1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001
2 @c Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Target Description Macros
8 @cindex machine description macros
9 @cindex target description macros
10 @cindex macros, target description
11 @cindex @file{tm.h} macros
13 In addition to the file @file{@var{machine}.md}, a machine description
14 includes a C header file conventionally given the name
15 @file{@var{machine}.h}. This header file defines numerous macros
16 that convey the information about the target machine that does not fit
17 into the scheme of the @file{.md} file. The file @file{tm.h} should be
18 a link to @file{@var{machine}.h}. The header file @file{config.h}
19 includes @file{tm.h} and most compiler source files include
23 * Driver:: Controlling how the driver runs the compilation passes.
24 * Run-time Target:: Defining @samp{-m} options like @samp{-m68000} and @samp{-m68020}.
25 * Per-Function Data:: Defining data structures for per-function information.
26 * Storage Layout:: Defining sizes and alignments of data.
27 * Type Layout:: Defining sizes and properties of basic user data types.
28 * Registers:: Naming and describing the hardware registers.
29 * Register Classes:: Defining the classes of hardware registers.
30 * Stack and Calling:: Defining which way the stack grows and by how much.
31 * Varargs:: Defining the varargs macros.
32 * Trampolines:: Code set up at run time to enter a nested function.
33 * Library Calls:: Controlling how library routines are implicitly called.
34 * Addressing Modes:: Defining addressing modes valid for memory operands.
35 * Condition Code:: Defining how insns update the condition code.
36 * Costs:: Defining relative costs of different operations.
37 * Sections:: Dividing storage into text, data, and other sections.
38 * PIC:: Macros for position independent code.
39 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
40 * Debugging Info:: Defining the format of debugging output.
41 * Cross-compilation:: Handling floating point for cross-compilers.
42 * Mode Switching:: Insertion of mode-switching instructions.
43 * Misc:: Everything else.
47 @section Controlling the Compilation Driver, @file{gcc}
49 @cindex controlling the compilation driver
51 @c prevent bad page break with this line
52 You can control the compilation driver.
55 @findex SWITCH_TAKES_ARG
56 @item SWITCH_TAKES_ARG (@var{char})
57 A C expression which determines whether the option @samp{-@var{char}}
58 takes arguments. The value should be the number of arguments that
59 option takes--zero, for many options.
61 By default, this macro is defined as
62 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
63 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
64 wish to add additional options which take arguments. Any redefinition
65 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
68 @findex WORD_SWITCH_TAKES_ARG
69 @item WORD_SWITCH_TAKES_ARG (@var{name})
70 A C expression which determines whether the option @samp{-@var{name}}
71 takes arguments. The value should be the number of arguments that
72 option takes--zero, for many options. This macro rather than
73 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
75 By default, this macro is defined as
76 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
77 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
78 wish to add additional options which take arguments. Any redefinition
79 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
82 @findex SWITCH_CURTAILS_COMPILATION
83 @item SWITCH_CURTAILS_COMPILATION (@var{char})
84 A C expression which determines whether the option @samp{-@var{char}}
85 stops compilation before the generation of an executable. The value is
86 boolean, non-zero if the option does stop an executable from being
87 generated, zero otherwise.
89 By default, this macro is defined as
90 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
91 options properly. You need not define
92 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
93 options which affect the generation of an executable. Any redefinition
94 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
95 for additional options.
97 @findex SWITCHES_NEED_SPACES
98 @item SWITCHES_NEED_SPACES
99 A string-valued C expression which enumerates the options for which
100 the linker needs a space between the option and its argument.
102 If this macro is not defined, the default value is @code{""}.
106 A C string constant that tells the GCC driver program options to
107 pass to CPP. It can also specify how to translate options you
108 give to GCC into options for GCC to pass to the CPP.
110 Do not define this macro if it does not need to do anything.
112 @findex CPLUSPLUS_CPP_SPEC
113 @item CPLUSPLUS_CPP_SPEC
114 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
115 than C. If you do not define this macro, then the value of
116 @code{CPP_SPEC} (if any) will be used instead.
118 @findex NO_BUILTIN_SIZE_TYPE
119 @item NO_BUILTIN_SIZE_TYPE
120 If this macro is defined, the preprocessor will not define the builtin macro
121 @code{__SIZE_TYPE__}. The macro @code{__SIZE_TYPE__} must then be defined
122 by @code{CPP_SPEC} instead.
124 This should be defined if @code{SIZE_TYPE} depends on target dependent flags
125 which are not accessible to the preprocessor. Otherwise, it should not
128 @findex NO_BUILTIN_PTRDIFF_TYPE
129 @item NO_BUILTIN_PTRDIFF_TYPE
130 If this macro is defined, the preprocessor will not define the builtin macro
131 @code{__PTRDIFF_TYPE__}. The macro @code{__PTRDIFF_TYPE__} must then be
132 defined by @code{CPP_SPEC} instead.
134 This should be defined if @code{PTRDIFF_TYPE} depends on target dependent flags
135 which are not accessible to the preprocessor. Otherwise, it should not
138 @findex NO_BUILTIN_WCHAR_TYPE
139 @item NO_BUILTIN_WCHAR_TYPE
140 If this macro is defined, the preprocessor will not define the builtin macro
141 @code{__WCHAR_TYPE__}. The macro @code{__WCHAR_TYPE__} must then be
142 defined by @code{CPP_SPEC} instead.
144 This should be defined if @code{WCHAR_TYPE} depends on target dependent flags
145 which are not accessible to the preprocessor. Otherwise, it should not
148 @findex NO_BUILTIN_WINT_TYPE
149 @item NO_BUILTIN_WINT_TYPE
150 If this macro is defined, the preprocessor will not define the builtin macro
151 @code{__WINT_TYPE__}. The macro @code{__WINT_TYPE__} must then be
152 defined by @code{CPP_SPEC} instead.
154 This should be defined if @code{WINT_TYPE} depends on target dependent flags
155 which are not accessible to the preprocessor. Otherwise, it should not
158 @findex SIGNED_CHAR_SPEC
159 @item SIGNED_CHAR_SPEC
160 A C string constant that tells the GCC driver program options to
161 pass to CPP. By default, this macro is defined to pass the option
162 @samp{-D__CHAR_UNSIGNED__} to CPP if @code{char} will be treated as
163 @code{unsigned char} by @code{cc1}.
165 Do not define this macro unless you need to override the default
170 A C string constant that tells the GCC driver program options to
171 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
173 It can also specify how to translate options you give to GCC into options
174 for GCC to pass to front ends..
176 Do not define this macro if it does not need to do anything.
180 A C string constant that tells the GCC driver program options to
181 pass to @code{cc1plus}. It can also specify how to translate options you
182 give to GCC into options for GCC to pass to the @code{cc1plus}.
184 Do not define this macro if it does not need to do anything.
185 Note that everything defined in CC1_SPEC is already passed to
186 @code{cc1plus} so there is no need to duplicate the contents of
187 CC1_SPEC in CC1PLUS_SPEC.
191 A C string constant that tells the GCC driver program options to
192 pass to the assembler. It can also specify how to translate options
193 you give to GCC into options for GCC to pass to the assembler.
194 See the file @file{sun3.h} for an example of this.
196 Do not define this macro if it does not need to do anything.
198 @findex ASM_FINAL_SPEC
200 A C string constant that tells the GCC driver program how to
201 run any programs which cleanup after the normal assembler.
202 Normally, this is not needed. See the file @file{mips.h} for
205 Do not define this macro if it does not need to do anything.
209 A C string constant that tells the GCC driver program options to
210 pass to the linker. It can also specify how to translate options you
211 give to GCC into options for GCC to pass to the linker.
213 Do not define this macro if it does not need to do anything.
217 Another C string constant used much like @code{LINK_SPEC}. The difference
218 between the two is that @code{LIB_SPEC} is used at the end of the
219 command given to the linker.
221 If this macro is not defined, a default is provided that
222 loads the standard C library from the usual place. See @file{gcc.c}.
