUND_EXPR} is used.@refill
        !          5797: 
        !          5798: @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
        !          5799: @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
        !          5800: Define this macro if the same instructions that convert a floating
        !          5801: point number to a signed fixed point number also convert validly to an
        !          5802: unsigned one.
        !          5803: 
        !          5804: @findex EASY_DIV_EXPR
        !          5805: @item EASY_DIV_EXPR
        !          5806: An alias for a tree code that is the easiest kind of division to
        !          5807: compile code for in the general case.  It may be
        !          5808: @code{TRUNC_DIV_EXPR}, @code{FLOOR_DIV_EXPR}, @code{CEIL_DIV_EXPR} or
        !          5809: @code{ROUND_DIV_EXPR}.  These four division operators differ in how
        !          5810: they round the result to an integer.  @code{EASY_DIV_EXPR} is used
        !          5811: when it is permissible to use any of those kinds of division and the
        !          5812: choice should be made on the basis of efficiency.@refill
        !          5813: 
        !          5814: @findex MOVE_MAX
        !          5815: @item MOVE_MAX
        !          5816: The maximum number of bytes that a single instruction can move quickly
        !          5817: from memory to memory.
        !          5818: 
        !          5819: @findex MAX_MOVE_MAX
        !          5820: @item MAX_MOVE_MAX
        !          5821: The maximum number of bytes that a single instruction can move quickly
        !          5822: from memory to memory.  If this is undefined, the default is
        !          5823: @code{MOVE_MAX}.  Otherwise, it is the constant value that is the
        !          5824: largest value that @code{MOVE_MAX} can have at run-time.
        !          5825: 
        !          5826: @findex SHIFT_COUNT_TRUNCATED
        !          5827: @item SHIFT_COUNT_TRUNCATED
        !          5828: A C expression that is nonzero if on this machine the number of bits
        !          5829: actually used for the count of a shift operation is equal to the number
        !          5830: of bits needed to represent the size of the object being shifted.  When
        !          5831: this macro is non-zero, the compiler will assume that it is safe to omit
        !          5832: a sign-extend, zero-extend, and certain bitwise `and' instructions that
        !          5833: truncates the count of a shift operation.  On machines that have
        !          5834: instructions that act on bitfields at variable positions, which may
        !          5835: include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
        !          5836: also enables deletion of truncations of the values that serve as
        !          5837: arguments to bitfield instructions.
        !          5838: 
        !          5839: If both types of instructions truncate the count (for shifts) and
        !          5840: position (for bitfield operations), or if no variable-position bitfield
        !          5841: instructions exist, you should define this macro.
        !          5842: 
        !          5843: However, on some machines, such as the 80386 and the 680x0, truncation
        !          5844: only applies to shift operations and not the (real or pretended)
        !          5845: bitfield operations.  Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
        !          5846: such machines.  Instead, add patterns to the @file{md} file that include
        !          5847: the implied truncation of the shift instructions.
        !          5848: 
        !          5849: You need not define this macro if it would always have the value of zero.
        !          5850: 
        !          5851: @findex TRULY_NOOP_TRUNCATION
        !          5852: @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
        !          5853: A C expression which is nonzero if on this machine it is safe to
        !          5854: ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
        !          5855: bits (where @var{outprec} is smaller than @var{inprec}) by merely
        !          5856: operating on it as if it had only @var{outprec} bits.
        !          5857: 
        !          5858: On many machines, this expression can be 1.
        !          5859: 
        !          5860: @c rearranged this, removed the phrase "it is reported that".  this was
        !          5861: @c to fix an overfull hbox.  --mew 10feb93
        !          5862: When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
        !          5863: modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
        !          5864: If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
        !          5865: such cases may improve things.
        !          5866: 
        !          5867: @findex STORE_FLAG_VALUE
        !          5868: @item STORE_FLAG_VALUE
        !          5869: A C expression describing the value returned by a comparison operator
        !          5870: with an integral mode and stored by a store-flag instruction
        !          5871: (@samp{s@var{cond}}) when the condition is true.  This description must
        !          5872: apply to @emph{all} the @samp{s@var{cond}} patterns and all the
        !          5873: comparison operators whose results have a @code{MODE_INT} mode.
