1 @c Copyright (C) 1988,89,92,93,94,96,97,98,1999 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
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
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
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.
44 @section Controlling the Compilation Driver, @file{gcc}
46 @cindex controlling the compilation driver
48 @c prevent bad page break with this line
49 You can control the compilation driver.
52 @findex SWITCH_TAKES_ARG
53 @item SWITCH_TAKES_ARG (@var{char})
54 A C expression which determines whether the option @samp{-@var{char}}
55 takes arguments. The value should be the number of arguments that
56 option takes--zero, for many options.
58 By default, this macro is defined as
59 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
60 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
61 wish to add additional options which take arguments. Any redefinition
62 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
65 @findex WORD_SWITCH_TAKES_ARG
66 @item WORD_SWITCH_TAKES_ARG (@var{name})
67 A C expression which determines whether the option @samp{-@var{name}}
68 takes arguments. The value should be the number of arguments that
69 option takes--zero, for many options. This macro rather than
70 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
72 By default, this macro is defined as
73 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
74 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
75 wish to add additional options which take arguments. Any redefinition
76 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
79 @findex SWITCH_CURTAILS_COMPILATION
80 @item SWITCH_CURTAILS_COMPILATION (@var{char})
81 A C expression which determines whether the option @samp{-@var{char}}
82 stops compilation before the generation of an executable. The value is
83 boolean, non-zero if the option does stop an executable from being
84 generated, zero otherwise.
86 By default, this macro is defined as
87 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
88 options properly. You need not define
89 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
90 options which affect the generation of an executable. Any redefinition
91 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
92 for additional options.
94 @findex SWITCHES_NEED_SPACES
95 @item SWITCHES_NEED_SPACES
96 A string-valued C expression which enumerates the options for which
97 the linker needs a space between the option and its argument.
99 If this macro is not defined, the default value is @code{""}.
103 A C string constant that tells the GNU CC driver program options to
104 pass to CPP. It can also specify how to translate options you
105 give to GNU CC into options for GNU CC to pass to the CPP.
107 Do not define this macro if it does not need to do anything.
109 @findex NO_BUILTIN_SIZE_TYPE
110 @item NO_BUILTIN_SIZE_TYPE
111 If this macro is defined, the preprocessor will not define the builtin macro
112 @code{__SIZE_TYPE__}. The macro @code{__SIZE_TYPE__} must then be defined
113 by @code{CPP_SPEC} instead.
115 This should be defined if @code{SIZE_TYPE} depends on target dependent flags
116 which are not accessible to the preprocessor. Otherwise, it should not
119 @findex NO_BUILTIN_PTRDIFF_TYPE
120 @item NO_BUILTIN_PTRDIFF_TYPE
121 If this macro is defined, the preprocessor will not define the builtin macro
122 @code{__PTRDIFF_TYPE__}. The macro @code{__PTRDIFF_TYPE__} must then be
123 defined by @code{CPP_SPEC} instead.
125 This should be defined if @code{PTRDIFF_TYPE} depends on target dependent flags
126 which are not accessible to the preprocessor. Otherwise, it should not
129 @findex SIGNED_CHAR_SPEC
130 @item SIGNED_CHAR_SPEC
131 A C string constant that tells the GNU CC driver program options to
132 pass to CPP. By default, this macro is defined to pass the option
133 @samp{-D__CHAR_UNSIGNED__} to CPP if @code{char} will be treated as
134 @code{unsigned char} by @code{cc1}.
136 Do not define this macro unless you need to override the default
141 A C string constant that tells the GNU CC driver program options to
142 pass to @code{cc1}. It can also specify how to translate options you
143 give to GNU CC into options for GNU CC to pass to the @code{cc1}.
145 Do not define this macro if it does not need to do anything.
149 A C string constant that tells the GNU CC driver program options to
150 pass to @code{cc1plus}. It can also specify how to translate options you
151 give to GNU CC into options for GNU CC to pass to the @code{cc1plus}.
153 Do not define this macro if it does not need to do anything.
157 A C string constant that tells the GNU CC driver program options to
158 pass to the assembler. It can also specify how to translate options
159 you give to GNU CC into options for GNU CC to pass to the assembler.
160 See the file @file{sun3.h} for an example of this.
162 Do not define this macro if it does not need to do anything.
164 @findex ASM_FINAL_SPEC
166 A C string constant that tells the GNU CC driver program how to
167 run any programs which cleanup after the normal assembler.
168 Normally, this is not needed. See the file @file{mips.h} for
171 Do not define this macro if it does not need to do anything.
175 A C string constant that tells the GNU CC driver program options to
176 pass to the linker. It can also specify how to translate options you
177 give to GNU CC into options for GNU CC to pass to the linker.
179 Do not define this macro if it does not need to do anything.
183 Another C string constant used much like @code{LINK_SPEC}. The difference
184 between the two is that @code{LIB_SPEC} is used at the end of the
185 command given to the linker.
187 If this macro is not defined, a default is provided that
188 loads the standard C library from the usual place. See @file{gcc.c}.
192 Another C string constant that tells the GNU CC driver program
193 how and when to place a reference to @file{libgcc.a} into the
194 linker command line. This constant is placed both before and after
195 the value of @code{LIB_SPEC}.
197 If this macro is not defined, the GNU CC driver provides a default that
198 passes the string @samp{-lgcc} to the linker unless the @samp{-shared}
201 @findex STARTFILE_SPEC
203 Another C string constant used much like @code{LINK_SPEC}. The
204 difference between the two is that @code{STARTFILE_SPEC} is used at
205 the very beginning of the command given to the linker.
207 If this macro is not defined, a default is provided that loads the
208 standard C startup file from the usual place. See @file{gcc.c}.
212 Another C string constant used much like @code{LINK_SPEC}. The
213 difference between the two is that @code{ENDFILE_SPEC} is used at
214 the very end of the command given to the linker.
216 Do not define this macro if it does not need to do anything.
220 Define this macro to provide additional specifications to put in the
221 @file{specs} file that can be used in various specifications like
224 The definition should be an initializer for an array of structures,
225 containing a string constant, that defines the specification name, and a
226 string constant that provides the specification.
228 Do not define this macro if it does not need to do anything.
230 @code{EXTRA_SPECS} is useful when an architecture contains several
231 related targets, which have various @code{..._SPECS} which are similar
232 to each other, and the maintainer would like one central place to keep
235 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
236 define either @code{_CALL_SYSV} when the System V calling sequence is
237 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
240 The @file{config/rs6000/rs6000.h} target file defines:
243 #define EXTRA_SPECS \
244 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
246 #define CPP_SYS_DEFAULT ""
249 The @file{config/rs6000/sysv.h} target file defines:
253 "%@{posix: -D_POSIX_SOURCE @} \
254 %@{mcall-sysv: -D_CALL_SYSV @} %@{mcall-aix: -D_CALL_AIX @} \
255 %@{!mcall-sysv: %@{!mcall-aix: %(cpp_sysv_default) @}@} \
256 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
258 #undef CPP_SYSV_DEFAULT
259 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
262 while the @file{config/rs6000/eabiaix.h} target file defines
263 @code{CPP_SYSV_DEFAULT} as:
266 #undef CPP_SYSV_DEFAULT
267 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
270 @findex LINK_LIBGCC_SPECIAL
271 @item LINK_LIBGCC_SPECIAL
272 Define this macro if the driver program should find the library
273 @file{libgcc.a} itself and should not pass @samp{-L} options to the
274 linker. If you do not define this macro, the driver program will pass
275 the argument @samp{-lgcc} to tell the linker to do the search and will
276 pass @samp{-L} options to it.
278 @findex LINK_LIBGCC_SPECIAL_1
279 @item LINK_LIBGCC_SPECIAL_1
280 Define this macro if the driver program should find the library
281 @file{libgcc.a}. If you do not define this macro, the driver program will pass
282 the argument @samp{-lgcc} to tell the linker to do the search.
283 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
284 not affect @samp{-L} options.
286 @findex LINK_COMMAND_SPEC
287 @item LINK_COMMAND_SPEC
288 A C string constant giving the complete command line need to execute the
289 linker. When you do this, you will need to update your port each time a
290 change is made to the link command line within @file{gcc.c}. Therefore,
291 define this macro only if you need to completely redefine the command
292 line for invoking the linker and there is no other way to accomplish
295 @findex MULTILIB_DEFAULTS
296 @item MULTILIB_DEFAULTS
297 Define this macro as a C expression for the initializer of an array of
298 string to tell the driver program which options are defaults for this
299 target and thus do not need to be handled specially when using
300 @code{MULTILIB_OPTIONS}.
302 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
303 the target makefile fragment or if none of the options listed in
304 @code{MULTILIB_OPTIONS} are set by default.
305 @xref{Target Fragment}.
307 @findex RELATIVE_PREFIX_NOT_LINKDIR
308 @item RELATIVE_PREFIX_NOT_LINKDIR
309 Define this macro to tell @code{gcc} that it should only translate
310 a @samp{-B} prefix into a @samp{-L} linker option if the prefix
311 indicates an absolute file name.
313 @findex STANDARD_EXEC_PREFIX
314 @item STANDARD_EXEC_PREFIX
315 Define this macro as a C string constant if you wish to override the
316 standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
317 try when searching for the executable files of the compiler.
319 @findex MD_EXEC_PREFIX
321 If defined, this macro is an additional prefix to try after
322 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
323 when the @samp{-b} option is used, or the compiler is built as a cross
326 @findex STANDARD_STARTFILE_PREFIX
327 @item STANDARD_STARTFILE_PREFIX
328 Define this macro as a C string constant if you wish to override the
329 standard choice of @file{/usr/local/lib/} as the default prefix to
330 try when searching for startup files such as @file{crt0.o}.
332 @findex MD_STARTFILE_PREFIX
333 @item MD_STARTFILE_PREFIX
334 If defined, this macro supplies an additional prefix to try after the
335 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
336 @samp{-b} option is used, or when the compiler is built as a cross
339 @findex MD_STARTFILE_PREFIX_1
340 @item MD_STARTFILE_PREFIX_1
341 If defined, this macro supplies yet another prefix to try after the
342 standard prefixes. It is not searched when the @samp{-b} option is
343 used, or when the compiler is built as a cross compiler.
345 @findex INIT_ENVIRONMENT
346 @item INIT_ENVIRONMENT
347 Define this macro as a C string constant if you wish to set environment
348 variables for programs called by the driver, such as the assembler and
349 loader. The driver passes the value of this macro to @code{putenv} to
350 initialize the necessary environment variables.
352 @findex LOCAL_INCLUDE_DIR
353 @item LOCAL_INCLUDE_DIR
354 Define this macro as a C string constant if you wish to override the
355 standard choice of @file{/usr/local/include} as the default prefix to
356 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
357 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
359 Cross compilers do not use this macro and do not search either
360 @file{/usr/local/include} or its replacement.
362 @findex SYSTEM_INCLUDE_DIR
363 @item SYSTEM_INCLUDE_DIR
364 Define this macro as a C string constant if you wish to specify a
365 system-specific directory to search for header files before the standard
366 directory. @code{SYSTEM_INCLUDE_DIR} comes before
367 @code{STANDARD_INCLUDE_DIR} in the search order.
369 Cross compilers do not use this macro and do not search the directory
372 @findex STANDARD_INCLUDE_DIR
373 @item STANDARD_INCLUDE_DIR
374 Define this macro as a C string constant if you wish to override the
375 standard choice of @file{/usr/include} as the default prefix to
376 try when searching for header files.
378 Cross compilers do not use this macro and do not search either
379 @file{/usr/include} or its replacement.
381 @findex STANDARD_INCLUDE_COMPONENT
382 @item STANDARD_INCLUDE_COMPONENT
383 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
384 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
385 If you do not define this macro, no component is used.
387 @findex INCLUDE_DEFAULTS
388 @item INCLUDE_DEFAULTS
389 Define this macro if you wish to override the entire default search path
390 for include files. For a native compiler, the default search path
391 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
392 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
393 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
394 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
395 and specify private search areas for GCC. The directory
396 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
398 The definition should be an initializer for an array of structures.
399 Each array element should have four elements: the directory name (a
400 string constant), the component name, and flag for C++-only directories,
401 and a flag showing that the includes in the directory don't need to be
402 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
403 the array with a null element.
405 The component name denotes what GNU package the include file is part of,
406 if any, in all upper-case letters. For example, it might be @samp{GCC}
407 or @samp{BINUTILS}. If the package is part of the a vendor-supplied
408 operating system, code the component name as @samp{0}.
411 For example, here is the definition used for VAX/VMS:
414 #define INCLUDE_DEFAULTS \
416 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
417 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
418 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
425 Here is the order of prefixes tried for exec files:
429 Any prefixes specified by the user with @samp{-B}.
432 The environment variable @code{GCC_EXEC_PREFIX}, if any.
435 The directories specified by the environment variable @code{COMPILER_PATH}.
438 The macro @code{STANDARD_EXEC_PREFIX}.
441 @file{/usr/lib/gcc/}.
444 The macro @code{MD_EXEC_PREFIX}, if any.
447 Here is the order of prefixes tried for startfiles:
451 Any prefixes specified by the user with @samp{-B}.
454 The environment variable @code{GCC_EXEC_PREFIX}, if any.
457 The directories specified by the environment variable @code{LIBRARY_PATH}
458 (native only, cross compilers do not use this).
461 The macro @code{STANDARD_EXEC_PREFIX}.
464 @file{/usr/lib/gcc/}.
467 The macro @code{MD_EXEC_PREFIX}, if any.
470 The macro @code{MD_STARTFILE_PREFIX}, if any.
473 The macro @code{STANDARD_STARTFILE_PREFIX}.
482 @node Run-time Target
483 @section Run-time Target Specification
484 @cindex run-time target specification
485 @cindex predefined macros
486 @cindex target specifications
488 @c prevent bad page break with this line
489 Here are run-time target specifications.
492 @findex CPP_PREDEFINES
494 Define this to be a string constant containing @samp{-D} options to
495 define the predefined macros that identify this machine and system.
496 These macros will be predefined unless the @samp{-ansi} option is
499 In addition, a parallel set of macros are predefined, whose names are
500 made by appending @samp{__} at the beginning and at the end. These
501 @samp{__} macros are permitted by the ANSI standard, so they are
502 predefined regardless of whether @samp{-ansi} is specified.
504 For example, on the Sun, one can use the following value:
507 "-Dmc68000 -Dsun -Dunix"
510 The result is to define the macros @code{__mc68000__}, @code{__sun__}
511 and @code{__unix__} unconditionally, and the macros @code{mc68000},
512 @code{sun} and @code{unix} provided @samp{-ansi} is not specified.
514 @findex extern int target_flags
515 @item extern int target_flags;
516 This declaration should be present.
518 @cindex optional hardware or system features
519 @cindex features, optional, in system conventions
521 This series of macros is to allow compiler command arguments to
522 enable or disable the use of optional features of the target machine.
523 For example, one machine description serves both the 68000 and
524 the 68020; a command argument tells the compiler whether it should
525 use 68020-only instructions or not. This command argument works
526 by means of a macro @code{TARGET_68020} that tests a bit in
529 Define a macro @code{TARGET_@var{featurename}} for each such option.
530 Its definition should test a bit in @code{target_flags}; for example:
533 #define TARGET_68020 (target_flags & 1)
536 One place where these macros are used is in the condition-expressions
537 of instruction patterns. Note how @code{TARGET_68020} appears
538 frequently in the 68000 machine description file, @file{m68k.md}.
539 Another place they are used is in the definitions of the other
540 macros in the @file{@var{machine}.h} file.
542 @findex TARGET_SWITCHES
543 @item TARGET_SWITCHES
544 This macro defines names of command options to set and clear
545 bits in @code{target_flags}. Its definition is an initializer
546 with a subgrouping for each command option.
548 Each subgrouping contains a string constant, that defines the option
549 name, a number, which contains the bits to set in
550 @code{target_flags}, and a second string which is the description
551 displayed by --help. If the number is negative then the bits specified
552 by the number are cleared instead of being set. If the description
553 string is present but empty, then no help information will be displayed
554 for that option, but it will not count as an undocumented option. The
555 actual option name is made by appending @samp{-m} to the specified name.
557 One of the subgroupings should have a null string. The number in
558 this grouping is the default value for @code{target_flags}. Any
559 target options act starting with that value.
561 Here is an example which defines @samp{-m68000} and @samp{-m68020}
562 with opposite meanings, and picks the latter as the default:
565 #define TARGET_SWITCHES \
566 @{ @{ "68020", 1, "" @}, \
567 @{ "68000", -1, "Compile for the 68000" @}, \
571 @findex TARGET_OPTIONS
573 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
574 options that have values. Its definition is an initializer with a
575 subgrouping for each command option.
577 Each subgrouping contains a string constant, that defines the fixed part
578 of the option name, the address of a variable, and a description string.
579 The variable, type @code{char *}, is set to the variable part of the
580 given option if the fixed part matches. The actual option name is made
581 by appending @samp{-m} to the specified name.
583 Here is an example which defines @samp{-mshort-data-@var{number}}. If the
584 given option is @samp{-mshort-data-512}, the variable @code{m88k_short_data}
585 will be set to the string @code{"512"}.
588 extern char *m88k_short_data;
589 #define TARGET_OPTIONS \
590 @{ @{ "short-data-", &m88k_short_data, "Specify the size of the short data section" @} @}
593 @findex TARGET_VERSION
595 This macro is a C statement to print on @code{stderr} a string
596 describing the particular machine description choice. Every machine
597 description should define @code{TARGET_VERSION}. For example:
601 #define TARGET_VERSION \
602 fprintf (stderr, " (68k, Motorola syntax)");
604 #define TARGET_VERSION \
605 fprintf (stderr, " (68k, MIT syntax)");
609 @findex OVERRIDE_OPTIONS
610 @item OVERRIDE_OPTIONS
611 Sometimes certain combinations of command options do not make sense on
612 a particular target machine. You can define a macro
613 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
614 defined, is executed once just after all the command options have been
617 Don't use this macro to turn on various extra optimizations for
618 @samp{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
620 @findex OPTIMIZATION_OPTIONS
621 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
622 Some machines may desire to change what optimizations are performed for
623 various optimization levels. This macro, if defined, is executed once
624 just after the optimization level is determined and before the remainder
625 of the command options have been parsed. Values set in this macro are
626 used as the default values for the other command line options.
628 @var{level} is the optimization level specified; 2 if @samp{-O2} is
629 specified, 1 if @samp{-O} is specified, and 0 if neither is specified.
631 @var{size} is non-zero if @samp{-Os} is specified and zero otherwise.
633 You should not use this macro to change options that are not
634 machine-specific. These should uniformly selected by the same
635 optimization level on all supported machines. Use this macro to enable
636 machine-specific optimizations.
638 @strong{Do not examine @code{write_symbols} in
639 this macro!} The debugging options are not supposed to alter the
642 @findex CAN_DEBUG_WITHOUT_FP
643 @item CAN_DEBUG_WITHOUT_FP
644 Define this macro if debugging can be performed even without a frame
645 pointer. If this macro is defined, GNU CC will turn on the
646 @samp{-fomit-frame-pointer} option whenever @samp{-O} is specified.
650 @section Storage Layout
651 @cindex storage layout
653 Note that the definitions of the macros in this table which are sizes or
654 alignments measured in bits do not need to be constant. They can be C
655 expressions that refer to static variables, such as the @code{target_flags}.
656 @xref{Run-time Target}.
659 @findex BITS_BIG_ENDIAN
660 @item BITS_BIG_ENDIAN
661 Define this macro to have the value 1 if the most significant bit in a
662 byte has the lowest number; otherwise define it to have the value zero.
663 This means that bit-field instructions count from the most significant
664 bit. If the machine has no bit-field instructions, then this must still
665 be defined, but it doesn't matter which value it is defined to. This
666 macro need not be a constant.
668 This macro does not affect the way structure fields are packed into
669 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
671 @findex BYTES_BIG_ENDIAN
672 @item BYTES_BIG_ENDIAN
673 Define this macro to have the value 1 if the most significant byte in a
674 word has the lowest number. This macro need not be a constant.
676 @findex WORDS_BIG_ENDIAN
677 @item WORDS_BIG_ENDIAN
678 Define this macro to have the value 1 if, in a multiword object, the
679 most significant word has the lowest number. This applies to both
680 memory locations and registers; GNU CC fundamentally assumes that the
681 order of words in memory is the same as the order in registers. This
682 macro need not be a constant.
684 @findex LIBGCC2_WORDS_BIG_ENDIAN
685 @item LIBGCC2_WORDS_BIG_ENDIAN
686 Define this macro if WORDS_BIG_ENDIAN is not constant. This must be a
687 constant value with the same meaning as WORDS_BIG_ENDIAN, which will be
688 used only when compiling libgcc2.c. Typically the value will be set
689 based on preprocessor defines.
691 @findex FLOAT_WORDS_BIG_ENDIAN
692 @item FLOAT_WORDS_BIG_ENDIAN
693 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
694 @code{TFmode} floating point numbers are stored in memory with the word
695 containing the sign bit at the lowest address; otherwise define it to
696 have the value 0. This macro need not be a constant.
698 You need not define this macro if the ordering is the same as for
701 @findex BITS_PER_UNIT
703 Define this macro to be the number of bits in an addressable storage
704 unit (byte); normally 8.
706 @findex BITS_PER_WORD
708 Number of bits in a word; normally 32.
710 @findex MAX_BITS_PER_WORD
711 @item MAX_BITS_PER_WORD
712 Maximum number of bits in a word. If this is undefined, the default is
713 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
714 largest value that @code{BITS_PER_WORD} can have at run-time.
716 @findex UNITS_PER_WORD
718 Number of storage units in a word; normally 4.
720 @findex MIN_UNITS_PER_WORD
721 @item MIN_UNITS_PER_WORD
722 Minimum number of units in a word. If this is undefined, the default is
723 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
724 smallest value that @code{UNITS_PER_WORD} can have at run-time.
728 Width of a pointer, in bits. You must specify a value no wider than the
729 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
730 you must define @code{POINTERS_EXTEND_UNSIGNED}.
732 @findex POINTERS_EXTEND_UNSIGNED
733 @item POINTERS_EXTEND_UNSIGNED
734 A C expression whose value is nonzero if pointers that need to be
735 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
736 be zero-extended and zero if they are to be sign-extended.
738 You need not define this macro if the @code{POINTER_SIZE} is equal
739 to the width of @code{Pmode}.
742 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
743 A macro to update @var{m} and @var{unsignedp} when an object whose type
744 is @var{type} and which has the specified mode and signedness is to be
745 stored in a register. This macro is only called when @var{type} is a
748 On most RISC machines, which only have operations that operate on a full
749 register, define this macro to set @var{m} to @code{word_mode} if
750 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
751 cases, only integer modes should be widened because wider-precision
752 floating-point operations are usually more expensive than their narrower
755 For most machines, the macro definition does not change @var{unsignedp}.
756 However, some machines, have instructions that preferentially handle
757 either signed or unsigned quantities of certain modes. For example, on
758 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
759 sign-extend the result to 64 bits. On such machines, set
760 @var{unsignedp} according to which kind of extension is more efficient.
762 Do not define this macro if it would never modify @var{m}.
764 @findex PROMOTE_FUNCTION_ARGS
765 @item PROMOTE_FUNCTION_ARGS
766 Define this macro if the promotion described by @code{PROMOTE_MODE}
767 should also be done for outgoing function arguments.
769 @findex PROMOTE_FUNCTION_RETURN
770 @item PROMOTE_FUNCTION_RETURN
771 Define this macro if the promotion described by @code{PROMOTE_MODE}
772 should also be done for the return value of functions.
774 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
775 promotions done by @code{PROMOTE_MODE}.
777 @findex PROMOTE_FOR_CALL_ONLY
778 @item PROMOTE_FOR_CALL_ONLY
779 Define this macro if the promotion described by @code{PROMOTE_MODE}
780 should @emph{only} be performed for outgoing function arguments or
781 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
782 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
784 @findex PARM_BOUNDARY
786 Normal alignment required for function parameters on the stack, in
787 bits. All stack parameters receive at least this much alignment
788 regardless of data type. On most machines, this is the same as the
791 @findex STACK_BOUNDARY
793 Define this macro if there is a guaranteed alignment for the stack
794 pointer on this machine. The definition is a C expression
795 for the desired alignment (measured in bits). This value is used as a
796 default if PREFERRED_STACK_BOUNDARY is not defined.
798 @findex PREFERRED_STACK_BOUNDARY
799 @item PREFERRED_STACK_BOUNDARY
800 Define this macro if you wish to preserve a certain alignment for
801 the stack pointer. The definition is a C expression
802 for the desired alignment (measured in bits). If STACK_BOUNDARY is
803 also defined, this macro must evaluate to a value equal to or larger
806 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
807 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
808 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
809 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
810 be momentarily unaligned while pushing arguments.
812 @findex FUNCTION_BOUNDARY
813 @item FUNCTION_BOUNDARY
814 Alignment required for a function entry point, in bits.
816 @findex BIGGEST_ALIGNMENT
817 @item BIGGEST_ALIGNMENT
818 Biggest alignment that any data type can require on this machine, in bits.
820 @findex MINIMUM_ATOMIC_ALIGNMENT
821 @item MINIMUM_ATOMIC_ALIGNMENT
822 If defined, the smallest alignment, in bits, that can be given to an
823 object that can be referenced in one operation, without disturbing any
824 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
825 on machines that don't have byte or half-word store operations.
827 @findex BIGGEST_FIELD_ALIGNMENT
828 @item BIGGEST_FIELD_ALIGNMENT
829 Biggest alignment that any structure field can require on this machine,
830 in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
831 structure fields only.
833 @findex ADJUST_FIELD_ALIGN
834 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
835 An expression for the alignment of a structure field @var{field} if the
836 alignment computed in the usual way is @var{computed}. GNU CC uses
837 this value instead of the value in @code{BIGGEST_ALIGNMENT} or
838 @code{BIGGEST_FIELD_ALIGNMENT}, if defined, for structure fields only.
840 @findex MAX_OFILE_ALIGNMENT
841 @item MAX_OFILE_ALIGNMENT
842 Biggest alignment supported by the object file format of this machine.
843 Use this macro to limit the alignment which can be specified using the
844 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
845 the default value is @code{BIGGEST_ALIGNMENT}.
847 @findex DATA_ALIGNMENT
848 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
849 If defined, a C expression to compute the alignment for a variables in
850 the static store. @var{type} is the data type, and @var{basic-align} is
851 the alignment that the object would ordinarily have. The value of this
852 macro is used instead of that alignment to align the object.
854 If this macro is not defined, then @var{basic-align} is used.
857 One use of this macro is to increase alignment of medium-size data to
858 make it all fit in fewer cache lines. Another is to cause character
859 arrays to be word-aligned so that @code{strcpy} calls that copy
860 constants to character arrays can be done inline.
862 @findex CONSTANT_ALIGNMENT
863 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
864 If defined, a C expression to compute the alignment given to a constant
865 that is being placed in memory. @var{constant} is the constant and
866 @var{basic-align} is the alignment that the object would ordinarily
867 have. The value of this macro is used instead of that alignment to
870 If this macro is not defined, then @var{basic-align} is used.
872 The typical use of this macro is to increase alignment for string
873 constants to be word aligned so that @code{strcpy} calls that copy
874 constants can be done inline.
876 @findex LOCAL_ALIGNMENT
877 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
878 If defined, a C expression to compute the alignment for a variables in
879 the local store. @var{type} is the data type, and @var{basic-align} is
880 the alignment that the object would ordinarily have. The value of this
881 macro is used instead of that alignment to align the object.
883 If this macro is not defined, then @var{basic-align} is used.
885 One use of this macro is to increase alignment of medium-size data to
886 make it all fit in fewer cache lines.
888 @findex EMPTY_FIELD_BOUNDARY
889 @item EMPTY_FIELD_BOUNDARY
890 Alignment in bits to be given to a structure bit field that follows an
891 empty field such as @code{int : 0;}.
893 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
894 that results from an empty field.
896 @findex STRUCTURE_SIZE_BOUNDARY
897 @item STRUCTURE_SIZE_BOUNDARY
898 Number of bits which any structure or union's size must be a multiple of.
899 Each structure or union's size is rounded up to a multiple of this.
901 If you do not define this macro, the default is the same as
902 @code{BITS_PER_UNIT}.
904 @findex STRICT_ALIGNMENT
905 @item STRICT_ALIGNMENT
906 Define this macro to be the value 1 if instructions will fail to work
907 if given data not on the nominal alignment. If instructions will merely
908 go slower in that case, define this macro as 0.
910 @findex PCC_BITFIELD_TYPE_MATTERS
911 @item PCC_BITFIELD_TYPE_MATTERS
912 Define this if you wish to imitate the way many other C compilers handle
913 alignment of bitfields and the structures that contain them.
915 The behavior is that the type written for a bitfield (@code{int},
916 @code{short}, or other integer type) imposes an alignment for the
917 entire structure, as if the structure really did contain an ordinary
918 field of that type. In addition, the bitfield is placed within the
919 structure so that it would fit within such a field, not crossing a
922 Thus, on most machines, a bitfield whose type is written as @code{int}
923 would not cross a four-byte boundary, and would force four-byte
924 alignment for the whole structure. (The alignment used may not be four
925 bytes; it is controlled by the other alignment parameters.)
927 If the macro is defined, its definition should be a C expression;
928 a nonzero value for the expression enables this behavior.
930 Note that if this macro is not defined, or its value is zero, some
931 bitfields may cross more than one alignment boundary. The compiler can
932 support such references if there are @samp{insv}, @samp{extv}, and
933 @samp{extzv} insns that can directly reference memory.
935 The other known way of making bitfields work is to define
936 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
937 Then every structure can be accessed with fullwords.
939 Unless the machine has bitfield instructions or you define
940 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
941 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
943 If your aim is to make GNU CC use the same conventions for laying out
944 bitfields as are used by another compiler, here is how to investigate
945 what the other compiler does. Compile and run this program:
964 printf ("Size of foo1 is %d\n",
965 sizeof (struct foo1));
966 printf ("Size of foo2 is %d\n",
967 sizeof (struct foo2));
972 If this prints 2 and 5, then the compiler's behavior is what you would
973 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
975 @findex BITFIELD_NBYTES_LIMITED
976 @item BITFIELD_NBYTES_LIMITED
977 Like PCC_BITFIELD_TYPE_MATTERS except that its effect is limited to
978 aligning a bitfield within the structure.
980 @findex ROUND_TYPE_SIZE
981 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
982 Define this macro as an expression for the overall size of a type
983 (given by @var{type} as a tree node) when the size computed in the
984 usual way is @var{computed} and the alignment is @var{specified}.
986 The default is to round @var{computed} up to a multiple of @var{specified}.
988 @findex ROUND_TYPE_ALIGN
989 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
990 Define this macro as an expression for the alignment of a type (given
991 by @var{type} as a tree node) if the alignment computed in the usual
992 way is @var{computed} and the alignment explicitly specified was
995 The default is to use @var{specified} if it is larger; otherwise, use
996 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
998 @findex MAX_FIXED_MODE_SIZE
999 @item MAX_FIXED_MODE_SIZE
1000 An integer expression for the size in bits of the largest integer
1001 machine mode that should actually be used. All integer machine modes of
1002 this size or smaller can be used for structures and unions with the
1003 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1004 (DImode)} is assumed.
1006 @findex STACK_SAVEAREA_MODE
1007 @item STACK_SAVEAREA_MODE (@var{save_level})
1008 If defined, an expression of type @code{enum machine_mode} that
1009 specifies the mode of the save area operand of a
1010 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1011 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1012 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1013 having its mode specified.
1015 You need not define this macro if it always returns @code{Pmode}. You
1016 would most commonly define this macro if the
1017 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1020 @findex STACK_SIZE_MODE
1021 @item STACK_SIZE_MODE
1022 If defined, an expression of type @code{enum machine_mode} that
1023 specifies the mode of the size increment operand of an
1024 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1026 You need not define this macro if it always returns @code{word_mode}.
1027 You would most commonly define this macro if the @code{allocate_stack}
1028 pattern needs to support both a 32- and a 64-bit mode.
1030 @findex CHECK_FLOAT_VALUE
1031 @item CHECK_FLOAT_VALUE (@var{mode}, @var{value}, @var{overflow})
1032 A C statement to validate the value @var{value} (of type
1033 @code{double}) for mode @var{mode}. This means that you check whether
1034 @var{value} fits within the possible range of values for mode
1035 @var{mode} on this target machine. The mode @var{mode} is always
1036 a mode of class @code{MODE_FLOAT}. @var{overflow} is nonzero if
1037 the value is already known to be out of range.
1039 If @var{value} is not valid or if @var{overflow} is nonzero, you should
1040 set @var{overflow} to 1 and then assign some valid value to @var{value}.
1041 Allowing an invalid value to go through the compiler can produce
1042 incorrect assembler code which may even cause Unix assemblers to crash.
1044 This macro need not be defined if there is no work for it to do.
1046 @findex TARGET_FLOAT_FORMAT
1047 @item TARGET_FLOAT_FORMAT
1048 A code distinguishing the floating point format of the target machine.
1049 There are three defined values:
1052 @findex IEEE_FLOAT_FORMAT
1053 @item IEEE_FLOAT_FORMAT
1054 This code indicates IEEE floating point. It is the default; there is no
1055 need to define this macro when the format is IEEE.
1057 @findex VAX_FLOAT_FORMAT
1058 @item VAX_FLOAT_FORMAT
1059 This code indicates the peculiar format used on the Vax.
1061 @findex UNKNOWN_FLOAT_FORMAT
1062 @item UNKNOWN_FLOAT_FORMAT
1063 This code indicates any other format.
1066 The value of this macro is compared with @code{HOST_FLOAT_FORMAT}
1067 (@pxref{Config}) to determine whether the target machine has the same
1068 format as the host machine. If any other formats are actually in use on
1069 supported machines, new codes should be defined for them.
1071 The ordering of the component words of floating point values stored in
1072 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN} for the target
1073 machine and @code{HOST_FLOAT_WORDS_BIG_ENDIAN} for the host.
1075 @findex DEFAULT_VTABLE_THUNKS
1076 @item DEFAULT_VTABLE_THUNKS
1077 GNU CC supports two ways of implementing C++ vtables: traditional or with
1078 so-called ``thunks''. The flag @samp{-fvtable-thunk} chooses between them.
1079 Define this macro to be a C expression for the default value of that flag.
1080 If @code{DEFAULT_VTABLE_THUNKS} is 0, GNU CC uses the traditional
1081 implementation by default. The ``thunk'' implementation is more efficient
1082 (especially if you have provided an implementation of
1083 @code{ASM_OUTPUT_MI_THUNK}, see @ref{Function Entry}), but is not binary
1084 compatible with code compiled using the traditional implementation.
1085 If you are writing a new ports, define @code{DEFAULT_VTABLE_THUNKS} to 1.
1087 If you do not define this macro, the default for @samp{-fvtable-thunk} is 0.
1091 @section Layout of Source Language Data Types
1093 These macros define the sizes and other characteristics of the standard
1094 basic data types used in programs being compiled. Unlike the macros in
1095 the previous section, these apply to specific features of C and related
1096 languages, rather than to fundamental aspects of storage layout.
1099 @findex INT_TYPE_SIZE
1101 A C expression for the size in bits of the type @code{int} on the
1102 target machine. If you don't define this, the default is one word.