226 Another C string constant that tells the GCC driver program
227 how and when to place a reference to @file{libgcc.a} into the
228 linker command line. This constant is placed both before and after
229 the value of @code{LIB_SPEC}.
231 If this macro is not defined, the GCC driver provides a default that
232 passes the string @samp{-lgcc} to the linker.
234 @findex STARTFILE_SPEC
236 Another C string constant used much like @code{LINK_SPEC}. The
237 difference between the two is that @code{STARTFILE_SPEC} is used at
238 the very beginning of the command given to the linker.
240 If this macro is not defined, a default is provided that loads the
241 standard C startup file from the usual place. See @file{gcc.c}.
245 Another C string constant used much like @code{LINK_SPEC}. The
246 difference between the two is that @code{ENDFILE_SPEC} is used at
247 the very end of the command given to the linker.
249 Do not define this macro if it does not need to do anything.
253 Define this macro to provide additional specifications to put in the
254 @file{specs} file that can be used in various specifications like
257 The definition should be an initializer for an array of structures,
258 containing a string constant, that defines the specification name, and a
259 string constant that provides the specification.
261 Do not define this macro if it does not need to do anything.
263 @code{EXTRA_SPECS} is useful when an architecture contains several
264 related targets, which have various @code{..._SPECS} which are similar
265 to each other, and the maintainer would like one central place to keep
268 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
269 define either @code{_CALL_SYSV} when the System V calling sequence is
270 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
273 The @file{config/rs6000/rs6000.h} target file defines:
276 #define EXTRA_SPECS \
277 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
279 #define CPP_SYS_DEFAULT ""
282 The @file{config/rs6000/sysv.h} target file defines:
286 "%@{posix: -D_POSIX_SOURCE @} \
287 %@{mcall-sysv: -D_CALL_SYSV @} %@{mcall-aix: -D_CALL_AIX @} \
288 %@{!mcall-sysv: %@{!mcall-aix: %(cpp_sysv_default) @}@} \
289 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
291 #undef CPP_SYSV_DEFAULT
292 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
295 while the @file{config/rs6000/eabiaix.h} target file defines
296 @code{CPP_SYSV_DEFAULT} as:
299 #undef CPP_SYSV_DEFAULT
300 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
303 @findex LINK_LIBGCC_SPECIAL
304 @item LINK_LIBGCC_SPECIAL
305 Define this macro if the driver program should find the library
306 @file{libgcc.a} itself and should not pass @samp{-L} options to the
307 linker. If you do not define this macro, the driver program will pass
308 the argument @samp{-lgcc} to tell the linker to do the search and will
309 pass @samp{-L} options to it.
311 @findex LINK_LIBGCC_SPECIAL_1
312 @item LINK_LIBGCC_SPECIAL_1
313 Define this macro if the driver program should find the library
314 @file{libgcc.a}. If you do not define this macro, the driver program will pass
315 the argument @samp{-lgcc} to tell the linker to do the search.
316 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
317 not affect @samp{-L} options.
319 @findex LINK_COMMAND_SPEC
320 @item LINK_COMMAND_SPEC
321 A C string constant giving the complete command line need to execute the
322 linker. When you do this, you will need to update your port each time a
323 change is made to the link command line within @file{gcc.c}. Therefore,
324 define this macro only if you need to completely redefine the command
325 line for invoking the linker and there is no other way to accomplish
328 @findex MULTILIB_DEFAULTS
329 @item MULTILIB_DEFAULTS
330 Define this macro as a C expression for the initializer of an array of
331 string to tell the driver program which options are defaults for this
332 target and thus do not need to be handled specially when using
333 @code{MULTILIB_OPTIONS}.
335 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
336 the target makefile fragment or if none of the options listed in
337 @code{MULTILIB_OPTIONS} are set by default.
338 @xref{Target Fragment}.
340 @findex RELATIVE_PREFIX_NOT_LINKDIR
341 @item RELATIVE_PREFIX_NOT_LINKDIR
342 Define this macro to tell @code{gcc} that it should only translate
343 a @samp{-B} prefix into a @samp{-L} linker option if the prefix
344 indicates an absolute file name.
346 @findex STANDARD_EXEC_PREFIX
347 @item STANDARD_EXEC_PREFIX
348 Define this macro as a C string constant if you wish to override the
349 standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
350 try when searching for the executable files of the compiler.
352 @findex MD_EXEC_PREFIX
354 If defined, this macro is an additional prefix to try after
355 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
356 when the @samp{-b} option is used, or the compiler is built as a cross
357 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
358 to the list of directories used to find the assembler in @file{configure.in}.
360 @findex STANDARD_STARTFILE_PREFIX
361 @item STANDARD_STARTFILE_PREFIX
362 Define this macro as a C string constant if you wish to override the
363 standard choice of @file{/usr/local/lib/} as the default prefix to
364 try when searching for startup files such as @file{crt0.o}.
366 @findex MD_STARTFILE_PREFIX
367 @item MD_STARTFILE_PREFIX
368 If defined, this macro supplies an additional prefix to try after the
369 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
370 @samp{-b} option is used, or when the compiler is built as a cross
373 @findex MD_STARTFILE_PREFIX_1
374 @item MD_STARTFILE_PREFIX_1
375 If defined, this macro supplies yet another prefix to try after the
376 standard prefixes. It is not searched when the @samp{-b} option is
377 used, or when the compiler is built as a cross compiler.
379 @findex INIT_ENVIRONMENT
380 @item INIT_ENVIRONMENT
381 Define this macro as a C string constant if you wish to set environment
382 variables for programs called by the driver, such as the assembler and
383 loader. The driver passes the value of this macro to @code{putenv} to
384 initialize the necessary environment variables.
386 @findex LOCAL_INCLUDE_DIR
387 @item LOCAL_INCLUDE_DIR
388 Define this macro as a C string constant if you wish to override the
389 standard choice of @file{/usr/local/include} as the default prefix to
390 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
391 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
393 Cross compilers do not use this macro and do not search either
394 @file{/usr/local/include} or its replacement.
396 @findex MODIFY_TARGET_NAME
397 @item MODIFY_TARGET_NAME
398 Define this macro if you with to define command-line switches that modify the
401 For each switch, you can include a string to be appended to the first
402 part of the configuration name or a string to be deleted from the
403 configuration name, if present. The definition should be an initializer
404 for an array of structures. Each array element should have three
405 elements: the switch name (a string constant, including the initial
406 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
407 indicate whether the string should be inserted or deleted, and the string
408 to be inserted or deleted (a string constant).
410 For example, on a machine where @samp{64} at the end of the
411 configuration name denotes a 64-bit target and you want the @samp{-32}
412 and @samp{-64} switches to select between 32- and 64-bit targets, you would
416 #define MODIFY_TARGET_NAME \
417 @{ @{ "-32", DELETE, "64"@}, \
418 @{"-64", ADD, "64"@}@}
422 @findex SYSTEM_INCLUDE_DIR
423 @item SYSTEM_INCLUDE_DIR
424 Define this macro as a C string constant if you wish to specify a
425 system-specific directory to search for header files before the standard
426 directory. @code{SYSTEM_INCLUDE_DIR} comes before
427 @code{STANDARD_INCLUDE_DIR} in the search order.
429 Cross compilers do not use this macro and do not search the directory
432 @findex STANDARD_INCLUDE_DIR
433 @item STANDARD_INCLUDE_DIR
434 Define this macro as a C string constant if you wish to override the
435 standard choice of @file{/usr/include} as the default prefix to
436 try when searching for header files.
438 Cross compilers do not use this macro and do not search either
439 @file{/usr/include} or its replacement.
441 @findex STANDARD_INCLUDE_COMPONENT
442 @item STANDARD_INCLUDE_COMPONENT
443 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
444 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
445 If you do not define this macro, no component is used.
447 @findex INCLUDE_DEFAULTS
448 @item INCLUDE_DEFAULTS
449 Define this macro if you wish to override the entire default search path
450 for include files. For a native compiler, the default search path
451 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
452 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
453 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
454 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
455 and specify private search areas for GCC. The directory
456 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
458 The definition should be an initializer for an array of structures.
459 Each array element should have four elements: the directory name (a
460 string constant), the component name (also a string constant), a flag
461 for C++-only directories,
462 and a flag showing that the includes in the directory don't need to be
463 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
464 the array with a null element.