        !          5874: 
        !          5875: A value of 1 or -1 means that the instruction implementing the
        !          5876: comparison operator returns exactly 1 or -1 when the comparison is true
        !          5877: and 0 when the comparison is false.  Otherwise, the value indicates
        !          5878: which bits of the result are guaranteed to be 1 when the comparison is
        !          5879: true.  This value is interpreted in the mode of the comparison
        !          5880: operation, which is given by the mode of the first operand in the
        !          5881: @samp{s@var{cond}} pattern.  Either the low bit or the sign bit of
        !          5882: @code{STORE_FLAG_VALUE} be on.  Presently, only those bits are used by
        !          5883: the compiler.
        !          5884: 
        !          5885: If @code{STORE_FLAG_VALUE} is neither 1 or -1, the compiler will
        !          5886: generate code that depends only on the specified bits.  It can also
        !          5887: replace comparison operators with equivalent operations if they cause
        !          5888: the required bits to be set, even if the remaining bits are undefined.
        !          5889: For example, on a machine whose comparison operators return an
        !          5890: @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
        !          5891: @samp{0x80000000}, saying that just the sign bit is relevant, the
        !          5892: expression
        !          5893: 
        !          5894: @smallexample
        !          5895: (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
        !          5896: @end smallexample
        !          5897: 
        !          5898: @noindent
        !          5899: can be converted to
        !          5900: 
        !          5901: @smallexample
        !          5902: (ashift:SI @var{x} (const_int @var{n}))
        !          5903: @end smallexample
        !          5904: 
        !          5905: @noindent
        !          5906: where @var{n} is the appropriate shift count to move the bit being
        !          5907: tested into the sign bit.
        !          5908: 
        !          5909: There is no way to describe a machine that always sets the low-order bit
        !          5910: for a true value, but does not guarantee the value of any other bits,
        !          5911: but we do not know of any machine that has such an instruction.  If you
        !          5912: are trying to port GNU CC to such a machine, include an instruction to
        !          5913: perform a logical-and of the result with 1 in the pattern for the
        !          5914: comparison operators and let us know
        !          5915: @ifset USING
        !          5916: (@pxref{Bug Reporting,,How to Report Bugs}).
        !          5917: @end ifset
        !          5918: @ifclear USING
        !          5919: (@pxref{Bug Reporting,,How to Report Bugs,gcc.info,Using GCC}).
        !          5920: @end ifclear
        !          5921: 
        !          5922: Often, a machine will have multiple instructions that obtain a value
        !          5923: from a comparison (or the condition codes).  Here are rules to guide the
        !          5924: choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
        !          5925: to be used:
        !          5926: 
        !          5927: @itemize @bullet
        !          5928: @item
        !          5929: Use the shortest sequence that yields a valid definition for
        !          5930: @code{STORE_FLAG_VALUE}.  It is more efficient for the compiler to
        !          5931: ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
        !          5932: comparison operators to do so because there may be opportunities to
        !          5933: combine the normalization with other operations.
        !          5934: 
        !          5935: @item
        !          5936: For equal-length sequences, use a value of 1 or -1, with -1 being
        !          5937: slightly preferred on machines with expensive jumps and 1 preferred on
        !          5938: other machines.
        !          5939: 
        !          5940: @item
        !          5941: As a second choice, choose a value of @samp{0x80000001} if instructions
        !          5942: exist that set both the sign and low-order bits but do not define the
        !          5943: others.
        !          5944: 
        !          5945: @item
        !          5946: Otherwise, use a value of @samp{0x80000000}.