1104 @findex MAX_INT_TYPE_SIZE
1105 @item MAX_INT_TYPE_SIZE
1106 Maximum number for the size in bits of the type @code{int} on the target
1107 machine. If this is undefined, the default is @code{INT_TYPE_SIZE}.
1108 Otherwise, it is the constant value that is the largest value that
1109 @code{INT_TYPE_SIZE} can have at run-time. This is used in @code{cpp}.
1111 @findex SHORT_TYPE_SIZE
1112 @item SHORT_TYPE_SIZE
1113 A C expression for the size in bits of the type @code{short} on the
1114 target machine. If you don't define this, the default is half a word.
1115 (If this would be less than one storage unit, it is rounded up to one
1118 @findex LONG_TYPE_SIZE
1119 @item LONG_TYPE_SIZE
1120 A C expression for the size in bits of the type @code{long} on the
1121 target machine. If you don't define this, the default is one word.
1123 @findex MAX_LONG_TYPE_SIZE
1124 @item MAX_LONG_TYPE_SIZE
1125 Maximum number for the size in bits of the type @code{long} on the
1126 target machine. If this is undefined, the default is
1127 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1128 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1131 @findex LONG_LONG_TYPE_SIZE
1132 @item LONG_LONG_TYPE_SIZE
1133 A C expression for the size in bits of the type @code{long long} on the
1134 target machine. If you don't define this, the default is two
1135 words. If you want to support GNU Ada on your machine, the value of
1136 macro must be at least 64.
1138 @findex CHAR_TYPE_SIZE
1139 @item CHAR_TYPE_SIZE
1140 A C expression for the size in bits of the type @code{char} on the
1141 target machine. If you don't define this, the default is one quarter
1142 of a word. (If this would be less than one storage unit, it is rounded up
1145 @findex MAX_CHAR_TYPE_SIZE
1146 @item MAX_CHAR_TYPE_SIZE
1147 Maximum number for the size in bits of the type @code{char} on the
1148 target machine. If this is undefined, the default is
1149 @code{CHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1150 largest value that @code{CHAR_TYPE_SIZE} can have at run-time. This is
1153 @findex FLOAT_TYPE_SIZE
1154 @item FLOAT_TYPE_SIZE
1155 A C expression for the size in bits of the type @code{float} on the
1156 target machine. If you don't define this, the default is one word.
1158 @findex DOUBLE_TYPE_SIZE
1159 @item DOUBLE_TYPE_SIZE
1160 A C expression for the size in bits of the type @code{double} on the
1161 target machine. If you don't define this, the default is two
1164 @findex LONG_DOUBLE_TYPE_SIZE
1165 @item LONG_DOUBLE_TYPE_SIZE
1166 A C expression for the size in bits of the type @code{long double} on
1167 the target machine. If you don't define this, the default is two
1170 @findex WIDEST_HARDWARE_FP_SIZE
1171 @item WIDEST_HARDWARE_FP_SIZE
1172 A C expression for the size in bits of the widest floating-point format
1173 supported by the hardware. If you define this macro, you must specify a
1174 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1175 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1178 @findex DEFAULT_SIGNED_CHAR
1179 @item DEFAULT_SIGNED_CHAR
1180 An expression whose value is 1 or 0, according to whether the type
1181 @code{char} should be signed or unsigned by default. The user can
1182 always override this default with the options @samp{-fsigned-char}
1183 and @samp{-funsigned-char}.
1185 @findex DEFAULT_SHORT_ENUMS
1186 @item DEFAULT_SHORT_ENUMS
1187 A C expression to determine whether to give an @code{enum} type
1188 only as many bytes as it takes to represent the range of possible values
1189 of that type. A nonzero value means to do that; a zero value means all
1190 @code{enum} types should be allocated like @code{int}.
1192 If you don't define the macro, the default is 0.
1196 A C expression for a string describing the name of the data type to use
1197 for size values. The typedef name @code{size_t} is defined using the
1198 contents of the string.
1200 The string can contain more than one keyword. If so, separate them with
1201 spaces, and write first any length keyword, then @code{unsigned} if
1202 appropriate, and finally @code{int}. The string must exactly match one
1203 of the data type names defined in the function
1204 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1205 omit @code{int} or change the order---that would cause the compiler to
1208 If you don't define this macro, the default is @code{"long unsigned
1211 @findex PTRDIFF_TYPE
1213 A C expression for a string describing the name of the data type to use
1214 for the result of subtracting two pointers. The typedef name
1215 @code{ptrdiff_t} is defined using the contents of the string. See
1216 @code{SIZE_TYPE} above for more information.
1218 If you don't define this macro, the default is @code{"long int"}.
1222 A C expression for a string describing the name of the data type to use
1223 for wide characters. The typedef name @code{wchar_t} is defined using
1224 the contents of the string. See @code{SIZE_TYPE} above for more
1227 If you don't define this macro, the default is @code{"int"}.
1229 @findex WCHAR_TYPE_SIZE
1230 @item WCHAR_TYPE_SIZE
1231 A C expression for the size in bits of the data type for wide
1232 characters. This is used in @code{cpp}, which cannot make use of
1235 @findex MAX_WCHAR_TYPE_SIZE
1236 @item MAX_WCHAR_TYPE_SIZE
1237 Maximum number for the size in bits of the data type for wide
1238 characters. If this is undefined, the default is
1239 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1240 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1243 @findex OBJC_INT_SELECTORS
1244 @item OBJC_INT_SELECTORS
1245 Define this macro if the type of Objective C selectors should be
1248 If this macro is not defined, then selectors should have the type
1249 @code{struct objc_selector *}.
1251 @findex OBJC_SELECTORS_WITHOUT_LABELS
1252 @item OBJC_SELECTORS_WITHOUT_LABELS
1253 Define this macro if the compiler can group all the selectors together
1254 into a vector and use just one label at the beginning of the vector.
1255 Otherwise, the compiler must give each selector its own assembler
1258 On certain machines, it is important to have a separate label for each
1259 selector because this enables the linker to eliminate duplicate selectors.
1263 A C constant expression for the integer value for escape sequence
1268 @findex TARGET_NEWLINE
1271 @itemx TARGET_NEWLINE
1272 C constant expressions for the integer values for escape sequences
1273 @samp{\b}, @samp{\t} and @samp{\n}.
1281 C constant expressions for the integer values for escape sequences
1282 @samp{\v}, @samp{\f} and @samp{\r}.
1286 @section Register Usage
1287 @cindex register usage
1289 This section explains how to describe what registers the target machine
1290 has, and how (in general) they can be used.
1292 The description of which registers a specific instruction can use is
1293 done with register classes; see @ref{Register Classes}. For information
1294 on using registers to access a stack frame, see @ref{Frame Registers}.
1295 For passing values in registers, see @ref{Register Arguments}.
1296 For returning values in registers, see @ref{Scalar Return}.
1299 * Register Basics:: Number and kinds of registers.
1300 * Allocation Order:: Order in which registers are allocated.
1301 * Values in Registers:: What kinds of values each reg can hold.
1302 * Leaf Functions:: Renumbering registers for leaf functions.
1303 * Stack Registers:: Handling a register stack such as 80387.
1304 * Obsolete Register Macros:: Macros formerly used for the 80387.
1307 @node Register Basics
1308 @subsection Basic Characteristics of Registers
1310 @c prevent bad page break with this line
1311 Registers have various characteristics.
1314 @findex FIRST_PSEUDO_REGISTER
1315 @item FIRST_PSEUDO_REGISTER
1316 Number of hardware registers known to the compiler. They receive
1317 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1318 pseudo register's number really is assigned the number
1319 @code{FIRST_PSEUDO_REGISTER}.
1321 @item FIXED_REGISTERS
1322 @findex FIXED_REGISTERS
1323 @cindex fixed register
1324 An initializer that says which registers are used for fixed purposes
1325 all throughout the compiled code and are therefore not available for
1326 general allocation. These would include the stack pointer, the frame
1327 pointer (except on machines where that can be used as a general
1328 register when no frame pointer is needed), the program counter on
1329 machines where that is considered one of the addressable registers,
1330 and any other numbered register with a standard use.
1332 This information is expressed as a sequence of numbers, separated by
1333 commas and surrounded by braces. The @var{n}th number is 1 if
1334 register @var{n} is fixed, 0 otherwise.
1336 The table initialized from this macro, and the table initialized by
1337 the following one, may be overridden at run time either automatically,
1338 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1339 the user with the command options @samp{-ffixed-@var{reg}},
1340 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}}.
1342 @findex CALL_USED_REGISTERS
1343 @item CALL_USED_REGISTERS
1344 @cindex call-used register
1345 @cindex call-clobbered register
1346 @cindex call-saved register
1347 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1348 clobbered (in general) by function calls as well as for fixed
1349 registers. This macro therefore identifies the registers that are not
1350 available for general allocation of values that must live across
1353 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1354 automatically saves it on function entry and restores it on function
1355 exit, if the register is used within the function.
1357 @findex HARD_REGNO_CALL_PART_CLOBBERED
1358 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1359 @cindex call-used register
1360 @cindex call-clobbered register
1361 @cindex call-saved register
1362 A C expression that is non-zero if it is not permissible to store a
1363 value of mode @var{mode} in hard register number @var{regno} across a
1364 call without some part of it being clobbered. For most machines this
1365 macro need not be defined. It is only required for machines that do not
1366 preserve the entire contents of a register across a call.
1368 @findex CONDITIONAL_REGISTER_USAGE
1370 @findex call_used_regs
1371 @item CONDITIONAL_REGISTER_USAGE
1372 Zero or more C statements that may conditionally modify four variables
1373 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs}
1374 (these three are of type @code{char []}) and @code{reg_class_contents}
1375 (of type @code{HARD_REG_SET}).
1376 Before the macro is called @code{fixed_regs}, @code{call_used_regs}
1377 and @code{reg_class_contents} have been initialized from
1378 @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS} and
1379 @code{REG_CLASS_CONTENTS}, respectively,
1380 @code{global_regs} has been cleared, and any @samp{-ffixed-@var{reg}},
1381 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}} command
1382 options have been applied.
1384 This is necessary in case the fixed or call-clobbered registers depend
1387 You need not define this macro if it has no work to do.
1389 @cindex disabling certain registers
1390 @cindex controlling register usage
1391 If the usage of an entire class of registers depends on the target
1392 flags, you may indicate this to GCC by using this macro to modify
1393 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1394 registers in the classes which should not be used by GCC. Also define
1395 the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1396 is called with a letter for a class that shouldn't be used.
1398 (However, if this class is not included in @code{GENERAL_REGS} and all
1399 of the insn patterns whose constraints permit this class are
1400 controlled by target switches, then GCC will automatically avoid using
1401 these registers when the target switches are opposed to them.)
1403 @findex NON_SAVING_SETJMP
1404 @item NON_SAVING_SETJMP
1405 If this macro is defined and has a nonzero value, it means that
1406 @code{setjmp} and related functions fail to save the registers, or that
1407 @code{longjmp} fails to restore them. To compensate, the compiler
1408 avoids putting variables in registers in functions that use
1411 @findex INCOMING_REGNO
1412 @item INCOMING_REGNO (@var{out})
1413 Define this macro if the target machine has register windows. This C
1414 expression returns the register number as seen by the called function
1415 corresponding to the register number @var{out} as seen by the calling
1416 function. Return @var{out} if register number @var{out} is not an
1419 @findex OUTGOING_REGNO
1420 @item OUTGOING_REGNO (@var{in})
1421 Define this macro if the target machine has register windows. This C
1422 expression returns the register number as seen by the calling function
1423 corresponding to the register number @var{in} as seen by the called
1424 function. Return @var{in} if register number @var{in} is not an inbound
1430 If the program counter has a register number, define this as that
1431 register number. Otherwise, do not define it.
1435 @node Allocation Order
1436 @subsection Order of Allocation of Registers
1437 @cindex order of register allocation
1438 @cindex register allocation order
1440 @c prevent bad page break with this line
1441 Registers are allocated in order.
1444 @findex REG_ALLOC_ORDER
1445 @item REG_ALLOC_ORDER
1446 If defined, an initializer for a vector of integers, containing the
1447 numbers of hard registers in the order in which GNU CC should prefer
1448 to use them (from most preferred to least).
1450 If this macro is not defined, registers are used lowest numbered first
1451 (all else being equal).
1453 One use of this macro is on machines where the highest numbered
1454 registers must always be saved and the save-multiple-registers
1455 instruction supports only sequences of consecutive registers. On such
1456 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1457 the highest numbered allocable register first.
1459 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1460 @item ORDER_REGS_FOR_LOCAL_ALLOC
1461 A C statement (sans semicolon) to choose the order in which to allocate
1462 hard registers for pseudo-registers local to a basic block.
1464 Store the desired register order in the array @code{reg_alloc_order}.
1465 Element 0 should be the register to allocate first; element 1, the next
1466 register; and so on.
1468 The macro body should not assume anything about the contents of
1469 @code{reg_alloc_order} before execution of the macro.
1471 On most machines, it is not necessary to define this macro.
1474 @node Values in Registers
1475 @subsection How Values Fit in Registers
1477 This section discusses the macros that describe which kinds of values
1478 (specifically, which machine modes) each register can hold, and how many
1479 consecutive registers are needed for a given mode.
1482 @findex HARD_REGNO_NREGS
1483 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1484 A C expression for the number of consecutive hard registers, starting
1485 at register number @var{regno}, required to hold a value of mode
1488 On a machine where all registers are exactly one word, a suitable
1489 definition of this macro is
1492 #define HARD_REGNO_NREGS(REGNO, MODE) \
1493 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1497 @findex ALTER_HARD_SUBREG
1498 @item ALTER_HARD_SUBREG (@var{tgt_mode}, @var{word}, @var{src_mode}, @var{regno})
1499 A C expression that returns an adjusted hard register number for
1502 (subreg:@var{tgt_mode} (reg:@var{src_mode} @var{regno}) @var{word})
1505 This may be needed if the target machine has mixed sized big-endian
1506 registers, like Sparc v9.
1508 @findex HARD_REGNO_MODE_OK
1509 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1510 A C expression that is nonzero if it is permissible to store a value
1511 of mode @var{mode} in hard register number @var{regno} (or in several
1512 registers starting with that one). For a machine where all registers
1513 are equivalent, a suitable definition is
1516 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1519 You need not include code to check for the numbers of fixed registers,
1520 because the allocation mechanism considers them to be always occupied.
1522 @cindex register pairs
1523 On some machines, double-precision values must be kept in even/odd
1524 register pairs. You can implement that by defining this macro to reject
1525 odd register numbers for such modes.
1527 The minimum requirement for a mode to be OK in a register is that the
1528 @samp{mov@var{mode}} instruction pattern support moves between the
1529 register and other hard register in the same class and that moving a
1530 value into the register and back out not alter it.
1532 Since the same instruction used to move @code{word_mode} will work for
1533 all narrower integer modes, it is not necessary on any machine for
1534 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1535 you define patterns @samp{movhi}, etc., to take advantage of this. This
1536 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1537 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1540 Many machines have special registers for floating point arithmetic.
1541 Often people assume that floating point machine modes are allowed only
1542 in floating point registers. This is not true. Any registers that
1543 can hold integers can safely @emph{hold} a floating point machine
1544 mode, whether or not floating arithmetic can be done on it in those
1545 registers. Integer move instructions can be used to move the values.
1547 On some machines, though, the converse is true: fixed-point machine
1548 modes may not go in floating registers. This is true if the floating
1549 registers normalize any value stored in them, because storing a
1550 non-floating value there would garble it. In this case,
1551 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1552 floating registers. But if the floating registers do not automatically
1553 normalize, if you can store any bit pattern in one and retrieve it
1554 unchanged without a trap, then any machine mode may go in a floating
1555 register, so you can define this macro to say so.
1557 The primary significance of special floating registers is rather that
1558 they are the registers acceptable in floating point arithmetic
1559 instructions. However, this is of no concern to
1560 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1561 constraints for those instructions.
1563 On some machines, the floating registers are especially slow to access,
1564 so that it is better to store a value in a stack frame than in such a
1565 register if floating point arithmetic is not being done. As long as the
1566 floating registers are not in class @code{GENERAL_REGS}, they will not
1567 be used unless some pattern's constraint asks for one.
1569 @findex MODES_TIEABLE_P
1570 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
1571 A C expression that is nonzero if a value of mode
1572 @var{mode1} is accessible in mode @var{mode2} without copying.
1574 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
1575 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
1576 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
1577 should be nonzero. If they differ for any @var{r}, you should define
1578 this macro to return zero unless some other mechanism ensures the
1579 accessibility of the value in a narrower mode.
1581 You should define this macro to return nonzero in as many cases as
1582 possible since doing so will allow GNU CC to perform better register
1585 @findex AVOID_CCMODE_COPIES
1586 @item AVOID_CCMODE_COPIES
1587 Define this macro if the compiler should avoid copies to/from @code{CCmode}
1588 registers. You should only define this macro if support fo copying to/from
1589 @code{CCmode} is incomplete.
1592 @node Leaf Functions
1593 @subsection Handling Leaf Functions
1595 @cindex leaf functions
1596 @cindex functions, leaf
1597 On some machines, a leaf function (i.e., one which makes no calls) can run
1598 more efficiently if it does not make its own register window. Often this
1599 means it is required to receive its arguments in the registers where they
1600 are passed by the caller, instead of the registers where they would
1603 The special treatment for leaf functions generally applies only when
1604 other conditions are met; for example, often they may use only those
1605 registers for its own variables and temporaries. We use the term ``leaf
1606 function'' to mean a function that is suitable for this special
1607 handling, so that functions with no calls are not necessarily ``leaf
1610 GNU CC assigns register numbers before it knows whether the function is
1611 suitable for leaf function treatment. So it needs to renumber the
1612 registers in order to output a leaf function. The following macros
1616 @findex LEAF_REGISTERS
1617 @item LEAF_REGISTERS
1618 A C initializer for a vector, indexed by hard register number, which
1619 contains 1 for a register that is allowable in a candidate for leaf
1622 If leaf function treatment involves renumbering the registers, then the
1623 registers marked here should be the ones before renumbering---those that
1624 GNU CC would ordinarily allocate. The registers which will actually be
1625 used in the assembler code, after renumbering, should not be marked with 1
1628 Define this macro only if the target machine offers a way to optimize
1629 the treatment of leaf functions.
1631 @findex LEAF_REG_REMAP
1632 @item LEAF_REG_REMAP (@var{regno})
1633 A C expression whose value is the register number to which @var{regno}
1634 should be renumbered, when a function is treated as a leaf function.
1636 If @var{regno} is a register number which should not appear in a leaf
1637 function before renumbering, then the expression should yield -1, which
1638 will cause the compiler to abort.
1640 Define this macro only if the target machine offers a way to optimize the
1641 treatment of leaf functions, and registers need to be renumbered to do
1645 @findex current_function_is_leaf
1646 @findex current_function_uses_only_leaf_regs
1647 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
1648 treat leaf functions specially. They can test the C variable
1649 @code{current_function_is_leaf} which is nonzero for leaf functions.
1650 @code{current_function_is_leaf} is set prior to local register allocation
1651 and is valid for the remaining compiler passes. They can also test the C
1652 variable @code{current_function_uses_only_leaf_regs} which is nonzero for
1653 leaf functions which only use leaf registers.
1654 @code{current_function_uses_only_leaf_regs} is valid after reload and is
1655 only useful if @code{LEAF_REGISTERS} is defined.
1656 @c changed this to fix overfull. ALSO: why the "it" at the beginning
1657 @c of the next paragraph?! --mew 2feb93
1659 @node Stack Registers
1660 @subsection Registers That Form a Stack
1662 There are special features to handle computers where some of the
1663 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
1664 Stack registers are normally written by pushing onto the stack, and are
1665 numbered relative to the top of the stack.
1667 Currently, GNU CC can only handle one group of stack-like registers, and
1668 they must be consecutively numbered.
1673 Define this if the machine has any stack-like registers.
1675 @findex FIRST_STACK_REG
1676 @item FIRST_STACK_REG
1677 The number of the first stack-like register. This one is the top
1680 @findex LAST_STACK_REG
1681 @item LAST_STACK_REG
1682 The number of the last stack-like register. This one is the bottom of
1686 @node Obsolete Register Macros
1687 @subsection Obsolete Macros for Controlling Register Usage
1689 These features do not work very well. They exist because they used to
1690 be required to generate correct code for the 80387 coprocessor of the
1691 80386. They are no longer used by that machine description and may be
1692 removed in a later version of the compiler. Don't use them!
1695 @findex OVERLAPPING_REGNO_P
1696 @item OVERLAPPING_REGNO_P (@var{regno})
1697 If defined, this is a C expression whose value is nonzero if hard
1698 register number @var{regno} is an overlapping register. This means a
1699 hard register which overlaps a hard register with a different number.
1700 (Such overlap is undesirable, but occasionally it allows a machine to
1701 be supported which otherwise could not be.) This macro must return
1702 nonzero for @emph{all} the registers which overlap each other. GNU CC
1703 can use an overlapping register only in certain limited ways. It can
1704 be used for allocation within a basic block, and may be spilled for
1705 reloading; that is all.
1707 If this macro is not defined, it means that none of the hard registers
1708 overlap each other. This is the usual situation.
1710 @findex INSN_CLOBBERS_REGNO_P
1711 @item INSN_CLOBBERS_REGNO_P (@var{insn}, @var{regno})
1712 If defined, this is a C expression whose value should be nonzero if
1713 the insn @var{insn} has the effect of mysteriously clobbering the
1714 contents of hard register number @var{regno}. By ``mysterious'' we
1715 mean that the insn's RTL expression doesn't describe such an effect.
1717 If this macro is not defined, it means that no insn clobbers registers
1718 mysteriously. This is the usual situation; all else being equal,
1719 it is best for the RTL expression to show all the activity.
1723 @node Register Classes
1724 @section Register Classes
1725 @cindex register class definitions
1726 @cindex class definitions, register
1728 On many machines, the numbered registers are not all equivalent.
1729 For example, certain registers may not be allowed for indexed addressing;
1730 certain registers may not be allowed in some instructions. These machine
1731 restrictions are described to the compiler using @dfn{register classes}.
1733 You define a number of register classes, giving each one a name and saying
1734 which of the registers belong to it. Then you can specify register classes
1735 that are allowed as operands to particular instruction patterns.
1739 In general, each register will belong to several classes. In fact, one
1740 class must be named @code{ALL_REGS} and contain all the registers. Another
1741 class must be named @code{NO_REGS} and contain no registers. Often the
1742 union of two classes will be another class; however, this is not required.
1744 @findex GENERAL_REGS
1745 One of the classes must be named @code{GENERAL_REGS}. There is nothing
1746 terribly special about the name, but the operand constraint letters
1747 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
1748 the same as @code{ALL_REGS}, just define it as a macro which expands
1751 Order the classes so that if class @var{x} is contained in class @var{y}
1752 then @var{x} has a lower class number than @var{y}.
1754 The way classes other than @code{GENERAL_REGS} are specified in operand
1755 constraints is through machine-dependent operand constraint letters.
1756 You can define such letters to correspond to various classes, then use
1757 them in operand constraints.
1759 You should define a class for the union of two classes whenever some
1760 instruction allows both classes. For example, if an instruction allows
1761 either a floating point (coprocessor) register or a general register for a
1762 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
1763 which includes both of them. Otherwise you will get suboptimal code.
1765 You must also specify certain redundant information about the register
1766 classes: for each class, which classes contain it and which ones are
1767 contained in it; for each pair of classes, the largest class contained
1770 When a value occupying several consecutive registers is expected in a
1771 certain class, all the registers used must belong to that class.
1772 Therefore, register classes cannot be used to enforce a requirement for
1773 a register pair to start with an even-numbered register. The way to
1774 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
1776 Register classes used for input-operands of bitwise-and or shift
1777 instructions have a special requirement: each such class must have, for
1778 each fixed-point machine mode, a subclass whose registers can transfer that
1779 mode to or from memory. For example, on some machines, the operations for
1780 single-byte values (@code{QImode}) are limited to certain registers. When
1781 this is so, each register class that is used in a bitwise-and or shift
1782 instruction must have a subclass consisting of registers from which
1783 single-byte values can be loaded or stored. This is so that
1784 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
1787 @findex enum reg_class
1788 @item enum reg_class
1789 An enumeral type that must be defined with all the register class names
1790 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
1791 must be the last register class, followed by one more enumeral value,
1792 @code{LIM_REG_CLASSES}, which is not a register class but rather
1793 tells how many classes there are.
1795 Each register class has a number, which is the value of casting
1796 the class name to type @code{int}. The number serves as an index
1797 in many of the tables described below.
1799 @findex N_REG_CLASSES
1801 The number of distinct register classes, defined as follows:
1804 #define N_REG_CLASSES (int) LIM_REG_CLASSES
1807 @findex REG_CLASS_NAMES
1808 @item REG_CLASS_NAMES
1809 An initializer containing the names of the register classes as C string
1810 constants. These names are used in writing some of the debugging dumps.
1812 @findex REG_CLASS_CONTENTS
1813 @item REG_CLASS_CONTENTS
1814 An initializer containing the contents of the register classes, as integers
1815 which are bit masks. The @var{n}th integer specifies the contents of class
1816 @var{n}. The way the integer @var{mask} is interpreted is that
1817 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
1819 When the machine has more than 32 registers, an integer does not suffice.
1820 Then the integers are replaced by sub-initializers, braced groupings containing
1821 several integers. Each sub-initializer must be suitable as an initializer
1822 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
1824 @findex REGNO_REG_CLASS
1825 @item REGNO_REG_CLASS (@var{regno})
1826 A C expression whose value is a register class containing hard register
1827 @var{regno}. In general there is more than one such class; choose a class
1828 which is @dfn{minimal}, meaning that no smaller class also contains the
1831 @findex BASE_REG_CLASS
1832 @item BASE_REG_CLASS
1833 A macro whose definition is the name of the class to which a valid
1834 base register must belong. A base register is one used in an address
1835 which is the register value plus a displacement.
1837 @findex INDEX_REG_CLASS
1838 @item INDEX_REG_CLASS
1839 A macro whose definition is the name of the class to which a valid
1840 index register must belong. An index register is one used in an
1841 address where its value is either multiplied by a scale factor or
1842 added to another register (as well as added to a displacement).
1844 @findex REG_CLASS_FROM_LETTER
1845 @item REG_CLASS_FROM_LETTER (@var{char})
1846 A C expression which defines the machine-dependent operand constraint
1847 letters for register classes. If @var{char} is such a letter, the
1848 value should be the register class corresponding to it. Otherwise,
1849 the value should be @code{NO_REGS}. The register letter @samp{r},
1850 corresponding to class @code{GENERAL_REGS}, will not be passed
1851 to this macro; you do not need to handle it.
1853 @findex REGNO_OK_FOR_BASE_P
1854 @item REGNO_OK_FOR_BASE_P (@var{num})
1855 A C expression which is nonzero if register number @var{num} is
1856 suitable for use as a base register in operand addresses. It may be
1857 either a suitable hard register or a pseudo register that has been
1858 allocated such a hard register.
1860 @findex REGNO_MODE_OK_FOR_BASE_P
1861 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
1862 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
1863 that expression may examine the mode of the memory reference in
1864 @var{mode}. You should define this macro if the mode of the memory
1865 reference affects whether a register may be used as a base register. If
1866 you define this macro, the compiler will use it instead of
1867 @code{REGNO_OK_FOR_BASE_P}.
1869 @findex REGNO_OK_FOR_INDEX_P
1870 @item REGNO_OK_FOR_INDEX_P (@var{num})
1871 A C expression which is nonzero if register number @var{num} is
1872 suitable for use as an index register in operand addresses. It may be
1873 either a suitable hard register or a pseudo register that has been
1874 allocated such a hard register.
1876 The difference between an index register and a base register is that
1877 the index register may be scaled. If an address involves the sum of
1878 two registers, neither one of them scaled, then either one may be
1879 labeled the ``base'' and the other the ``index''; but whichever
1880 labeling is used must fit the machine's constraints of which registers
1881 may serve in each capacity. The compiler will try both labelings,
1882 looking for one that is valid, and will reload one or both registers
1883 only if neither labeling works.
1885 @findex PREFERRED_RELOAD_CLASS
1886 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
1887 A C expression that places additional restrictions on the register class
1888 to use when it is necessary to copy value @var{x} into a register in class
1889 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
1890 another, smaller class. On many machines, the following definition is
1894 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
1897 Sometimes returning a more restrictive class makes better code. For
1898 example, on the 68000, when @var{x} is an integer constant that is in range
1899 for a @samp{moveq} instruction, the value of this macro is always
1900 @code{DATA_REGS} as long as @var{class} includes the data registers.
1901 Requiring a data register guarantees that a @samp{moveq} will be used.
1903 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
1904 you can force @var{x} into a memory constant. This is useful on
1905 certain machines where immediate floating values cannot be loaded into
1906 certain kinds of registers.
1908 @findex PREFERRED_OUTPUT_RELOAD_CLASS
1909 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
1910 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
1911 input reloads. If you don't define this macro, the default is to use
1912 @var{class}, unchanged.
1914 @findex LIMIT_RELOAD_CLASS
1915 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
1916 A C expression that places additional restrictions on the register class
1917 to use when it is necessary to be able to hold a value of mode
1918 @var{mode} in a reload register for which class @var{class} would
1921 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
1922 there are certain modes that simply can't go in certain reload classes.
1924 The value is a register class; perhaps @var{class}, or perhaps another,
1927 Don't define this macro unless the target machine has limitations which
1928 require the macro to do something nontrivial.
1930 @findex SECONDARY_RELOAD_CLASS
1931 @findex SECONDARY_INPUT_RELOAD_CLASS
1932 @findex SECONDARY_OUTPUT_RELOAD_CLASS
1933 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1934 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1935 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1936 Many machines have some registers that cannot be copied directly to or
1937 from memory or even from other types of registers. An example is the
1938 @samp{MQ} register, which on most machines, can only be copied to or
1939 from general registers, but not memory. Some machines allow copying all
1940 registers to and from memory, but require a scratch register for stores
1941 to some memory locations (e.g., those with symbolic address on the RT,
1942 and those with certain symbolic address on the Sparc when compiling
1943 PIC). In some cases, both an intermediate and a scratch register are
1946 You should define these macros to indicate to the reload phase that it may
1947 need to allocate at least one register for a reload in addition to the
1948 register to contain the data. Specifically, if copying @var{x} to a
1949 register @var{class} in @var{mode} requires an intermediate register,
1950 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
1951 largest register class all of whose registers can be used as
1952 intermediate registers or scratch registers.
1954 If copying a register @var{class} in @var{mode} to @var{x} requires an
1955 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
1956 should be defined to return the largest register class required. If the
1957 requirements for input and output reloads are the same, the macro
1958 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
1961 The values returned by these macros are often @code{GENERAL_REGS}.
1962 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
1963 can be directly copied to or from a register of @var{class} in
1964 @var{mode} without requiring a scratch register. Do not define this
1965 macro if it would always return @code{NO_REGS}.
1967 If a scratch register is required (either with or without an
1968 intermediate register), you should define patterns for
1969 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
1970 (@pxref{Standard Names}. These patterns, which will normally be
1971 implemented with a @code{define_expand}, should be similar to the
1972 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
1975 Define constraints for the reload register and scratch register that
1976 contain a single register class. If the original reload register (whose
1977 class is @var{class}) can meet the constraint given in the pattern, the
1978 value returned by these macros is used for the class of the scratch
1979 register. Otherwise, two additional reload registers are required.
1980 Their classes are obtained from the constraints in the insn pattern.
1982 @var{x} might be a pseudo-register or a @code{subreg} of a
1983 pseudo-register, which could either be in a hard register or in memory.
1984 Use @code{true_regnum} to find out; it will return -1 if the pseudo is
1985 in memory and the hard register number if it is in a register.
1987 These macros should not be used in the case where a particular class of
1988 registers can only be copied to memory and not to another class of
1989 registers. In that case, secondary reload registers are not needed and
1990 would not be helpful. Instead, a stack location must be used to perform
1991 the copy and the @code{mov@var{m}} pattern should use memory as a
1992 intermediate storage. This case often occurs between floating-point and
1995 @findex SECONDARY_MEMORY_NEEDED
1996 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
1997 Certain machines have the property that some registers cannot be copied
1998 to some other registers without using memory. Define this macro on
1999 those machines to be a C expression that is non-zero if objects of mode
2000 @var{m} in registers of @var{class1} can only be copied to registers of
2001 class @var{class2} by storing a register of @var{class1} into memory
2002 and loading that memory location into a register of @var{class2}.
2004 Do not define this macro if its value would always be zero.
2006 @findex SECONDARY_MEMORY_NEEDED_RTX
2007 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2008 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2009 allocates a stack slot for a memory location needed for register copies.
2010 If this macro is defined, the compiler instead uses the memory location
2011 defined by this macro.
2013 Do not define this macro if you do not define
2014 @code{SECONDARY_MEMORY_NEEDED}.
2016 @findex SECONDARY_MEMORY_NEEDED_MODE
2017 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2018 When the compiler needs a secondary memory location to copy between two
2019 registers of mode @var{mode}, it normally allocates sufficient memory to
2020 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2021 load operations in a mode that many bits wide and whose class is the
2022 same as that of @var{mode}.
2024 This is right thing to do on most machines because it ensures that all
2025 bits of the register are copied and prevents accesses to the registers
2026 in a narrower mode, which some machines prohibit for floating-point
2029 However, this default behavior is not correct on some machines, such as
2030 the DEC Alpha, that store short integers in floating-point registers
2031 differently than in integer registers. On those machines, the default
2032 widening will not work correctly and you must define this macro to
2033 suppress that widening in some cases. See the file @file{alpha.h} for
2036 Do not define this macro if you do not define
2037 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2038 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2040 @findex SMALL_REGISTER_CLASSES
2041 @item SMALL_REGISTER_CLASSES
2042 On some machines, it is risky to let hard registers live across arbitrary
2043 insns. Typically, these machines have instructions that require values
2044 to be in specific registers (like an accumulator), and reload will fail
2045 if the required hard register is used for another purpose across such an
2048 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a non-zero
2049 value on these machines. When this macro has a non-zero value, the
2050 compiler will try to minimize the lifetime of hard registers.
2052 It is always safe to define this macro with a non-zero value, but if you
2053 unnecessarily define it, you will reduce the amount of optimizations
2054 that can be performed in some cases. If you do not define this macro
2055 with a non-zero value when it is required, the compiler will run out of
2056 spill registers and print a fatal error message. For most machines, you
2057 should not define this macro at all.
2059 @findex CLASS_LIKELY_SPILLED_P
2060 @item CLASS_LIKELY_SPILLED_P (@var{class})
2061 A C expression whose value is nonzero if pseudos that have been assigned
2062 to registers of class @var{class} would likely be spilled because
2063 registers of @var{class} are needed for spill registers.
2065 The default value of this macro returns 1 if @var{class} has exactly one
2066 register and zero otherwise. On most machines, this default should be
2067 used. Only define this macro to some other expression if pseudos
2068 allocated by @file{local-alloc.c} end up in memory because their hard
2069 registers were needed for spill registers. If this macro returns nonzero
2070 for those classes, those pseudos will only be allocated by
2071 @file{global.c}, which knows how to reallocate the pseudo to another
2072 register. If there would not be another register available for
2073 reallocation, you should not change the definition of this macro since
2074 the only effect of such a definition would be to slow down register
2077 @findex CLASS_MAX_NREGS
2078 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2079 A C expression for the maximum number of consecutive registers
2080 of class @var{class} needed to hold a value of mode @var{mode}.