466 The component name denotes what GNU package the include file is part of,
467 if any, in all upper-case letters. For example, it might be @samp{GCC}
468 or @samp{BINUTILS}. If the package is part of a vendor-supplied
469 operating system, code the component name as @samp{0}.
471 For example, here is the definition used for VAX/VMS:
474 #define INCLUDE_DEFAULTS \
476 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
477 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
478 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
485 Here is the order of prefixes tried for exec files:
489 Any prefixes specified by the user with @samp{-B}.
492 The environment variable @code{GCC_EXEC_PREFIX}, if any.
495 The directories specified by the environment variable @code{COMPILER_PATH}.
498 The macro @code{STANDARD_EXEC_PREFIX}.
501 @file{/usr/lib/gcc/}.
504 The macro @code{MD_EXEC_PREFIX}, if any.
507 Here is the order of prefixes tried for startfiles:
511 Any prefixes specified by the user with @samp{-B}.
514 The environment variable @code{GCC_EXEC_PREFIX}, if any.
517 The directories specified by the environment variable @code{LIBRARY_PATH}
518 (or port-specific name; native only, cross compilers do not use this).
521 The macro @code{STANDARD_EXEC_PREFIX}.
524 @file{/usr/lib/gcc/}.
527 The macro @code{MD_EXEC_PREFIX}, if any.
530 The macro @code{MD_STARTFILE_PREFIX}, if any.
533 The macro @code{STANDARD_STARTFILE_PREFIX}.
542 @node Run-time Target
543 @section Run-time Target Specification
544 @cindex run-time target specification
545 @cindex predefined macros
546 @cindex target specifications
548 @c prevent bad page break with this line
549 Here are run-time target specifications.
552 @findex CPP_PREDEFINES
554 Define this to be a string constant containing @samp{-D} options to
555 define the predefined macros that identify this machine and system.
556 These macros will be predefined unless the @option{-ansi} option (or a
557 @option{-std} option for strict ISO C conformance) is specified.
559 In addition, a parallel set of macros are predefined, whose names are
560 made by appending @samp{__} at the beginning and at the end. These
561 @samp{__} macros are permitted by the ISO standard, so they are
562 predefined regardless of whether @option{-ansi} or a @option{-std} option
565 For example, on the Sun, one can use the following value:
568 "-Dmc68000 -Dsun -Dunix"
571 The result is to define the macros @code{__mc68000__}, @code{__sun__}
572 and @code{__unix__} unconditionally, and the macros @code{mc68000},
573 @code{sun} and @code{unix} provided @samp{-ansi} is not specified.
575 @findex extern int target_flags
576 @item extern int target_flags;
577 This declaration should be present.
579 @cindex optional hardware or system features
580 @cindex features, optional, in system conventions
582 This series of macros is to allow compiler command arguments to
583 enable or disable the use of optional features of the target machine.
584 For example, one machine description serves both the 68000 and
585 the 68020; a command argument tells the compiler whether it should
586 use 68020-only instructions or not. This command argument works
587 by means of a macro @code{TARGET_68020} that tests a bit in
590 Define a macro @code{TARGET_@var{featurename}} for each such option.
591 Its definition should test a bit in @code{target_flags}. It is
592 recommended that a helper macro @code{TARGET_MASK_@var{featurename}}
593 is defined for each bit-value to test, and used in
594 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
598 #define TARGET_MASK_68020 1
599 #define TARGET_68020 (target_flags & TARGET_MASK_68020)
602 One place where these macros are used is in the condition-expressions
603 of instruction patterns. Note how @code{TARGET_68020} appears
604 frequently in the 68000 machine description file, @file{m68k.md}.
605 Another place they are used is in the definitions of the other
606 macros in the @file{@var{machine}.h} file.
608 @findex TARGET_SWITCHES
609 @item TARGET_SWITCHES
610 This macro defines names of command options to set and clear
611 bits in @code{target_flags}. Its definition is an initializer
612 with a subgrouping for each command option.
614 Each subgrouping contains a string constant, that defines the option
615 name, a number, which contains the bits to set in
616 @code{target_flags}, and a second string which is the description
617 displayed by --help. If the number is negative then the bits specified
618 by the number are cleared instead of being set. If the description
619 string is present but empty, then no help information will be displayed
620 for that option, but it will not count as an undocumented option. The
621 actual option name is made by appending @samp{-m} to the specified name.
623 One of the subgroupings should have a null string. The number in
624 this grouping is the default value for @code{target_flags}. Any
625 target options act starting with that value.
627 Here is an example which defines @samp{-m68000} and @samp{-m68020}
628 with opposite meanings, and picks the latter as the default:
631 #define TARGET_SWITCHES \
632 @{ @{ "68020", TARGET_MASK_68020, "" @}, \
633 @{ "68000", -TARGET_MASK_68020, "Compile for the 68000" @}, \
634 @{ "", TARGET_MASK_68020, "" @}@}
637 @findex TARGET_OPTIONS
639 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
640 options that have values. Its definition is an initializer with a
641 subgrouping for each command option.
643 Each subgrouping contains a string constant, that defines the fixed part
644 of the option name, the address of a variable, and a description string.
645 The variable, type @code{char *}, is set to the variable part of the
646 given option if the fixed part matches. The actual option name is made
647 by appending @samp{-m} to the specified name.
649 Here is an example which defines @samp{-mshort-data-@var{number}}. If the
650 given option is @samp{-mshort-data-512}, the variable @code{m88k_short_data}
651 will be set to the string @code{"512"}.
654 extern char *m88k_short_data;
655 #define TARGET_OPTIONS \
656 @{ @{ "short-data-", &m88k_short_data, "Specify the size of the short data section" @} @}
659 @findex TARGET_VERSION
661 This macro is a C statement to print on @code{stderr} a string
662 describing the particular machine description choice. Every machine
663 description should define @code{TARGET_VERSION}. For example:
667 #define TARGET_VERSION \
668 fprintf (stderr, " (68k, Motorola syntax)");
670 #define TARGET_VERSION \
671 fprintf (stderr, " (68k, MIT syntax)");
675 @findex OVERRIDE_OPTIONS
676 @item OVERRIDE_OPTIONS
677 Sometimes certain combinations of command options do not make sense on
678 a particular target machine. You can define a macro
679 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
680 defined, is executed once just after all the command options have been
683 Don't use this macro to turn on various extra optimizations for
684 @samp{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
686 @findex OPTIMIZATION_OPTIONS
687 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
688 Some machines may desire to change what optimizations are performed for
689 various optimization levels. This macro, if defined, is executed once
690 just after the optimization level is determined and before the remainder
691 of the command options have been parsed. Values set in this macro are
692 used as the default values for the other command line options.
694 @var{level} is the optimization level specified; 2 if @samp{-O2} is
695 specified, 1 if @samp{-O} is specified, and 0 if neither is specified.
697 @var{size} is non-zero if @samp{-Os} is specified and zero otherwise.
699 You should not use this macro to change options that are not
700 machine-specific. These should uniformly selected by the same
701 optimization level on all supported machines. Use this macro to enable
702 machine-specific optimizations.
704 @strong{Do not examine @code{write_symbols} in
705 this macro!} The debugging options are not supposed to alter the
708 @findex CAN_DEBUG_WITHOUT_FP
709 @item CAN_DEBUG_WITHOUT_FP
710 Define this macro if debugging can be performed even without a frame
711 pointer. If this macro is defined, GCC will turn on the
712 @samp{-fomit-frame-pointer} option whenever @samp{-O} is specified.
715 @node Per-Function Data
716 @section Defining data structures for per-function information.