        !          5947: @end itemize
        !          5948: 
        !          5949: Many machines can produce both the value chosen for
        !          5950: @code{STORE_FLAG_VALUE} and its negation in the same number of
        !          5951: instructions.  On those machines, you should also define a pattern for
        !          5952: those cases, e.g., one matching
        !          5953: 
        !          5954: @smallexample
        !          5955: (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
        !          5956: @end smallexample
        !          5957: 
        !          5958: Some machines can also perform @code{and} or @code{plus} operations on
        !          5959: condition code values with less instructions than the corresponding
        !          5960: @samp{s@var{cond}} insn followed by @code{and} or @code{plus}.  On those
        !          5961: machines, define the appropriate patterns.  Use the names @code{incscc}
        !          5962: and @code{decscc}, respectively, for the the patterns which perform
        !          5963: @code{plus} or @code{minus} operations on condition code values.  See
        !          5964: @file{rs6000.md} for some examples.  The GNU Superoptizer can be used to
        !          5965: find such instruction sequences on other machines.
        !          5966: 
        !          5967: You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
        !          5968: instructions.
        !          5969: 
        !          5970: @findex FLOAT_STORE_FLAG_VALUE
        !          5971: @item FLOAT_STORE_FLAG_VALUE
        !          5972: A C expression that gives a non-zero floating point value that is
        !          5973: returned when comparison operators with floating-point results are true.
        !          5974: Define this macro on machine that have comparison operations that return
        !          5975: floating-point values.  If there are no such operations, do not define
        !          5976: this macro.
        !          5977: 
        !          5978: @findex Pmode
        !          5979: @item Pmode
        !          5980: An alias for the machine mode for pointers.  Normally the definition
        !          5981: can be
        !          5982: 
        !          5983: @smallexample
        !          5984: #define Pmode SImode
        !          5985: @end smallexample
        !          5986: 
        !          5987: @findex FUNCTION_MODE
        !          5988: @item FUNCTION_MODE
        !          5989: An alias for the machine mode used for memory references to functions
        !          5990: being called, in @code{call} RTL expressions.  On most machines this
        !          5991: should be @code{QImode}.
        !          5992: 
        !          5993: @findex INTEGRATE_THRESHOLD
        !          5994: @item INTEGRATE_THRESHOLD (@var{decl})
        !          5995: A C expression for the maximum number of instructions above which the
        !          5996: function @var{decl} should not be inlined.  @var{decl} is a
        !          5997: @code{FUNCTION_DECL} node.
        !          5998: 
        !          5999: The default definition of this macro is 64 plus 8 times the number of
        !          6000: arguments that the function accepts.  Some people think a larger
        !          6001: threshold should be used on RISC machines.
        !          6002: 
        !          6003: @findex SCCS_DIRECTIVE
        !          6004: @item SCCS_DIRECTIVE
        !          6005: Define this if the preprocessor should ignore @code{#sccs} directives
        !          6006: and print no error message.
        !          6007: 
        !          6008: @findex NO_IMPLICIT_EXTERN_C
        !          6009: @item NO_IMPLICIT_EXTERN_C
        !          6010: Define this macro if the system header files support C++ as well as C.
        !          6011: This macro inhibits the usual method of using system header files in
        !          6012: C++, which is to pretend that the file's contents are enclosed in
        !          6013: @samp{extern "C" @{@dots{}@}}.
        !          6014: 
        !          6015: @findex HANDLE_PRAGMA
        !          6016: @findex #pragma
        !          6017: @findex pragma
        !          6018: @item HANDLE_PRAGMA (@var{stream})
        !          6019: Define this macro if you want to implement any pragmas.  If defined, it
        !          6020: should be a C statement to be executed when @code{#pragma} is seen.  The
        !          6021: argument @var{stream} is the stdio input stream from which the source
        !          6022: text can be read.
        !          6023: 
        !          6024: It is generally a bad idea to implement new uses of @code{#pragma}.  The
        !          6025: only reason to define this macro is for compatibility with other
        !          6026: compilers that do support @code{#pragma} for the sake of any user
        !          6027: programs which already use it.
        !          6028: 
        !          6029: @findex DOLLARS_IN_IDENTIFIERS
        !          6030: @item DOLLARS_IN_IDENTIFIERS
        !          6031: Define this macro to control use of the character @samp{$} in identifier
        !          6032: names.  The value should be 0, 1, or 2.  0 means @samp{$} is not allowed
        !          6033: by default; 1 means it is allowed by default if @samp{-traditional} is
        !          6034: used; 2 means it is allowed by default provided @samp{-ansi} is not used.