2082 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2083 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2084 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2085 @var{mode})} for all @var{regno} values in the class @var{class}.
2087 This macro helps control the handling of multiple-word values
2090 @item CLASS_CANNOT_CHANGE_SIZE
2091 If defined, a C expression for a class that contains registers which the
2092 compiler must always access in a mode that is the same size as the mode
2093 in which it loaded the register.
2095 For the example, loading 32-bit integer or floating-point objects into
2096 floating-point registers on the Alpha extends them to 64-bits.
2097 Therefore loading a 64-bit object and then storing it as a 32-bit object
2098 does not store the low-order 32-bits, as would be the case for a normal
2099 register. Therefore, @file{alpha.h} defines this macro as
2103 Three other special macros describe which operands fit which constraint
2107 @findex CONST_OK_FOR_LETTER_P
2108 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2109 A C expression that defines the machine-dependent operand constraint
2110 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2111 particular ranges of integer values. If @var{c} is one of those
2112 letters, the expression should check that @var{value}, an integer, is in
2113 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2114 not one of those letters, the value should be 0 regardless of
2117 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2118 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2119 A C expression that defines the machine-dependent operand constraint
2120 letters that specify particular ranges of @code{const_double} values
2121 (@samp{G} or @samp{H}).
2123 If @var{c} is one of those letters, the expression should check that
2124 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2125 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2126 letters, the value should be 0 regardless of @var{value}.
2128 @code{const_double} is used for all floating-point constants and for
2129 @code{DImode} fixed-point constants. A given letter can accept either
2130 or both kinds of values. It can use @code{GET_MODE} to distinguish
2131 between these kinds.
2133 @findex EXTRA_CONSTRAINT
2134 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2135 A C expression that defines the optional machine-dependent constraint
2136 letters (@samp{Q}, @samp{R}, @samp{S}, @samp{T}, @samp{U}) that can
2137 be used to segregate specific types of operands, usually memory
2138 references, for the target machine. Normally this macro will not be
2139 defined. If it is required for a particular target machine, it should
2140 return 1 if @var{value} corresponds to the operand type represented by
2141 the constraint letter @var{c}. If @var{c} is not defined as an extra
2142 constraint, the value returned should be 0 regardless of @var{value}.
2144 For example, on the ROMP, load instructions cannot have their output in r0 if
2145 the memory reference contains a symbolic address. Constraint letter
2146 @samp{Q} is defined as representing a memory address that does
2147 @emph{not} contain a symbolic address. An alternative is specified with
2148 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2149 alternative specifies @samp{m} on the input and a register class that
2150 does not include r0 on the output.
2153 @node Stack and Calling
2154 @section Stack Layout and Calling Conventions
2155 @cindex calling conventions
2157 @c prevent bad page break with this line
2158 This describes the stack layout and calling conventions.
2166 * Register Arguments::
2168 * Aggregate Return::
2175 @subsection Basic Stack Layout
2176 @cindex stack frame layout
2177 @cindex frame layout
2179 @c prevent bad page break with this line
2180 Here is the basic stack layout.
2183 @findex STACK_GROWS_DOWNWARD
2184 @item STACK_GROWS_DOWNWARD
2185 Define this macro if pushing a word onto the stack moves the stack
2186 pointer to a smaller address.
2188 When we say, ``define this macro if @dots{},'' it means that the
2189 compiler checks this macro only with @code{#ifdef} so the precise
2190 definition used does not matter.
2192 @findex FRAME_GROWS_DOWNWARD
2193 @item FRAME_GROWS_DOWNWARD
2194 Define this macro if the addresses of local variable slots are at negative
2195 offsets from the frame pointer.
2197 @findex ARGS_GROW_DOWNWARD
2198 @item ARGS_GROW_DOWNWARD
2199 Define this macro if successive arguments to a function occupy decreasing
2200 addresses on the stack.
2202 @findex STARTING_FRAME_OFFSET
2203 @item STARTING_FRAME_OFFSET
2204 Offset from the frame pointer to the first local variable slot to be allocated.
2206 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2207 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2208 Otherwise, it is found by adding the length of the first slot to the
2209 value @code{STARTING_FRAME_OFFSET}.
2210 @c i'm not sure if the above is still correct.. had to change it to get
2211 @c rid of an overfull. --mew 2feb93
2213 @findex STACK_POINTER_OFFSET
2214 @item STACK_POINTER_OFFSET
2215 Offset from the stack pointer register to the first location at which
2216 outgoing arguments are placed. If not specified, the default value of
2217 zero is used. This is the proper value for most machines.
2219 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2220 the first location at which outgoing arguments are placed.
2222 @findex FIRST_PARM_OFFSET
2223 @item FIRST_PARM_OFFSET (@var{fundecl})
2224 Offset from the argument pointer register to the first argument's
2225 address. On some machines it may depend on the data type of the
2228 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2229 the first argument's address.
2231 @findex STACK_DYNAMIC_OFFSET
2232 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2233 Offset from the stack pointer register to an item dynamically allocated
2234 on the stack, e.g., by @code{alloca}.
2236 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2237 length of the outgoing arguments. The default is correct for most
2238 machines. See @file{function.c} for details.
2240 @findex DYNAMIC_CHAIN_ADDRESS
2241 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2242 A C expression whose value is RTL representing the address in a stack
2243 frame where the pointer to the caller's frame is stored. Assume that
2244 @var{frameaddr} is an RTL expression for the address of the stack frame
2247 If you don't define this macro, the default is to return the value
2248 of @var{frameaddr}---that is, the stack frame address is also the
2249 address of the stack word that points to the previous frame.
2251 @findex SETUP_FRAME_ADDRESSES
2252 @item SETUP_FRAME_ADDRESSES
2253 If defined, a C expression that produces the machine-specific code to
2254 setup the stack so that arbitrary frames can be accessed. For example,
2255 on the Sparc, we must flush all of the register windows to the stack
2256 before we can access arbitrary stack frames. You will seldom need to
2259 @findex BUILTIN_SETJMP_FRAME_VALUE
2260 @item BUILTIN_SETJMP_FRAME_VALUE
2261 If defined, a C expression that contains an rtx that is used to store
2262 the address of the current frame into the built in @code{setjmp} buffer.
2263 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2264 machines. One reason you may need to define this macro is if
2265 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2267 @findex RETURN_ADDR_RTX
2268 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2269 A C expression whose value is RTL representing the value of the return
2270 address for the frame @var{count} steps up from the current frame, after
2271 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2272 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2273 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2275 The value of the expression must always be the correct address when
2276 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2277 determine the return address of other frames.
2279 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2280 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2281 Define this if the return address of a particular stack frame is accessed
2282 from the frame pointer of the previous stack frame.
2284 @findex INCOMING_RETURN_ADDR_RTX
2285 @item INCOMING_RETURN_ADDR_RTX
2286 A C expression whose value is RTL representing the location of the
2287 incoming return address at the beginning of any function, before the
2288 prologue. This RTL is either a @code{REG}, indicating that the return
2289 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2292 You only need to define this macro if you want to support call frame
2293 debugging information like that provided by DWARF 2.
2295 @findex INCOMING_FRAME_SP_OFFSET
2296 @item INCOMING_FRAME_SP_OFFSET
2297 A C expression whose value is an integer giving the offset, in bytes,
2298 from the value of the stack pointer register to the top of the stack
2299 frame at the beginning of any function, before the prologue. The top of
2300 the frame is defined to be the value of the stack pointer in the
2301 previous frame, just before the call instruction.
2303 You only need to define this macro if you want to support call frame
2304 debugging information like that provided by DWARF 2.
2306 @findex ARG_POINTER_CFA_OFFSET
2307 @item ARG_POINTER_CFA_OFFSET
2308 A C expression whose value is an integer giving the offset, in bytes,
2309 from the argument pointer to the canonical frame address (cfa). The
2310 final value should coincide with that calculated by
2311 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2312 during virtual register instantiation.
2314 You only need to define this macro if you want to support call frame
2315 debugging information like that provided by DWARF 2.
2318 @node Stack Checking
2319 @subsection Specifying How Stack Checking is Done
2321 GNU CC will check that stack references are within the boundaries of
2322 the stack, if the @samp{-fstack-check} is specified, in one of three ways:
2326 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GNU CC
2327 will assume that you have arranged for stack checking to be done at
2328 appropriate places in the configuration files, e.g., in
2329 @code{FUNCTION_PROLOGUE}. GNU CC will do not other special processing.
2332 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
2333 called @code{check_stack} in your @file{md} file, GNU CC will call that
2334 pattern with one argument which is the address to compare the stack
2335 value against. You must arrange for this pattern to report an error if
2336 the stack pointer is out of range.
2339 If neither of the above are true, GNU CC will generate code to periodically
2340 ``probe'' the stack pointer using the values of the macros defined below.
2343 Normally, you will use the default values of these macros, so GNU CC
2344 will use the third approach.
2347 @findex STACK_CHECK_BUILTIN
2348 @item STACK_CHECK_BUILTIN
2349 A nonzero value if stack checking is done by the configuration files in a
2350 machine-dependent manner. You should define this macro if stack checking
2351 is require by the ABI of your machine or if you would like to have to stack
2352 checking in some more efficient way than GNU CC's portable approach.
2353 The default value of this macro is zero.
2355 @findex STACK_CHECK_PROBE_INTERVAL
2356 @item STACK_CHECK_PROBE_INTERVAL
2357 An integer representing the interval at which GNU CC must generate stack
2358 probe instructions. You will normally define this macro to be no larger
2359 than the size of the ``guard pages'' at the end of a stack area. The
2360 default value of 4096 is suitable for most systems.
2362 @findex STACK_CHECK_PROBE_LOAD
2363 @item STACK_CHECK_PROBE_LOAD
2364 A integer which is nonzero if GNU CC should perform the stack probe
2365 as a load instruction and zero if GNU CC should use a store instruction.
2366 The default is zero, which is the most efficient choice on most systems.
2368 @findex STACK_CHECK_PROTECT
2369 @item STACK_CHECK_PROTECT
2370 The number of bytes of stack needed to recover from a stack overflow,
2371 for languages where such a recovery is supported. The default value of
2372 75 words should be adequate for most machines.
2374 @findex STACK_CHECK_MAX_FRAME_SIZE
2375 @item STACK_CHECK_MAX_FRAME_SIZE
2376 The maximum size of a stack frame, in bytes. GNU CC will generate probe
2377 instructions in non-leaf functions to ensure at least this many bytes of
2378 stack are available. If a stack frame is larger than this size, stack
2379 checking will not be reliable and GNU CC will issue a warning. The
2380 default is chosen so that GNU CC only generates one instruction on most
2381 systems. You should normally not change the default value of this macro.
2383 @findex STACK_CHECK_FIXED_FRAME_SIZE
2384 @item STACK_CHECK_FIXED_FRAME_SIZE
2385 GNU CC uses this value to generate the above warning message. It
2386 represents the amount of fixed frame used by a function, not including
2387 space for any callee-saved registers, temporaries and user variables.
2388 You need only specify an upper bound for this amount and will normally
2389 use the default of four words.
2391 @findex STACK_CHECK_MAX_VAR_SIZE
2392 @item STACK_CHECK_MAX_VAR_SIZE
2393 The maximum size, in bytes, of an object that GNU CC will place in the
2394 fixed area of the stack frame when the user specifies
2395 @samp{-fstack-check}.
2396 GNU CC computed the default from the values of the above macros and you will
2397 normally not need to override that default.
2401 @node Frame Registers
2402 @subsection Registers That Address the Stack Frame
2404 @c prevent bad page break with this line
2405 This discusses registers that address the stack frame.
2408 @findex STACK_POINTER_REGNUM
2409 @item STACK_POINTER_REGNUM
2410 The register number of the stack pointer register, which must also be a
2411 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2412 the hardware determines which register this is.
2414 @findex FRAME_POINTER_REGNUM
2415 @item FRAME_POINTER_REGNUM
2416 The register number of the frame pointer register, which is used to
2417 access automatic variables in the stack frame. On some machines, the
2418 hardware determines which register this is. On other machines, you can
2419 choose any register you wish for this purpose.
2421 @findex HARD_FRAME_POINTER_REGNUM
2422 @item HARD_FRAME_POINTER_REGNUM
2423 On some machines the offset between the frame pointer and starting
2424 offset of the automatic variables is not known until after register
2425 allocation has been done (for example, because the saved registers are
2426 between these two locations). On those machines, define
2427 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
2428 be used internally until the offset is known, and define
2429 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
2430 used for the frame pointer.
2432 You should define this macro only in the very rare circumstances when it
2433 is not possible to calculate the offset between the frame pointer and
2434 the automatic variables until after register allocation has been
2435 completed. When this macro is defined, you must also indicate in your
2436 definition of @code{ELIMINABLE_REGS} how to eliminate
2437 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
2438 or @code{STACK_POINTER_REGNUM}.
2440 Do not define this macro if it would be the same as
2441 @code{FRAME_POINTER_REGNUM}.
2443 @findex ARG_POINTER_REGNUM
2444 @item ARG_POINTER_REGNUM
2445 The register number of the arg pointer register, which is used to access
2446 the function's argument list. On some machines, this is the same as the
2447 frame pointer register. On some machines, the hardware determines which
2448 register this is. On other machines, you can choose any register you
2449 wish for this purpose. If this is not the same register as the frame
2450 pointer register, then you must mark it as a fixed register according to
2451 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2452 (@pxref{Elimination}).
2454 @findex RETURN_ADDRESS_POINTER_REGNUM
2455 @item RETURN_ADDRESS_POINTER_REGNUM
2456 The register number of the return address pointer register, which is used to
2457 access the current function's return address from the stack. On some
2458 machines, the return address is not at a fixed offset from the frame
2459 pointer or stack pointer or argument pointer. This register can be defined
2460 to point to the return address on the stack, and then be converted by
2461 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
2463 Do not define this macro unless there is no other way to get the return
2464 address from the stack.
2466 @findex STATIC_CHAIN_REGNUM
2467 @findex STATIC_CHAIN_INCOMING_REGNUM
2468 @item STATIC_CHAIN_REGNUM
2469 @itemx STATIC_CHAIN_INCOMING_REGNUM
2470 Register numbers used for passing a function's static chain pointer. If
2471 register windows are used, the register number as seen by the called
2472 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
2473 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
2474 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
2475 not be defined.@refill
2477 The static chain register need not be a fixed register.
2479 If the static chain is passed in memory, these macros should not be
2480 defined; instead, the next two macros should be defined.
2482 @findex STATIC_CHAIN
2483 @findex STATIC_CHAIN_INCOMING
2485 @itemx STATIC_CHAIN_INCOMING
2486 If the static chain is passed in memory, these macros provide rtx giving
2487 @code{mem} expressions that denote where they are stored.
2488 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
2489 as seen by the calling and called functions, respectively. Often the former
2490 will be at an offset from the stack pointer and the latter at an offset from
2491 the frame pointer.@refill
2493 @findex stack_pointer_rtx
2494 @findex frame_pointer_rtx
2495 @findex arg_pointer_rtx
2496 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
2497 @code{arg_pointer_rtx} will have been initialized prior to the use of these
2498 macros and should be used to refer to those items.
2500 If the static chain is passed in a register, the two previous macros should
2505 @subsection Eliminating Frame Pointer and Arg Pointer
2507 @c prevent bad page break with this line
2508 This is about eliminating the frame pointer and arg pointer.
2511 @findex FRAME_POINTER_REQUIRED
2512 @item FRAME_POINTER_REQUIRED
2513 A C expression which is nonzero if a function must have and use a frame
2514 pointer. This expression is evaluated in the reload pass. If its value is
2515 nonzero the function will have a frame pointer.
2517 The expression can in principle examine the current function and decide
2518 according to the facts, but on most machines the constant 0 or the
2519 constant 1 suffices. Use 0 when the machine allows code to be generated
2520 with no frame pointer, and doing so saves some time or space. Use 1
2521 when there is no possible advantage to avoiding a frame pointer.
2523 In certain cases, the compiler does not know how to produce valid code
2524 without a frame pointer. The compiler recognizes those cases and
2525 automatically gives the function a frame pointer regardless of what
2526 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
2529 In a function that does not require a frame pointer, the frame pointer
2530 register can be allocated for ordinary usage, unless you mark it as a
2531 fixed register. See @code{FIXED_REGISTERS} for more information.
2533 @findex INITIAL_FRAME_POINTER_OFFSET
2534 @findex get_frame_size
2535 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
2536 A C statement to store in the variable @var{depth-var} the difference
2537 between the frame pointer and the stack pointer values immediately after
2538 the function prologue. The value would be computed from information
2539 such as the result of @code{get_frame_size ()} and the tables of
2540 registers @code{regs_ever_live} and @code{call_used_regs}.
2542 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
2543 need not be defined. Otherwise, it must be defined even if
2544 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
2545 case, you may set @var{depth-var} to anything.
2547 @findex ELIMINABLE_REGS
2548 @item ELIMINABLE_REGS
2549 If defined, this macro specifies a table of register pairs used to
2550 eliminate unneeded registers that point into the stack frame. If it is not
2551 defined, the only elimination attempted by the compiler is to replace
2552 references to the frame pointer with references to the stack pointer.
2554 The definition of this macro is a list of structure initializations, each
2555 of which specifies an original and replacement register.
2557 On some machines, the position of the argument pointer is not known until
2558 the compilation is completed. In such a case, a separate hard register
2559 must be used for the argument pointer. This register can be eliminated by
2560 replacing it with either the frame pointer or the argument pointer,
2561 depending on whether or not the frame pointer has been eliminated.
2563 In this case, you might specify:
2565 #define ELIMINABLE_REGS \
2566 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
2567 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
2568 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
2571 Note that the elimination of the argument pointer with the stack pointer is
2572 specified first since that is the preferred elimination.
2574 @findex CAN_ELIMINATE
2575 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
2576 A C expression that returns non-zero if the compiler is allowed to try
2577 to replace register number @var{from-reg} with register number
2578 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
2579 is defined, and will usually be the constant 1, since most of the cases
2580 preventing register elimination are things that the compiler already
2583 @findex INITIAL_ELIMINATION_OFFSET
2584 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
2585 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
2586 specifies the initial difference between the specified pair of
2587 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
2590 @findex LONGJMP_RESTORE_FROM_STACK
2591 @item LONGJMP_RESTORE_FROM_STACK
2592 Define this macro if the @code{longjmp} function restores registers from
2593 the stack frames, rather than from those saved specifically by
2594 @code{setjmp}. Certain quantities must not be kept in registers across
2595 a call to @code{setjmp} on such machines.
2598 @node Stack Arguments
2599 @subsection Passing Function Arguments on the Stack
2600 @cindex arguments on stack
2601 @cindex stack arguments
2603 The macros in this section control how arguments are passed
2604 on the stack. See the following section for other macros that
2605 control passing certain arguments in registers.
2608 @findex PROMOTE_PROTOTYPES
2609 @item PROMOTE_PROTOTYPES
2610 Define this macro if an argument declared in a prototype as an
2611 integral type smaller than @code{int} should actually be passed as an
2612 @code{int}. In addition to avoiding errors in certain cases of
2613 mismatch, it also makes for better code on certain machines.
2615 @findex PUSH_ROUNDING
2616 @item PUSH_ROUNDING (@var{npushed})
2617 A C expression that is the number of bytes actually pushed onto the
2618 stack when an instruction attempts to push @var{npushed} bytes.
2620 If the target machine does not have a push instruction, do not define
2621 this macro. That directs GNU CC to use an alternate strategy: to
2622 allocate the entire argument block and then store the arguments into
2625 On some machines, the definition
2628 #define PUSH_ROUNDING(BYTES) (BYTES)
2632 will suffice. But on other machines, instructions that appear
2633 to push one byte actually push two bytes in an attempt to maintain
2634 alignment. Then the definition should be
2637 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
2640 @findex ACCUMULATE_OUTGOING_ARGS
2641 @findex current_function_outgoing_args_size
2642 @item ACCUMULATE_OUTGOING_ARGS
2643 If defined, the maximum amount of space required for outgoing arguments
2644 will be computed and placed into the variable
2645 @code{current_function_outgoing_args_size}. No space will be pushed
2646 onto the stack for each call; instead, the function prologue should
2647 increase the stack frame size by this amount.
2649 Defining both @code{PUSH_ROUNDING} and @code{ACCUMULATE_OUTGOING_ARGS}
2652 @findex REG_PARM_STACK_SPACE
2653 @item REG_PARM_STACK_SPACE (@var{fndecl})
2654 Define this macro if functions should assume that stack space has been
2655 allocated for arguments even when their values are passed in
2658 The value of this macro is the size, in bytes, of the area reserved for
2659 arguments passed in registers for the function represented by @var{fndecl},
2660 which can be zero if GNU CC is calling a library function.
2662 This space can be allocated by the caller, or be a part of the
2663 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
2665 @c above is overfull. not sure what to do. --mew 5feb93 did
2666 @c something, not sure if it looks good. --mew 10feb93
2668 @findex MAYBE_REG_PARM_STACK_SPACE
2669 @findex FINAL_REG_PARM_STACK_SPACE
2670 @item MAYBE_REG_PARM_STACK_SPACE
2671 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
2672 Define these macros in addition to the one above if functions might
2673 allocate stack space for arguments even when their values are passed
2674 in registers. These should be used when the stack space allocated
2675 for arguments in registers is not a simple constant independent of the
2676 function declaration.
2678 The value of the first macro is the size, in bytes, of the area that
2679 we should initially assume would be reserved for arguments passed in registers.
2681 The value of the second macro is the actual size, in bytes, of the area
2682 that will be reserved for arguments passed in registers. This takes two
2683 arguments: an integer representing the number of bytes of fixed sized
2684 arguments on the stack, and a tree representing the number of bytes of
2685 variable sized arguments on the stack.
2687 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
2688 called for libcall functions, the current function, or for a function
2689 being called when it is known that such stack space must be allocated.
2690 In each case this value can be easily computed.
2692 When deciding whether a called function needs such stack space, and how
2693 much space to reserve, GNU CC uses these two macros instead of
2694 @code{REG_PARM_STACK_SPACE}.
2696 @findex OUTGOING_REG_PARM_STACK_SPACE
2697 @item OUTGOING_REG_PARM_STACK_SPACE
2698 Define this if it is the responsibility of the caller to allocate the area
2699 reserved for arguments passed in registers.
2701 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
2702 whether the space for these arguments counts in the value of
2703 @code{current_function_outgoing_args_size}.
2705 @findex STACK_PARMS_IN_REG_PARM_AREA
2706 @item STACK_PARMS_IN_REG_PARM_AREA
2707 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
2708 stack parameters don't skip the area specified by it.
2709 @c i changed this, makes more sens and it should have taken care of the
2710 @c overfull.. not as specific, tho. --mew 5feb93
2712 Normally, when a parameter is not passed in registers, it is placed on the
2713 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
2714 suppresses this behavior and causes the parameter to be passed on the
2715 stack in its natural location.
2717 @findex RETURN_POPS_ARGS
2718 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
2719 A C expression that should indicate the number of bytes of its own
2720 arguments that a function pops on returning, or 0 if the
2721 function pops no arguments and the caller must therefore pop them all
2722 after the function returns.
2724 @var{fundecl} is a C variable whose value is a tree node that describes
2725 the function in question. Normally it is a node of type
2726 @code{FUNCTION_DECL} that describes the declaration of the function.
2727 From this you can obtain the DECL_MACHINE_ATTRIBUTES of the function.
2729 @var{funtype} is a C variable whose value is a tree node that
2730 describes the function in question. Normally it is a node of type
2731 @code{FUNCTION_TYPE} that describes the data type of the function.
2732 From this it is possible to obtain the data types of the value and
2733 arguments (if known).
2735 When a call to a library function is being considered, @var{fundecl}
2736 will contain an identifier node for the library function. Thus, if
2737 you need to distinguish among various library functions, you can do so
2738 by their names. Note that ``library function'' in this context means
2739 a function used to perform arithmetic, whose name is known specially
2740 in the compiler and was not mentioned in the C code being compiled.
2742 @var{stack-size} is the number of bytes of arguments passed on the
2743 stack. If a variable number of bytes is passed, it is zero, and
2744 argument popping will always be the responsibility of the calling function.
2746 On the Vax, all functions always pop their arguments, so the definition
2747 of this macro is @var{stack-size}. On the 68000, using the standard
2748 calling convention, no functions pop their arguments, so the value of
2749 the macro is always 0 in this case. But an alternative calling
2750 convention is available in which functions that take a fixed number of
2751 arguments pop them but other functions (such as @code{printf}) pop
2752 nothing (the caller pops all). When this convention is in use,
2753 @var{funtype} is examined to determine whether a function takes a fixed
2754 number of arguments.
2757 @node Register Arguments
2758 @subsection Passing Arguments in Registers
2759 @cindex arguments in registers
2760 @cindex registers arguments
2762 This section describes the macros which let you control how various
2763 types of arguments are passed in registers or how they are arranged in
2767 @findex FUNCTION_ARG
2768 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2769 A C expression that controls whether a function argument is passed
2770 in a register, and which register.
2772 The arguments are @var{cum}, which summarizes all the previous
2773 arguments; @var{mode}, the machine mode of the argument; @var{type},
2774 the data type of the argument as a tree node or 0 if that is not known
2775 (which happens for C support library functions); and @var{named},
2776 which is 1 for an ordinary argument and 0 for nameless arguments that
2777 correspond to @samp{@dots{}} in the called function's prototype.
2779 The value of the expression is usually either a @code{reg} RTX for the
2780 hard register in which to pass the argument, or zero to pass the
2781 argument on the stack.
2783 For machines like the Vax and 68000, where normally all arguments are
2784 pushed, zero suffices as a definition.
2786 The value of the expression can also be a @code{parallel} RTX. This is
2787 used when an argument is passed in multiple locations. The mode of the
2788 of the @code{parallel} should be the mode of the entire argument. The
2789 @code{parallel} holds any number of @code{expr_list} pairs; each one
2790 describes where part of the argument is passed. In each
2791 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
2792 register in which to pass this part of the argument, and the mode of the
2793 register RTX indicates how large this part of the argument is. The
2794 second operand of the @code{expr_list} is a @code{const_int} which gives
2795 the offset in bytes into the entire argument of where this part starts.
2796 As a special exception the first @code{expr_list} in the @code{parallel}
2797 RTX may have a first operand of zero. This indicates that the bytes
2798 starting from the second operand of that @code{expr_list} are stored on
2799 the stack and not held in a register.
2801 @cindex @file{stdarg.h} and register arguments
2802 The usual way to make the ANSI library @file{stdarg.h} work on a machine
2803 where some arguments are usually passed in registers, is to cause
2804 nameless arguments to be passed on the stack instead. This is done
2805 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
2807 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
2808 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
2809 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
2810 in the definition of this macro to determine if this argument is of a
2811 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
2812 is not defined and @code{FUNCTION_ARG} returns non-zero for such an
2813 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
2814 defined, the argument will be computed in the stack and then loaded into
2817 @findex MUST_PASS_IN_STACK
2818 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
2819 Define as a C expression that evaluates to nonzero if we do not know how
2820 to pass TYPE solely in registers. The file @file{expr.h} defines a
2821 definition that is usually appropriate, refer to @file{expr.h} for additional
2824 @findex FUNCTION_INCOMING_ARG
2825 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2826 Define this macro if the target machine has ``register windows'', so
2827 that the register in which a function sees an arguments is not
2828 necessarily the same as the one in which the caller passed the
2831 For such machines, @code{FUNCTION_ARG} computes the register in which
2832 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
2833 be defined in a similar fashion to tell the function being called
2834 where the arguments will arrive.
2836 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
2837 serves both purposes.@refill
2839 @findex FUNCTION_ARG_PARTIAL_NREGS
2840 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
2841 A C expression for the number of words, at the beginning of an
2842 argument, must be put in registers. The value must be zero for
2843 arguments that are passed entirely in registers or that are entirely
2844 pushed on the stack.
2846 On some machines, certain arguments must be passed partially in
2847 registers and partially in memory. On these machines, typically the
2848 first @var{n} words of arguments are passed in registers, and the rest
2849 on the stack. If a multi-word argument (a @code{double} or a
2850 structure) crosses that boundary, its first few words must be passed
2851 in registers and the rest must be pushed. This macro tells the
2852 compiler when this occurs, and how many of the words should go in
2855 @code{FUNCTION_ARG} for these arguments should return the first
2856 register to be used by the caller for this argument; likewise
2857 @code{FUNCTION_INCOMING_ARG}, for the called function.
2859 @findex FUNCTION_ARG_PASS_BY_REFERENCE
2860 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2861 A C expression that indicates when an argument must be passed by reference.
2862 If nonzero for an argument, a copy of that argument is made in memory and a
2863 pointer to the argument is passed instead of the argument itself.
2864 The pointer is passed in whatever way is appropriate for passing a pointer
2867 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
2868 definition of this macro might be
2870 #define FUNCTION_ARG_PASS_BY_REFERENCE\
2871 (CUM, MODE, TYPE, NAMED) \
2872 MUST_PASS_IN_STACK (MODE, TYPE)
2874 @c this is *still* too long. --mew 5feb93
2876 @findex FUNCTION_ARG_CALLEE_COPIES
2877 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
2878 If defined, a C expression that indicates when it is the called function's
2879 responsibility to make a copy of arguments passed by invisible reference.
2880 Normally, the caller makes a copy and passes the address of the copy to the
2881 routine being called. When FUNCTION_ARG_CALLEE_COPIES is defined and is
2882 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
2883 ``live'' value. The called function must not modify this value. If it can be
2884 determined that the value won't be modified, it need not make a copy;
2885 otherwise a copy must be made.
2887 @findex CUMULATIVE_ARGS
2888 @item CUMULATIVE_ARGS
2889 A C type for declaring a variable that is used as the first argument of
2890 @code{FUNCTION_ARG} and other related values. For some target machines,
2891 the type @code{int} suffices and can hold the number of bytes of
2894 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
2895 arguments that have been passed on the stack. The compiler has other
2896 variables to keep track of that. For target machines on which all
2897 arguments are passed on the stack, there is no need to store anything in
2898 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
2899 should not be empty, so use @code{int}.
2901 @findex INIT_CUMULATIVE_ARGS
2902 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
2903 A C statement (sans semicolon) for initializing the variable @var{cum}
2904 for the state at the beginning of the argument list. The variable has
2905 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
2906 for the data type of the function which will receive the args, or 0
2907 if the args are to a compiler support library function. The value of
2908 @var{indirect} is nonzero when processing an indirect call, for example
2909 a call through a function pointer. The value of @var{indirect} is zero
2910 for a call to an explicitly named function, a library function call, or when
2911 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
2914 When processing a call to a compiler support library function,
2915 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
2916 contains the name of the function, as a string. @var{libname} is 0 when
2917 an ordinary C function call is being processed. Thus, each time this
2918 macro is called, either @var{libname} or @var{fntype} is nonzero, but
2919 never both of them at once.
2921 @findex INIT_CUMULATIVE_INCOMING_ARGS
2922 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
2923 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
2924 finding the arguments for the function being compiled. If this macro is
2925 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
2927 The value passed for @var{libname} is always 0, since library routines
2928 with special calling conventions are never compiled with GNU CC. The
2929 argument @var{libname} exists for symmetry with
2930 @code{INIT_CUMULATIVE_ARGS}.
2931 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
2932 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
2934 @findex FUNCTION_ARG_ADVANCE
2935 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2936 A C statement (sans semicolon) to update the summarizer variable
2937 @var{cum} to advance past an argument in the argument list. The
2938 values @var{mode}, @var{type} and @var{named} describe that argument.
2939 Once this is done, the variable @var{cum} is suitable for analyzing
2940 the @emph{following} argument with @code{FUNCTION_ARG}, etc.@refill
2942 This macro need not do anything if the argument in question was passed
2943 on the stack. The compiler knows how to track the amount of stack space
2944 used for arguments without any special help.
2946 @findex FUNCTION_ARG_PADDING
2947 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
2948 If defined, a C expression which determines whether, and in which direction,
2949 to pad out an argument with extra space. The value should be of type
2950 @code{enum direction}: either @code{upward} to pad above the argument,
2951 @code{downward} to pad below, or @code{none} to inhibit padding.
2953 The @emph{amount} of padding is always just enough to reach the next
2954 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
2957 This macro has a default definition which is right for most systems.
2958 For little-endian machines, the default is to pad upward. For
2959 big-endian machines, the default is to pad downward for an argument of
2960 constant size shorter than an @code{int}, and upward otherwise.
2962 @findex FUNCTION_ARG_BOUNDARY
2963 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
2964 If defined, a C expression that gives the alignment boundary, in bits,
2965 of an argument with the specified mode and type. If it is not defined,
2966 @code{PARM_BOUNDARY} is used for all arguments.
2968 @findex FUNCTION_ARG_REGNO_P
2969 @item FUNCTION_ARG_REGNO_P (@var{regno})
2970 A C expression that is nonzero if @var{regno} is the number of a hard
2971 register in which function arguments are sometimes passed. This does
2972 @emph{not} include implicit arguments such as the static chain and
2973 the structure-value address. On many machines, no registers can be
2974 used for this purpose since all function arguments are pushed on the
2977 @findex LOAD_ARGS_REVERSED
2978 @item LOAD_ARGS_REVERSED
2979 If defined, the order in which arguments are loaded into their
2980 respective argument registers is reversed so that the last
2981 argument is loaded first. This macro only effects arguments
2982 passed in registers.
2987 @subsection How Scalar Function Values Are Returned
2988 @cindex return values in registers
2989 @cindex values, returned by functions
2990 @cindex scalars, returned as values
2992 This section discusses the macros that control returning scalars as
2993 values---values that can fit in registers.
2996 @findex TRADITIONAL_RETURN_FLOAT
2997 @item TRADITIONAL_RETURN_FLOAT
2998 Define this macro if @samp{-traditional} should not cause functions
2999 declared to return @code{float} to convert the value to @code{double}.
3001 @findex FUNCTION_VALUE
3002 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3003 A C expression to create an RTX representing the place where a
3004 function returns a value of data type @var{valtype}. @var{valtype} is
3005 a tree node representing a data type. Write @code{TYPE_MODE
3006 (@var{valtype})} to get the machine mode used to represent that type.
3007 On many machines, only the mode is relevant. (Actually, on most
3008 machines, scalar values are returned in the same place regardless of
3011 The value of the expression is usually a @code{reg} RTX for the hard
3012 register where the return value is stored. The value can also be a
3013 @code{parallel} RTX, if the return value is in multiple places. See
3014 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3016 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3017 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3020 If the precise function being called is known, @var{func} is a tree
3021 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3022 pointer. This makes it possible to use a different value-returning
3023 convention for specific functions when all their calls are
3026 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3027 types, because these are returned in another way. See
3028 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3030 @findex FUNCTION_OUTGOING_VALUE
3031 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3032 Define this macro if the target machine has ``register windows''
3033 so that the register in which a function returns its value is not
3034 the same as the one in which the caller sees the value.