717 @cindex per-function data
718 @cindex data structures
720 If the target needs to store information on a per-function basis, GCC
721 provides a macro and a couple of variables to allow this. Note, just
722 using statics to store the information is a bad idea, since GCC supports
723 nested functions, so you can be halfway through encoding one function
724 when another one comes along.
726 GCC defines a data structure called @code{struct function} which
727 contains all of the data specific to an individual function. This
728 structure contains a field called @code{machine} whose type is
729 @code{struct machine_function *}, which can be used by targets to point
730 to their own specific data.
732 If a target needs per-function specific data it should define the type
733 @code{struct machine_function} and also the macro
734 @code{INIT_EXPANDERS}. This macro should be used to initialise some or
735 all of the function pointers @code{init_machine_status},
736 @code{free_machine_status} and @code{mark_machine_status}. These
737 pointers are explained below.
739 One typical use of per-function, target specific data is to create an
740 RTX to hold the register containing the function's return address. This
741 RTX can then be used to implement the @code{__builtin_return_address}
742 function, for level 0.
744 Note - earlier implementations of GCC used a single data area to hold
745 all of the per-function information. Thus when processing of a nested
746 function began the old per-function data had to be pushed onto a
747 stack, and when the processing was finished, it had to be popped off the
748 stack. GCC used to provide function pointers called
749 @code{save_machine_status} and @code{restore_machine_status} to handle
750 the saving and restoring of the target specific information. Since the
751 single data area approach is no longer used, these pointers are no
754 The macro and function pointers are described below.
757 @findex INIT_EXPANDERS
759 Macro called to initialise any target specific information. This macro
760 is called once per function, before generation of any RTL has begun.
761 The intention of this macro is to allow the initialisation of the
762 function pointers below.
764 @findex init_machine_status
765 @item init_machine_status
766 This is a @code{void (*)(struct function *)} function pointer. If this
767 pointer is non-NULL it will be called once per function, before function
768 compilation starts, in order to allow the target to perform any target
769 specific initialisation of the @code{struct function} structure. It is
770 intended that this would be used to initialise the @code{machine} of
773 @findex free_machine_status
774 @item free_machine_status
775 This is a @code{void (*)(struct function *)} function pointer. If this
776 pointer is non-NULL it will be called once per function, after the
777 function has been compiled, in order to allow any memory allocated
778 during the @code{init_machine_status} function call to be freed.
780 @findex mark_machine_status
781 @item mark_machine_status
782 This is a @code{void (*)(struct function *)} function pointer. If this
783 pointer is non-NULL it will be called once per function in order to mark
784 any data items in the @code{struct machine_function} structure which
785 need garbage collection.
790 @section Storage Layout
791 @cindex storage layout
793 Note that the definitions of the macros in this table which are sizes or
794 alignments measured in bits do not need to be constant. They can be C
795 expressions that refer to static variables, such as the @code{target_flags}.
796 @xref{Run-time Target}.
799 @findex BITS_BIG_ENDIAN
800 @item BITS_BIG_ENDIAN
801 Define this macro to have the value 1 if the most significant bit in a
802 byte has the lowest number; otherwise define it to have the value zero.
803 This means that bit-field instructions count from the most significant
804 bit. If the machine has no bit-field instructions, then this must still
805 be defined, but it doesn't matter which value it is defined to. This
806 macro need not be a constant.
808 This macro does not affect the way structure fields are packed into
809 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
811 @findex BYTES_BIG_ENDIAN
812 @item BYTES_BIG_ENDIAN
813 Define this macro to have the value 1 if the most significant byte in a
814 word has the lowest number. This macro need not be a constant.
816 @findex WORDS_BIG_ENDIAN
817 @item WORDS_BIG_ENDIAN
818 Define this macro to have the value 1 if, in a multiword object, the
819 most significant word has the lowest number. This applies to both
820 memory locations and registers; GCC fundamentally assumes that the
821 order of words in memory is the same as the order in registers. This
822 macro need not be a constant.
824 @findex LIBGCC2_WORDS_BIG_ENDIAN
825 @item LIBGCC2_WORDS_BIG_ENDIAN
826 Define this macro if WORDS_BIG_ENDIAN is not constant. This must be a
827 constant value with the same meaning as WORDS_BIG_ENDIAN, which will be
828 used only when compiling libgcc2.c. Typically the value will be set
829 based on preprocessor defines.
831 @findex FLOAT_WORDS_BIG_ENDIAN
832 @item FLOAT_WORDS_BIG_ENDIAN
833 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
834 @code{TFmode} floating point numbers are stored in memory with the word
835 containing the sign bit at the lowest address; otherwise define it to
836 have the value 0. This macro need not be a constant.
838 You need not define this macro if the ordering is the same as for
841 @findex BITS_PER_UNIT
843 Define this macro to be the number of bits in an addressable storage
844 unit (byte); normally 8.
846 @findex BITS_PER_WORD
848 Number of bits in a word; normally 32.
850 @findex MAX_BITS_PER_WORD
851 @item MAX_BITS_PER_WORD
852 Maximum number of bits in a word. If this is undefined, the default is
853 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
854 largest value that @code{BITS_PER_WORD} can have at run-time.
856 @findex UNITS_PER_WORD
858 Number of storage units in a word; normally 4.
860 @findex MIN_UNITS_PER_WORD
861 @item MIN_UNITS_PER_WORD
862 Minimum number of units in a word. If this is undefined, the default is
863 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
864 smallest value that @code{UNITS_PER_WORD} can have at run-time.
868 Width of a pointer, in bits. You must specify a value no wider than the
869 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
870 you must define @code{POINTERS_EXTEND_UNSIGNED}.
872 @findex POINTERS_EXTEND_UNSIGNED
873 @item POINTERS_EXTEND_UNSIGNED
874 A C expression whose value is nonzero if pointers that need to be
875 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
876 be zero-extended and zero if they are to be sign-extended.
878 You need not define this macro if the @code{POINTER_SIZE} is equal
879 to the width of @code{Pmode}.
882 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
883 A macro to update @var{m} and @var{unsignedp} when an object whose type
884 is @var{type} and which has the specified mode and signedness is to be
885 stored in a register. This macro is only called when @var{type} is a
888 On most RISC machines, which only have operations that operate on a full
889 register, define this macro to set @var{m} to @code{word_mode} if
890 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
891 cases, only integer modes should be widened because wider-precision
892 floating-point operations are usually more expensive than their narrower
895 For most machines, the macro definition does not change @var{unsignedp}.
896 However, some machines, have instructions that preferentially handle
897 either signed or unsigned quantities of certain modes. For example, on
898 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
899 sign-extend the result to 64 bits. On such machines, set
900 @var{unsignedp} according to which kind of extension is more efficient.
902 Do not define this macro if it would never modify @var{m}.
904 @findex PROMOTE_FUNCTION_ARGS
905 @item PROMOTE_FUNCTION_ARGS
906 Define this macro if the promotion described by @code{PROMOTE_MODE}
907 should also be done for outgoing function arguments.
909 @findex PROMOTE_FUNCTION_RETURN
910 @item PROMOTE_FUNCTION_RETURN
911 Define this macro if the promotion described by @code{PROMOTE_MODE}
912 should also be done for the return value of functions.
914 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
915 promotions done by @code{PROMOTE_MODE}.
917 @findex PROMOTE_FOR_CALL_ONLY
918 @item PROMOTE_FOR_CALL_ONLY
919 Define this macro if the promotion described by @code{PROMOTE_MODE}
920 should @emph{only} be performed for outgoing function arguments or
921 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
922 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
924 @findex PARM_BOUNDARY
926 Normal alignment required for function parameters on the stack, in
927 bits. All stack parameters receive at least this much alignment
928 regardless of data type. On most machines, this is the same as the
931 @findex STACK_BOUNDARY
933 Define this macro if there is a guaranteed alignment for the stack
934 pointer on this machine. The definition is a C expression
935 for the desired alignment (measured in bits). This value is used as a
936 default if PREFERRED_STACK_BOUNDARY is not defined.