        !          6035: 1 is the default; there is no need to define this macro in that case.
        !          6036: 
        !          6037: @findex NO_DOLLAR_IN_LABEL
        !          6038: @item NO_DOLLAR_IN_LABEL
        !          6039: Define this macro if the assembler does not accept the character
        !          6040: @samp{$} in label names.  By default constructors and destructors in
        !          6041: G++ have @samp{$} in the identifiers.  If this macro is defined,
        !          6042: @samp{.} is used instead.
        !          6043: 
        !          6044: @findex NO_DOT_IN_LABEL
        !          6045: @item NO_DOT_IN_LABEL
        !          6046: Define this macro if the assembler does not accept the character
        !          6047: @samp{.} in label names.  By default constructors and destructors in G++
        !          6048: have names that use @samp{.}.  If this macro is defined, these names
        !          6049: are rewritten to avoid @samp{.}.
        !          6050: 
        !          6051: @findex DEFAULT_MAIN_RETURN
        !          6052: @item DEFAULT_MAIN_RETURN
        !          6053: Define this macro if the target system expects every program's @code{main}
        !          6054: function to return a standard ``success'' value by default (if no other
        !          6055: value is explicitly returned).
        !          6056: 
        !          6057: The definition should be a C statement (sans semicolon) to generate the
        !          6058: appropriate rtl instructions.  It is used only when compiling the end of
        !          6059: @code{main}.
        !          6060: 
        !          6061: @item HAVE_ATEXIT
        !          6062: @findex HAVE_ATEXIT
        !          6063: Define this if the target system supports the function
        !          6064: @code{atexit} from the ANSI C standard.  If this is not defined,
        !          6065: and @code{INIT_SECTION_ASM_OP} is not defined, a default
        !          6066: @code{exit} function will be provided to support C++.
        !          6067: 
        !          6068: @item EXIT_BODY
        !          6069: @findex EXIT_BODY
        !          6070: Define this if your @code{exit} function needs to do something
        !          6071: besides calling an external function @code{_cleanup} before
        !          6072: terminating with @code{_exit}.  The @code{EXIT_BODY} macro is
        !          6073: only needed if netiher @code{HAVE_ATEXIT} nor
        !          6074: @code{INIT_SECTION_ASM_OP} are defined.
        !          6075: 
        !          6076: @findex INSN_SETS_ARE_DELAYED
        !          6077: @item INSN_SETS_ARE_DELAYED (@var{insn})
        !          6078: Define this macro as a C expression that is nonzero if it is safe for the
        !          6079: delay slot scheduler to place instructions in the delay slot of @var{insn},
        !          6080: even if they appear to use a resource set or clobbered in @var{insn}.
        !          6081: @var{insn} is always a @code{jump_insn} or an @code{insn}; GNU CC knows that 
        !          6082: every @code{call_insn} has this behavior.  On machines where some @code{insn} 
        !          6083: or @code{jump_insn} is really a function call and hence has this behavior, 
        !          6084: you should define this macro.
        !          6085: 
        !          6086: You need not define this macro if it would always return zero.
        !          6087: 
        !          6088: @findex INSN_REFERENCES_ARE_DELAYED
        !          6089: @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
        !          6090: Define this macro as a C expression that is nonzero if it is safe for the
        !          6091: delay slot scheduler to place instructions in the delay slot of @var{insn},
        !          6092: even if they appear to set or clobber a resource referenced in @var{insn}.  
        !          6093: @var{insn} is always a @code{jump_insn} or an @code{insn}.  On machines where
        !          6094: some @code{insn} or @code{jump_insn} is really a function call and its operands
        !          6095: are registers whose use is actually in the subroutine it calls, you should
        !          6096: define this macro.  Doing so allows the delay slot scheduler to move 
        !          6097: instructions which copy arguments into the argument registers into the delay 
        !          6098: slot of @var{insn}.
        !          6099: 
        !          6100: You need not define this macro if it would always return zero.
        !          6101: @end table

This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.