3036 For such machines, @code{FUNCTION_VALUE} computes the register in which
3037 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3038 defined in a similar fashion to tell the function where to put the
3041 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3042 @code{FUNCTION_VALUE} serves both purposes.@refill
3044 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3045 aggregate data types, because these are returned in another way. See
3046 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3048 @findex LIBCALL_VALUE
3049 @item LIBCALL_VALUE (@var{mode})
3050 A C expression to create an RTX representing the place where a library
3051 function returns a value of mode @var{mode}. If the precise function
3052 being called is known, @var{func} is a tree node
3053 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3054 pointer. This makes it possible to use a different value-returning
3055 convention for specific functions when all their calls are
3058 Note that ``library function'' in this context means a compiler
3059 support routine, used to perform arithmetic, whose name is known
3060 specially by the compiler and was not mentioned in the C code being
3063 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3064 data types, because none of the library functions returns such types.
3066 @findex FUNCTION_VALUE_REGNO_P
3067 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3068 A C expression that is nonzero if @var{regno} is the number of a hard
3069 register in which the values of called function may come back.
3071 A register whose use for returning values is limited to serving as the
3072 second of a pair (for a value of type @code{double}, say) need not be
3073 recognized by this macro. So for most machines, this definition
3077 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3080 If the machine has register windows, so that the caller and the called
3081 function use different registers for the return value, this macro
3082 should recognize only the caller's register numbers.
3084 @findex APPLY_RESULT_SIZE
3085 @item APPLY_RESULT_SIZE
3086 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3087 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3088 saving and restoring an arbitrary return value.
3091 @node Aggregate Return
3092 @subsection How Large Values Are Returned
3093 @cindex aggregates as return values
3094 @cindex large return values
3095 @cindex returning aggregate values
3096 @cindex structure value address
3098 When a function value's mode is @code{BLKmode} (and in some other
3099 cases), the value is not returned according to @code{FUNCTION_VALUE}
3100 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3101 block of memory in which the value should be stored. This address
3102 is called the @dfn{structure value address}.
3104 This section describes how to control returning structure values in
3108 @findex RETURN_IN_MEMORY
3109 @item RETURN_IN_MEMORY (@var{type})
3110 A C expression which can inhibit the returning of certain function
3111 values in registers, based on the type of value. A nonzero value says
3112 to return the function value in memory, just as large structures are
3113 always returned. Here @var{type} will be a C expression of type
3114 @code{tree}, representing the data type of the value.
3116 Note that values of mode @code{BLKmode} must be explicitly handled
3117 by this macro. Also, the option @samp{-fpcc-struct-return}
3118 takes effect regardless of this macro. On most systems, it is
3119 possible to leave the macro undefined; this causes a default
3120 definition to be used, whose value is the constant 1 for @code{BLKmode}
3121 values, and 0 otherwise.
3123 Do not use this macro to indicate that structures and unions should always
3124 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3127 @findex DEFAULT_PCC_STRUCT_RETURN
3128 @item DEFAULT_PCC_STRUCT_RETURN
3129 Define this macro to be 1 if all structure and union return values must be
3130 in memory. Since this results in slower code, this should be defined
3131 only if needed for compatibility with other compilers or with an ABI.
3132 If you define this macro to be 0, then the conventions used for structure
3133 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3135 If not defined, this defaults to the value 1.
3137 @findex STRUCT_VALUE_REGNUM
3138 @item STRUCT_VALUE_REGNUM
3139 If the structure value address is passed in a register, then
3140 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3142 @findex STRUCT_VALUE
3144 If the structure value address is not passed in a register, define
3145 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3146 where the address is passed. If it returns 0, the address is passed as
3147 an ``invisible'' first argument.
3149 @findex STRUCT_VALUE_INCOMING_REGNUM
3150 @item STRUCT_VALUE_INCOMING_REGNUM
3151 On some architectures the place where the structure value address
3152 is found by the called function is not the same place that the
3153 caller put it. This can be due to register windows, or it could
3154 be because the function prologue moves it to a different place.
3156 If the incoming location of the structure value address is in a
3157 register, define this macro as the register number.
3159 @findex STRUCT_VALUE_INCOMING
3160 @item STRUCT_VALUE_INCOMING
3161 If the incoming location is not a register, then you should define
3162 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3163 called function should find the value. If it should find the value on
3164 the stack, define this to create a @code{mem} which refers to the frame
3165 pointer. A definition of 0 means that the address is passed as an
3166 ``invisible'' first argument.
3168 @findex PCC_STATIC_STRUCT_RETURN
3169 @item PCC_STATIC_STRUCT_RETURN
3170 Define this macro if the usual system convention on the target machine
3171 for returning structures and unions is for the called function to return
3172 the address of a static variable containing the value.
3174 Do not define this if the usual system convention is for the caller to
3175 pass an address to the subroutine.
3177 This macro has effect in @samp{-fpcc-struct-return} mode, but it does
3178 nothing when you use @samp{-freg-struct-return} mode.
3182 @subsection Caller-Saves Register Allocation
3184 If you enable it, GNU CC can save registers around function calls. This
3185 makes it possible to use call-clobbered registers to hold variables that
3186 must live across calls.
3189 @findex DEFAULT_CALLER_SAVES
3190 @item DEFAULT_CALLER_SAVES
3191 Define this macro if function calls on the target machine do not preserve
3192 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3193 for all registers. When defined, this macro enables @samp{-fcaller-saves}
3194 by default for all optimization levels. It has no effect for optimization
3195 levels 2 and higher, where @samp{-fcaller-saves} is the default.
3197 @findex CALLER_SAVE_PROFITABLE
3198 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3199 A C expression to determine whether it is worthwhile to consider placing
3200 a pseudo-register in a call-clobbered hard register and saving and
3201 restoring it around each function call. The expression should be 1 when
3202 this is worth doing, and 0 otherwise.
3204 If you don't define this macro, a default is used which is good on most
3205 machines: @code{4 * @var{calls} < @var{refs}}.
3207 @findex HARD_REGNO_CALLER_SAVE_MODE
3208 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3209 A C expression specifying which mode is required for saving @var{nregs}
3210 of a pseudo-register in call-clobbered hard register @var{regno}. If
3211 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3212 returned. For most machines this macro need not be defined since GCC
3213 will select the smallest suitable mode.
3216 @node Function Entry
3217 @subsection Function Entry and Exit
3218 @cindex function entry and exit
3222 This section describes the macros that output function entry
3223 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3226 @findex FUNCTION_PROLOGUE
3227 @item FUNCTION_PROLOGUE (@var{file}, @var{size})
3228 A C compound statement that outputs the assembler code for entry to a
3229 function. The prologue is responsible for setting up the stack frame,
3230 initializing the frame pointer register, saving registers that must be
3231 saved, and allocating @var{size} additional bytes of storage for the
3232 local variables. @var{size} is an integer. @var{file} is a stdio
3233 stream to which the assembler code should be output.
3235 The label for the beginning of the function need not be output by this
3236 macro. That has already been done when the macro is run.
3238 @findex regs_ever_live
3239 To determine which registers to save, the macro can refer to the array
3240 @code{regs_ever_live}: element @var{r} is nonzero if hard register
3241 @var{r} is used anywhere within the function. This implies the function
3242 prologue should save register @var{r}, provided it is not one of the
3243 call-used registers. (@code{FUNCTION_EPILOGUE} must likewise use
3244 @code{regs_ever_live}.)
3246 On machines that have ``register windows'', the function entry code does
3247 not save on the stack the registers that are in the windows, even if
3248 they are supposed to be preserved by function calls; instead it takes
3249 appropriate steps to ``push'' the register stack, if any non-call-used
3250 registers are used in the function.
3252 @findex frame_pointer_needed
3253 On machines where functions may or may not have frame-pointers, the
3254 function entry code must vary accordingly; it must set up the frame
3255 pointer if one is wanted, and not otherwise. To determine whether a
3256 frame pointer is in wanted, the macro can refer to the variable
3257 @code{frame_pointer_needed}. The variable's value will be 1 at run
3258 time in a function that needs a frame pointer. @xref{Elimination}.
3260 The function entry code is responsible for allocating any stack space
3261 required for the function. This stack space consists of the regions
3262 listed below. In most cases, these regions are allocated in the
3263 order listed, with the last listed region closest to the top of the
3264 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
3265 the highest address if it is not defined). You can use a different order
3266 for a machine if doing so is more convenient or required for
3267 compatibility reasons. Except in cases where required by standard
3268 or by a debugger, there is no reason why the stack layout used by GCC
3269 need agree with that used by other compilers for a machine.
3273 @findex current_function_pretend_args_size
3274 A region of @code{current_function_pretend_args_size} bytes of
3275 uninitialized space just underneath the first argument arriving on the
3276 stack. (This may not be at the very start of the allocated stack region
3277 if the calling sequence has pushed anything else since pushing the stack
3278 arguments. But usually, on such machines, nothing else has been pushed
3279 yet, because the function prologue itself does all the pushing.) This
3280 region is used on machines where an argument may be passed partly in
3281 registers and partly in memory, and, in some cases to support the
3282 features in @file{varargs.h} and @file{stdargs.h}.
3285 An area of memory used to save certain registers used by the function.
3286 The size of this area, which may also include space for such things as
3287 the return address and pointers to previous stack frames, is
3288 machine-specific and usually depends on which registers have been used
3289 in the function. Machines with register windows often do not require
3293 A region of at least @var{size} bytes, possibly rounded up to an allocation
3294 boundary, to contain the local variables of the function. On some machines,
3295 this region and the save area may occur in the opposite order, with the
3296 save area closer to the top of the stack.
3299 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3300 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3301 @code{current_function_outgoing_args_size} bytes to be used for outgoing
3302 argument lists of the function. @xref{Stack Arguments}.
3305 Normally, it is necessary for the macros @code{FUNCTION_PROLOGUE} and
3306 @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C
3307 variable @code{current_function_is_leaf} is nonzero for such a function.
3309 @findex EXIT_IGNORE_STACK
3310 @item EXIT_IGNORE_STACK
3311 Define this macro as a C expression that is nonzero if the return
3312 instruction or the function epilogue ignores the value of the stack
3313 pointer; in other words, if it is safe to delete an instruction to
3314 adjust the stack pointer before a return from the function.
3316 Note that this macro's value is relevant only for functions for which
3317 frame pointers are maintained. It is never safe to delete a final
3318 stack adjustment in a function that has no frame pointer, and the
3319 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3321 @findex EPILOGUE_USES
3322 @item EPILOGUE_USES (@var{regno})
3323 Define this macro as a C expression that is nonzero for registers are
3324 used by the epilogue or the @samp{return} pattern. The stack and frame
3325 pointer registers are already be assumed to be used as needed.
3327 @findex FUNCTION_EPILOGUE
3328 @item FUNCTION_EPILOGUE (@var{file}, @var{size})
3329 A C compound statement that outputs the assembler code for exit from a
3330 function. The epilogue is responsible for restoring the saved
3331 registers and stack pointer to their values when the function was
3332 called, and returning control to the caller. This macro takes the
3333 same arguments as the macro @code{FUNCTION_PROLOGUE}, and the
3334 registers to restore are determined from @code{regs_ever_live} and
3335 @code{CALL_USED_REGISTERS} in the same way.
3337 On some machines, there is a single instruction that does all the work
3338 of returning from the function. On these machines, give that
3339 instruction the name @samp{return} and do not define the macro
3340 @code{FUNCTION_EPILOGUE} at all.
3342 Do not define a pattern named @samp{return} if you want the
3343 @code{FUNCTION_EPILOGUE} to be used. If you want the target switches
3344 to control whether return instructions or epilogues are used, define a
3345 @samp{return} pattern with a validity condition that tests the target
3346 switches appropriately. If the @samp{return} pattern's validity
3347 condition is false, epilogues will be used.
3349 On machines where functions may or may not have frame-pointers, the
3350 function exit code must vary accordingly. Sometimes the code for these
3351 two cases is completely different. To determine whether a frame pointer
3352 is wanted, the macro can refer to the variable
3353 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
3354 a function that needs a frame pointer.
3356 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
3357 treat leaf functions specially. The C variable @code{current_function_is_leaf}
3358 is nonzero for such a function. @xref{Leaf Functions}.
3360 On some machines, some functions pop their arguments on exit while
3361 others leave that for the caller to do. For example, the 68020 when
3362 given @samp{-mrtd} pops arguments in functions that take a fixed
3363 number of arguments.
3365 @findex current_function_pops_args
3366 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
3367 functions pop their own arguments. @code{FUNCTION_EPILOGUE} needs to
3368 know what was decided. The variable that is called
3369 @code{current_function_pops_args} is the number of bytes of its
3370 arguments that a function should pop. @xref{Scalar Return}.
3371 @c what is the "its arguments" in the above sentence referring to, pray
3372 @c tell? --mew 5feb93
3374 @findex DELAY_SLOTS_FOR_EPILOGUE
3375 @item DELAY_SLOTS_FOR_EPILOGUE
3376 Define this macro if the function epilogue contains delay slots to which
3377 instructions from the rest of the function can be ``moved''. The
3378 definition should be a C expression whose value is an integer
3379 representing the number of delay slots there.
3381 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
3382 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
3383 A C expression that returns 1 if @var{insn} can be placed in delay
3384 slot number @var{n} of the epilogue.
3386 The argument @var{n} is an integer which identifies the delay slot now
3387 being considered (since different slots may have different rules of
3388 eligibility). It is never negative and is always less than the number
3389 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
3390 If you reject a particular insn for a given delay slot, in principle, it
3391 may be reconsidered for a subsequent delay slot. Also, other insns may
3392 (at least in principle) be considered for the so far unfilled delay
3395 @findex current_function_epilogue_delay_list
3396 @findex final_scan_insn
3397 The insns accepted to fill the epilogue delay slots are put in an RTL
3398 list made with @code{insn_list} objects, stored in the variable
3399 @code{current_function_epilogue_delay_list}. The insn for the first
3400 delay slot comes first in the list. Your definition of the macro
3401 @code{FUNCTION_EPILOGUE} should fill the delay slots by outputting the
3402 insns in this list, usually by calling @code{final_scan_insn}.
3404 You need not define this macro if you did not define
3405 @code{DELAY_SLOTS_FOR_EPILOGUE}.
3407 @findex ASM_OUTPUT_MI_THUNK
3408 @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function})
3409 A C compound statement that outputs the assembler code for a thunk
3410 function, used to implement C++ virtual function calls with multiple
3411 inheritance. The thunk acts as a wrapper around a virtual function,
3412 adjusting the implicit object parameter before handing control off to
3415 First, emit code to add the integer @var{delta} to the location that
3416 contains the incoming first argument. Assume that this argument
3417 contains a pointer, and is the one used to pass the @code{this} pointer
3418 in C++. This is the incoming argument @emph{before} the function prologue,
3419 e.g. @samp{%o0} on a sparc. The addition must preserve the values of
3420 all other incoming arguments.
3422 After the addition, emit code to jump to @var{function}, which is a
3423 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
3424 not touch the return address. Hence returning from @var{FUNCTION} will
3425 return to whoever called the current @samp{thunk}.
3427 The effect must be as if @var{function} had been called directly with
3428 the adjusted first argument. This macro is responsible for emitting all
3429 of the code for a thunk function; @code{FUNCTION_PROLOGUE} and
3430 @code{FUNCTION_EPILOGUE} are not invoked.
3432 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
3433 have already been extracted from it.) It might possibly be useful on
3434 some targets, but probably not.
3436 If you do not define this macro, the target-independent code in the C++
3437 frontend will generate a less efficient heavyweight thunk that calls
3438 @var{function} instead of jumping to it. The generic approach does
3439 not support varargs.
3443 @subsection Generating Code for Profiling
3444 @cindex profiling, code generation
3446 These macros will help you generate code for profiling.
3449 @findex FUNCTION_PROFILER
3450 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
3451 A C statement or compound statement to output to @var{file} some
3452 assembler code to call the profiling subroutine @code{mcount}.
3453 Before calling, the assembler code must load the address of a
3454 counter variable into a register where @code{mcount} expects to
3455 find the address. The name of this variable is @samp{LP} followed
3456 by the number @var{labelno}, so you would generate the name using
3457 @samp{LP%d} in a @code{fprintf}.
3460 The details of how the address should be passed to @code{mcount} are
3461 determined by your operating system environment, not by GNU CC. To
3462 figure them out, compile a small program for profiling using the
3463 system's installed C compiler and look at the assembler code that
3466 @findex PROFILE_BEFORE_PROLOGUE
3467 @item PROFILE_BEFORE_PROLOGUE
3468 Define this macro if the code for function profiling should come before
3469 the function prologue. Normally, the profiling code comes after.
3471 @findex FUNCTION_BLOCK_PROFILER
3472 @vindex profile_block_flag
3473 @item FUNCTION_BLOCK_PROFILER (@var{file}, @var{labelno})
3474 A C statement or compound statement to output to @var{file} some
3475 assembler code to initialize basic-block profiling for the current
3476 object module. The global compile flag @code{profile_block_flag}
3477 distinguishes two profile modes.
3480 @findex __bb_init_func
3481 @item profile_block_flag != 2
3482 Output code to call the subroutine @code{__bb_init_func} once per
3483 object module, passing it as its sole argument the address of a block
3484 allocated in the object module.
3486 The name of the block is a local symbol made with this statement:
3489 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3492 Of course, since you are writing the definition of
3493 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3494 can take a short cut in the definition of this macro and use the name
3495 that you know will result.
3497 The first word of this block is a flag which will be nonzero if the
3498 object module has already been initialized. So test this word first,
3499 and do not call @code{__bb_init_func} if the flag is
3500 nonzero. BLOCK_OR_LABEL contains a unique number which may be used to
3501 generate a label as a branch destination when @code{__bb_init_func}
3504 Described in assembler language, the code to be output looks like:
3514 @findex __bb_init_trace_func
3515 @item profile_block_flag == 2
3516 Output code to call the subroutine @code{__bb_init_trace_func}
3517 and pass two parameters to it. The first parameter is the same as
3518 for @code{__bb_init_func}. The second parameter is the number of the
3519 first basic block of the function as given by BLOCK_OR_LABEL. Note
3520 that @code{__bb_init_trace_func} has to be called, even if the object
3521 module has been initialized already.
3523 Described in assembler language, the code to be output looks like:
3526 parameter2 <- BLOCK_OR_LABEL
3527 call __bb_init_trace_func
3531 @findex BLOCK_PROFILER
3532 @vindex profile_block_flag
3533 @item BLOCK_PROFILER (@var{file}, @var{blockno})
3534 A C statement or compound statement to output to @var{file} some
3535 assembler code to increment the count associated with the basic
3536 block number @var{blockno}. The global compile flag
3537 @code{profile_block_flag} distinguishes two profile modes.
3540 @item profile_block_flag != 2
3541 Output code to increment the counter directly. Basic blocks are
3542 numbered separately from zero within each compilation. The count
3543 associated with block number @var{blockno} is at index
3544 @var{blockno} in a vector of words; the name of this array is a local
3545 symbol made with this statement:
3548 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 2);
3551 @c This paragraph is the same as one a few paragraphs up.
3552 @c That is not an error.
3553 Of course, since you are writing the definition of
3554 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3555 can take a short cut in the definition of this macro and use the name
3556 that you know will result.
3558 Described in assembler language, the code to be output looks like:
3561 inc (LPBX2+4*BLOCKNO)
3565 @findex __bb_trace_func
3566 @item profile_block_flag == 2
3567 Output code to initialize the global structure @code{__bb} and
3568 call the function @code{__bb_trace_func}, which will increment the
3571 @code{__bb} consists of two words. In the first word, the current
3572 basic block number, as given by BLOCKNO, has to be stored. In
3573 the second word, the address of a block allocated in the object
3574 module has to be stored. The address is given by the label created
3575 with this statement:
3578 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3581 Described in assembler language, the code to be output looks like:
3583 move BLOCKNO -> (__bb)
3584 move LPBX0 -> (__bb+4)
3585 call __bb_trace_func
3589 @findex FUNCTION_BLOCK_PROFILER_EXIT
3590 @findex __bb_trace_ret
3591 @vindex profile_block_flag
3592 @item FUNCTION_BLOCK_PROFILER_EXIT (@var{file})
3593 A C statement or compound statement to output to @var{file}
3594 assembler code to call function @code{__bb_trace_ret}. The
3595 assembler code should only be output
3596 if the global compile flag @code{profile_block_flag} == 2. This
3597 macro has to be used at every place where code for returning from
3598 a function is generated (e.g. @code{FUNCTION_EPILOGUE}). Although
3599 you have to write the definition of @code{FUNCTION_EPILOGUE}
3600 as well, you have to define this macro to tell the compiler, that
3601 the proper call to @code{__bb_trace_ret} is produced.
3603 @findex MACHINE_STATE_SAVE
3604 @findex __bb_init_trace_func
3605 @findex __bb_trace_func
3606 @findex __bb_trace_ret
3607 @item MACHINE_STATE_SAVE (@var{id})
3608 A C statement or compound statement to save all registers, which may
3609 be clobbered by a function call, including condition codes. The
3610 @code{asm} statement will be mostly likely needed to handle this
3611 task. Local labels in the assembler code can be concatenated with the
3612 string @var{id}, to obtain a unique lable name.
3614 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3615 @code{FUNCTION_EPILOGUE} must be saved in the macros
3616 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3617 @code{BLOCK_PROFILER} prior calling @code{__bb_init_trace_func},
3618 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3620 @findex MACHINE_STATE_RESTORE
3621 @findex __bb_init_trace_func
3622 @findex __bb_trace_func
3623 @findex __bb_trace_ret
3624 @item MACHINE_STATE_RESTORE (@var{id})
3625 A C statement or compound statement to restore all registers, including
3626 condition codes, saved by @code{MACHINE_STATE_SAVE}.
3628 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3629 @code{FUNCTION_EPILOGUE} must be restored in the macros
3630 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3631 @code{BLOCK_PROFILER} after calling @code{__bb_init_trace_func},
3632 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3634 @findex BLOCK_PROFILER_CODE
3635 @item BLOCK_PROFILER_CODE
3636 A C function or functions which are needed in the library to
3637 support block profiling.
3641 @section Implementing the Varargs Macros
3642 @cindex varargs implementation
3644 GNU CC comes with an implementation of @file{varargs.h} and
3645 @file{stdarg.h} that work without change on machines that pass arguments
3646 on the stack. Other machines require their own implementations of
3647 varargs, and the two machine independent header files must have
3648 conditionals to include it.
3650 ANSI @file{stdarg.h} differs from traditional @file{varargs.h} mainly in
3651 the calling convention for @code{va_start}. The traditional
3652 implementation takes just one argument, which is the variable in which
3653 to store the argument pointer. The ANSI implementation of
3654 @code{va_start} takes an additional second argument. The user is
3655 supposed to write the last named argument of the function here.
3657 However, @code{va_start} should not use this argument. The way to find
3658 the end of the named arguments is with the built-in functions described
3662 @findex __builtin_saveregs
3663 @item __builtin_saveregs ()
3664 Use this built-in function to save the argument registers in memory so
3665 that the varargs mechanism can access them. Both ANSI and traditional
3666 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3667 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
3669 On some machines, @code{__builtin_saveregs} is open-coded under the
3670 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
3671 it calls a routine written in assembler language, found in
3674 Code generated for the call to @code{__builtin_saveregs} appears at the
3675 beginning of the function, as opposed to where the call to
3676 @code{__builtin_saveregs} is written, regardless of what the code is.
3677 This is because the registers must be saved before the function starts
3678 to use them for its own purposes.
3679 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3682 @findex __builtin_args_info
3683 @item __builtin_args_info (@var{category})
3684 Use this built-in function to find the first anonymous arguments in
3687 In general, a machine may have several categories of registers used for
3688 arguments, each for a particular category of data types. (For example,
3689 on some machines, floating-point registers are used for floating-point
3690 arguments while other arguments are passed in the general registers.)
3691 To make non-varargs functions use the proper calling convention, you
3692 have defined the @code{CUMULATIVE_ARGS} data type to record how many
3693 registers in each category have been used so far
3695 @code{__builtin_args_info} accesses the same data structure of type
3696 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
3697 with it, with @var{category} specifying which word to access. Thus, the
3698 value indicates the first unused register in a given category.
3700 Normally, you would use @code{__builtin_args_info} in the implementation
3701 of @code{va_start}, accessing each category just once and storing the
3702 value in the @code{va_list} object. This is because @code{va_list} will
3703 have to update the values, and there is no way to alter the
3704 values accessed by @code{__builtin_args_info}.
3706 @findex __builtin_next_arg
3707 @item __builtin_next_arg (@var{lastarg})
3708 This is the equivalent of @code{__builtin_args_info}, for stack
3709 arguments. It returns the address of the first anonymous stack
3710 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
3711 returns the address of the location above the first anonymous stack
3712 argument. Use it in @code{va_start} to initialize the pointer for
3713 fetching arguments from the stack. Also use it in @code{va_start} to
3714 verify that the second parameter @var{lastarg} is the last named argument
3715 of the current function.
3717 @findex __builtin_classify_type
3718 @item __builtin_classify_type (@var{object})
3719 Since each machine has its own conventions for which data types are
3720 passed in which kind of register, your implementation of @code{va_arg}
3721 has to embody these conventions. The easiest way to categorize the
3722 specified data type is to use @code{__builtin_classify_type} together
3723 with @code{sizeof} and @code{__alignof__}.
3725 @code{__builtin_classify_type} ignores the value of @var{object},
3726 considering only its data type. It returns an integer describing what
3727 kind of type that is---integer, floating, pointer, structure, and so on.
3729 The file @file{typeclass.h} defines an enumeration that you can use to
3730 interpret the values of @code{__builtin_classify_type}.
3733 These machine description macros help implement varargs:
3736 @findex EXPAND_BUILTIN_SAVEREGS
3737 @item EXPAND_BUILTIN_SAVEREGS (@var{args})
3738 If defined, is a C expression that produces the machine-specific code
3739 for a call to @code{__builtin_saveregs}. This code will be moved to the
3740 very beginning of the function, before any parameter access are made.
3741 The return value of this function should be an RTX that contains the
3742 value to use as the return of @code{__builtin_saveregs}.
3744 The argument @var{args} is a @code{tree_list} containing the arguments
3745 that were passed to @code{__builtin_saveregs}.
3747 If this macro is not defined, the compiler will output an ordinary
3748 call to the library function @samp{__builtin_saveregs}.
3750 @findex SETUP_INCOMING_VARARGS
3751 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
3752 This macro offers an alternative to using @code{__builtin_saveregs} and
3753 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
3754 anonymous register arguments into the stack so that all the arguments
3755 appear to have been passed consecutively on the stack. Once this is
3756 done, you can use the standard implementation of varargs that works for
3757 machines that pass all their arguments on the stack.
3759 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
3760 structure, containing the values that obtain after processing of the
3761 named arguments. The arguments @var{mode} and @var{type} describe the
3762 last named argument---its machine mode and its data type as a tree node.
3764 The macro implementation should do two things: first, push onto the
3765 stack all the argument registers @emph{not} used for the named
3766 arguments, and second, store the size of the data thus pushed into the
3767 @code{int}-valued variable whose name is supplied as the argument
3768 @var{pretend_args_size}. The value that you store here will serve as
3769 additional offset for setting up the stack frame.
3771 Because you must generate code to push the anonymous arguments at
3772 compile time without knowing their data types,
3773 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
3774 a single category of argument register and use it uniformly for all data
3777 If the argument @var{second_time} is nonzero, it means that the
3778 arguments of the function are being analyzed for the second time. This
3779 happens for an inline function, which is not actually compiled until the
3780 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
3781 not generate any instructions in this case.
3783 @findex STRICT_ARGUMENT_NAMING
3784 @item STRICT_ARGUMENT_NAMING
3785 Define this macro to be a nonzero value if the location where a function
3786 argument is passed depends on whether or not it is a named argument.
3788 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
3789 is set for varargs and stdarg functions. If this macro returns a
3790 nonzero value, the @var{named} argument is always true for named
3791 arguments, and false for unnamed arguments. If it returns a value of
3792 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
3793 are treated as named. Otherwise, all named arguments except the last
3794 are treated as named.
3796 You need not define this macro if it always returns zero.
3798 @findex PRETEND_OUTGOING_VARARGS_NAMED
3799 @item PRETEND_OUTGOING_VARARGS_NAMED
3800 If you need to conditionally change ABIs so that one works with
3801 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
3802 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
3803 defined, then define this macro to return nonzero if
3804 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
3805 Otherwise, you should not define this macro.
3809 @section Trampolines for Nested Functions
3810 @cindex trampolines for nested functions
3811 @cindex nested functions, trampolines for
3813 A @dfn{trampoline} is a small piece of code that is created at run time
3814 when the address of a nested function is taken. It normally resides on
3815 the stack, in the stack frame of the containing function. These macros
3816 tell GNU CC how to generate code to allocate and initialize a
3819 The instructions in the trampoline must do two things: load a constant
3820 address into the static chain register, and jump to the real address of
3821 the nested function. On CISC machines such as the m68k, this requires
3822 two instructions, a move immediate and a jump. Then the two addresses
3823 exist in the trampoline as word-long immediate operands. On RISC
3824 machines, it is often necessary to load each address into a register in
3825 two parts. Then pieces of each address form separate immediate
3828 The code generated to initialize the trampoline must store the variable
3829 parts---the static chain value and the function address---into the
3830 immediate operands of the instructions. On a CISC machine, this is
3831 simply a matter of copying each address to a memory reference at the
3832 proper offset from the start of the trampoline. On a RISC machine, it
3833 may be necessary to take out pieces of the address and store them
3837 @findex TRAMPOLINE_TEMPLATE
3838 @item TRAMPOLINE_TEMPLATE (@var{file})
3839 A C statement to output, on the stream @var{file}, assembler code for a
3840 block of data that contains the constant parts of a trampoline. This
3841 code should not include a label---the label is taken care of
3844 If you do not define this macro, it means no template is needed
3845 for the target. Do not define this macro on systems where the block move
3846 code to copy the trampoline into place would be larger than the code
3847 to generate it on the spot.
3849 @findex TRAMPOLINE_SECTION
3850 @item TRAMPOLINE_SECTION
3851 The name of a subroutine to switch to the section in which the
3852 trampoline template is to be placed (@pxref{Sections}). The default is
3853 a value of @samp{readonly_data_section}, which places the trampoline in
3854 the section containing read-only data.
3856 @findex TRAMPOLINE_SIZE
3857 @item TRAMPOLINE_SIZE
3858 A C expression for the size in bytes of the trampoline, as an integer.
3860 @findex TRAMPOLINE_ALIGNMENT
3861 @item TRAMPOLINE_ALIGNMENT
3862 Alignment required for trampolines, in bits.
3864 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
3865 is used for aligning trampolines.
3867 @findex INITIALIZE_TRAMPOLINE
3868 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
3869 A C statement to initialize the variable parts of a trampoline.
3870 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
3871 an RTX for the address of the nested function; @var{static_chain} is an
3872 RTX for the static chain value that should be passed to the function
3875 @findex ALLOCATE_TRAMPOLINE
3876 @item ALLOCATE_TRAMPOLINE (@var{fp})
3877 A C expression to allocate run-time space for a trampoline. The
3878 expression value should be an RTX representing a memory reference to the
3879 space for the trampoline.
3881 @cindex @code{FUNCTION_EPILOGUE} and trampolines
3882 @cindex @code{FUNCTION_PROLOGUE} and trampolines
3883 If this macro is not defined, by default the trampoline is allocated as
3884 a stack slot. This default is right for most machines. The exceptions
3885 are machines where it is impossible to execute instructions in the stack
3886 area. On such machines, you may have to implement a separate stack,
3887 using this macro in conjunction with @code{FUNCTION_PROLOGUE} and
3888 @code{FUNCTION_EPILOGUE}.
3890 @var{fp} points to a data structure, a @code{struct function}, which
3891 describes the compilation status of the immediate containing function of
3892 the function which the trampoline is for. Normally (when
3893 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
3894 trampoline is in the stack frame of this containing function. Other
3895 allocation strategies probably must do something analogous with this
3899 Implementing trampolines is difficult on many machines because they have
3900 separate instruction and data caches. Writing into a stack location
3901 fails to clear the memory in the instruction cache, so when the program
3902 jumps to that location, it executes the old contents.
3904 Here are two possible solutions. One is to clear the relevant parts of
3905 the instruction cache whenever a trampoline is set up. The other is to
3906 make all trampolines identical, by having them jump to a standard
3907 subroutine. The former technique makes trampoline execution faster; the
3908 latter makes initialization faster.
3910 To clear the instruction cache when a trampoline is initialized, define
3911 the following macros which describe the shape of the cache.
3914 @findex INSN_CACHE_SIZE
3915 @item INSN_CACHE_SIZE
3916 The total size in bytes of the cache.
3918 @findex INSN_CACHE_LINE_WIDTH
3919 @item INSN_CACHE_LINE_WIDTH
3920 The length in bytes of each cache line. The cache is divided into cache
3921 lines which are disjoint slots, each holding a contiguous chunk of data
3922 fetched from memory. Each time data is brought into the cache, an
3923 entire line is read at once. The data loaded into a cache line is
3924 always aligned on a boundary equal to the line size.
3926 @findex INSN_CACHE_DEPTH
3927 @item INSN_CACHE_DEPTH
3928 The number of alternative cache lines that can hold any particular memory
3932 Alternatively, if the machine has system calls or instructions to clear
3933 the instruction cache directly, you can define the following macro.
3936 @findex CLEAR_INSN_CACHE
3937 @item CLEAR_INSN_CACHE (@var{BEG}, @var{END})
3938 If defined, expands to a C expression clearing the @emph{instruction
3939 cache} in the specified interval. If it is not defined, and the macro
3940 INSN_CACHE_SIZE is defined, some generic code is generated to clear the
3941 cache. The definition of this macro would typically be a series of
3942 @code{asm} statements. Both @var{BEG} and @var{END} are both pointer
3946 To use a standard subroutine, define the following macro. In addition,
3947 you must make sure that the instructions in a trampoline fill an entire
3948 cache line with identical instructions, or else ensure that the
3949 beginning of the trampoline code is always aligned at the same point in
3950 its cache line. Look in @file{m68k.h} as a guide.
3953 @findex TRANSFER_FROM_TRAMPOLINE
3954 @item TRANSFER_FROM_TRAMPOLINE
3955 Define this macro if trampolines need a special subroutine to do their
3956 work. The macro should expand to a series of @code{asm} statements
3957 which will be compiled with GNU CC. They go in a library function named
3958 @code{__transfer_from_trampoline}.
3960 If you need to avoid executing the ordinary prologue code of a compiled
3961 C function when you jump to the subroutine, you can do so by placing a
3962 special label of your own in the assembler code. Use one @code{asm}
3963 statement to generate an assembler label, and another to make the label
3964 global. Then trampolines can use that label to jump directly to your
3965 special assembler code.
3969 @section Implicit Calls to Library Routines
3970 @cindex library subroutine names
3971 @cindex @file{libgcc.a}
3973 @c prevent bad page break with this line
3974 Here is an explanation of implicit calls to library routines.
3977 @findex MULSI3_LIBCALL
3978 @item MULSI3_LIBCALL
3979 A C string constant giving the name of the function to call for
3980 multiplication of one signed full-word by another. If you do not
3981 define this macro, the default name is used, which is @code{__mulsi3},
3982 a function defined in @file{libgcc.a}.