938 @findex PREFERRED_STACK_BOUNDARY
939 @item PREFERRED_STACK_BOUNDARY
940 Define this macro if you wish to preserve a certain alignment for
941 the stack pointer. The definition is a C expression
942 for the desired alignment (measured in bits). If STACK_BOUNDARY is
943 also defined, this macro must evaluate to a value equal to or larger
946 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
947 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
948 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
949 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
950 be momentarily unaligned while pushing arguments.
952 @findex FUNCTION_BOUNDARY
953 @item FUNCTION_BOUNDARY
954 Alignment required for a function entry point, in bits.
956 @findex BIGGEST_ALIGNMENT
957 @item BIGGEST_ALIGNMENT
958 Biggest alignment that any data type can require on this machine, in bits.
960 @findex MINIMUM_ATOMIC_ALIGNMENT
961 @item MINIMUM_ATOMIC_ALIGNMENT
962 If defined, the smallest alignment, in bits, that can be given to an
963 object that can be referenced in one operation, without disturbing any
964 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
965 on machines that don't have byte or half-word store operations.
967 @findex BIGGEST_FIELD_ALIGNMENT
968 @item BIGGEST_FIELD_ALIGNMENT
969 Biggest alignment that any structure or union field can require on this
970 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
971 structure and union fields only, unless the field alignment has been set
972 by the @code{__attribute__ ((aligned (@var{n})))} construct.
974 @findex ADJUST_FIELD_ALIGN
975 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
976 An expression for the alignment of a structure field @var{field} if the
977 alignment computed in the usual way is @var{computed}. GCC uses
978 this value instead of the value in @code{BIGGEST_ALIGNMENT} or
979 @code{BIGGEST_FIELD_ALIGNMENT}, if defined, for structure fields only.
981 @findex MAX_OFILE_ALIGNMENT
982 @item MAX_OFILE_ALIGNMENT
983 Biggest alignment supported by the object file format of this machine.
984 Use this macro to limit the alignment which can be specified using the
985 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
986 the default value is @code{BIGGEST_ALIGNMENT}.
988 @findex DATA_ALIGNMENT
989 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
990 If defined, a C expression to compute the alignment for a variable in
991 the static store. @var{type} is the data type, and @var{basic-align} is
992 the alignment that the object would ordinarily have. The value of this
993 macro is used instead of that alignment to align the object.
995 If this macro is not defined, then @var{basic-align} is used.
998 One use of this macro is to increase alignment of medium-size data to
999 make it all fit in fewer cache lines. Another is to cause character
1000 arrays to be word-aligned so that @code{strcpy} calls that copy
1001 constants to character arrays can be done inline.
1003 @findex CONSTANT_ALIGNMENT
1004 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1005 If defined, a C expression to compute the alignment given to a constant
1006 that is being placed in memory. @var{constant} is the constant and
1007 @var{basic-align} is the alignment that the object would ordinarily
1008 have. The value of this macro is used instead of that alignment to
1011 If this macro is not defined, then @var{basic-align} is used.
1013 The typical use of this macro is to increase alignment for string
1014 constants to be word aligned so that @code{strcpy} calls that copy
1015 constants can be done inline.
1017 @findex LOCAL_ALIGNMENT
1018 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1019 If defined, a C expression to compute the alignment for a variable in
1020 the local store. @var{type} is the data type, and @var{basic-align} is
1021 the alignment that the object would ordinarily have. The value of this
1022 macro is used instead of that alignment to align the object.
1024 If this macro is not defined, then @var{basic-align} is used.
1026 One use of this macro is to increase alignment of medium-size data to
1027 make it all fit in fewer cache lines.
1029 @findex EMPTY_FIELD_BOUNDARY
1030 @item EMPTY_FIELD_BOUNDARY
1031 Alignment in bits to be given to a structure bit field that follows an
1032 empty field such as @code{int : 0;}.
1034 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
1035 that results from an empty field.
1037 @findex STRUCTURE_SIZE_BOUNDARY
1038 @item STRUCTURE_SIZE_BOUNDARY
1039 Number of bits which any structure or union's size must be a multiple of.
1040 Each structure or union's size is rounded up to a multiple of this.
1042 If you do not define this macro, the default is the same as
1043 @code{BITS_PER_UNIT}.
1045 @findex STRICT_ALIGNMENT
1046 @item STRICT_ALIGNMENT
1047 Define this macro to be the value 1 if instructions will fail to work
1048 if given data not on the nominal alignment. If instructions will merely
1049 go slower in that case, define this macro as 0.
1051 @findex PCC_BITFIELD_TYPE_MATTERS
1052 @item PCC_BITFIELD_TYPE_MATTERS
1053 Define this if you wish to imitate the way many other C compilers handle
1054 alignment of bitfields and the structures that contain them.
1056 The behavior is that the type written for a bitfield (@code{int},
1057 @code{short}, or other integer type) imposes an alignment for the
1058 entire structure, as if the structure really did contain an ordinary
1059 field of that type. In addition, the bitfield is placed within the
1060 structure so that it would fit within such a field, not crossing a
1063 Thus, on most machines, a bitfield whose type is written as @code{int}
1064 would not cross a four-byte boundary, and would force four-byte
1065 alignment for the whole structure. (The alignment used may not be four
1066 bytes; it is controlled by the other alignment parameters.)
1068 If the macro is defined, its definition should be a C expression;
1069 a nonzero value for the expression enables this behavior.
1071 Note that if this macro is not defined, or its value is zero, some
1072 bitfields may cross more than one alignment boundary. The compiler can
1073 support such references if there are @samp{insv}, @samp{extv}, and
1074 @samp{extzv} insns that can directly reference memory.
1076 The other known way of making bitfields work is to define
1077 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1078 Then every structure can be accessed with fullwords.
1080 Unless the machine has bitfield instructions or you define
1081 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1082 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1084 If your aim is to make GCC use the same conventions for laying out
1085 bitfields as are used by another compiler, here is how to investigate
1086 what the other compiler does. Compile and run this program:
1105 printf ("Size of foo1 is %d\n",
1106 sizeof (struct foo1));
1107 printf ("Size of foo2 is %d\n",
1108 sizeof (struct foo2));
1113 If this prints 2 and 5, then the compiler's behavior is what you would
1114 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1116 @findex BITFIELD_NBYTES_LIMITED
1117 @item BITFIELD_NBYTES_LIMITED
1118 Like PCC_BITFIELD_TYPE_MATTERS except that its effect is limited to
1119 aligning a bitfield within the structure.
1121 @findex MEMBER_TYPE_FORCES_BLK
1122 @item MEMBER_TYPE_FORCES_BLK (@var{field})
1123 Return 1 if a structure or array containing @var{field} should be accessed using
1126 Normally, this is not needed. See the file @file{c4x.h} for an example
1127 of how to use this macro to prevent a structure having a floating point
1128 field from being accessed in an integer mode.
1130 @findex ROUND_TYPE_SIZE
1131 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
1132 Define this macro as an expression for the overall size of a type
1133 (given by @var{type} as a tree node) when the size computed in the
1134 usual way is @var{computed} and the alignment is @var{specified}.
1136 The default is to round @var{computed} up to a multiple of @var{specified}.
1138 @findex ROUND_TYPE_SIZE_UNIT
1139 @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified})
1140 Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are
1141 specified in units (bytes). If you define @code{ROUND_TYPE_SIZE},
1142 you must also define this macro and they must be defined consistently
1145 @findex ROUND_TYPE_ALIGN
1146 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1147 Define this macro as an expression for the alignment of a type (given
1148 by @var{type} as a tree node) if the alignment computed in the usual
1149 way is @var{computed} and the alignment explicitly specified was
1152 The default is to use @var{specified} if it is larger; otherwise, use
1153 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1155 @findex MAX_FIXED_MODE_SIZE
1156 @item MAX_FIXED_MODE_SIZE
1157 An integer expression for the size in bits of the largest integer
1158 machine mode that should actually be used. All integer machine modes of
1159 this size or smaller can be used for structures and unions with the
1160 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1161 (DImode)} is assumed.