3984 @findex DIVSI3_LIBCALL
3985 @item DIVSI3_LIBCALL
3986 A C string constant giving the name of the function to call for
3987 division of one signed full-word by another. If you do not define
3988 this macro, the default name is used, which is @code{__divsi3}, a
3989 function defined in @file{libgcc.a}.
3991 @findex UDIVSI3_LIBCALL
3992 @item UDIVSI3_LIBCALL
3993 A C string constant giving the name of the function to call for
3994 division of one unsigned full-word by another. If you do not define
3995 this macro, the default name is used, which is @code{__udivsi3}, a
3996 function defined in @file{libgcc.a}.
3998 @findex MODSI3_LIBCALL
3999 @item MODSI3_LIBCALL
4000 A C string constant giving the name of the function to call for the
4001 remainder in division of one signed full-word by another. If you do
4002 not define this macro, the default name is used, which is
4003 @code{__modsi3}, a function defined in @file{libgcc.a}.
4005 @findex UMODSI3_LIBCALL
4006 @item UMODSI3_LIBCALL
4007 A C string constant giving the name of the function to call for the
4008 remainder in division of one unsigned full-word by another. If you do
4009 not define this macro, the default name is used, which is
4010 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4012 @findex MULDI3_LIBCALL
4013 @item MULDI3_LIBCALL
4014 A C string constant giving the name of the function to call for
4015 multiplication of one signed double-word by another. If you do not
4016 define this macro, the default name is used, which is @code{__muldi3},
4017 a function defined in @file{libgcc.a}.
4019 @findex DIVDI3_LIBCALL
4020 @item DIVDI3_LIBCALL
4021 A C string constant giving the name of the function to call for
4022 division of one signed double-word by another. If you do not define
4023 this macro, the default name is used, which is @code{__divdi3}, a
4024 function defined in @file{libgcc.a}.
4026 @findex UDIVDI3_LIBCALL
4027 @item UDIVDI3_LIBCALL
4028 A C string constant giving the name of the function to call for
4029 division of one unsigned full-word by another. If you do not define
4030 this macro, the default name is used, which is @code{__udivdi3}, a
4031 function defined in @file{libgcc.a}.
4033 @findex MODDI3_LIBCALL
4034 @item MODDI3_LIBCALL
4035 A C string constant giving the name of the function to call for the
4036 remainder in division of one signed double-word by another. If you do
4037 not define this macro, the default name is used, which is
4038 @code{__moddi3}, a function defined in @file{libgcc.a}.
4040 @findex UMODDI3_LIBCALL
4041 @item UMODDI3_LIBCALL
4042 A C string constant giving the name of the function to call for the
4043 remainder in division of one unsigned full-word by another. If you do
4044 not define this macro, the default name is used, which is
4045 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4047 @findex INIT_TARGET_OPTABS
4048 @item INIT_TARGET_OPTABS
4049 Define this macro as a C statement that declares additional library
4050 routines renames existing ones. @code{init_optabs} calls this macro after
4051 initializing all the normal library routines.
4054 @cindex @code{EDOM}, implicit usage
4056 The value of @code{EDOM} on the target machine, as a C integer constant
4057 expression. If you don't define this macro, GNU CC does not attempt to
4058 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4059 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4062 If you do not define @code{TARGET_EDOM}, then compiled code reports
4063 domain errors by calling the library function and letting it report the
4064 error. If mathematical functions on your system use @code{matherr} when
4065 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4066 that @code{matherr} is used normally.
4068 @findex GEN_ERRNO_RTX
4069 @cindex @code{errno}, implicit usage
4071 Define this macro as a C expression to create an rtl expression that
4072 refers to the global ``variable'' @code{errno}. (On certain systems,
4073 @code{errno} may not actually be a variable.) If you don't define this
4074 macro, a reasonable default is used.
4076 @findex TARGET_MEM_FUNCTIONS
4077 @cindex @code{bcopy}, implicit usage
4078 @cindex @code{memcpy}, implicit usage
4079 @cindex @code{bzero}, implicit usage
4080 @cindex @code{memset}, implicit usage
4081 @item TARGET_MEM_FUNCTIONS
4082 Define this macro if GNU CC should generate calls to the System V
4083 (and ANSI C) library functions @code{memcpy} and @code{memset}
4084 rather than the BSD functions @code{bcopy} and @code{bzero}.
4086 @findex LIBGCC_NEEDS_DOUBLE
4087 @item LIBGCC_NEEDS_DOUBLE
4088 Define this macro if only @code{float} arguments cannot be passed to
4089 library routines (so they must be converted to @code{double}). This
4090 macro affects both how library calls are generated and how the library
4091 routines in @file{libgcc1.c} accept their arguments. It is useful on
4092 machines where floating and fixed point arguments are passed
4093 differently, such as the i860.
4095 @findex FLOAT_ARG_TYPE
4096 @item FLOAT_ARG_TYPE
4097 Define this macro to override the type used by the library routines to
4098 pick up arguments of type @code{float}. (By default, they use a union
4099 of @code{float} and @code{int}.)
4101 The obvious choice would be @code{float}---but that won't work with
4102 traditional C compilers that expect all arguments declared as @code{float}
4103 to arrive as @code{double}. To avoid this conversion, the library routines
4104 ask for the value as some other type and then treat it as a @code{float}.
4106 On some systems, no other type will work for this. For these systems,
4107 you must use @code{LIBGCC_NEEDS_DOUBLE} instead, to force conversion of
4108 the values @code{double} before they are passed.
4111 @item FLOATIFY (@var{passed-value})
4112 Define this macro to override the way library routines redesignate a
4113 @code{float} argument as a @code{float} instead of the type it was
4114 passed as. The default is an expression which takes the @code{float}
4117 @findex FLOAT_VALUE_TYPE
4118 @item FLOAT_VALUE_TYPE
4119 Define this macro to override the type used by the library routines to
4120 return values that ought to have type @code{float}. (By default, they
4123 The obvious choice would be @code{float}---but that won't work with
4124 traditional C compilers gratuitously convert values declared as
4125 @code{float} into @code{double}.
4128 @item INTIFY (@var{float-value})
4129 Define this macro to override the way the value of a
4130 @code{float}-returning library routine should be packaged in order to
4131 return it. These functions are actually declared to return type
4132 @code{FLOAT_VALUE_TYPE} (normally @code{int}).
4134 These values can't be returned as type @code{float} because traditional
4135 C compilers would gratuitously convert the value to a @code{double}.
4137 A local variable named @code{intify} is always available when the macro
4138 @code{INTIFY} is used. It is a union of a @code{float} field named
4139 @code{f} and a field named @code{i} whose type is
4140 @code{FLOAT_VALUE_TYPE} or @code{int}.
4142 If you don't define this macro, the default definition works by copying
4143 the value through that union.
4145 @findex nongcc_SI_type
4146 @item nongcc_SI_type
4147 Define this macro as the name of the data type corresponding to
4148 @code{SImode} in the system's own C compiler.
4150 You need not define this macro if that type is @code{long int}, as it usually
4153 @findex nongcc_word_type
4154 @item nongcc_word_type
4155 Define this macro as the name of the data type corresponding to the
4156 word_mode in the system's own C compiler.
4158 You need not define this macro if that type is @code{long int}, as it usually
4161 @findex perform_@dots{}
4162 @item perform_@dots{}
4163 Define these macros to supply explicit C statements to carry out various
4164 arithmetic operations on types @code{float} and @code{double} in the
4165 library routines in @file{libgcc1.c}. See that file for a full list
4166 of these macros and their arguments.
4168 On most machines, you don't need to define any of these macros, because
4169 the C compiler that comes with the system takes care of doing them.
4171 @findex NEXT_OBJC_RUNTIME
4172 @item NEXT_OBJC_RUNTIME
4173 Define this macro to generate code for Objective C message sending using
4174 the calling convention of the NeXT system. This calling convention
4175 involves passing the object, the selector and the method arguments all
4176 at once to the method-lookup library function.
4178 The default calling convention passes just the object and the selector
4179 to the lookup function, which returns a pointer to the method.
4182 @node Addressing Modes
4183 @section Addressing Modes
4184 @cindex addressing modes
4186 @c prevent bad page break with this line
4187 This is about addressing modes.
4190 @findex HAVE_POST_INCREMENT
4191 @item HAVE_POST_INCREMENT
4192 A C expression that is nonzero the machine supports post-increment addressing.
4194 @findex HAVE_PRE_INCREMENT
4195 @findex HAVE_POST_DECREMENT
4196 @findex HAVE_PRE_DECREMENT
4197 @item HAVE_PRE_INCREMENT
4198 @itemx HAVE_POST_DECREMENT
4199 @itemx HAVE_PRE_DECREMENT
4200 Similar for other kinds of addressing.
4202 @findex CONSTANT_ADDRESS_P
4203 @item CONSTANT_ADDRESS_P (@var{x})
4204 A C expression that is 1 if the RTX @var{x} is a constant which
4205 is a valid address. On most machines, this can be defined as
4206 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4207 in which constant addresses are supported.
4210 @code{CONSTANT_P} accepts integer-values expressions whose values are
4211 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4212 @code{high} expressions and @code{const} arithmetic expressions, in
4213 addition to @code{const_int} and @code{const_double} expressions.
4215 @findex MAX_REGS_PER_ADDRESS
4216 @item MAX_REGS_PER_ADDRESS
4217 A number, the maximum number of registers that can appear in a valid
4218 memory address. Note that it is up to you to specify a value equal to
4219 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4222 @findex GO_IF_LEGITIMATE_ADDRESS
4223 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4224 A C compound statement with a conditional @code{goto @var{label};}
4225 executed if @var{x} (an RTX) is a legitimate memory address on the
4226 target machine for a memory operand of mode @var{mode}.
4228 It usually pays to define several simpler macros to serve as
4229 subroutines for this one. Otherwise it may be too complicated to
4232 This macro must exist in two variants: a strict variant and a
4233 non-strict one. The strict variant is used in the reload pass. It
4234 must be defined so that any pseudo-register that has not been
4235 allocated a hard register is considered a memory reference. In
4236 contexts where some kind of register is required, a pseudo-register
4237 with no hard register must be rejected.
4239 The non-strict variant is used in other passes. It must be defined to
4240 accept all pseudo-registers in every context where some kind of
4241 register is required.
4243 @findex REG_OK_STRICT
4244 Compiler source files that want to use the strict variant of this
4245 macro define the macro @code{REG_OK_STRICT}. You should use an
4246 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4247 in that case and the non-strict variant otherwise.
4249 Subroutines to check for acceptable registers for various purposes (one
4250 for base registers, one for index registers, and so on) are typically
4251 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4252 Then only these subroutine macros need have two variants; the higher
4253 levels of macros may be the same whether strict or not.@refill
4255 Normally, constant addresses which are the sum of a @code{symbol_ref}
4256 and an integer are stored inside a @code{const} RTX to mark them as
4257 constant. Therefore, there is no need to recognize such sums
4258 specifically as legitimate addresses. Normally you would simply
4259 recognize any @code{const} as legitimate.
4261 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4262 sums that are not marked with @code{const}. It assumes that a naked
4263 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4264 naked constant sums as illegitimate addresses, so that none of them will
4265 be given to @code{PRINT_OPERAND_ADDRESS}.
4267 @cindex @code{ENCODE_SECTION_INFO} and address validation
4268 On some machines, whether a symbolic address is legitimate depends on
4269 the section that the address refers to. On these machines, define the
4270 macro @code{ENCODE_SECTION_INFO} to store the information into the
4271 @code{symbol_ref}, and then check for it here. When you see a
4272 @code{const}, you will have to look inside it to find the
4273 @code{symbol_ref} in order to determine the section. @xref{Assembler
4276 @findex saveable_obstack
4277 The best way to modify the name string is by adding text to the
4278 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4279 the new name in @code{saveable_obstack}. You will have to modify
4280 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4281 output the name accordingly, and define @code{STRIP_NAME_ENCODING} to
4282 access the original name string.
4284 You can check the information stored here into the @code{symbol_ref} in
4285 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4286 @code{PRINT_OPERAND_ADDRESS}.
4288 @findex REG_OK_FOR_BASE_P
4289 @item REG_OK_FOR_BASE_P (@var{x})
4290 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4291 RTX) is valid for use as a base register. For hard registers, it
4292 should always accept those which the hardware permits and reject the
4293 others. Whether the macro accepts or rejects pseudo registers must be
4294 controlled by @code{REG_OK_STRICT} as described above. This usually
4295 requires two variant definitions, of which @code{REG_OK_STRICT}
4296 controls the one actually used.
4298 @findex REG_MODE_OK_FOR_BASE_P
4299 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4300 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4301 that expression may examine the mode of the memory reference in
4302 @var{mode}. You should define this macro if the mode of the memory
4303 reference affects whether a register may be used as a base register. If
4304 you define this macro, the compiler will use it instead of
4305 @code{REG_OK_FOR_BASE_P}.
4307 @findex REG_OK_FOR_INDEX_P
4308 @item REG_OK_FOR_INDEX_P (@var{x})
4309 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4310 RTX) is valid for use as an index register.
4312 The difference between an index register and a base register is that
4313 the index register may be scaled. If an address involves the sum of
4314 two registers, neither one of them scaled, then either one may be
4315 labeled the ``base'' and the other the ``index''; but whichever
4316 labeling is used must fit the machine's constraints of which registers
4317 may serve in each capacity. The compiler will try both labelings,
4318 looking for one that is valid, and will reload one or both registers
4319 only if neither labeling works.
4321 @findex LEGITIMIZE_ADDRESS
4322 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4323 A C compound statement that attempts to replace @var{x} with a valid
4324 memory address for an operand of mode @var{mode}. @var{win} will be a
4325 C statement label elsewhere in the code; the macro definition may use
4328 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4332 to avoid further processing if the address has become legitimate.
4334 @findex break_out_memory_refs
4335 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4336 and @var{oldx} will be the operand that was given to that function to produce
4339 The code generated by this macro should not alter the substructure of
4340 @var{x}. If it transforms @var{x} into a more legitimate form, it
4341 should assign @var{x} (which will always be a C variable) a new value.
4343 It is not necessary for this macro to come up with a legitimate
4344 address. The compiler has standard ways of doing so in all cases. In
4345 fact, it is safe for this macro to do nothing. But often a
4346 machine-dependent strategy can generate better code.
4348 @findex LEGITIMIZE_RELOAD_ADDRESS
4349 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4350 A C compound statement that attempts to replace @var{x}, which is an address
4351 that needs reloading, with a valid memory address for an operand of mode
4352 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4353 It is not necessary to define this macro, but it might be useful for
4354 performance reasons.
4356 For example, on the i386, it is sometimes possible to use a single
4357 reload register instead of two by reloading a sum of two pseudo
4358 registers into a register. On the other hand, for number of RISC
4359 processors offsets are limited so that often an intermediate address
4360 needs to be generated in order to address a stack slot. By defining
4361 LEGITIMIZE_RELOAD_ADDRESS appropriately, the intermediate addresses
4362 generated for adjacent some stack slots can be made identical, and thus
4365 @emph{Note}: This macro should be used with caution. It is necessary
4366 to know something of how reload works in order to effectively use this,
4367 and it is quite easy to produce macros that build in too much knowledge
4368 of reload internals.
4370 @emph{Note}: This macro must be able to reload an address created by a
4371 previous invocation of this macro. If it fails to handle such addresses
4372 then the compiler may generate incorrect code or abort.
4375 The macro definition should use @code{push_reload} to indicate parts that
4376 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4377 suitable to be passed unaltered to @code{push_reload}.
4379 The code generated by this macro must not alter the substructure of
4380 @var{x}. If it transforms @var{x} into a more legitimate form, it
4381 should assign @var{x} (which will always be a C variable) a new value.
4382 This also applies to parts that you change indirectly by calling
4385 @findex strict_memory_address_p
4386 The macro definition may use @code{strict_memory_address_p} to test if
4387 the address has become legitimate.
4390 If you want to change only a part of @var{x}, one standard way of doing
4391 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4392 single level of rtl. Thus, if the part to be changed is not at the
4393 top level, you'll need to replace first the top leve
4394 It is not necessary for this macro to come up with a legitimate
4395 address; but often a machine-dependent strategy can generate better code.
4397 @findex GO_IF_MODE_DEPENDENT_ADDRESS
4398 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4399 A C statement or compound statement with a conditional @code{goto
4400 @var{label};} executed if memory address @var{x} (an RTX) can have
4401 different meanings depending on the machine mode of the memory
4402 reference it is used for or if the address is valid for some modes
4405 Autoincrement and autodecrement addresses typically have mode-dependent
4406 effects because the amount of the increment or decrement is the size
4407 of the operand being addressed. Some machines have other mode-dependent
4408 addresses. Many RISC machines have no mode-dependent addresses.
4410 You may assume that @var{addr} is a valid address for the machine.
4412 @findex LEGITIMATE_CONSTANT_P
4413 @item LEGITIMATE_CONSTANT_P (@var{x})
4414 A C expression that is nonzero if @var{x} is a legitimate constant for
4415 an immediate operand on the target machine. You can assume that
4416 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4417 @samp{1} is a suitable definition for this macro on machines where
4418 anything @code{CONSTANT_P} is valid.@refill
4421 @node Condition Code
4422 @section Condition Code Status
4423 @cindex condition code status
4425 @c prevent bad page break with this line
4426 This describes the condition code status.
4429 The file @file{conditions.h} defines a variable @code{cc_status} to
4430 describe how the condition code was computed (in case the interpretation of
4431 the condition code depends on the instruction that it was set by). This
4432 variable contains the RTL expressions on which the condition code is
4433 currently based, and several standard flags.
4435 Sometimes additional machine-specific flags must be defined in the machine
4436 description header file. It can also add additional machine-specific
4437 information by defining @code{CC_STATUS_MDEP}.
4440 @findex CC_STATUS_MDEP
4441 @item CC_STATUS_MDEP
4442 C code for a data type which is used for declaring the @code{mdep}
4443 component of @code{cc_status}. It defaults to @code{int}.
4445 This macro is not used on machines that do not use @code{cc0}.
4447 @findex CC_STATUS_MDEP_INIT
4448 @item CC_STATUS_MDEP_INIT
4449 A C expression to initialize the @code{mdep} field to ``empty''.
4450 The default definition does nothing, since most machines don't use
4451 the field anyway. If you want to use the field, you should probably
4452 define this macro to initialize it.
4454 This macro is not used on machines that do not use @code{cc0}.
4456 @findex NOTICE_UPDATE_CC
4457 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4458 A C compound statement to set the components of @code{cc_status}
4459 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4460 this macro's responsibility to recognize insns that set the condition
4461 code as a byproduct of other activity as well as those that explicitly
4464 This macro is not used on machines that do not use @code{cc0}.
4466 If there are insns that do not set the condition code but do alter
4467 other machine registers, this macro must check to see whether they
4468 invalidate the expressions that the condition code is recorded as
4469 reflecting. For example, on the 68000, insns that store in address
4470 registers do not set the condition code, which means that usually
4471 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4472 insns. But suppose that the previous insn set the condition code
4473 based on location @samp{a4@@(102)} and the current insn stores a new
4474 value in @samp{a4}. Although the condition code is not changed by
4475 this, it will no longer be true that it reflects the contents of
4476 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4477 @code{cc_status} in this case to say that nothing is known about the
4478 condition code value.
4480 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4481 with the results of peephole optimization: insns whose patterns are
4482 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4483 constants which are just the operands. The RTL structure of these
4484 insns is not sufficient to indicate what the insns actually do. What
4485 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4486 @code{CC_STATUS_INIT}.
4488 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4489 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4490 @samp{cc}. This avoids having detailed information about patterns in
4491 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4493 @findex EXTRA_CC_MODES
4494 @item EXTRA_CC_MODES
4495 A list of names to be used for additional modes for condition code
4496 values in registers (@pxref{Jump Patterns}). These names are added
4497 to @code{enum machine_mode} and all have class @code{MODE_CC}. By
4498 convention, they should start with @samp{CC} and end with @samp{mode}.
4500 You should only define this macro if your machine does not use @code{cc0}
4501 and only if additional modes are required.
4503 @findex EXTRA_CC_NAMES
4504 @item EXTRA_CC_NAMES
4505 A list of C strings giving the names for the modes listed in
4506 @code{EXTRA_CC_MODES}. For example, the Sparc defines this macro and
4507 @code{EXTRA_CC_MODES} as
4510 #define EXTRA_CC_MODES CC_NOOVmode, CCFPmode, CCFPEmode
4511 #define EXTRA_CC_NAMES "CC_NOOV", "CCFP", "CCFPE"
4514 This macro is not required if @code{EXTRA_CC_MODES} is not defined.
4516 @findex SELECT_CC_MODE
4517 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4518 Returns a mode from class @code{MODE_CC} to be used when comparison
4519 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
4520 example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
4521 @pxref{Jump Patterns} for a description of the reason for this
4525 #define SELECT_CC_MODE(OP,X,Y) \
4526 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4527 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
4528 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4529 || GET_CODE (X) == NEG) \
4530 ? CC_NOOVmode : CCmode))
4533 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
4535 @findex CANONICALIZE_COMPARISON
4536 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
4537 One some machines not all possible comparisons are defined, but you can
4538 convert an invalid comparison into a valid one. For example, the Alpha
4539 does not have a @code{GT} comparison, but you can use an @code{LT}
4540 comparison instead and swap the order of the operands.
4542 On such machines, define this macro to be a C statement to do any
4543 required conversions. @var{code} is the initial comparison code
4544 and @var{op0} and @var{op1} are the left and right operands of the
4545 comparison, respectively. You should modify @var{code}, @var{op0}, and
4546 @var{op1} as required.
4548 GNU CC will not assume that the comparison resulting from this macro is
4549 valid but will see if the resulting insn matches a pattern in the
4552 You need not define this macro if it would never change the comparison
4555 @findex REVERSIBLE_CC_MODE
4556 @item REVERSIBLE_CC_MODE (@var{mode})
4557 A C expression whose value is one if it is always safe to reverse a
4558 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4559 can ever return @var{mode} for a floating-point inequality comparison,
4560 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4562 You need not define this macro if it would always returns zero or if the
4563 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4564 For example, here is the definition used on the Sparc, where floating-point
4565 inequality comparisons are always given @code{CCFPEmode}:
4568 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
4574 @section Describing Relative Costs of Operations
4575 @cindex costs of instructions
4576 @cindex relative costs
4577 @cindex speed of instructions
4579 These macros let you describe the relative speed of various operations
4580 on the target machine.
4584 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
4585 A part of a C @code{switch} statement that describes the relative costs
4586 of constant RTL expressions. It must contain @code{case} labels for
4587 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
4588 @code{label_ref} and @code{const_double}. Each case must ultimately
4589 reach a @code{return} statement to return the relative cost of the use
4590 of that kind of constant value in an expression. The cost may depend on
4591 the precise value of the constant, which is available for examination in
4592 @var{x}, and the rtx code of the expression in which it is contained,
4593 found in @var{outer_code}.
4595 @var{code} is the expression code---redundant, since it can be
4596 obtained with @code{GET_CODE (@var{x})}.
4599 @findex COSTS_N_INSNS
4600 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4601 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
4602 This can be used, for example, to indicate how costly a multiply
4603 instruction is. In writing this macro, you can use the construct
4604 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
4605 instructions. @var{outer_code} is the code of the expression in which
4606 @var{x} is contained.
4608 This macro is optional; do not define it if the default cost assumptions
4609 are adequate for the target machine.
4611 @findex DEFAULT_RTX_COSTS
4612 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4613 This macro, if defined, is called for any case not handled by the
4614 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
4615 to put case labels into the macro, but the code, or any functions it
4616 calls, must assume that the RTL in @var{x} could be of any type that has
4617 not already been handled. The arguments are the same as for
4618 @code{RTX_COSTS}, and the macro should execute a return statement giving
4619 the cost of any RTL expressions that it can handle. The default cost
4620 calculation is used for any RTL for which this macro does not return a
4623 This macro is optional; do not define it if the default cost assumptions
4624 are adequate for the target machine.
4626 @findex ADDRESS_COST
4627 @item ADDRESS_COST (@var{address})
4628 An expression giving the cost of an addressing mode that contains
4629 @var{address}. If not defined, the cost is computed from
4630 the @var{address} expression and the @code{CONST_COSTS} values.
4632 For most CISC machines, the default cost is a good approximation of the
4633 true cost of the addressing mode. However, on RISC machines, all
4634 instructions normally have the same length and execution time. Hence
4635 all addresses will have equal costs.
4637 In cases where more than one form of an address is known, the form with
4638 the lowest cost will be used. If multiple forms have the same, lowest,
4639 cost, the one that is the most complex will be used.
4641 For example, suppose an address that is equal to the sum of a register
4642 and a constant is used twice in the same basic block. When this macro
4643 is not defined, the address will be computed in a register and memory
4644 references will be indirect through that register. On machines where
4645 the cost of the addressing mode containing the sum is no higher than
4646 that of a simple indirect reference, this will produce an additional
4647 instruction and possibly require an additional register. Proper
4648 specification of this macro eliminates this overhead for such machines.
4650 Similar use of this macro is made in strength reduction of loops.
4652 @var{address} need not be valid as an address. In such a case, the cost
4653 is not relevant and can be any value; invalid addresses need not be
4654 assigned a different cost.
4656 On machines where an address involving more than one register is as
4657 cheap as an address computation involving only one register, defining
4658 @code{ADDRESS_COST} to reflect this can cause two registers to be live
4659 over a region of code where only one would have been if
4660 @code{ADDRESS_COST} were not defined in that manner. This effect should
4661 be considered in the definition of this macro. Equivalent costs should
4662 probably only be given to addresses with different numbers of registers
4663 on machines with lots of registers.
4665 This macro will normally either not be defined or be defined as a
4668 @findex REGISTER_MOVE_COST
4669 @item REGISTER_MOVE_COST (@var{from}, @var{to})
4670 A C expression for the cost of moving data from a register in class
4671 @var{from} to one in class @var{to}. The classes are expressed using
4672 the enumeration values such as @code{GENERAL_REGS}. A value of 2 is the
4673 default; other values are interpreted relative to that.
4675 It is not required that the cost always equal 2 when @var{from} is the
4676 same as @var{to}; on some machines it is expensive to move between
4677 registers if they are not general registers.
4679 If reload sees an insn consisting of a single @code{set} between two
4680 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4681 classes returns a value of 2, reload does not check to ensure that the
4682 constraints of the insn are met. Setting a cost of other than 2 will
4683 allow reload to verify that the constraints are met. You should do this
4684 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4686 @findex MEMORY_MOVE_COST
4687 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4688 A C expression for the cost of moving data of mode @var{mode} between a
4689 register of class @var{class} and memory; @var{in} is zero if the value
4690 is to be written to memory, non-zero if it is to be read in. This cost
4691 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
4692 registers and memory is more expensive than between two registers, you
4693 should define this macro to express the relative cost.
4695 If you do not define this macro, GNU CC uses a default cost of 4 plus
4696 the cost of copying via a secondary reload register, if one is
4697 needed. If your machine requires a secondary reload register to copy
4698 between memory and a register of @var{class} but the reload mechanism is
4699 more complex than copying via an intermediate, define this macro to
4700 reflect the actual cost of the move.
4702 GNU CC defines the function @code{memory_move_secondary_cost} if
4703 secondary reloads are needed. It computes the costs due to copying via
4704 a secondary register. If your machine copies from memory using a
4705 secondary register in the conventional way but the default base value of
4706 4 is not correct for your machine, define this macro to add some other
4707 value to the result of that function. The arguments to that function
4708 are the same as to this macro.
4712 A C expression for the cost of a branch instruction. A value of 1 is
4713 the default; other values are interpreted relative to that.
4716 Here are additional macros which do not specify precise relative costs,
4717 but only that certain actions are more expensive than GNU CC would
4721 @findex SLOW_BYTE_ACCESS
4722 @item SLOW_BYTE_ACCESS
4723 Define this macro as a C expression which is nonzero if accessing less
4724 than a word of memory (i.e. a @code{char} or a @code{short}) is no
4725 faster than accessing a word of memory, i.e., if such access
4726 require more than one instruction or if there is no difference in cost
4727 between byte and (aligned) word loads.
4729 When this macro is not defined, the compiler will access a field by
4730 finding the smallest containing object; when it is defined, a fullword
4731 load will be used if alignment permits. Unless bytes accesses are
4732 faster than word accesses, using word accesses is preferable since it
4733 may eliminate subsequent memory access if subsequent accesses occur to
4734 other fields in the same word of the structure, but to different bytes.
4736 @findex SLOW_ZERO_EXTEND
4737 @item SLOW_ZERO_EXTEND
4738 Define this macro if zero-extension (of a @code{char} or @code{short}
4739 to an @code{int}) can be done faster if the destination is a register
4740 that is known to be zero.
4742 If you define this macro, you must have instruction patterns that
4743 recognize RTL structures like this:
4746 (set (strict_low_part (subreg:QI (reg:SI @dots{}) 0)) @dots{})
4750 and likewise for @code{HImode}.
4752 @findex SLOW_UNALIGNED_ACCESS
4753 @item SLOW_UNALIGNED_ACCESS
4754 Define this macro to be the value 1 if unaligned accesses have a cost
4755 many times greater than aligned accesses, for example if they are
4756 emulated in a trap handler.
4758 When this macro is non-zero, the compiler will act as if
4759 @code{STRICT_ALIGNMENT} were non-zero when generating code for block
4760 moves. This can cause significantly more instructions to be produced.
4761 Therefore, do not set this macro non-zero if unaligned accesses only add a
4762 cycle or two to the time for a memory access.
4764 If the value of this macro is always zero, it need not be defined.
4766 @findex DONT_REDUCE_ADDR
4767 @item DONT_REDUCE_ADDR
4768 Define this macro to inhibit strength reduction of memory addresses.
4769 (On some machines, such strength reduction seems to do harm rather
4774 The threshold of number of scalar memory-to-memory move insns, @emph{below}
4775 which a sequence of insns should be generated instead of a
4776 string move insn or a library call. Increasing the value will always
4777 make code faster, but eventually incurs high cost in increased code size.
4779 Note that on machines with no memory-to-memory move insns, this macro denotes
4780 the corresponding number of memory-to-memory @emph{sequences}.
4782 If you don't define this, a reasonable default is used.
4784 @findex MOVE_BY_PIECES_P
4785 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
4786 A C expression used to determine whether @code{move_by_pieces} will be used to
4787 copy a chunk of memory, or whether some other block move mechanism
4788 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
4789 than @code{MOVE_RATIO}.
4791 @findex MOVE_MAX_PIECES
4792 @item MOVE_MAX_PIECES
4793 A C expression used by @code{move_by_pieces} to determine the largest unit
4794 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
4796 @findex USE_LOAD_POST_INCREMENT
4797 @item USE_LOAD_POST_INCREMENT (@var{mode})
4798 A C expression used to determine whether a load postincrement is a good
4799 thing to use for a given mode. Defaults to the value of
4800 @code{HAVE_POST_INCREMENT}.
4802 @findex USE_LOAD_POST_DECREMENT
4803 @item USE_LOAD_POST_DECREMENT (@var{mode})
4804 A C expression used to determine whether a load postdecrement is a good
4805 thing to use for a given mode. Defaults to the value of
4806 @code{HAVE_POST_DECREMENT}.
4808 @findex USE_LOAD_PRE_INCREMENT
4809 @item USE_LOAD_PRE_INCREMENT (@var{mode})
4810 A C expression used to determine whether a load preincrement is a good
4811 thing to use for a given mode. Defaults to the value of
4812 @code{HAVE_PRE_INCREMENT}.
4814 @findex USE_LOAD_PRE_DECREMENT
4815 @item USE_LOAD_PRE_DECREMENT (@var{mode})
4816 A C expression used to determine whether a load predecrement is a good
4817 thing to use for a given mode. Defaults to the value of
4818 @code{HAVE_PRE_DECREMENT}.
4820 @findex USE_STORE_POST_INCREMENT
4821 @item USE_STORE_POST_INCREMENT (@var{mode})
4822 A C expression used to determine whether a store postincrement is a good
4823 thing to use for a given mode. Defaults to the value of
4824 @code{HAVE_POST_INCREMENT}.
4826 @findex USE_STORE_POST_DECREMENT
4827 @item USE_STORE_POST_DECREMENT (@var{mode})
4828 A C expression used to determine whether a store postdeccrement is a good
4829 thing to use for a given mode. Defaults to the value of
4830 @code{HAVE_POST_DECREMENT}.
4832 @findex USE_STORE_PRE_INCREMENT
4833 @item USE_STORE_PRE_INCREMENT (@var{mode})
4834 This macro is used to determine whether a store preincrement is a good
4835 thing to use for a given mode. Defaults to the value of
4836 @code{HAVE_PRE_INCREMENT}.
4838 @findex USE_STORE_PRE_DECREMENT
4839 @item USE_STORE_PRE_DECREMENT (@var{mode})
4840 This macro is used to determine whether a store predecrement is a good
4841 thing to use for a given mode. Defaults to the value of
4842 @code{HAVE_PRE_DECREMENT}.
4844 @findex NO_FUNCTION_CSE
4845 @item NO_FUNCTION_CSE
4846 Define this macro if it is as good or better to call a constant
4847 function address than to call an address kept in a register.
4849 @findex NO_RECURSIVE_FUNCTION_CSE
4850 @item NO_RECURSIVE_FUNCTION_CSE
4851 Define this macro if it is as good or better for a function to call
4852 itself with an explicit address than to call an address kept in a
4856 @item ADJUST_COST (@var{insn}, @var{link}, @var{dep_insn}, @var{cost})
4857 A C statement (sans semicolon) to update the integer variable @var{cost}
4858 based on the relationship between @var{insn} that is dependent on
4859 @var{dep_insn} through the dependence @var{link}. The default is to
4860 make no adjustment to @var{cost}. This can be used for example to
4861 specify to the scheduler that an output- or anti-dependence does not
4862 incur the same cost as a data-dependence.
4864 @findex ADJUST_PRIORITY
4865 @item ADJUST_PRIORITY (@var{insn})
4866 A C statement (sans semicolon) to update the integer scheduling
4867 priority @code{INSN_PRIORITY(@var{insn})}. Reduce the priority
4868 to execute the @var{insn} earlier, increase the priority to execute
4869 @var{insn} later. Do not define this macro if you do not need to
4870 adjust the scheduling priorities of insns.
4874 @section Dividing the Output into Sections (Texts, Data, @dots{})
4875 @c the above section title is WAY too long. maybe cut the part between
4876 @c the (...)? --mew 10feb93
4878 An object file is divided into sections containing different types of
4879 data. In the most common case, there are three sections: the @dfn{text
4880 section}, which holds instructions and read-only data; the @dfn{data
4881 section}, which holds initialized writable data; and the @dfn{bss
4882 section}, which holds uninitialized data. Some systems have other kinds
4885 The compiler must tell the assembler when to switch sections. These
4886 macros control what commands to output to tell the assembler this. You
4887 can also define additional sections.
4890 @findex TEXT_SECTION_ASM_OP
4891 @item TEXT_SECTION_ASM_OP
4892 A C expression whose value is a string containing the assembler
4893 operation that should precede instructions and read-only data. Normally
4894 @code{".text"} is right.
4896 @findex DATA_SECTION_ASM_OP
4897 @item DATA_SECTION_ASM_OP
4898 A C expression whose value is a string containing the assembler
4899 operation to identify the following data as writable initialized data.