1163 @findex VECTOR_MODE_SUPPORTED_P
1164 @item VECTOR_MODE_SUPPORTED_P(@var{mode})
1165 Define this macro to be nonzero if the port is prepared to handle insns
1166 involving vector mode @var{mode}. At the very least, it must have move
1167 patterns for this mode.
1169 @findex STACK_SAVEAREA_MODE
1170 @item STACK_SAVEAREA_MODE (@var{save_level})
1171 If defined, an expression of type @code{enum machine_mode} that
1172 specifies the mode of the save area operand of a
1173 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1174 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1175 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1176 having its mode specified.
1178 You need not define this macro if it always returns @code{Pmode}. You
1179 would most commonly define this macro if the
1180 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1183 @findex STACK_SIZE_MODE
1184 @item STACK_SIZE_MODE
1185 If defined, an expression of type @code{enum machine_mode} that
1186 specifies the mode of the size increment operand of an
1187 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1189 You need not define this macro if it always returns @code{word_mode}.
1190 You would most commonly define this macro if the @code{allocate_stack}
1191 pattern needs to support both a 32- and a 64-bit mode.
1193 @findex CHECK_FLOAT_VALUE
1194 @item CHECK_FLOAT_VALUE (@var{mode}, @var{value}, @var{overflow})
1195 A C statement to validate the value @var{value} (of type
1196 @code{double}) for mode @var{mode}. This means that you check whether
1197 @var{value} fits within the possible range of values for mode
1198 @var{mode} on this target machine. The mode @var{mode} is always
1199 a mode of class @code{MODE_FLOAT}. @var{overflow} is nonzero if
1200 the value is already known to be out of range.
1202 If @var{value} is not valid or if @var{overflow} is nonzero, you should
1203 set @var{overflow} to 1 and then assign some valid value to @var{value}.
1204 Allowing an invalid value to go through the compiler can produce
1205 incorrect assembler code which may even cause Unix assemblers to crash.
1207 This macro need not be defined if there is no work for it to do.
1209 @findex TARGET_FLOAT_FORMAT
1210 @item TARGET_FLOAT_FORMAT
1211 A code distinguishing the floating point format of the target machine.
1212 There are three defined values:
1215 @findex IEEE_FLOAT_FORMAT
1216 @item IEEE_FLOAT_FORMAT
1217 This code indicates IEEE floating point. It is the default; there is no
1218 need to define this macro when the format is IEEE.
1220 @findex VAX_FLOAT_FORMAT
1221 @item VAX_FLOAT_FORMAT
1222 This code indicates the peculiar format used on the Vax.
1224 @findex UNKNOWN_FLOAT_FORMAT
1225 @item UNKNOWN_FLOAT_FORMAT
1226 This code indicates any other format.
1229 The value of this macro is compared with @code{HOST_FLOAT_FORMAT}
1230 (@pxref{Config}) to determine whether the target machine has the same
1231 format as the host machine. If any other formats are actually in use on
1232 supported machines, new codes should be defined for them.
1234 The ordering of the component words of floating point values stored in
1235 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN} for the target
1236 machine and @code{HOST_FLOAT_WORDS_BIG_ENDIAN} for the host.
1238 @findex DEFAULT_VTABLE_THUNKS
1239 @item DEFAULT_VTABLE_THUNKS
1240 GCC supports two ways of implementing C++ vtables: traditional or with
1241 so-called ``thunks''. The flag @samp{-fvtable-thunk} chooses between them.
1242 Define this macro to be a C expression for the default value of that flag.
1243 If @code{DEFAULT_VTABLE_THUNKS} is 0, GCC uses the traditional
1244 implementation by default. The ``thunk'' implementation is more efficient
1245 (especially if you have provided an implementation of
1246 @code{ASM_OUTPUT_MI_THUNK}, see @ref{Function Entry}), but is not binary
1247 compatible with code compiled using the traditional implementation.
1248 If you are writing a new port, define @code{DEFAULT_VTABLE_THUNKS} to 1.
1250 If you do not define this macro, the default for @samp{-fvtable-thunk} is 0.
1254 @section Layout of Source Language Data Types
1256 These macros define the sizes and other characteristics of the standard
1257 basic data types used in programs being compiled. Unlike the macros in
1258 the previous section, these apply to specific features of C and related
1259 languages, rather than to fundamental aspects of storage layout.
1262 @findex INT_TYPE_SIZE
1264 A C expression for the size in bits of the type @code{int} on the
1265 target machine. If you don't define this, the default is one word.
1267 @findex MAX_INT_TYPE_SIZE
1268 @item MAX_INT_TYPE_SIZE
1269 Maximum number for the size in bits of the type @code{int} on the target
1270 machine. If this is undefined, the default is @code{INT_TYPE_SIZE}.
1271 Otherwise, it is the constant value that is the largest value that
1272 @code{INT_TYPE_SIZE} can have at run-time. This is used in @code{cpp}.
1274 @findex SHORT_TYPE_SIZE
1275 @item SHORT_TYPE_SIZE
1276 A C expression for the size in bits of the type @code{short} on the
1277 target machine. If you don't define this, the default is half a word.
1278 (If this would be less than one storage unit, it is rounded up to one
1281 @findex LONG_TYPE_SIZE
1282 @item LONG_TYPE_SIZE
1283 A C expression for the size in bits of the type @code{long} on the
1284 target machine. If you don't define this, the default is one word.
1286 @findex MAX_LONG_TYPE_SIZE
1287 @item MAX_LONG_TYPE_SIZE
1288 Maximum number for the size in bits of the type @code{long} on the
1289 target machine. If this is undefined, the default is
1290 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1291 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1294 @findex LONG_LONG_TYPE_SIZE
1295 @item LONG_LONG_TYPE_SIZE
1296 A C expression for the size in bits of the type @code{long long} on the
1297 target machine. If you don't define this, the default is two
1298 words. If you want to support GNU Ada on your machine, the value of this
1299 macro must be at least 64.
1301 @findex CHAR_TYPE_SIZE
1302 @item CHAR_TYPE_SIZE
1303 A C expression for the size in bits of the type @code{char} on the
1304 target machine. If you don't define this, the default is
1305 @code{BITS_PER_UNIT}.
1307 @findex MAX_CHAR_TYPE_SIZE
1308 @item MAX_CHAR_TYPE_SIZE
1309 Maximum number for the size in bits of the type @code{char} on the
1310 target machine. If this is undefined, the default is
1311 @code{CHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1312 largest value that @code{CHAR_TYPE_SIZE} can have at run-time. This is
1315 @findex FLOAT_TYPE_SIZE
1316 @item FLOAT_TYPE_SIZE
1317 A C expression for the size in bits of the type @code{float} on the
1318 target machine. If you don't define this, the default is one word.
1320 @findex DOUBLE_TYPE_SIZE
1321 @item DOUBLE_TYPE_SIZE
1322 A C expression for the size in bits of the type @code{double} on the
1323 target machine. If you don't define this, the default is two
1326 @findex LONG_DOUBLE_TYPE_SIZE
1327 @item LONG_DOUBLE_TYPE_SIZE
1328 A C expression for the size in bits of the type @code{long double} on
1329 the target machine. If you don't define this, the default is two
1332 @findex WIDEST_HARDWARE_FP_SIZE
1333 @item WIDEST_HARDWARE_FP_SIZE
1334 A C expression for the size in bits of the widest floating-point format
1335 supported by the hardware. If you define this macro, you must specify a
1336 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1337 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1340 @findex DEFAULT_SIGNED_CHAR
1341 @item DEFAULT_SIGNED_CHAR
1342 An expression whose value is 1 or 0, according to whether the type
1343 @code{char} should be signed or unsigned by default. The user can
1344 always override this default with the options @samp{-fsigned-char}
1345 and @samp{-funsigned-char}.