4900 Normally @code{".data"} is right.
4902 @findex SHARED_SECTION_ASM_OP
4903 @item SHARED_SECTION_ASM_OP
4904 If defined, a C expression whose value is a string containing the
4905 assembler operation to identify the following data as shared data. If
4906 not defined, @code{DATA_SECTION_ASM_OP} will be used.
4908 @findex BSS_SECTION_ASM_OP
4909 @item BSS_SECTION_ASM_OP
4910 If defined, a C expression whose value is a string containing the
4911 assembler operation to identify the following data as uninitialized global
4912 data. If not defined, and neither @code{ASM_OUTPUT_BSS} nor
4913 @code{ASM_OUTPUT_ALIGNED_BSS} are defined, uninitialized global data will be
4914 output in the data section if @samp{-fno-common} is passed, otherwise
4915 @code{ASM_OUTPUT_COMMON} will be used.
4917 @findex SHARED_BSS_SECTION_ASM_OP
4918 @item SHARED_BSS_SECTION_ASM_OP
4919 If defined, a C expression whose value is a string containing the
4920 assembler operation to identify the following data as uninitialized global
4921 shared data. If not defined, and @code{BSS_SECTION_ASM_OP} is, the latter
4924 @findex INIT_SECTION_ASM_OP
4925 @item INIT_SECTION_ASM_OP
4926 If defined, a C expression whose value is a string containing the
4927 assembler operation to identify the following data as initialization
4928 code. If not defined, GNU CC will assume such a section does not
4931 @findex EXTRA_SECTIONS
4934 @item EXTRA_SECTIONS
4935 A list of names for sections other than the standard two, which are
4936 @code{in_text} and @code{in_data}. You need not define this macro
4937 on a system with no other sections (that GCC needs to use).
4939 @findex EXTRA_SECTION_FUNCTIONS
4940 @findex text_section
4941 @findex data_section
4942 @item EXTRA_SECTION_FUNCTIONS
4943 One or more functions to be defined in @file{varasm.c}. These
4944 functions should do jobs analogous to those of @code{text_section} and
4945 @code{data_section}, for your additional sections. Do not define this
4946 macro if you do not define @code{EXTRA_SECTIONS}.
4948 @findex READONLY_DATA_SECTION
4949 @item READONLY_DATA_SECTION
4950 On most machines, read-only variables, constants, and jump tables are
4951 placed in the text section. If this is not the case on your machine,
4952 this macro should be defined to be the name of a function (either
4953 @code{data_section} or a function defined in @code{EXTRA_SECTIONS}) that
4954 switches to the section to be used for read-only items.
4956 If these items should be placed in the text section, this macro should
4959 @findex SELECT_SECTION
4960 @item SELECT_SECTION (@var{exp}, @var{reloc})
4961 A C statement or statements to switch to the appropriate section for
4962 output of @var{exp}. You can assume that @var{exp} is either a
4963 @code{VAR_DECL} node or a constant of some sort. @var{reloc}
4964 indicates whether the initial value of @var{exp} requires link-time
4965 relocations. Select the section by calling @code{text_section} or one
4966 of the alternatives for other sections.
4968 Do not define this macro if you put all read-only variables and
4969 constants in the read-only data section (usually the text section).
4971 @findex SELECT_RTX_SECTION
4972 @item SELECT_RTX_SECTION (@var{mode}, @var{rtx})
4973 A C statement or statements to switch to the appropriate section for
4974 output of @var{rtx} in mode @var{mode}. You can assume that @var{rtx}
4975 is some kind of constant in RTL. The argument @var{mode} is redundant
4976 except in the case of a @code{const_int} rtx. Select the section by
4977 calling @code{text_section} or one of the alternatives for other
4980 Do not define this macro if you put all constants in the read-only
4983 @findex JUMP_TABLES_IN_TEXT_SECTION
4984 @item JUMP_TABLES_IN_TEXT_SECTION
4985 Define this macro to be an expression with a non-zero value if jump
4986 tables (for @code{tablejump} insns) should be output in the text
4987 section, along with the assembler instructions. Otherwise, the
4988 readonly data section is used.
4990 This macro is irrelevant if there is no separate readonly data section.
4992 @findex ENCODE_SECTION_INFO
4993 @item ENCODE_SECTION_INFO (@var{decl})
4994 Define this macro if references to a symbol must be treated differently
4995 depending on something about the variable or function named by the
4996 symbol (such as what section it is in).
4998 The macro definition, if any, is executed immediately after the rtl for
4999 @var{decl} has been created and stored in @code{DECL_RTL (@var{decl})}.
5000 The value of the rtl will be a @code{mem} whose address is a
5003 @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO}
5004 The usual thing for this macro to do is to record a flag in the
5005 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
5006 modified name string in the @code{symbol_ref} (if one bit is not enough
5009 @findex STRIP_NAME_ENCODING
5010 @item STRIP_NAME_ENCODING (@var{var}, @var{sym_name})
5011 Decode @var{sym_name} and store the real name part in @var{var}, sans
5012 the characters that encode section info. Define this macro if
5013 @code{ENCODE_SECTION_INFO} alters the symbol's name string.
5015 @findex UNIQUE_SECTION_P
5016 @item UNIQUE_SECTION_P (@var{decl})
5017 A C expression which evaluates to true if @var{decl} should be placed
5018 into a unique section for some target-specific reason. If you do not
5019 define this macro, the default is @samp{0}. Note that the flag
5020 @samp{-ffunction-sections} will also cause functions to be placed into
5023 @findex UNIQUE_SECTION
5024 @item UNIQUE_SECTION (@var{decl}, @var{reloc})
5025 A C statement to build up a unique section name, expressed as a
5026 STRING_CST node, and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5027 @var{reloc} indicates whether the initial value of @var{exp} requires
5028 link-time relocations. If you do not define this macro, GNU CC will use
5029 the symbol name prefixed by @samp{.} as the section name.
5033 @section Position Independent Code
5034 @cindex position independent code
5037 This section describes macros that help implement generation of position
5038 independent code. Simply defining these macros is not enough to
5039 generate valid PIC; you must also add support to the macros
5040 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5041 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5042 @samp{movsi} to do something appropriate when the source operand
5043 contains a symbolic address. You may also need to alter the handling of
5044 switch statements so that they use relative addresses.
5045 @c i rearranged the order of the macros above to try to force one of
5046 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5049 @findex PIC_OFFSET_TABLE_REGNUM
5050 @item PIC_OFFSET_TABLE_REGNUM
5051 The register number of the register used to address a table of static
5052 data addresses in memory. In some cases this register is defined by a
5053 processor's ``application binary interface'' (ABI). When this macro
5054 is defined, RTL is generated for this register once, as with the stack
5055 pointer and frame pointer registers. If this macro is not defined, it
5056 is up to the machine-dependent files to allocate such a register (if
5059 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5060 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5061 Define this macro if the register defined by
5062 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
5063 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5065 @findex FINALIZE_PIC
5067 By generating position-independent code, when two different programs (A
5068 and B) share a common library (libC.a), the text of the library can be
5069 shared whether or not the library is linked at the same address for both
5070 programs. In some of these environments, position-independent code
5071 requires not only the use of different addressing modes, but also
5072 special code to enable the use of these addressing modes.
5074 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
5075 codes once the function is being compiled into assembly code, but not
5076 before. (It is not done before, because in the case of compiling an
5077 inline function, it would lead to multiple PIC prologues being
5078 included in functions which used inline functions and were compiled to
5081 @findex LEGITIMATE_PIC_OPERAND_P
5082 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
5083 A C expression that is nonzero if @var{x} is a legitimate immediate
5084 operand on the target machine when generating position independent code.
5085 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5086 check this. You can also assume @var{flag_pic} is true, so you need not
5087 check it either. You need not define this macro if all constants
5088 (including @code{SYMBOL_REF}) can be immediate operands when generating
5089 position independent code.
5092 @node Assembler Format
5093 @section Defining the Output Assembler Language
5095 This section describes macros whose principal purpose is to describe how
5096 to write instructions in assembler language--rather than what the
5100 * File Framework:: Structural information for the assembler file.
5101 * Data Output:: Output of constants (numbers, strings, addresses).
5102 * Uninitialized Data:: Output of uninitialized variables.
5103 * Label Output:: Output and generation of labels.
5104 * Initialization:: General principles of initialization
5105 and termination routines.
5106 * Macros for Initialization::
5107 Specific macros that control the handling of
5108 initialization and termination routines.
5109 * Instruction Output:: Output of actual instructions.
5110 * Dispatch Tables:: Output of jump tables.
5111 * Exception Region Output:: Output of exception region code.
5112 * Alignment Output:: Pseudo ops for alignment and skipping data.
5115 @node File Framework
5116 @subsection The Overall Framework of an Assembler File
5117 @cindex assembler format
5118 @cindex output of assembler code
5120 @c prevent bad page break with this line
5121 This describes the overall framework of an assembler file.
5124 @findex ASM_FILE_START
5125 @item ASM_FILE_START (@var{stream})
5126 A C expression which outputs to the stdio stream @var{stream}
5127 some appropriate text to go at the start of an assembler file.
5129 Normally this macro is defined to output a line containing
5130 @samp{#NO_APP}, which is a comment that has no effect on most
5131 assemblers but tells the GNU assembler that it can save time by not
5132 checking for certain assembler constructs.
5134 On systems that use SDB, it is necessary to output certain commands;
5135 see @file{attasm.h}.
5137 @findex ASM_FILE_END
5138 @item ASM_FILE_END (@var{stream})
5139 A C expression which outputs to the stdio stream @var{stream}
5140 some appropriate text to go at the end of an assembler file.
5142 If this macro is not defined, the default is to output nothing
5143 special at the end of the file. Most systems don't require any
5146 On systems that use SDB, it is necessary to output certain commands;
5147 see @file{attasm.h}.
5149 @findex ASM_IDENTIFY_GCC
5150 @item ASM_IDENTIFY_GCC (@var{file})
5151 A C statement to output assembler commands which will identify
5152 the object file as having been compiled with GNU CC (or another
5155 If you don't define this macro, the string @samp{gcc_compiled.:}
5156 is output. This string is calculated to define a symbol which,
5157 on BSD systems, will never be defined for any other reason.
5158 GDB checks for the presence of this symbol when reading the
5159 symbol table of an executable.
5161 On non-BSD systems, you must arrange communication with GDB in
5162 some other fashion. If GDB is not used on your system, you can
5163 define this macro with an empty body.
5165 @findex ASM_COMMENT_START
5166 @item ASM_COMMENT_START
5167 A C string constant describing how to begin a comment in the target
5168 assembler language. The compiler assumes that the comment will end at
5169 the end of the line.
5173 A C string constant for text to be output before each @code{asm}
5174 statement or group of consecutive ones. Normally this is
5175 @code{"#APP"}, which is a comment that has no effect on most
5176 assemblers but tells the GNU assembler that it must check the lines
5177 that follow for all valid assembler constructs.
5181 A C string constant for text to be output after each @code{asm}
5182 statement or group of consecutive ones. Normally this is
5183 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5184 time-saving assumptions that are valid for ordinary compiler output.
5186 @findex ASM_OUTPUT_SOURCE_FILENAME
5187 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5188 A C statement to output COFF information or DWARF debugging information
5189 which indicates that filename @var{name} is the current source file to
5190 the stdio stream @var{stream}.
5192 This macro need not be defined if the standard form of output
5193 for the file format in use is appropriate.
5195 @findex OUTPUT_QUOTED_STRING
5196 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{name})
5197 A C statement to output the string @var{string} to the stdio stream
5198 @var{stream}. If you do not call the function @code{output_quoted_string}
5199 in your config files, GNU CC will only call it to output filenames to
5200 the assembler source. So you can use it to canonicalize the format
5201 of the filename using this macro.
5203 @findex ASM_OUTPUT_SOURCE_LINE
5204 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
5205 A C statement to output DBX or SDB debugging information before code
5206 for line number @var{line} of the current source file to the
5207 stdio stream @var{stream}.
5209 This macro need not be defined if the standard form of debugging
5210 information for the debugger in use is appropriate.
5212 @findex ASM_OUTPUT_IDENT
5213 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
5214 A C statement to output something to the assembler file to handle a
5215 @samp{#ident} directive containing the text @var{string}. If this
5216 macro is not defined, nothing is output for a @samp{#ident} directive.
5218 @findex ASM_OUTPUT_SECTION_NAME
5219 @item ASM_OUTPUT_SECTION_NAME (@var{stream}, @var{decl}, @var{name}, @var{reloc})
5220 A C statement to output something to the assembler file to switch to section
5221 @var{name} for object @var{decl} which is either a @code{FUNCTION_DECL}, a
5222 @code{VAR_DECL} or @code{NULL_TREE}. @var{reloc}
5223 indicates whether the initial value of @var{exp} requires link-time
5224 relocations. Some target formats do not support
5225 arbitrary sections. Do not define this macro in such cases.
5227 At present this macro is only used to support section attributes.
5228 When this macro is undefined, section attributes are disabled.
5230 @findex OBJC_PROLOGUE
5232 A C statement to output any assembler statements which are required to
5233 precede any Objective C object definitions or message sending. The
5234 statement is executed only when compiling an Objective C program.
5239 @subsection Output of Data
5241 @c prevent bad page break with this line
5242 This describes data output.
5245 @findex ASM_OUTPUT_LONG_DOUBLE
5246 @findex ASM_OUTPUT_DOUBLE
5247 @findex ASM_OUTPUT_FLOAT
5248 @item ASM_OUTPUT_LONG_DOUBLE (@var{stream}, @var{value})
5249 @itemx ASM_OUTPUT_DOUBLE (@var{stream}, @var{value})
5250 @itemx ASM_OUTPUT_FLOAT (@var{stream}, @var{value})
5251 @itemx ASM_OUTPUT_THREE_QUARTER_FLOAT (@var{stream}, @var{value})
5252 @itemx ASM_OUTPUT_SHORT_FLOAT (@var{stream}, @var{value})
5253 @itemx ASM_OUTPUT_BYTE_FLOAT (@var{stream}, @var{value})
5254 A C statement to output to the stdio stream @var{stream} an assembler
5255 instruction to assemble a floating-point constant of @code{TFmode},
5256 @code{DFmode}, @code{SFmode}, @code{TQFmode}, @code{HFmode}, or
5257 @code{QFmode}, respectively, whose value is @var{value}. @var{value}
5258 will be a C expression of type @code{REAL_VALUE_TYPE}. Macros such as
5259 @code{REAL_VALUE_TO_TARGET_DOUBLE} are useful for writing these
5262 @findex ASM_OUTPUT_QUADRUPLE_INT
5263 @findex ASM_OUTPUT_DOUBLE_INT
5264 @findex ASM_OUTPUT_INT
5265 @findex ASM_OUTPUT_SHORT
5266 @findex ASM_OUTPUT_CHAR
5267 @findex output_addr_const
5268 @item ASM_OUTPUT_QUADRUPLE_INT (@var{stream}, @var{exp})
5269 @itemx ASM_OUTPUT_DOUBLE_INT (@var{stream}, @var{exp})
5270 @itemx ASM_OUTPUT_INT (@var{stream}, @var{exp})
5271 @itemx ASM_OUTPUT_SHORT (@var{stream}, @var{exp})
5272 @itemx ASM_OUTPUT_CHAR (@var{stream}, @var{exp})
5273 A C statement to output to the stdio stream @var{stream} an assembler
5274 instruction to assemble an integer of 16, 8, 4, 2 or 1 bytes,
5275 respectively, whose value is @var{value}. The argument @var{exp} will
5276 be an RTL expression which represents a constant value. Use
5277 @samp{output_addr_const (@var{stream}, @var{exp})} to output this value
5278 as an assembler expression.@refill
5280 For sizes larger than @code{UNITS_PER_WORD}, if the action of a macro
5281 would be identical to repeatedly calling the macro corresponding to
5282 a size of @code{UNITS_PER_WORD}, once for each word, you need not define
5285 @findex ASM_OUTPUT_BYTE
5286 @item ASM_OUTPUT_BYTE (@var{stream}, @var{value})
5287 A C statement to output to the stdio stream @var{stream} an assembler
5288 instruction to assemble a single byte containing the number @var{value}.
5292 A C string constant giving the pseudo-op to use for a sequence of
5293 single-byte constants. If this macro is not defined, the default is
5296 @findex ASM_OUTPUT_ASCII
5297 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5298 A C statement to output to the stdio stream @var{stream} an assembler
5299 instruction to assemble a string constant containing the @var{len}
5300 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5301 @code{char *} and @var{len} a C expression of type @code{int}.
5303 If the assembler has a @code{.ascii} pseudo-op as found in the
5304 Berkeley Unix assembler, do not define the macro
5305 @code{ASM_OUTPUT_ASCII}.
5307 @findex CONSTANT_POOL_BEFORE_FUNCTION
5308 @item CONSTANT_POOL_BEFORE_FUNCTION
5309 You may define this macro as a C expression. You should define the
5310 expression to have a non-zero value if GNU CC should output the constant
5311 pool for a function before the code for the function, or a zero value if
5312 GNU CC should output the constant pool after the function. If you do
5313 not define this macro, the usual case, GNU CC will output the constant
5314 pool before the function.
5316 @findex ASM_OUTPUT_POOL_PROLOGUE
5317 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5318 A C statement to output assembler commands to define the start of the
5319 constant pool for a function. @var{funname} is a string giving
5320 the name of the function. Should the return type of the function
5321 be required, it can be obtained via @var{fundecl}. @var{size}
5322 is the size, in bytes, of the constant pool that will be written
5323 immediately after this call.
5325 If no constant-pool prefix is required, the usual case, this macro need
5328 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
5329 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5330 A C statement (with or without semicolon) to output a constant in the
5331 constant pool, if it needs special treatment. (This macro need not do
5332 anything for RTL expressions that can be output normally.)
5334 The argument @var{file} is the standard I/O stream to output the
5335 assembler code on. @var{x} is the RTL expression for the constant to
5336 output, and @var{mode} is the machine mode (in case @var{x} is a
5337 @samp{const_int}). @var{align} is the required alignment for the value
5338 @var{x}; you should output an assembler directive to force this much
5341 The argument @var{labelno} is a number to use in an internal label for
5342 the address of this pool entry. The definition of this macro is
5343 responsible for outputting the label definition at the proper place.
5344 Here is how to do this:
5347 ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
5350 When you output a pool entry specially, you should end with a
5351 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5352 entry from being output a second time in the usual manner.
5354 You need not define this macro if it would do nothing.
5356 @findex CONSTANT_AFTER_FUNCTION_P
5357 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
5358 Define this macro as a C expression which is nonzero if the constant
5359 @var{exp}, of type @code{tree}, should be output after the code for a
5360 function. The compiler will normally output all constants before the
5361 function; you need not define this macro if this is OK.
5363 @findex ASM_OUTPUT_POOL_EPILOGUE
5364 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5365 A C statement to output assembler commands to at the end of the constant
5366 pool for a function. @var{funname} is a string giving the name of the
5367 function. Should the return type of the function be required, you can
5368 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5369 constant pool that GNU CC wrote immediately before this call.
5371 If no constant-pool epilogue is required, the usual case, you need not
5374 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
5375 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
5376 Define this macro as a C expression which is nonzero if @var{C} is
5377 used as a logical line separator by the assembler.
5379 If you do not define this macro, the default is that only
5380 the character @samp{;} is treated as a logical line separator.
5383 @findex ASM_OPEN_PAREN
5384 @findex ASM_CLOSE_PAREN
5385 @item ASM_OPEN_PAREN
5386 @itemx ASM_CLOSE_PAREN
5387 These macros are defined as C string constant, describing the syntax
5388 in the assembler for grouping arithmetic expressions. The following
5389 definitions are correct for most assemblers:
5392 #define ASM_OPEN_PAREN "("
5393 #define ASM_CLOSE_PAREN ")"
5397 These macros are provided by @file{real.h} for writing the definitions
5398 of @code{ASM_OUTPUT_DOUBLE} and the like:
5401 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5402 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5403 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5404 @findex REAL_VALUE_TO_TARGET_SINGLE
5405 @findex REAL_VALUE_TO_TARGET_DOUBLE
5406 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
5407 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
5408 floating point representation, and store its bit pattern in the array of
5409 @code{long int} whose address is @var{l}. The number of elements in the
5410 output array is determined by the size of the desired target floating
5411 point data type: 32 bits of it go in each @code{long int} array
5412 element. Each array element holds 32 bits of the result, even if
5413 @code{long int} is wider than 32 bits on the host machine.
5415 The array element values are designed so that you can print them out
5416 using @code{fprintf} in the order they should appear in the target
5419 @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
5420 @findex REAL_VALUE_TO_DECIMAL
5421 This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
5422 decimal number and stores it as a string into @var{string}.
5423 You must pass, as @var{string}, the address of a long enough block
5424 of space to hold the result.
5426 The argument @var{format} is a @code{printf}-specification that serves
5427 as a suggestion for how to format the output string.
5430 @node Uninitialized Data
5431 @subsection Output of Uninitialized Variables
5433 Each of the macros in this section is used to do the whole job of
5434 outputting a single uninitialized variable.
5437 @findex ASM_OUTPUT_COMMON
5438 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5439 A C statement (sans semicolon) to output to the stdio stream
5440 @var{stream} the assembler definition of a common-label named
5441 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5442 is the size rounded up to whatever alignment the caller wants.
5444 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5445 output the name itself; before and after that, output the additional
5446 assembler syntax for defining the name, and a newline.
5448 This macro controls how the assembler definitions of uninitialized
5449 common global variables are output.
5451 @findex ASM_OUTPUT_ALIGNED_COMMON
5452 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5453 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5454 separate, explicit argument. If you define this macro, it is used in
5455 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5456 handling the required alignment of the variable. The alignment is specified
5457 as the number of bits.
5459 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
5460 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5461 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5462 variable to be output, if there is one, or @code{NULL_TREE} if there
5463 is not corresponding variable. If you define this macro, GNU CC wil use it
5464 in place of both @code{ASM_OUTPUT_COMMON} and
5465 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5466 the variable's decl in order to chose what to output.
5468 @findex ASM_OUTPUT_SHARED_COMMON
5469 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5470 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
5471 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
5474 @findex ASM_OUTPUT_BSS
5475 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5476 A C statement (sans semicolon) to output to the stdio stream
5477 @var{stream} the assembler definition of uninitialized global @var{decl} named
5478 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5479 is the size rounded up to whatever alignment the caller wants.
5481 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
5482 defining this macro. If unable, use the expression
5483 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5484 before and after that, output the additional assembler syntax for defining
5485 the name, and a newline.
5487 This macro controls how the assembler definitions of uninitialized global
5488 variables are output. This macro exists to properly support languages like
5489 @code{c++} which do not have @code{common} data. However, this macro currently
5490 is not defined for all targets. If this macro and
5491 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
5492 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
5493 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
5495 @findex ASM_OUTPUT_ALIGNED_BSS
5496 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5497 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
5498 separate, explicit argument. If you define this macro, it is used in
5499 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
5500 handling the required alignment of the variable. The alignment is specified
5501 as the number of bits.
5503 Try to use function @code{asm_output_aligned_bss} defined in file
5504 @file{varasm.c} when defining this macro.
5506 @findex ASM_OUTPUT_SHARED_BSS
5507 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5508 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
5509 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
5512 @findex ASM_OUTPUT_LOCAL
5513 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5514 A C statement (sans semicolon) to output to the stdio stream
5515 @var{stream} the assembler definition of a local-common-label named
5516 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5517 is the size rounded up to whatever alignment the caller wants.
5519 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5520 output the name itself; before and after that, output the additional
5521 assembler syntax for defining the name, and a newline.
5523 This macro controls how the assembler definitions of uninitialized
5524 static variables are output.
5526 @findex ASM_OUTPUT_ALIGNED_LOCAL
5527 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5528 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5529 separate, explicit argument. If you define this macro, it is used in
5530 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5531 handling the required alignment of the variable. The alignment is specified
5532 as the number of bits.
5534 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
5535 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5536 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5537 variable to be output, if there is one, or @code{NULL_TREE} if there
5538 is not corresponding variable. If you define this macro, GNU CC wil use it
5539 in place of both @code{ASM_OUTPUT_DECL} and
5540 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5541 the variable's decl in order to chose what to output.
5544 @findex ASM_OUTPUT_SHARED_LOCAL
5545 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5546 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
5547 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
5552 @subsection Output and Generation of Labels
5554 @c prevent bad page break with this line
5555 This is about outputting labels.
5558 @findex ASM_OUTPUT_LABEL
5559 @findex assemble_name
5560 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5561 A C statement (sans semicolon) to output to the stdio stream
5562 @var{stream} the assembler definition of a label named @var{name}.
5563 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5564 output the name itself; before and after that, output the additional
5565 assembler syntax for defining the name, and a newline.
5567 @findex ASM_DECLARE_FUNCTION_NAME
5568 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5569 A C statement (sans semicolon) to output to the stdio stream
5570 @var{stream} any text necessary for declaring the name @var{name} of a
5571 function which is being defined. This macro is responsible for
5572 outputting the label definition (perhaps using
5573 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
5574 @code{FUNCTION_DECL} tree node representing the function.
5576 If this macro is not defined, then the function name is defined in the
5577 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5579 @findex ASM_DECLARE_FUNCTION_SIZE
5580 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5581 A C statement (sans semicolon) to output to the stdio stream
5582 @var{stream} any text necessary for declaring the size of a function
5583 which is being defined. The argument @var{name} is the name of the
5584 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5585 representing the function.
5587 If this macro is not defined, then the function size is not defined.
5589 @findex ASM_DECLARE_OBJECT_NAME
5590 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5591 A C statement (sans semicolon) to output to the stdio stream
5592 @var{stream} any text necessary for declaring the name @var{name} of an
5593 initialized variable which is being defined. This macro must output the
5594 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
5595 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
5597 If this macro is not defined, then the variable name is defined in the
5598 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5600 @findex ASM_FINISH_DECLARE_OBJECT
5601 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5602 A C statement (sans semicolon) to finish up declaring a variable name
5603 once the compiler has processed its initializer fully and thus has had a
5604 chance to determine the size of an array when controlled by an
5605 initializer. This is used on systems where it's necessary to declare
5606 something about the size of the object.
5608 If you don't define this macro, that is equivalent to defining it to do
5611 @findex ASM_GLOBALIZE_LABEL
5612 @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
5613 A C statement (sans semicolon) to output to the stdio stream
5614 @var{stream} some commands that will make the label @var{name} global;
5615 that is, available for reference from other files. Use the expression
5616 @code{assemble_name (@var{stream}, @var{name})} to output the name
5617 itself; before and after that, output the additional assembler syntax
5618 for making that name global, and a newline.
5620 @findex ASM_WEAKEN_LABEL
5621 @item ASM_WEAKEN_LABEL
5622 A C statement (sans semicolon) to output to the stdio stream
5623 @var{stream} some commands that will make the label @var{name} weak;
5624 that is, available for reference from other files but only used if
5625 no other definition is available. Use the expression
5626 @code{assemble_name (@var{stream}, @var{name})} to output the name
5627 itself; before and after that, output the additional assembler syntax
5628 for making that name weak, and a newline.
5630 If you don't define this macro, GNU CC will not support weak
5631 symbols and you should not define the @code{SUPPORTS_WEAK} macro.
5633 @findex SUPPORTS_WEAK
5635 A C expression which evaluates to true if the target supports weak symbols.
5637 If you don't define this macro, @file{defaults.h} provides a default
5638 definition. If @code{ASM_WEAKEN_LABEL} is defined, the default
5639 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5640 you want to control weak symbol support with a compiler flag such as
5643 @findex MAKE_DECL_ONE_ONLY (@var{decl})
5644 @item MAKE_DECL_ONE_ONLY
5645 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5646 public symbol such that extra copies in multiple translation units will
5647 be discarded by the linker. Define this macro if your object file
5648 format provides support for this concept, such as the @samp{COMDAT}
5649 section flags in the Microsoft Windows PE/COFF format, and this support
5650 requires changes to @var{decl}, such as putting it in a separate section.
5652 @findex SUPPORTS_ONE_ONLY
5653 @item SUPPORTS_ONE_ONLY
5654 A C expression which evaluates to true if the target supports one-only
5657 If you don't define this macro, @file{varasm.c} provides a default
5658 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
5659 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5660 you want to control one-only symbol support with a compiler flag, or if
5661 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
5662 be emitted as one-only.
5664 @findex ASM_OUTPUT_EXTERNAL
5665 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
5666 A C statement (sans semicolon) to output to the stdio stream
5667 @var{stream} any text necessary for declaring the name of an external
5668 symbol named @var{name} which is referenced in this compilation but
5669 not defined. The value of @var{decl} is the tree node for the
5672 This macro need not be defined if it does not need to output anything.
5673 The GNU assembler and most Unix assemblers don't require anything.
5675 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
5676 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
5677 A C statement (sans semicolon) to output on @var{stream} an assembler
5678 pseudo-op to declare a library function name external. The name of the
5679 library function is given by @var{symref}, which has type @code{rtx} and
5680 is a @code{symbol_ref}.
5682 This macro need not be defined if it does not need to output anything.
5683 The GNU assembler and most Unix assemblers don't require anything.
5685 @findex ASM_OUTPUT_LABELREF
5686 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
5687 A C statement (sans semicolon) to output to the stdio stream
5688 @var{stream} a reference in assembler syntax to a label named
5689 @var{name}. This should add @samp{_} to the front of the name, if that
5690 is customary on your operating system, as it is in most Berkeley Unix
5691 systems. This macro is used in @code{assemble_name}.
5693 @ignore @c Seems not to exist anymore.
5694 @findex ASM_OUTPUT_LABELREF_AS_INT
5695 @item ASM_OUTPUT_LABELREF_AS_INT (@var{file}, @var{label})
5696 Define this macro for systems that use the program @code{collect2}.
5697 The definition should be a C statement to output a word containing
5698 a reference to the label @var{label}.
5701 @findex ASM_OUTPUT_INTERNAL_LABEL
5702 @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
5703 A C statement to output to the stdio stream @var{stream} a label whose
5704 name is made from the string @var{prefix} and the number @var{num}.
5706 It is absolutely essential that these labels be distinct from the labels
5707 used for user-level functions and variables. Otherwise, certain programs
5708 will have name conflicts with internal labels.
5710 It is desirable to exclude internal labels from the symbol table of the
5711 object file. Most assemblers have a naming convention for labels that
5712 should be excluded; on many systems, the letter @samp{L} at the
5713 beginning of a label has this effect. You should find out what
5714 convention your system uses, and follow it.
5716 The usual definition of this macro is as follows:
5719 fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
5722 @findex ASM_GENERATE_INTERNAL_LABEL
5723 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
5724 A C statement to store into the string @var{string} a label whose name
5725 is made from the string @var{prefix} and the number @var{num}.
5727 This string, when output subsequently by @code{assemble_name}, should
5728 produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
5729 with the same @var{prefix} and @var{num}.
5731 If the string begins with @samp{*}, then @code{assemble_name} will
5732 output the rest of the string unchanged. It is often convenient for
5733 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
5734 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
5735 to output the string, and may change it. (Of course,
5736 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
5737 you should know what it does on your machine.)
5739 @findex ASM_FORMAT_PRIVATE_NAME
5740 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
5741 A C expression to assign to @var{outvar} (which is a variable of type
5742 @code{char *}) a newly allocated string made from the string
5743 @var{name} and the number @var{number}, with some suitable punctuation
5744 added. Use @code{alloca} to get space for the string.
5746 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
5747 produce an assembler label for an internal static variable whose name is
5748 @var{name}. Therefore, the string must be such as to result in valid
5749 assembler code. The argument @var{number} is different each time this
5750 macro is executed; it prevents conflicts between similarly-named
5751 internal static variables in different scopes.
5753 Ideally this string should not be a valid C identifier, to prevent any
5754 conflict with the user's own symbols. Most assemblers allow periods
5755 or percent signs in assembler symbols; putting at least one of these
5756 between the name and the number will suffice.
5758 @findex ASM_OUTPUT_DEF
5759 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
5760 A C statement to output to the stdio stream @var{stream} assembler code
5761 which defines (equates) the symbol @var{name} to have the value @var{value}.
5763 If SET_ASM_OP is defined, a default definition is provided which is
5764 correct for most systems.
5766 @findex ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL
5767 @item ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL (@var{stream}, @var{symbol}, @var{high}, @var{low})
5768 A C statement to output to the stdio stream @var{stream} assembler code
5769 which defines (equates) the symbol @var{symbol} to have a value equal to
5770 the difference of the two symbols @var{high} and @var{low}, i.e.
5771 @var{high} minus @var{low}. GNU CC guarantees that the symbols @var{high}
5772 and @var{low} are already known by the assembler so that the difference
5773 resolves into a constant.
5775 If SET_ASM_OP is defined, a default definition is provided which is
5776 correct for most systems.
5778 @findex ASM_OUTPUT_WEAK_ALIAS
5779 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
5780 A C statement to output to the stdio stream @var{stream} assembler code
5781 which defines (equates) the weak symbol @var{name} to have the value
5784 Define this macro if the target only supports weak aliases; define
5785 ASM_OUTPUT_DEF instead if possible.
5787 @findex OBJC_GEN_METHOD_LABEL
5788 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
5789 Define this macro to override the default assembler names used for
5790 Objective C methods.
5792 The default name is a unique method number followed by the name of the
5793 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
5794 the category is also included in the assembler name (e.g.@:
5797 These names are safe on most systems, but make debugging difficult since
5798 the method's selector is not present in the name. Therefore, particular
5799 systems define other ways of computing names.
5801 @var{buf} is an expression of type @code{char *} which gives you a
5802 buffer in which to store the name; its length is as long as
5803 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
5804 50 characters extra.
5806 The argument @var{is_inst} specifies whether the method is an instance
5807 method or a class method; @var{class_name} is the name of the class;
5808 @var{cat_name} is the name of the category (or NULL if the method is not
5809 in a category); and @var{sel_name} is the name of the selector.
5811 On systems where the assembler can handle quoted names, you can use this
5812 macro to provide more human-readable names.
5815 @node Initialization
5816 @subsection How Initialization Functions Are Handled
5817 @cindex initialization routines
5818 @cindex termination routines
5819 @cindex constructors, output of
5820 @cindex destructors, output of
5822 The compiled code for certain languages includes @dfn{constructors}
5823 (also called @dfn{initialization routines})---functions to initialize
5824 data in the program when the program is started. These functions need
5825 to be called before the program is ``started''---that is to say, before
5826 @code{main} is called.
5828 Compiling some languages generates @dfn{destructors} (also called
5829 @dfn{termination routines}) that should be called when the program
5832 To make the initialization and termination functions work, the compiler
5833 must output something in the assembler code to cause those functions to
5834 be called at the appropriate time. When you port the compiler to a new
5835 system, you need to specify how to do this.
5837 There are two major ways that GCC currently supports the execution of
5838 initialization and termination functions. Each way has two variants.
5839 Much of the structure is common to all four variations.
5841 @findex __CTOR_LIST__
5842 @findex __DTOR_LIST__
5843 The linker must build two lists of these functions---a list of
5844 initialization functions, called @code{__CTOR_LIST__}, and a list of
5845 termination functions, called @code{__DTOR_LIST__}.
5847 Each list always begins with an ignored function pointer (which may hold
5848 0, @minus{}1, or a count of the function pointers after it, depending on
5849 the environment). This is followed by a series of zero or more function
5850 pointers to constructors (or destructors), followed by a function
5851 pointer containing zero.
5853 Depending on the operating system and its executable file format, either
5854 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
5855 time and exit time. Constructors are called in reverse order of the
5856 list; destructors in forward order.
5858 The best way to handle static constructors works only for object file
5859 formats which provide arbitrarily-named sections. A section is set
5860 aside for a list of constructors, and another for a list of destructors.
5861 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
5862 object file that defines an initialization function also puts a word in
5863 the constructor section to point to that function. The linker
5864 accumulates all these words into one contiguous @samp{.ctors} section.
5865 Termination functions are handled similarly.
5867 To use this method, you need appropriate definitions of the macros
5868 @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}. Usually
5869 you can get them by including @file{svr4.h}.
5871 When arbitrary sections are available, there are two variants, depending
5872 upon how the code in @file{crtstuff.c} is called. On systems that
5873 support an @dfn{init} section which is executed at program startup,
5874 parts of @file{crtstuff.c} are compiled into that section. The
5875 program is linked by the @code{gcc} driver like this:
5878 ld -o @var{output_file} crtbegin.o @dots{} crtend.o -lgcc
5881 The head of a function (@code{__do_global_ctors}) appears in the init
5882 section of @file{crtbegin.o}; the remainder of the function appears in
5883 the init section of @file{crtend.o}. The linker will pull these two
5884 parts of the section together, making a whole function. If any of the
5885 user's object files linked into the middle of it contribute code, then that
5886 code will be executed as part of the body of @code{__do_global_ctors}.
5888 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
5891 If no init section is available, do not define
5892 @code{INIT_SECTION_ASM_OP}. Then @code{__do_global_ctors} is built into
5893 the text section like all other functions, and resides in
5894 @file{libgcc.a}. When GCC compiles any function called @code{main}, it
5895 inserts a procedure call to @code{__main} as the first executable code
5896 after the function prologue. The @code{__main} function, also defined
5897 in @file{libgcc2.c}, simply calls @file{__do_global_ctors}.
5899 In file formats that don't support arbitrary sections, there are again
5900 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
5901 and an `a.out' format must be used. In this case,
5902 @code{ASM_OUTPUT_CONSTRUCTOR} is defined to produce a @code{.stabs}
5903 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
5904 and with the address of the void function containing the initialization
5905 code as its value. The GNU linker recognizes this as a request to add
5906 the value to a ``set''; the values are accumulated, and are eventually
5907 placed in the executable as a vector in the format described above, with
5908 a leading (ignored) count and a trailing zero element.
5909 @code{ASM_OUTPUT_DESTRUCTOR} is handled similarly. Since no init
5910 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
5911 the compilation of @code{main} to call @code{__main} as above, starting
5912 the initialization process.
5914 The last variant uses neither arbitrary sections nor the GNU linker.
5915 This is preferable when you want to do dynamic linking and when using
5916 file formats which the GNU linker does not support, such as `ECOFF'. In
5917 this case, @code{ASM_OUTPUT_CONSTRUCTOR} does not produce an
5918 @code{N_SETT} symbol; initialization and termination functions are
5919 recognized simply by their names. This requires an extra program in the
5920 linkage step, called @code{collect2}. This program pretends to be the
5921 linker, for use with GNU CC; it does its job by running the ordinary
5922 linker, but also arranges to include the vectors of initialization and
5923 termination functions. These functions are called via @code{__main} as
5926 Choosing among these configuration options has been simplified by a set
5927 of operating-system-dependent files in the @file{config} subdirectory.
5928 These files define all of the relevant parameters. Usually it is
5929 sufficient to include one into your specific machine-dependent
5930 configuration file. These files are:
5934 For operating systems using the `a.out' format.
5937 For operating systems using the `MachO' format.
5940 For System V Release 3 and similar systems using `COFF' format.
5943 For System V Release 4 and similar systems using `ELF' format.
5946 For the VMS operating system.
5950 The following section describes the specific macros that control and
5951 customize the handling of initialization and termination functions.
5954 @node Macros for Initialization
5955 @subsection Macros Controlling Initialization Routines
5957 Here are the macros that control how the compiler handles initialization
5958 and termination functions:
5961 @findex INIT_SECTION_ASM_OP
5962 @item INIT_SECTION_ASM_OP
5963 If defined, a C string constant for the assembler operation to identify
5964 the following data as initialization code. If not defined, GNU CC will
5965 assume such a section does not exist. When you are using special
5966 sections for initialization and termination functions, this macro also
5967 controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to run the
5968 initialization functions.
5970 @item HAS_INIT_SECTION
5971 @findex HAS_INIT_SECTION
5972 If defined, @code{main} will not call @code{__main} as described above.
5973 This macro should be defined for systems that control the contents of the
5974 init section on a symbol-by-symbol basis, such as OSF/1, and should not
5975 be defined explicitly for systems that support
5976 @code{INIT_SECTION_ASM_OP}.
5978 @item LD_INIT_SWITCH
5979 @findex LD_INIT_SWITCH
5980 If defined, a C string constant for a switch that tells the linker that
5981 the following symbol is an initialization routine.
5983 @item LD_FINI_SWITCH
5984 @findex LD_FINI_SWITCH
5985 If defined, a C string constant for a switch that tells the linker that
5986 the following symbol is a finalization routine.
5989 @findex INVOKE__main
5990 If defined, @code{main} will call @code{__main} despite the presence of
5991 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
5992 where the init section is not actually run automatically, but is still
5993 useful for collecting the lists of constructors and destructors.
5995 @item ASM_OUTPUT_CONSTRUCTOR (@var{stream}, @var{name})
5996 @findex ASM_OUTPUT_CONSTRUCTOR
5997 Define this macro as a C statement to output on the stream @var{stream}
5998 the assembler code to arrange to call the function named @var{name} at
5999 initialization time.
6001 Assume that @var{name} is the name of a C function generated
6002 automatically by the compiler. This function takes no arguments. Use
6003 the function @code{assemble_name} to output the name @var{name}; this
6004 performs any system-specific syntactic transformations such as adding an
6007 If you don't define this macro, nothing special is output to arrange to
6008 call the function. This is correct when the function will be called in
6009 some other manner---for example, by means of the @code{collect2} program,
6010 which looks through the symbol table to find these functions by their
6013 @item ASM_OUTPUT_DESTRUCTOR (@var{stream}, @var{name})
6014 @findex ASM_OUTPUT_DESTRUCTOR
6015 This is like @code{ASM_OUTPUT_CONSTRUCTOR} but used for termination
6016 functions rather than initialization functions.
6018 When @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR} are
6019 defined, the initializaiton routine generated for the generated object
6020 file will have static linkage.
6023 If your system uses @code{collect2} as the means of processing
6024 constructors, then that program normally uses @code{nm} to scan an
6025 object file for constructor functions to be called. On such systems you
6026 must not define @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}
6027 as the object file's initialization routine must have global scope.
6029 On certain kinds of systems, you can define these macros to make
6030 @code{collect2} work faster (and, in some cases, make it work at all):
6033 @findex OBJECT_FORMAT_COFF
6034 @item OBJECT_FORMAT_COFF
6035 Define this macro if the system uses COFF (Common Object File Format)
6036 object files, so that @code{collect2} can assume this format and scan
6037 object files directly for dynamic constructor/destructor functions.
6039 @findex OBJECT_FORMAT_ROSE
6040 @item OBJECT_FORMAT_ROSE
6041 Define this macro if the system uses ROSE format object files, so that
6042 @code{collect2} can assume this format and scan object files directly
6043 for dynamic constructor/destructor functions.
6045 These macros are effective only in a native compiler; @code{collect2} as
6046 part of a cross compiler always uses @code{nm} for the target machine.
6048 @findex REAL_NM_FILE_NAME
6049 @item REAL_NM_FILE_NAME
6050 Define this macro as a C string constant containing the file name to use
6051 to execute @code{nm}. The default is to search the path normally for
6054 If your system supports shared libraries and has a program to list the
6055 dynamic dependencies of a given library or executable, you can define
6056 these macros to enable support for running initialization and
6057 termination functions in shared libraries:
6061 Define this macro to a C string constant containing the name of the
6062 program which lists dynamic dependencies, like @code{"ldd"} under SunOS 4.
6064 @findex PARSE_LDD_OUTPUT
6065 @item PARSE_LDD_OUTPUT (@var{PTR})
6066 Define this macro to be C code that extracts filenames from the output
6067 of the program denoted by @code{LDD_SUFFIX}. @var{PTR} is a variable
6068 of type @code{char *} that points to the beginning of a line of output
6069 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6070 code must advance @var{PTR} to the beginning of the filename on that
6071 line. Otherwise, it must set @var{PTR} to @code{NULL}.
6075 @node Instruction Output
6076 @subsection Output of Assembler Instructions
6078 @c prevent bad page break with this line
6079 This describes assembler instruction output.
6082 @findex REGISTER_NAMES
6083 @item REGISTER_NAMES
6084 A C initializer containing the assembler's names for the machine
6085 registers, each one as a C string constant. This is what translates
6086 register numbers in the compiler into assembler language.
6088 @findex ADDITIONAL_REGISTER_NAMES
6089 @item ADDITIONAL_REGISTER_NAMES
6090 If defined, a C initializer for an array of structures containing a name
6091 and a register number. This macro defines additional names for hard
6092 registers, thus allowing the @code{asm} option in declarations to refer
6093 to registers using alternate names.
6095 @findex ASM_OUTPUT_OPCODE
6096 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6097 Define this macro if you are using an unusual assembler that
6098 requires different names for the machine instructions.
6100 The definition is a C statement or statements which output an
6101 assembler instruction opcode to the stdio stream @var{stream}. The
6102 macro-operand @var{ptr} is a variable of type @code{char *} which
6103 points to the opcode name in its ``internal'' form---the form that is
6104 written in the machine description. The definition should output the
6105 opcode name to @var{stream}, performing any translation you desire, and
6106 increment the variable @var{ptr} to point at the end of the opcode
6107 so that it will not be output twice.
6109 In fact, your macro definition may process less than the entire opcode
6110 name, or more than the opcode name; but if you want to process text
6111 that includes @samp{%}-sequences to substitute operands, you must take
6112 care of the substitution yourself. Just be sure to increment
6113 @var{ptr} over whatever text should not be output normally.
6115 @findex recog_operand
6116 If you need to look at the operand values, they can be found as the
6117 elements of @code{recog_operand}.
6119 If the macro definition does nothing, the instruction is output
6122 @findex FINAL_PRESCAN_INSN
6123 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6124 If defined, a C statement to be executed just prior to the output of
6125 assembler code for @var{insn}, to modify the extracted operands so
6126 they will be output differently.
6128 Here the argument @var{opvec} is the vector containing the operands
6129 extracted from @var{insn}, and @var{noperands} is the number of
6130 elements of the vector which contain meaningful data for this insn.
6131 The contents of this vector are what will be used to convert the insn
6132 template into assembler code, so you can change the assembler output
6133 by changing the contents of the vector.
6135 This macro is useful when various assembler syntaxes share a single
6136 file of instruction patterns; by defining this macro differently, you
6137 can cause a large class of instructions to be output differently (such
6138 as with rearranged operands). Naturally, variations in assembler
6139 syntax affecting individual insn patterns ought to be handled by
6140 writing conditional output routines in those patterns.
6142 If this macro is not defined, it is equivalent to a null statement.
6144 @findex FINAL_PRESCAN_LABEL
6145 @item FINAL_PRESCAN_LABEL
6146 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
6147 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
6148 @var{noperands} will be zero.
6150 @findex PRINT_OPERAND
6151 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6152 A C compound statement to output to stdio stream @var{stream} the
6153 assembler syntax for an instruction operand @var{x}. @var{x} is an
6156 @var{code} is a value that can be used to specify one of several ways
6157 of printing the operand. It is used when identical operands must be
6158 printed differently depending on the context. @var{code} comes from
6159 the @samp{%} specification that was used to request printing of the
6160 operand. If the specification was just @samp{%@var{digit}} then
6161 @var{code} is 0; if the specification was @samp{%@var{ltr}
6162 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6165 If @var{x} is a register, this macro should print the register's name.
6166 The names can be found in an array @code{reg_names} whose type is
6167 @code{char *[]}. @code{reg_names} is initialized from
6168 @code{REGISTER_NAMES}.
6170 When the machine description has a specification @samp{%@var{punct}}
6171 (a @samp{%} followed by a punctuation character), this macro is called
6172 with a null pointer for @var{x} and the punctuation character for
6175 @findex PRINT_OPERAND_PUNCT_VALID_P
6176 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6177 A C expression which evaluates to true if @var{code} is a valid
6178 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6179 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6180 punctuation characters (except for the standard one, @samp{%}) are used
6183 @findex PRINT_OPERAND_ADDRESS
6184 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6185 A C compound statement to output to stdio stream @var{stream} the
6186 assembler syntax for an instruction operand that is a memory reference
6187 whose address is @var{x}. @var{x} is an RTL expression.
6189 @cindex @code{ENCODE_SECTION_INFO} usage
6190 On some machines, the syntax for a symbolic address depends on the
6191 section that the address refers to. On these machines, define the macro
6192 @code{ENCODE_SECTION_INFO} to store the information into the
6193 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
6195 @findex DBR_OUTPUT_SEQEND
6196 @findex dbr_sequence_length
6197 @item DBR_OUTPUT_SEQEND(@var{file})
6198 A C statement, to be executed after all slot-filler instructions have
6199 been output. If necessary, call @code{dbr_sequence_length} to
6200 determine the number of slots filled in a sequence (zero if not
6201 currently outputting a sequence), to decide how many no-ops to output,
6204 Don't define this macro if it has nothing to do, but it is helpful in
6205 reading assembly output if the extent of the delay sequence is made
6206 explicit (e.g. with white space).
6208 @findex final_sequence
6209 Note that output routines for instructions with delay slots must be
6210 prepared to deal with not being output as part of a sequence (i.e.
6211 when the scheduling pass is not run, or when no slot fillers could be
6212 found.) The variable @code{final_sequence} is null when not
6213 processing a sequence, otherwise it contains the @code{sequence} rtx
6216 @findex REGISTER_PREFIX
6217 @findex LOCAL_LABEL_PREFIX
6218 @findex USER_LABEL_PREFIX
6219 @findex IMMEDIATE_PREFIX
6221 @item REGISTER_PREFIX
6222 @itemx LOCAL_LABEL_PREFIX
6223 @itemx USER_LABEL_PREFIX
6224 @itemx IMMEDIATE_PREFIX
6225 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6226 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6227 @file{final.c}). These are useful when a single @file{md} file must
6228 support multiple assembler formats. In that case, the various @file{tm.h}
6229 files can define these macros differently.
6231 @findex ASSEMBLER_DIALECT
6232 @item ASSEMBLER_DIALECT
6233 If your target supports multiple dialects of assembler language (such as
6234 different opcodes), define this macro as a C expression that gives the
6235 numeric index of the assembler language dialect to use, with zero as the
6238 If this macro is defined, you may use constructs of the form
6239 @samp{@{option0|option1|option2@dots{}@}} in the output
6240 templates of patterns (@pxref{Output Template}) or in the first argument
6241 of @code{asm_fprintf}. This construct outputs @samp{option0},
6242 @samp{option1} or @samp{option2}, etc., if the value of
6243 @code{ASSEMBLER_DIALECT} is zero, one or two, etc. Any special
6244 characters within these strings retain their usual meaning.
6246 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6247 @samp{@}} do not have any special meaning when used in templates or
6248 operands to @code{asm_fprintf}.
6250 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6251 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6252 the variations in assembler language syntax with that mechanism. Define
6253 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6254 if the syntax variant are larger and involve such things as different
6255 opcodes or operand order.
6257 @findex ASM_OUTPUT_REG_PUSH
6258 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6259 A C expression to output to @var{stream} some assembler code
6260 which will push hard register number @var{regno} onto the stack.
6261 The code need not be optimal, since this macro is used only when
6264 @findex ASM_OUTPUT_REG_POP
6265 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6266 A C expression to output to @var{stream} some assembler code
6267 which will pop hard register number @var{regno} off of the stack.
6268 The code need not be optimal, since this macro is used only when
6272 @node Dispatch Tables
6273 @subsection Output of Dispatch Tables
6275 @c prevent bad page break with this line
6276 This concerns dispatch tables.
6279 @cindex dispatch table
6280 @findex ASM_OUTPUT_ADDR_DIFF_ELT
6281 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6282 A C statement to output to the stdio stream @var{stream} an assembler
6283 pseudo-instruction to generate a difference between two labels.
6284 @var{value} and @var{rel} are the numbers of two internal labels. The
6285 definitions of these labels are output using
6286 @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
6287 way here. For example,
6290 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6291 @var{value}, @var{rel})
6294 You must provide this macro on machines where the addresses in a
6295 dispatch table are relative to the table's own address. If defined, GNU
6296 CC will also use this macro on all machines when producing PIC.
6297 @var{body} is the body of the ADDR_DIFF_VEC; it is provided so that the
6298 mode and flags can be read.
6300 @findex ASM_OUTPUT_ADDR_VEC_ELT
6301 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6302 This macro should be provided on machines where the addresses
6303 in a dispatch table are absolute.
6305 The definition should be a C statement to output to the stdio stream
6306 @var{stream} an assembler pseudo-instruction to generate a reference to
6307 a label. @var{value} is the number of an internal label whose
6308 definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
6312 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6315 @findex ASM_OUTPUT_CASE_LABEL
6316 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6317 Define this if the label before a jump-table needs to be output
6318 specially. The first three arguments are the same as for
6319 @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
6320 jump-table which follows (a @code{jump_insn} containing an
6321 @code{addr_vec} or @code{addr_diff_vec}).
6323 This feature is used on system V to output a @code{swbeg} statement
6326 If this macro is not defined, these labels are output with
6327 @code{ASM_OUTPUT_INTERNAL_LABEL}.
6329 @findex ASM_OUTPUT_CASE_END
6330 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6331 Define this if something special must be output at the end of a
6332 jump-table. The definition should be a C statement to be executed
6333 after the assembler code for the table is written. It should write
6334 the appropriate code to stdio stream @var{stream}. The argument
6335 @var{table} is the jump-table insn, and @var{num} is the label-number
6336 of the preceding label.
6338 If this macro is not defined, nothing special is output at the end of
6342 @node Exception Region Output
6343 @subsection Assembler Commands for Exception Regions
6345 @c prevent bad page break with this line
6347 This describes commands marking the start and the end of an exception
6351 @findex ASM_OUTPUT_EH_REGION_BEG
6352 @item ASM_OUTPUT_EH_REGION_BEG ()
6353 A C expression to output text to mark the start of an exception region.
6355 This macro need not be defined on most platforms.
6357 @findex ASM_OUTPUT_EH_REGION_END
6358 @item ASM_OUTPUT_EH_REGION_END ()
6359 A C expression to output text to mark the end of an exception region.
6361 This macro need not be defined on most platforms.
6363 @findex EXCEPTION_SECTION
6364 @item EXCEPTION_SECTION ()
6365 A C expression to switch to the section in which the main
6366 exception table is to be placed (@pxref{Sections}). The default is a
6367 section named @code{.gcc_except_table} on machines that support named
6368 sections via @code{ASM_OUTPUT_SECTION_NAME}, otherwise if @samp{-fpic}
6369 or @samp{-fPIC} is in effect, the @code{data_section}, otherwise the
6370 @code{readonly_data_section}.
6372 @findex EH_FRAME_SECTION_ASM_OP
6373 @item EH_FRAME_SECTION_ASM_OP
6374 If defined, a C string constant for the assembler operation to switch to
6375 the section for exception handling frame unwind information. If not
6376 defined, GNU CC will provide a default definition if the target supports
6377 named sections. @file{crtstuff.c} uses this macro to switch to the
6378 appropriate section.
6380 You should define this symbol if your target supports DWARF 2 frame
6381 unwind information and the default definition does not work.
6383 @findex OMIT_EH_TABLE
6384 @item OMIT_EH_TABLE ()
6385 A C expression that is nonzero if the normal exception table output
6388 This macro need not be defined on most platforms.
6390 @findex EH_TABLE_LOOKUP
6391 @item EH_TABLE_LOOKUP ()
6392 Alternate runtime support for looking up an exception at runtime and
6393 finding the associated handler, if the default method won't work.
6395 This macro need not be defined on most platforms.
6397 @findex DOESNT_NEED_UNWINDER
6398 @item DOESNT_NEED_UNWINDER
6399 A C expression that decides whether or not the current function needs to
6400 have a function unwinder generated for it. See the file @code{except.c}
6401 for details on when to define this, and how.
6403 @findex MASK_RETURN_ADDR
6404 @item MASK_RETURN_ADDR
6405 An rtx used to mask the return address found via RETURN_ADDR_RTX, so
6406 that it does not contain any extraneous set bits in it.
6408 @findex DWARF2_UNWIND_INFO
6409 @item DWARF2_UNWIND_INFO
6410 Define this macro to 0 if your target supports DWARF 2 frame unwind
6411 information, but it does not yet work with exception handling.
6412 Otherwise, if your target supports this information (if it defines
6413 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
6414 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
6417 If this macro is defined to 1, the DWARF 2 unwinder will be the default
6418 exception handling mechanism; otherwise, setjmp/longjmp will be used by
6421 If this macro is defined to anything, the DWARF 2 unwinder will be used
6422 instead of inline unwinders and __unwind_function in the non-setjmp case.
6426 @node Alignment Output
6427 @subsection Assembler Commands for Alignment
6429 @c prevent bad page break with this line
6430 This describes commands for alignment.
6433 @findex LABEL_ALIGN_AFTER_BARRIER
6434 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
6435 The alignment (log base 2) to put in front of @var{label}, which follows
6438 This macro need not be defined if you don't want any special alignment
6439 to be done at such a time. Most machine descriptions do not currently
6443 @item LOOP_ALIGN (@var{label})
6444 The alignment (log base 2) to put in front of @var{label}, which follows
6445 a NOTE_INSN_LOOP_BEG note.
6447 This macro need not be defined if you don't want any special alignment
6448 to be done at such a time. Most machine descriptions do not currently
6452 @item LABEL_ALIGN (@var{label})
6453 The alignment (log base 2) to put in front of @var{label}.
6454 If LABEL_ALIGN_AFTER_BARRIER / LOOP_ALIGN specify a different alignment,
6455 the maximum of the specified values is used.
6457 @findex ASM_OUTPUT_SKIP
6458 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6459 A C statement to output to the stdio stream @var{stream} an assembler
6460 instruction to advance the location counter by @var{nbytes} bytes.
6461 Those bytes should be zero when loaded. @var{nbytes} will be a C
6462 expression of type @code{int}.
6464 @findex ASM_NO_SKIP_IN_TEXT
6465 @item ASM_NO_SKIP_IN_TEXT
6466 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6467 text section because it fails to put zeros in the bytes that are skipped.
6468 This is true on many Unix systems, where the pseudo--op to skip bytes
6469 produces no-op instructions rather than zeros when used in the text
6472 @findex ASM_OUTPUT_ALIGN
6473 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6474 A C statement to output to the stdio stream @var{stream} an assembler
6475 command to advance the location counter to a multiple of 2 to the
6476 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
6478 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
6479 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
6480 A C statement to output to the stdio stream @var{stream} an assembler
6481 command to advance the location counter to a multiple of 2 to the
6482 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
6483 satisfy the alignment request. @var{power} and @var{max_skip} will be
6484 a C expression of type @code{int}.
6488 @node Debugging Info
6489 @section Controlling Debugging Information Format
6491 @c prevent bad page break with this line
6492 This describes how to specify debugging information.
6495 * All Debuggers:: Macros that affect all debugging formats uniformly.
6496 * DBX Options:: Macros enabling specific options in DBX format.
6497 * DBX Hooks:: Hook macros for varying DBX format.
6498 * File Names and DBX:: Macros controlling output of file names in DBX format.
6499 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
6503 @subsection Macros Affecting All Debugging Formats
6505 @c prevent bad page break with this line
6506 These macros affect all debugging formats.
6509 @findex DBX_REGISTER_NUMBER
6510 @item DBX_REGISTER_NUMBER (@var{regno})
6511 A C expression that returns the DBX register number for the compiler
6512 register number @var{regno}. In simple cases, the value of this
6513 expression may be @var{regno} itself. But sometimes there are some
6514 registers that the compiler knows about and DBX does not, or vice
6515 versa. In such cases, some register may need to have one number in
6516 the compiler and another for DBX.
6518 If two registers have consecutive numbers inside GNU CC, and they can be
6519 used as a pair to hold a multiword value, then they @emph{must} have
6520 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
6521 Otherwise, debuggers will be unable to access such a pair, because they
6522 expect register pairs to be consecutive in their own numbering scheme.
6524 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
6525 does not preserve register pairs, then what you must do instead is
6526 redefine the actual register numbering scheme.
6528 @findex DEBUGGER_AUTO_OFFSET
6529 @item DEBUGGER_AUTO_OFFSET (@var{x})
6530 A C expression that returns the integer offset value for an automatic
6531 variable having address @var{x} (an RTL expression). The default
6532 computation assumes that @var{x} is based on the frame-pointer and
6533 gives the offset from the frame-pointer. This is required for targets
6534 that produce debugging output for DBX or COFF-style debugging output
6535 for SDB and allow the frame-pointer to be eliminated when the
6536 @samp{-g} options is used.
6538 @findex DEBUGGER_ARG_OFFSET
6539 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6540 A C expression that returns the integer offset value for an argument
6541 having address @var{x} (an RTL expression). The nominal offset is
6544 @findex PREFERRED_DEBUGGING_TYPE
6545 @item PREFERRED_DEBUGGING_TYPE
6546 A C expression that returns the type of debugging output GNU CC should
6547 produce when the user specifies just @samp{-g}. Define
6548 this if you have arranged for GNU CC to support more than one format of
6549 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
6550 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG}, and
6553 When the user specifies @samp{-ggdb}, GNU CC normally also uses the
6554 value of this macro to select the debugging output format, but with two
6555 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
6556 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GNU CC uses the
6557 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
6558 defined, GNU CC uses @code{DBX_DEBUG}.
6560 The value of this macro only affects the default debugging output; the
6561 user can always get a specific type of output by using @samp{-gstabs},
6562 @samp{-gcoff}, @samp{-gdwarf-1}, @samp{-gdwarf-2}, or @samp{-gxcoff}.
6566 @subsection Specific Options for DBX Output
6568 @c prevent bad page break with this line
6569 These are specific options for DBX output.
6572 @findex DBX_DEBUGGING_INFO
6573 @item DBX_DEBUGGING_INFO
6574 Define this macro if GNU CC should produce debugging output for DBX
6575 in response to the @samp{-g} option.
6577 @findex XCOFF_DEBUGGING_INFO
6578 @item XCOFF_DEBUGGING_INFO
6579 Define this macro if GNU CC should produce XCOFF format debugging output
6580 in response to the @samp{-g} option. This is a variant of DBX format.
6582 @findex DEFAULT_GDB_EXTENSIONS
6583 @item DEFAULT_GDB_EXTENSIONS
6584 Define this macro to control whether GNU CC should by default generate
6585 GDB's extended version of DBX debugging information (assuming DBX-format
6586 debugging information is enabled at all). If you don't define the
6587 macro, the default is 1: always generate the extended information
6588 if there is any occasion to.
6590 @findex DEBUG_SYMS_TEXT
6591 @item DEBUG_SYMS_TEXT
6592 Define this macro if all @code{.stabs} commands should be output while
6593 in the text section.
6595 @findex ASM_STABS_OP
6597 A C string constant naming the assembler pseudo op to use instead of
6598 @code{.stabs} to define an ordinary debugging symbol. If you don't
6599 define this macro, @code{.stabs} is used. This macro applies only to
6600 DBX debugging information format.
6602 @findex ASM_STABD_OP
6604 A C string constant naming the assembler pseudo op to use instead of
6605 @code{.stabd} to define a debugging symbol whose value is the current
6606 location. If you don't define this macro, @code{.stabd} is used.
6607 This macro applies only to DBX debugging information format.
6609 @findex ASM_STABN_OP
6611 A C string constant naming the assembler pseudo op to use instead of
6612 @code{.stabn} to define a debugging symbol with no name. If you don't
6613 define this macro, @code{.stabn} is used. This macro applies only to
6614 DBX debugging information format.
6616 @findex DBX_NO_XREFS
6618 Define this macro if DBX on your system does not support the construct
6619 @samp{xs@var{tagname}}. On some systems, this construct is used to
6620 describe a forward reference to a structure named @var{tagname}.
6621 On other systems, this construct is not supported at all.
6623 @findex DBX_CONTIN_LENGTH
6624 @item DBX_CONTIN_LENGTH
6625 A symbol name in DBX-format debugging information is normally
6626 continued (split into two separate @code{.stabs} directives) when it
6627 exceeds a certain length (by default, 80 characters). On some
6628 operating systems, DBX requires this splitting; on others, splitting
6629 must not be done. You can inhibit splitting by defining this macro
6630 with the value zero. You can override the default splitting-length by
6631 defining this macro as an expression for the length you desire.
6633 @findex DBX_CONTIN_CHAR
6634 @item DBX_CONTIN_CHAR
6635 Normally continuation is indicated by adding a @samp{\} character to
6636 the end of a @code{.stabs} string when a continuation follows. To use
6637 a different character instead, define this macro as a character
6638 constant for the character you want to use. Do not define this macro
6639 if backslash is correct for your system.
6641 @findex DBX_STATIC_STAB_DATA_SECTION
6642 @item DBX_STATIC_STAB_DATA_SECTION
6643 Define this macro if it is necessary to go to the data section before
6644 outputting the @samp{.stabs} pseudo-op for a non-global static
6647 @findex DBX_TYPE_DECL_STABS_CODE
6648 @item DBX_TYPE_DECL_STABS_CODE
6649 The value to use in the ``code'' field of the @code{.stabs} directive
6650 for a typedef. The default is @code{N_LSYM}.
6652 @findex DBX_STATIC_CONST_VAR_CODE
6653 @item DBX_STATIC_CONST_VAR_CODE
6654 The value to use in the ``code'' field of the @code{.stabs} directive
6655 for a static variable located in the text section. DBX format does not
6656 provide any ``right'' way to do this. The default is @code{N_FUN}.
6658 @findex DBX_REGPARM_STABS_CODE
6659 @item DBX_REGPARM_STABS_CODE
6660 The value to use in the ``code'' field of the @code{.stabs} directive
6661 for a parameter passed in registers. DBX format does not provide any
6662 ``right'' way to do this. The default is @code{N_RSYM}.
6664 @findex DBX_REGPARM_STABS_LETTER
6665 @item DBX_REGPARM_STABS_LETTER
6666 The letter to use in DBX symbol data to identify a symbol as a parameter
6667 passed in registers. DBX format does not customarily provide any way to
6668 do this. The default is @code{'P'}.
6670 @findex DBX_MEMPARM_STABS_LETTER
6671 @item DBX_MEMPARM_STABS_LETTER
6672 The letter to use in DBX symbol data to identify a symbol as a stack
6673 parameter. The default is @code{'p'}.
6675 @findex DBX_FUNCTION_FIRST
6676 @item DBX_FUNCTION_FIRST
6677 Define this macro if the DBX information for a function and its
6678 arguments should precede the assembler code for the function. Normally,
6679 in DBX format, the debugging information entirely follows the assembler
6682 @findex DBX_LBRAC_FIRST
6683 @item DBX_LBRAC_FIRST
6684 Define this macro if the @code{N_LBRAC} symbol for a block should
6685 precede the debugging information for variables and functions defined in
6686 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
6689 @findex DBX_BLOCKS_FUNCTION_RELATIVE
6690 @item DBX_BLOCKS_FUNCTION_RELATIVE
6691 Define this macro if the value of a symbol describing the scope of a
6692 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
6693 of the enclosing function. Normally, GNU C uses an absolute address.
6695 @findex DBX_USE_BINCL
6697 Define this macro if GNU C should generate @code{N_BINCL} and
6698 @code{N_EINCL} stabs for included header files, as on Sun systems. This
6699 macro also directs GNU C to output a type number as a pair of a file
6700 number and a type number within the file. Normally, GNU C does not
6701 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
6702 number for a type number.
6706 @subsection Open-Ended Hooks for DBX Format
6708 @c prevent bad page break with this line
6709 These are hooks for DBX format.
6712 @findex DBX_OUTPUT_LBRAC
6713 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
6714 Define this macro to say how to output to @var{stream} the debugging
6715 information for the start of a scope level for variable names. The
6716 argument @var{name} is the name of an assembler symbol (for use with
6717 @code{assemble_name}) whose value is the address where the scope begins.
6719 @findex DBX_OUTPUT_RBRAC
6720 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
6721 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
6723 @findex DBX_OUTPUT_ENUM
6724 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
6725 Define this macro if the target machine requires special handling to
6726 output an enumeration type. The definition should be a C statement
6727 (sans semicolon) to output the appropriate information to @var{stream}
6728 for the type @var{type}.
6730 @findex DBX_OUTPUT_FUNCTION_END
6731 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
6732 Define this macro if the target machine requires special output at the
6733 end of the debugging information for a function. The definition should
6734 be a C statement (sans semicolon) to output the appropriate information
6735 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
6738 @findex DBX_OUTPUT_STANDARD_TYPES
6739 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
6740 Define this macro if you need to control the order of output of the
6741 standard data types at the beginning of compilation. The argument
6742 @var{syms} is a @code{tree} which is a chain of all the predefined
6743 global symbols, including names of data types.
6745 Normally, DBX output starts with definitions of the types for integers
6746 and characters, followed by all the other predefined types of the
6747 particular language in no particular order.
6749 On some machines, it is necessary to output different particular types
6750 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
6751 those symbols in the necessary order. Any predefined types that you
6752 don't explicitly output will be output afterward in no particular order.
6754 Be careful not to define this macro so that it works only for C. There
6755 are no global variables to access most of the built-in types, because
6756 another language may have another set of types. The way to output a
6757 particular type is to look through @var{syms} to see if you can find it.
6763 for (decl = syms; decl; decl = TREE_CHAIN (decl))
6764 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
6766 dbxout_symbol (decl);
6772 This does nothing if the expected type does not exist.
6774 See the function @code{init_decl_processing} in @file{c-decl.c} to find
6775 the names to use for all the built-in C types.
6777 Here is another way of finding a particular type:
6779 @c this is still overfull. --mew 10feb93
6783 for (decl = syms; decl; decl = TREE_CHAIN (decl))
6784 if (TREE_CODE (decl) == TYPE_DECL
6785 && (TREE_CODE (TREE_TYPE (decl))
6787 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
6788 && TYPE_UNSIGNED (TREE_TYPE (decl)))
6790 /* @r{This must be @code{unsigned short}.} */
6791 dbxout_symbol (decl);
6797 @findex NO_DBX_FUNCTION_END
6798 @item NO_DBX_FUNCTION_END
6799 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
6800 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extention construct.
6801 On those machines, define this macro to turn this feature off without
6802 disturbing the rest of the gdb extensions.
6806 @node File Names and DBX
6807 @subsection File Names in DBX Format
6809 @c prevent bad page break with this line
6810 This describes file names in DBX format.
6813 @findex DBX_WORKING_DIRECTORY
6814 @item DBX_WORKING_DIRECTORY
6815 Define this if DBX wants to have the current directory recorded in each
6818 Note that the working directory is always recorded if GDB extensions are
6821 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
6822 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
6823 A C statement to output DBX debugging information to the stdio stream
6824 @var{stream} which indicates that file @var{name} is the main source
6825 file---the file specified as the input file for compilation.