1347 @findex DEFAULT_SHORT_ENUMS
1348 @item DEFAULT_SHORT_ENUMS
1349 A C expression to determine whether to give an @code{enum} type
1350 only as many bytes as it takes to represent the range of possible values
1351 of that type. A nonzero value means to do that; a zero value means all
1352 @code{enum} types should be allocated like @code{int}.
1354 If you don't define the macro, the default is 0.
1358 A C expression for a string describing the name of the data type to use
1359 for size values. The typedef name @code{size_t} is defined using the
1360 contents of the string.
1362 The string can contain more than one keyword. If so, separate them with
1363 spaces, and write first any length keyword, then @code{unsigned} if
1364 appropriate, and finally @code{int}. The string must exactly match one
1365 of the data type names defined in the function
1366 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1367 omit @code{int} or change the order---that would cause the compiler to
1370 If you don't define this macro, the default is @code{"long unsigned
1373 @findex PTRDIFF_TYPE
1375 A C expression for a string describing the name of the data type to use
1376 for the result of subtracting two pointers. The typedef name
1377 @code{ptrdiff_t} is defined using the contents of the string. See
1378 @code{SIZE_TYPE} above for more information.
1380 If you don't define this macro, the default is @code{"long int"}.
1384 A C expression for a string describing the name of the data type to use
1385 for wide characters. The typedef name @code{wchar_t} is defined using
1386 the contents of the string. See @code{SIZE_TYPE} above for more
1389 If you don't define this macro, the default is @code{"int"}.
1391 @findex WCHAR_TYPE_SIZE
1392 @item WCHAR_TYPE_SIZE
1393 A C expression for the size in bits of the data type for wide
1394 characters. This is used in @code{cpp}, which cannot make use of
1397 @findex MAX_WCHAR_TYPE_SIZE
1398 @item MAX_WCHAR_TYPE_SIZE
1399 Maximum number for the size in bits of the data type for wide
1400 characters. If this is undefined, the default is
1401 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1402 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1407 A C expression for a string describing the name of the data type to
1408 use for wide characters passed to @code{printf} and returned from
1409 @code{getwc}. The typedef name @code{wint_t} is defined using the
1410 contents of the string. See @code{SIZE_TYPE} above for more
1413 If you don't define this macro, the default is @code{"unsigned int"}.
1417 A C expression for a string describing the name of the data type that
1418 can represent any value of any standard or extended signed integer type.
1419 The typedef name @code{intmax_t} is defined using the contents of the
1420 string. See @code{SIZE_TYPE} above for more information.
1422 If you don't define this macro, the default is the first of
1423 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1424 much precision as @code{long long int}.
1426 @findex UINTMAX_TYPE
1428 A C expression for a string describing the name of the data type that
1429 can represent any value of any standard or extended unsigned integer
1430 type. The typedef name @code{uintmax_t} is defined using the contents
1431 of the string. See @code{SIZE_TYPE} above for more information.
1433 If you don't define this macro, the default is the first of
1434 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1435 unsigned int"} that has as much precision as @code{long long unsigned
1438 @findex OBJC_SELECTORS_WITHOUT_LABELS
1439 @item OBJC_SELECTORS_WITHOUT_LABELS
1440 Define this macro if the compiler can group all the selectors together
1441 into a vector and use just one label at the beginning of the vector.
1442 Otherwise, the compiler must give each selector its own assembler
1445 On certain machines, it is important to have a separate label for each
1446 selector because this enables the linker to eliminate duplicate selectors.
1450 A C constant expression for the integer value for escape sequence
1455 @findex TARGET_NEWLINE
1458 @itemx TARGET_NEWLINE
1459 C constant expressions for the integer values for escape sequences
1460 @samp{\b}, @samp{\t} and @samp{\n}.
1468 C constant expressions for the integer values for escape sequences
1469 @samp{\v}, @samp{\f} and @samp{\r}.
1473 @section Register Usage
1474 @cindex register usage
1476 This section explains how to describe what registers the target machine
1477 has, and how (in general) they can be used.
1479 The description of which registers a specific instruction can use is
1480 done with register classes; see @ref{Register Classes}. For information
1481 on using registers to access a stack frame, see @ref{Frame Registers}.
1482 For passing values in registers, see @ref{Register Arguments}.
1483 For returning values in registers, see @ref{Scalar Return}.
1486 * Register Basics:: Number and kinds of registers.
1487 * Allocation Order:: Order in which registers are allocated.
1488 * Values in Registers:: What kinds of values each reg can hold.
1489 * Leaf Functions:: Renumbering registers for leaf functions.
1490 * Stack Registers:: Handling a register stack such as 80387.
1493 @node Register Basics
1494 @subsection Basic Characteristics of Registers
1496 @c prevent bad page break with this line
1497 Registers have various characteristics.
1500 @findex FIRST_PSEUDO_REGISTER
1501 @item FIRST_PSEUDO_REGISTER
1502 Number of hardware registers known to the compiler. They receive
1503 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1504 pseudo register's number really is assigned the number
1505 @code{FIRST_PSEUDO_REGISTER}.
1507 @item FIXED_REGISTERS
1508 @findex FIXED_REGISTERS
1509 @cindex fixed register
1510 An initializer that says which registers are used for fixed purposes
1511 all throughout the compiled code and are therefore not available for
1512 general allocation. These would include the stack pointer, the frame
1513 pointer (except on machines where that can be used as a general
1514 register when no frame pointer is needed), the program counter on
1515 machines where that is considered one of the addressable registers,
1516 and any other numbered register with a standard use.
1518 This information is expressed as a sequence of numbers, separated by
1519 commas and surrounded by braces. The @var{n}th number is 1 if
1520 register @var{n} is fixed, 0 otherwise.
1522 The table initialized from this macro, and the table initialized by
1523 the following one, may be overridden at run time either automatically,
1524 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1525 the user with the command options @samp{-ffixed-@var{reg}},
1526 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}}.
1528 @findex CALL_USED_REGISTERS
1529 @item CALL_USED_REGISTERS
1530 @cindex call-used register
1531 @cindex call-clobbered register
1532 @cindex call-saved register
1533 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1534 clobbered (in general) by function calls as well as for fixed
1535 registers. This macro therefore identifies the registers that are not
1536 available for general allocation of values that must live across
1539 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1540 automatically saves it on function entry and restores it on function
1541 exit, if the register is used within the function.
1543 @findex HARD_REGNO_CALL_PART_CLOBBERED
1544 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1545 @cindex call-used register
1546 @cindex call-clobbered register
1547 @cindex call-saved register
1548 A C expression that is non-zero if it is not permissible to store a
1549 value of mode @var{mode} in hard register number @var{regno} across a
1550 call without some part of it being clobbered. For most machines this
1551 macro need not be defined. It is only required for machines that do not
1552 preserve the entire contents of a register across a call.
1554 @findex CONDITIONAL_REGISTER_USAGE
1556 @findex call_used_regs
1557 @item CONDITIONAL_REGISTER_USAGE
1558 Zero or more C statements that may conditionally modify five variables
1559 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1560 (these three are of type @code{char []}), @code{reg_names} (of type
1561 @code{const char * []}) and @code{reg_class_contents} (of type
1562 @code{HARD_REG_SET}).
1563 Before the macro is called @code{fixed_regs}, @code{call_used_regs}
1564 @code{reg_class_contents} and @code{reg_names} have been initialized
1565 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1566 @code{REG_CLASS_CONTENTS} and @code{REGISTER_NAMES}, respectively,
1567 @code{global_regs} has been cleared, and any @samp{-ffixed-@var{reg}},
1568 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}} command
1569 options have been applied.
1571 This is necessary in case the fixed or call-clobbered registers depend
1574 You need not define this macro if it has no work to do.
1576 @cindex disabling certain registers
1577 @cindex controlling register usage
1578 If the usage of an entire class of registers depends on the target
1579 flags, you may indicate this to GCC by using this macro to modify
1580 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1581 registers in the classes which should not be used by GCC. Also define
1582 the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1583 is called with a letter for a class that shouldn't be used.