6826 This macro is called only once, at the beginning of compilation.
6828 This macro need not be defined if the standard form of output
6829 for DBX debugging information is appropriate.
6831 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
6832 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
6833 A C statement to output DBX debugging information to the stdio stream
6834 @var{stream} which indicates that the current directory during
6835 compilation is named @var{name}.
6837 This macro need not be defined if the standard form of output
6838 for DBX debugging information is appropriate.
6840 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
6841 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
6842 A C statement to output DBX debugging information at the end of
6843 compilation of the main source file @var{name}.
6845 If you don't define this macro, nothing special is output at the end
6846 of compilation, which is correct for most machines.
6848 @findex DBX_OUTPUT_SOURCE_FILENAME
6849 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6850 A C statement to output DBX debugging information to the stdio stream
6851 @var{stream} which indicates that file @var{name} is the current source
6852 file. This output is generated each time input shifts to a different
6853 source file as a result of @samp{#include}, the end of an included file,
6854 or a @samp{#line} command.
6856 This macro need not be defined if the standard form of output
6857 for DBX debugging information is appropriate.
6862 @subsection Macros for SDB and DWARF Output
6864 @c prevent bad page break with this line
6865 Here are macros for SDB and DWARF output.
6868 @findex SDB_DEBUGGING_INFO
6869 @item SDB_DEBUGGING_INFO
6870 Define this macro if GNU CC should produce COFF-style debugging output
6871 for SDB in response to the @samp{-g} option.
6873 @findex DWARF_DEBUGGING_INFO
6874 @item DWARF_DEBUGGING_INFO
6875 Define this macro if GNU CC should produce dwarf format debugging output
6876 in response to the @samp{-g} option.
6878 @findex DWARF2_DEBUGGING_INFO
6879 @item DWARF2_DEBUGGING_INFO
6880 Define this macro if GNU CC should produce dwarf version 2 format
6881 debugging output in response to the @samp{-g} option.
6883 To support optional call frame debugging information, you must also
6884 define @code{INCOMING_RETURN_ADDR_RTX} and either set
6885 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
6886 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
6887 as appropriate from @code{FUNCTION_PROLOGUE} if you don't.
6889 @findex DWARF2_FRAME_INFO
6890 @item DWARF2_FRAME_INFO
6891 Define this macro to a nonzero value if GNU CC should always output
6892 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
6893 (@pxref{Exception Region Output} is nonzero, GNU CC will output this
6894 information not matter how you define @code{DWARF2_FRAME_INFO}.
6896 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
6897 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
6898 Define this macro if the linker does not work with Dwarf version 2.
6899 Normally, if the user specifies only @samp{-ggdb} GNU CC will use Dwarf
6900 version 2 if available; this macro disables this. See the description
6901 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
6903 @findex PUT_SDB_@dots{}
6904 @item PUT_SDB_@dots{}
6905 Define these macros to override the assembler syntax for the special
6906 SDB assembler directives. See @file{sdbout.c} for a list of these
6907 macros and their arguments. If the standard syntax is used, you need
6908 not define them yourself.
6912 Some assemblers do not support a semicolon as a delimiter, even between
6913 SDB assembler directives. In that case, define this macro to be the
6914 delimiter to use (usually @samp{\n}). It is not necessary to define
6915 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
6918 @findex SDB_GENERATE_FAKE
6919 @item SDB_GENERATE_FAKE
6920 Define this macro to override the usual method of constructing a dummy
6921 name for anonymous structure and union types. See @file{sdbout.c} for
6924 @findex SDB_ALLOW_UNKNOWN_REFERENCES
6925 @item SDB_ALLOW_UNKNOWN_REFERENCES
6926 Define this macro to allow references to unknown structure,
6927 union, or enumeration tags to be emitted. Standard COFF does not
6928 allow handling of unknown references, MIPS ECOFF has support for
6931 @findex SDB_ALLOW_FORWARD_REFERENCES
6932 @item SDB_ALLOW_FORWARD_REFERENCES
6933 Define this macro to allow references to structure, union, or
6934 enumeration tags that have not yet been seen to be handled. Some
6935 assemblers choke if forward tags are used, while some require it.
6938 @node Cross-compilation
6939 @section Cross Compilation and Floating Point
6940 @cindex cross compilation and floating point
6941 @cindex floating point and cross compilation
6943 While all modern machines use 2's complement representation for integers,
6944 there are a variety of representations for floating point numbers. This
6945 means that in a cross-compiler the representation of floating point numbers
6946 in the compiled program may be different from that used in the machine
6947 doing the compilation.
6950 Because different representation systems may offer different amounts of
6951 range and precision, the cross compiler cannot safely use the host
6952 machine's floating point arithmetic. Therefore, floating point constants
6953 must be represented in the target machine's format. This means that the
6954 cross compiler cannot use @code{atof} to parse a floating point constant;
6955 it must have its own special routine to use instead. Also, constant
6956 folding must emulate the target machine's arithmetic (or must not be done
6959 The macros in the following table should be defined only if you are cross
6960 compiling between different floating point formats.
6962 Otherwise, don't define them. Then default definitions will be set up which
6963 use @code{double} as the data type, @code{==} to test for equality, etc.
6965 You don't need to worry about how many times you use an operand of any
6966 of these macros. The compiler never uses operands which have side effects.
6969 @findex REAL_VALUE_TYPE
6970 @item REAL_VALUE_TYPE
6971 A macro for the C data type to be used to hold a floating point value
6972 in the target machine's format. Typically this would be a
6973 @code{struct} containing an array of @code{int}.
6975 @findex REAL_VALUES_EQUAL
6976 @item REAL_VALUES_EQUAL (@var{x}, @var{y})
6977 A macro for a C expression which compares for equality the two values,
6978 @var{x} and @var{y}, both of type @code{REAL_VALUE_TYPE}.
6980 @findex REAL_VALUES_LESS
6981 @item REAL_VALUES_LESS (@var{x}, @var{y})
6982 A macro for a C expression which tests whether @var{x} is less than
6983 @var{y}, both values being of type @code{REAL_VALUE_TYPE} and
6984 interpreted as floating point numbers in the target machine's
6987 @findex REAL_VALUE_LDEXP
6989 @item REAL_VALUE_LDEXP (@var{x}, @var{scale})
6990 A macro for a C expression which performs the standard library
6991 function @code{ldexp}, but using the target machine's floating point
6992 representation. Both @var{x} and the value of the expression have
6993 type @code{REAL_VALUE_TYPE}. The second argument, @var{scale}, is an
6996 @findex REAL_VALUE_FIX
6997 @item REAL_VALUE_FIX (@var{x})
6998 A macro whose definition is a C expression to convert the target-machine
6999 floating point value @var{x} to a signed integer. @var{x} has type
7000 @code{REAL_VALUE_TYPE}.
7002 @findex REAL_VALUE_UNSIGNED_FIX
7003 @item REAL_VALUE_UNSIGNED_FIX (@var{x})
7004 A macro whose definition is a C expression to convert the target-machine
7005 floating point value @var{x} to an unsigned integer. @var{x} has type
7006 @code{REAL_VALUE_TYPE}.
7008 @findex REAL_VALUE_RNDZINT
7009 @item REAL_VALUE_RNDZINT (@var{x})
7010 A macro whose definition is a C expression to round the target-machine
7011 floating point value @var{x} towards zero to an integer value (but still
7012 as a floating point number). @var{x} has type @code{REAL_VALUE_TYPE},
7013 and so does the value.
7015 @findex REAL_VALUE_UNSIGNED_RNDZINT
7016 @item REAL_VALUE_UNSIGNED_RNDZINT (@var{x})
7017 A macro whose definition is a C expression to round the target-machine
7018 floating point value @var{x} towards zero to an unsigned integer value
7019 (but still represented as a floating point number). @var{x} has type
7020 @code{REAL_VALUE_TYPE}, and so does the value.
7022 @findex REAL_VALUE_ATOF
7023 @item REAL_VALUE_ATOF (@var{string}, @var{mode})
7024 A macro for a C expression which converts @var{string}, an expression of
7025 type @code{char *}, into a floating point number in the target machine's
7026 representation for mode @var{mode}. The value has type
7027 @code{REAL_VALUE_TYPE}.
7029 @findex REAL_INFINITY
7031 Define this macro if infinity is a possible floating point value, and
7032 therefore division by 0 is legitimate.
7034 @findex REAL_VALUE_ISINF
7036 @item REAL_VALUE_ISINF (@var{x})
7037 A macro for a C expression which determines whether @var{x}, a floating
7038 point value, is infinity. The value has type @code{int}.
7039 By default, this is defined to call @code{isinf}.
7041 @findex REAL_VALUE_ISNAN
7043 @item REAL_VALUE_ISNAN (@var{x})
7044 A macro for a C expression which determines whether @var{x}, a floating
7045 point value, is a ``nan'' (not-a-number). The value has type
7046 @code{int}. By default, this is defined to call @code{isnan}.
7049 @cindex constant folding and floating point
7050 Define the following additional macros if you want to make floating
7051 point constant folding work while cross compiling. If you don't
7052 define them, cross compilation is still possible, but constant folding
7053 will not happen for floating point values.
7056 @findex REAL_ARITHMETIC
7057 @item REAL_ARITHMETIC (@var{output}, @var{code}, @var{x}, @var{y})
7058 A macro for a C statement which calculates an arithmetic operation of
7059 the two floating point values @var{x} and @var{y}, both of type
7060 @code{REAL_VALUE_TYPE} in the target machine's representation, to
7061 produce a result of the same type and representation which is stored
7062 in @var{output} (which will be a variable).
7064 The operation to be performed is specified by @var{code}, a tree code
7065 which will always be one of the following: @code{PLUS_EXPR},
7066 @code{MINUS_EXPR}, @code{MULT_EXPR}, @code{RDIV_EXPR},
7067 @code{MAX_EXPR}, @code{MIN_EXPR}.@refill
7069 @cindex overflow while constant folding
7070 The expansion of this macro is responsible for checking for overflow.
7071 If overflow happens, the macro expansion should execute the statement
7072 @code{return 0;}, which indicates the inability to perform the
7073 arithmetic operation requested.
7075 @findex REAL_VALUE_NEGATE
7076 @item REAL_VALUE_NEGATE (@var{x})
7077 A macro for a C expression which returns the negative of the floating
7078 point value @var{x}. Both @var{x} and the value of the expression
7079 have type @code{REAL_VALUE_TYPE} and are in the target machine's
7080 floating point representation.
7082 There is no way for this macro to report overflow, since overflow
7083 can't happen in the negation operation.
7085 @findex REAL_VALUE_TRUNCATE
7086 @item REAL_VALUE_TRUNCATE (@var{mode}, @var{x})
7087 A macro for a C expression which converts the floating point value
7088 @var{x} to mode @var{mode}.
7090 Both @var{x} and the value of the expression are in the target machine's
7091 floating point representation and have type @code{REAL_VALUE_TYPE}.
7092 However, the value should have an appropriate bit pattern to be output
7093 properly as a floating constant whose precision accords with mode
7096 There is no way for this macro to report overflow.
7098 @findex REAL_VALUE_TO_INT
7099 @item REAL_VALUE_TO_INT (@var{low}, @var{high}, @var{x})
7100 A macro for a C expression which converts a floating point value
7101 @var{x} into a double-precision integer which is then stored into
7102 @var{low} and @var{high}, two variables of type @var{int}.
7104 @item REAL_VALUE_FROM_INT (@var{x}, @var{low}, @var{high}, @var{mode})
7105 @findex REAL_VALUE_FROM_INT
7106 A macro for a C expression which converts a double-precision integer
7107 found in @var{low} and @var{high}, two variables of type @var{int},
7108 into a floating point value which is then stored into @var{x}.
7109 The value is in the target machine's representation for mode @var{mode}
7110 and has the type @code{REAL_VALUE_TYPE}.
7114 @section Miscellaneous Parameters
7115 @cindex parameters, miscellaneous
7117 @c prevent bad page break with this line
7118 Here are several miscellaneous parameters.
7121 @item PREDICATE_CODES
7122 @findex PREDICATE_CODES
7123 Define this if you have defined special-purpose predicates in the file
7124 @file{@var{machine}.c}. This macro is called within an initializer of an
7125 array of structures. The first field in the structure is the name of a
7126 predicate and the second field is an array of rtl codes. For each
7127 predicate, list all rtl codes that can be in expressions matched by the
7128 predicate. The list should have a trailing comma. Here is an example
7129 of two entries in the list for a typical RISC machine:
7132 #define PREDICATE_CODES \
7133 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
7134 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
7137 Defining this macro does not affect the generated code (however,
7138 incorrect definitions that omit an rtl code that may be matched by the
7139 predicate can cause the compiler to malfunction). Instead, it allows
7140 the table built by @file{genrecog} to be more compact and efficient,
7141 thus speeding up the compiler. The most important predicates to include
7142 in the list specified by this macro are those used in the most insn
7145 @findex CASE_VECTOR_MODE
7146 @item CASE_VECTOR_MODE
7147 An alias for a machine mode name. This is the machine mode that
7148 elements of a jump-table should have.
7150 @findex CASE_VECTOR_SHORTEN_MODE
7151 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7152 Optional: return the preferred mode for an @code{addr_diff_vec}
7153 when the minimum and maximum offset are known. If you define this,
7154 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7155 To make this work, you also have to define INSN_ALIGN and
7156 make the alignment for @code{addr_diff_vec} explicit.
7157 The @var{body} argument is provided so that the offset_unsigned and scale
7158 flags can be updated.
7160 @findex CASE_VECTOR_PC_RELATIVE
7161 @item CASE_VECTOR_PC_RELATIVE
7162 Define this macro to be a C expression to indicate when jump-tables
7163 should contain relative addresses. If jump-tables never contain
7164 relative addresses, then you need not define this macro.
7166 @findex CASE_DROPS_THROUGH
7167 @item CASE_DROPS_THROUGH
7168 Define this if control falls through a @code{case} insn when the index
7169 value is out of range. This means the specified default-label is
7170 actually ignored by the @code{case} insn proper.
7172 @findex CASE_VALUES_THRESHOLD
7173 @item CASE_VALUES_THRESHOLD
7174 Define this to be the smallest number of different values for which it
7175 is best to use a jump-table instead of a tree of conditional branches.
7176 The default is four for machines with a @code{casesi} instruction and
7177 five otherwise. This is best for most machines.
7179 @findex WORD_REGISTER_OPERATIONS
7180 @item WORD_REGISTER_OPERATIONS
7181 Define this macro if operations between registers with integral mode
7182 smaller than a word are always performed on the entire register.
7183 Most RISC machines have this property and most CISC machines do not.
7185 @findex LOAD_EXTEND_OP
7186 @item LOAD_EXTEND_OP (@var{mode})
7187 Define this macro to be a C expression indicating when insns that read
7188 memory in @var{mode}, an integral mode narrower than a word, set the
7189 bits outside of @var{mode} to be either the sign-extension or the
7190 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7191 of @var{mode} for which the
7192 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7193 @code{NIL} for other modes.
7195 This macro is not called with @var{mode} non-integral or with a width
7196 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7197 value in this case. Do not define this macro if it would always return
7198 @code{NIL}. On machines where this macro is defined, you will normally
7199 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7201 @findex SHORT_IMMEDIATES_SIGN_EXTEND
7202 @item SHORT_IMMEDIATES_SIGN_EXTEND
7203 Define this macro if loading short immediate values into registers sign
7206 @findex IMPLICIT_FIX_EXPR
7207 @item IMPLICIT_FIX_EXPR
7208 An alias for a tree code that should be used by default for conversion
7209 of floating point values to fixed point. Normally,
7210 @code{FIX_ROUND_EXPR} is used.@refill
7212 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
7213 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
7214 Define this macro if the same instructions that convert a floating
7215 point number to a signed fixed point number also convert validly to an
7218 @findex EASY_DIV_EXPR
7220 An alias for a tree code that is the easiest kind of division to
7221 compile code for in the general case. It may be
7222 @code{TRUNC_DIV_EXPR}, @code{FLOOR_DIV_EXPR}, @code{CEIL_DIV_EXPR} or
7223 @code{ROUND_DIV_EXPR}. These four division operators differ in how
7224 they round the result to an integer. @code{EASY_DIV_EXPR} is used
7225 when it is permissible to use any of those kinds of division and the
7226 choice should be made on the basis of efficiency.@refill
7230 The maximum number of bytes that a single instruction can move quickly
7231 between memory and registers or between two memory locations.
7233 @findex MAX_MOVE_MAX
7235 The maximum number of bytes that a single instruction can move quickly
7236 between memory and registers or between two memory locations. If this
7237 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7238 constant value that is the largest value that @code{MOVE_MAX} can have
7241 @findex SHIFT_COUNT_TRUNCATED
7242 @item SHIFT_COUNT_TRUNCATED
7243 A C expression that is nonzero if on this machine the number of bits
7244 actually used for the count of a shift operation is equal to the number
7245 of bits needed to represent the size of the object being shifted. When
7246 this macro is non-zero, the compiler will assume that it is safe to omit
7247 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7248 truncates the count of a shift operation. On machines that have
7249 instructions that act on bitfields at variable positions, which may
7250 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7251 also enables deletion of truncations of the values that serve as
7252 arguments to bitfield instructions.
7254 If both types of instructions truncate the count (for shifts) and
7255 position (for bitfield operations), or if no variable-position bitfield
7256 instructions exist, you should define this macro.
7258 However, on some machines, such as the 80386 and the 680x0, truncation
7259 only applies to shift operations and not the (real or pretended)
7260 bitfield operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7261 such machines. Instead, add patterns to the @file{md} file that include
7262 the implied truncation of the shift instructions.
7264 You need not define this macro if it would always have the value of zero.
7266 @findex TRULY_NOOP_TRUNCATION
7267 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
7268 A C expression which is nonzero if on this machine it is safe to
7269 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
7270 bits (where @var{outprec} is smaller than @var{inprec}) by merely
7271 operating on it as if it had only @var{outprec} bits.
7273 On many machines, this expression can be 1.
7275 @c rearranged this, removed the phrase "it is reported that". this was
7276 @c to fix an overfull hbox. --mew 10feb93
7277 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
7278 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
7279 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
7280 such cases may improve things.
7282 @findex STORE_FLAG_VALUE
7283 @item STORE_FLAG_VALUE
7284 A C expression describing the value returned by a comparison operator
7285 with an integral mode and stored by a store-flag instruction
7286 (@samp{s@var{cond}}) when the condition is true. This description must
7287 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
7288 comparison operators whose results have a @code{MODE_INT} mode.
7290 A value of 1 or -1 means that the instruction implementing the
7291 comparison operator returns exactly 1 or -1 when the comparison is true
7292 and 0 when the comparison is false. Otherwise, the value indicates
7293 which bits of the result are guaranteed to be 1 when the comparison is
7294 true. This value is interpreted in the mode of the comparison
7295 operation, which is given by the mode of the first operand in the
7296 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
7297 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7300 If @code{STORE_FLAG_VALUE} is neither 1 or -1, the compiler will
7301 generate code that depends only on the specified bits. It can also
7302 replace comparison operators with equivalent operations if they cause
7303 the required bits to be set, even if the remaining bits are undefined.
7304 For example, on a machine whose comparison operators return an
7305 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7306 @samp{0x80000000}, saying that just the sign bit is relevant, the
7310 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7317 (ashift:SI @var{x} (const_int @var{n}))
7321 where @var{n} is the appropriate shift count to move the bit being
7322 tested into the sign bit.
7324 There is no way to describe a machine that always sets the low-order bit
7325 for a true value, but does not guarantee the value of any other bits,
7326 but we do not know of any machine that has such an instruction. If you
7327 are trying to port GNU CC to such a machine, include an instruction to
7328 perform a logical-and of the result with 1 in the pattern for the
7329 comparison operators and let us know
7331 (@pxref{Bug Reporting,,How to Report Bugs}).
7334 (@pxref{Bug Reporting,,How to Report Bugs,gcc.info,Using GCC}).
7337 Often, a machine will have multiple instructions that obtain a value
7338 from a comparison (or the condition codes). Here are rules to guide the
7339 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7344 Use the shortest sequence that yields a valid definition for
7345 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7346 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7347 comparison operators to do so because there may be opportunities to
7348 combine the normalization with other operations.
7351 For equal-length sequences, use a value of 1 or -1, with -1 being
7352 slightly preferred on machines with expensive jumps and 1 preferred on
7356 As a second choice, choose a value of @samp{0x80000001} if instructions
7357 exist that set both the sign and low-order bits but do not define the
7361 Otherwise, use a value of @samp{0x80000000}.
7364 Many machines can produce both the value chosen for
7365 @code{STORE_FLAG_VALUE} and its negation in the same number of
7366 instructions. On those machines, you should also define a pattern for
7367 those cases, e.g., one matching
7370 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7373 Some machines can also perform @code{and} or @code{plus} operations on
7374 condition code values with less instructions than the corresponding
7375 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
7376 machines, define the appropriate patterns. Use the names @code{incscc}
7377 and @code{decscc}, respectively, for the patterns which perform
7378 @code{plus} or @code{minus} operations on condition code values. See
7379 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
7380 find such instruction sequences on other machines.
7382 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7385 @findex FLOAT_STORE_FLAG_VALUE
7386 @item FLOAT_STORE_FLAG_VALUE
7387 A C expression that gives a non-zero floating point value that is
7388 returned when comparison operators with floating-point results are true.
7389 Define this macro on machine that have comparison operations that return
7390 floating-point values. If there are no such operations, do not define
7395 An alias for the machine mode for pointers. On most machines, define
7396 this to be the integer mode corresponding to the width of a hardware
7397 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7398 On some machines you must define this to be one of the partial integer
7399 modes, such as @code{PSImode}.
7401 The width of @code{Pmode} must be at least as large as the value of
7402 @code{POINTER_SIZE}. If it is not equal, you must define the macro
7403 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7406 @findex FUNCTION_MODE
7408 An alias for the machine mode used for memory references to functions
7409 being called, in @code{call} RTL expressions. On most machines this
7410 should be @code{QImode}.
7412 @findex INTEGRATE_THRESHOLD
7413 @item INTEGRATE_THRESHOLD (@var{decl})
7414 A C expression for the maximum number of instructions above which the
7415 function @var{decl} should not be inlined. @var{decl} is a
7416 @code{FUNCTION_DECL} node.
7418 The default definition of this macro is 64 plus 8 times the number of
7419 arguments that the function accepts. Some people think a larger
7420 threshold should be used on RISC machines.
7422 @findex SCCS_DIRECTIVE
7423 @item SCCS_DIRECTIVE
7424 Define this if the preprocessor should ignore @code{#sccs} directives
7425 and print no error message.
7427 @findex NO_IMPLICIT_EXTERN_C
7428 @item NO_IMPLICIT_EXTERN_C
7429 Define this macro if the system header files support C++ as well as C.
7430 This macro inhibits the usual method of using system header files in
7431 C++, which is to pretend that the file's contents are enclosed in
7432 @samp{extern "C" @{@dots{}@}}.
7434 @findex HANDLE_PRAGMA
7437 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
7438 Define this macro if you want to implement any pragmas. If defined, it
7439 is a C expression whose value is 1 if the pragma was handled by the
7440 macro, zero otherwise. The argument @var{getc} is a function of type
7441 @samp{int (*)(void)} which will return the next character in the input
7442 stream, or EOF if no characters are left. The argument @var{ungetc} is
7443 a function of type @samp{void (*)(int)} which will push a character back
7444 into the input stream. The argument @var{name} is the word following
7445 #pragma in the input stream. The input stream pointer will be pointing
7446 just beyond the end of this word. The input stream should be left
7447 undistrubed if the expression returns zero, otherwise it should be
7448 pointing at the next character after the end of the pragma. Any
7449 characters remaining on the line will be ignored.
7451 It is generally a bad idea to implement new uses of @code{#pragma}. The
7452 only reason to define this macro is for compatibility with other
7453 compilers that do support @code{#pragma} for the sake of any user
7454 programs which already use it.
7456 If the pragma can be implemented by atttributes then the macro
7457 @samp{INSERT_ATTRIBUTES} might be a useful one to define as well.
7459 Note: older versions of this macro only had two arguments: @var{stream}
7460 and @var{token}. The macro was changed in order to allow it to work
7461 when gcc is built both with and without a cpp library.
7463 @findex HANDLE_SYSV_PRAGMA
7466 @item HANDLE_SYSV_PRAGMA
7467 Define this macro (to a value of 1) if you want the System V style
7468 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
7469 [=<value>]} to be supported by gcc.
7471 The pack pragma specifies the maximum alignment (in bytes) of fields
7472 within a structure, in much the same way as the @samp{__aligned__} and
7473 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
7474 the behaviour to the default.
7476 The weak pragma only works if @code{SUPPORTS_WEAK} and
7477 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
7478 of specifically named weak labels, optionally with a value.
7480 @findex HANDLE_PRAGMA_PACK_PUSH_POP
7483 @item HANDLE_PRAGMA_PACK_PUSH_POP
7484 Define this macro (to a value of 1) if you want to support the Win32
7485 style pragmas @samp{#pragma pack(push,<n>)} and @samp{#pragma
7486 pack(pop)}. The pack(push,<n>) pragma specifies the maximum alignment
7487 (in bytes) of fields within a structure, in much the same way as the
7488 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
7489 pack value of zero resets the behaviour to the default. Successive
7490 invocations of this pragma cause the previous values to be stacked, so
7491 that invocations of @samp{#pragma pack(pop)} will return to the previous
7494 @findex VALID_MACHINE_DECL_ATTRIBUTE
7495 @item VALID_MACHINE_DECL_ATTRIBUTE (@var{decl}, @var{attributes}, @var{identifier}, @var{args})
7496 If defined, a C expression whose value is nonzero if @var{identifier} with
7497 arguments @var{args} is a valid machine specific attribute for @var{decl}.
7498 The attributes in @var{attributes} have previously been assigned to @var{decl}.
7500 @findex VALID_MACHINE_TYPE_ATTRIBUTE
7501 @item VALID_MACHINE_TYPE_ATTRIBUTE (@var{type}, @var{attributes}, @var{identifier}, @var{args})
7502 If defined, a C expression whose value is nonzero if @var{identifier} with
7503 arguments @var{args} is a valid machine specific attribute for @var{type}.
7504 The attributes in @var{attributes} have previously been assigned to @var{type}.
7506 @findex COMP_TYPE_ATTRIBUTES
7507 @item COMP_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
7508 If defined, a C expression whose value is zero if the attributes on
7509 @var{type1} and @var{type2} are incompatible, one if they are compatible,
7510 and two if they are nearly compatible (which causes a warning to be
7513 @findex SET_DEFAULT_TYPE_ATTRIBUTES
7514 @item SET_DEFAULT_TYPE_ATTRIBUTES (@var{type})
7515 If defined, a C statement that assigns default attributes to
7516 newly defined @var{type}.
7518 @findex MERGE_MACHINE_TYPE_ATTRIBUTES
7519 @item MERGE_MACHINE_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
7520 Define this macro if the merging of type attributes needs special handling.
7521 If defined, the result is a list of the combined TYPE_ATTRIBUTES of
7522 @var{type1} and @var{type2}. It is assumed that comptypes has already been
7523 called and returned 1.
7525 @findex MERGE_MACHINE_DECL_ATTRIBUTES
7526 @item MERGE_MACHINE_DECL_ATTRIBUTES (@var{olddecl}, @var{newdecl})
7527 Define this macro if the merging of decl attributes needs special handling.
7528 If defined, the result is a list of the combined DECL_MACHINE_ATTRIBUTES of
7529 @var{olddecl} and @var{newdecl}. @var{newdecl} is a duplicate declaration
7530 of @var{olddecl}. Examples of when this is needed are when one attribute
7531 overrides another, or when an attribute is nullified by a subsequent
7534 @findex INSERT_ATTRIBUTES
7535 @item INSERT_ATTRIBUTES (@var{node}, @var{attr_ptr}, @var{prefix_ptr})
7536 Define this macro if you want to be able to add attributes to a decl
7537 when it is being created. This is normally useful for backends which
7538 wish to implement a pragma by using the attributes which correspond to
7539 the pragma's effect. The @var{node} argument is the decl which is being
7540 created. The @var{attr_ptr} argument is a pointer to the attribute list
7541 for this decl. The @var{prefix_ptr} is a pointer to the list of
7542 attributes that have appeared after the specifiers and modifiers of the
7543 declaration, but before the declaration proper.
7545 @findex SET_DEFAULT_DECL_ATTRIBUTES
7546 @item SET_DEFAULT_DECL_ATTRIBUTES (@var{decl}, @var{attributes})
7547 If defined, a C statement that assigns default attributes to
7548 newly defined @var{decl}.
7550 @findex DOLLARS_IN_IDENTIFIERS
7551 @item DOLLARS_IN_IDENTIFIERS
7552 Define this macro to control use of the character @samp{$} in identifier
7553 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
7554 1 is the default; there is no need to define this macro in that case.
7555 This macro controls the compiler proper; it does not affect the preprocessor.
7557 @findex NO_DOLLAR_IN_LABEL
7558 @item NO_DOLLAR_IN_LABEL
7559 Define this macro if the assembler does not accept the character
7560 @samp{$} in label names. By default constructors and destructors in
7561 G++ have @samp{$} in the identifiers. If this macro is defined,
7562 @samp{.} is used instead.
7564 @findex NO_DOT_IN_LABEL
7565 @item NO_DOT_IN_LABEL
7566 Define this macro if the assembler does not accept the character
7567 @samp{.} in label names. By default constructors and destructors in G++
7568 have names that use @samp{.}. If this macro is defined, these names
7569 are rewritten to avoid @samp{.}.
7571 @findex DEFAULT_MAIN_RETURN
7572 @item DEFAULT_MAIN_RETURN
7573 Define this macro if the target system expects every program's @code{main}
7574 function to return a standard ``success'' value by default (if no other
7575 value is explicitly returned).
7577 The definition should be a C statement (sans semicolon) to generate the
7578 appropriate rtl instructions. It is used only when compiling the end of
7583 Define this if the target system supports the function
7584 @code{atexit} from the ANSI C standard. If this is not defined,
7585 and @code{INIT_SECTION_ASM_OP} is not defined, a default
7586 @code{exit} function will be provided to support C++.
7590 Define this if your @code{exit} function needs to do something
7591 besides calling an external function @code{_cleanup} before
7592 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
7593 only needed if neither @code{HAVE_ATEXIT} nor
7594 @code{INIT_SECTION_ASM_OP} are defined.
7596 @findex INSN_SETS_ARE_DELAYED
7597 @item INSN_SETS_ARE_DELAYED (@var{insn})
7598 Define this macro as a C expression that is nonzero if it is safe for the
7599 delay slot scheduler to place instructions in the delay slot of @var{insn},
7600 even if they appear to use a resource set or clobbered in @var{insn}.
7601 @var{insn} is always a @code{jump_insn} or an @code{insn}; GNU CC knows that
7602 every @code{call_insn} has this behavior. On machines where some @code{insn}
7603 or @code{jump_insn} is really a function call and hence has this behavior,
7604 you should define this macro.
7606 You need not define this macro if it would always return zero.
7608 @findex INSN_REFERENCES_ARE_DELAYED
7609 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
7610 Define this macro as a C expression that is nonzero if it is safe for the
7611 delay slot scheduler to place instructions in the delay slot of @var{insn},
7612 even if they appear to set or clobber a resource referenced in @var{insn}.
7613 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
7614 some @code{insn} or @code{jump_insn} is really a function call and its operands
7615 are registers whose use is actually in the subroutine it calls, you should
7616 define this macro. Doing so allows the delay slot scheduler to move
7617 instructions which copy arguments into the argument registers into the delay
7620 You need not define this macro if it would always return zero.
7622 @findex MACHINE_DEPENDENT_REORG
7623 @item MACHINE_DEPENDENT_REORG (@var{insn})
7624 In rare cases, correct code generation requires extra machine
7625 dependent processing between the second jump optimization pass and
7626 delayed branch scheduling. On those machines, define this macro as a C
7627 statement to act on the code starting at @var{insn}.
7629 @findex MULTIPLE_SYMBOL_SPACES
7630 @item MULTIPLE_SYMBOL_SPACES
7631 Define this macro if in some cases global symbols from one translation
7632 unit may not be bound to undefined symbols in another translation unit
7633 without user intervention. For instance, under Microsoft Windows
7634 symbols must be explicitly imported from shared libraries (DLLs).
7638 A C expression that returns how many instructions can be issued at the
7639 same time if the machine is a superscalar machine. This is only used by
7640 the @samp{Haifa} scheduler, and not the traditional scheduler.
7642 @findex MD_SCHED_INIT
7643 @item MD_SCHED_INIT (@var{file}, @var{verbose})
7644 A C statement which is executed by the @samp{Haifa} scheduler at the
7645 beginning of each block of instructions that are to be scheduled.
7646 @var{file} is either a null pointer, or a stdio stream to write any
7647 debug output to. @var{verbose} is the verbose level provided by
7648 @samp{-fsched-verbose-}@var{n}.
7650 @findex MD_SCHED_REORDER
7651 @item MD_SCHED_REORDER (@var{file}, @var{verbose}, @var{ready}, @var{n_ready})
7652 A C statement which is executed by the @samp{Haifa} scheduler after it
7653 has scheduled the ready list to allow the machine description to reorder
7654 it (for example to combine two small instructions together on
7655 @samp{VLIW} machines). @var{file} is either a null pointer, or a stdio
7656 stream to write any debug output to. @var{verbose} is the verbose level
7657 provided by @samp{-fsched-verbose-}@var{n}. @var{ready} is a pointer to
7658 the ready list of instructions that are ready to be scheduled.
7659 @var{n_ready} is the number of elements in the ready list. The
7660 scheduler reads the ready list in reverse order, starting with
7661 @var{ready}[@var{n_ready}-1] and going to @var{ready}[0].
7663 @findex MD_SCHED_VARIABLE_ISSUE
7664 @item MD_SCHED_VARIABLE_ISSUE (@var{file}, @var{verbose}, @var{insn}, @var{more})
7665 A C statement which is executed by the @samp{Haifa} scheduler after it
7666 has scheduled an insn from the ready list. @var{file} is either a null
7667 pointer, or a stdio stream to write any debug output to. @var{verbose}
7668 is the verbose level provided by @samp{-fsched-verbose-}@var{n}.
7669 @var{insn} is the instruction that was scheduled. @var{more} is the
7670 number of instructions that can be issued in the current cycle. The
7671 @samp{MD_SCHED_VARIABLE_ISSUE} macro is responsible for updating the
7672 value of @var{more} (typically by @var{more}--).
7674 @findex MAX_INTEGER_COMPUTATION_MODE
7675 @item MAX_INTEGER_COMPUTATION_MODE
7676 Define this to the largest integer machine mode which can be used for
7677 operations other than load, store and copy operations.
7679 You need only define this macro if the target holds values larger than
7680 @code{word_mode} in general purpose registers. Most targets should not define
7683 @findex MATH_LIBRARY
7685 Define this macro as a C string constant for the linker argument to link
7686 in the system math library, or @samp{""} if the target does not have a
7687 separate math library.
7689 You need only define this macro if the default of @samp{"-lm"} is wrong.