1585 (However, if this class is not included in @code{GENERAL_REGS} and all
1586 of the insn patterns whose constraints permit this class are
1587 controlled by target switches, then GCC will automatically avoid using
1588 these registers when the target switches are opposed to them.)
1590 @findex NON_SAVING_SETJMP
1591 @item NON_SAVING_SETJMP
1592 If this macro is defined and has a nonzero value, it means that
1593 @code{setjmp} and related functions fail to save the registers, or that
1594 @code{longjmp} fails to restore them. To compensate, the compiler
1595 avoids putting variables in registers in functions that use
1598 @findex INCOMING_REGNO
1599 @item INCOMING_REGNO (@var{out})
1600 Define this macro if the target machine has register windows. This C
1601 expression returns the register number as seen by the called function
1602 corresponding to the register number @var{out} as seen by the calling
1603 function. Return @var{out} if register number @var{out} is not an
1606 @findex OUTGOING_REGNO
1607 @item OUTGOING_REGNO (@var{in})
1608 Define this macro if the target machine has register windows. This C
1609 expression returns the register number as seen by the calling function
1610 corresponding to the register number @var{in} as seen by the called
1611 function. Return @var{in} if register number @var{in} is not an inbound
1615 @item LOCAL_REGNO (@var{regno})
1616 Define this macro if the target machine has register windows. This C
1617 expression returns true if the register is call-saved but is in the
1618 register window. Unlike most call-saved registers, such registers
1619 need not be explicitly restored on function exit or during non-local
1625 If the program counter has a register number, define this as that
1626 register number. Otherwise, do not define it.
1630 @node Allocation Order
1631 @subsection Order of Allocation of Registers
1632 @cindex order of register allocation
1633 @cindex register allocation order
1635 @c prevent bad page break with this line
1636 Registers are allocated in order.
1639 @findex REG_ALLOC_ORDER
1640 @item REG_ALLOC_ORDER
1641 If defined, an initializer for a vector of integers, containing the
1642 numbers of hard registers in the order in which GCC should prefer
1643 to use them (from most preferred to least).
1645 If this macro is not defined, registers are used lowest numbered first
1646 (all else being equal).
1648 One use of this macro is on machines where the highest numbered
1649 registers must always be saved and the save-multiple-registers
1650 instruction supports only sequences of consecutive registers. On such
1651 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1652 the highest numbered allocable register first.
1654 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1655 @item ORDER_REGS_FOR_LOCAL_ALLOC
1656 A C statement (sans semicolon) to choose the order in which to allocate
1657 hard registers for pseudo-registers local to a basic block.
1659 Store the desired register order in the array @code{reg_alloc_order}.
1660 Element 0 should be the register to allocate first; element 1, the next
1661 register; and so on.
1663 The macro body should not assume anything about the contents of
1664 @code{reg_alloc_order} before execution of the macro.
1666 On most machines, it is not necessary to define this macro.
1669 @node Values in Registers
1670 @subsection How Values Fit in Registers
1672 This section discusses the macros that describe which kinds of values
1673 (specifically, which machine modes) each register can hold, and how many
1674 consecutive registers are needed for a given mode.
1677 @findex HARD_REGNO_NREGS
1678 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1679 A C expression for the number of consecutive hard registers, starting
1680 at register number @var{regno}, required to hold a value of mode
1683 On a machine where all registers are exactly one word, a suitable
1684 definition of this macro is
1687 #define HARD_REGNO_NREGS(REGNO, MODE) \
1688 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1692 @findex HARD_REGNO_MODE_OK
1693 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1694 A C expression that is nonzero if it is permissible to store a value
1695 of mode @var{mode} in hard register number @var{regno} (or in several
1696 registers starting with that one). For a machine where all registers
1697 are equivalent, a suitable definition is
1700 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1703 You need not include code to check for the numbers of fixed registers,
1704 because the allocation mechanism considers them to be always occupied.
1706 @cindex register pairs
1707 On some machines, double-precision values must be kept in even/odd
1708 register pairs. You can implement that by defining this macro to reject
1709 odd register numbers for such modes.
1711 The minimum requirement for a mode to be OK in a register is that the
1712 @samp{mov@var{mode}} instruction pattern support moves between the
1713 register and other hard register in the same class and that moving a
1714 value into the register and back out not alter it.
1716 Since the same instruction used to move @code{word_mode} will work for
1717 all narrower integer modes, it is not necessary on any machine for
1718 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1719 you define patterns @samp{movhi}, etc., to take advantage of this. This
1720 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1721 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1724 Many machines have special registers for floating point arithmetic.
1725 Often people assume that floating point machine modes are allowed only
1726 in floating point registers. This is not true. Any registers that
1727 can hold integers can safely @emph{hold} a floating point machine
1728 mode, whether or not floating arithmetic can be done on it in those
1729 registers. Integer move instructions can be used to move the values.
1731 On some machines, though, the converse is true: fixed-point machine
1732 modes may not go in floating registers. This is true if the floating
1733 registers normalize any value stored in them, because storing a
1734 non-floating value there would garble it. In this case,
1735 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1736 floating registers. But if the floating registers do not automatically
1737 normalize, if you can store any bit pattern in one and retrieve it
1738 unchanged without a trap, then any machine mode may go in a floating
1739 register, so you can define this macro to say so.
1741 The primary significance of special floating registers is rather that
1742 they are the registers acceptable in floating point arithmetic
1743 instructions. However, this is of no concern to
1744 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1745 constraints for those instructions.
1747 On some machines, the floating registers are especially slow to access,
1748 so that it is better to store a value in a stack frame than in such a
1749 register if floating point arithmetic is not being done. As long as the
1750 floating registers are not in class @code{GENERAL_REGS}, they will not
1751 be used unless some pattern's constraint asks for one.
1753 @findex MODES_TIEABLE_P
1754 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
1755 A C expression that is nonzero if a value of mode
1756 @var{mode1} is accessible in mode @var{mode2} without copying.
1758 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
1759 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
1760 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
1761 should be nonzero. If they differ for any @var{r}, you should define
1762 this macro to return zero unless some other mechanism ensures the
1763 accessibility of the value in a narrower mode.
1765 You should define this macro to return nonzero in as many cases as
1766 possible since doing so will allow GCC to perform better register
1769 @findex AVOID_CCMODE_COPIES
1770 @item AVOID_CCMODE_COPIES
1771 Define this macro if the compiler should avoid copies to/from @code{CCmode}
1772 registers. You should only define this macro if support for copying to/from
1773 @code{CCmode} is incomplete.
1775 @findex SUBREG_REGNO_OFFSET
1776 @item SUBREG_REGNO_OFFSET
1777 Define this macro if the compiler needs to handle subregs in a non-standard
1778 way. The macro returns the correct regno offset for mode @code{YMODE} given
1779 a subreg of type @code{XMODE}.
1780 This macro takes 4 parameters:
1781 @code{XREGNO} - A regno of an inner hard subreg_reg (or what will become one).
1782 @code{XMODE} - The mode of xregno.
1783 @code{OFFSET} - The byte offset.
1784 @code{YMODE} - The mode of a top level SUBREG (or what may become one).
1785 The default function can be found in rtlanal.c, function
1786 @code{subreg_regno_offset}. Normally this does not need to be defined.
1789 @node Leaf Functions
1790 @subsection Handling Leaf Functions
1792 @cindex leaf functions
1793 @cindex functions, leaf
1794 On some machines, a leaf function (i.e., one which makes no calls) can run
1795 more efficiently if it does not make its own register window. Often this
1796 means it is required to receive its arguments in the registers where they
1797 are passed by the caller, instead of the registers where they would
1800 The special treatment for leaf functions generally applies only when
1801 other conditions are met; for example, often they may use only those
1802 registers for its own variables and temporaries. We use the term ``leaf
1803 function'' to mean a function that is suitable for this special
1804 handling, so that functions with no calls are not necessarily ``leaf
1807 GCC assigns register numbers before it knows whether the function is
1808 suitable for leaf function treatment. So it needs to renumber the
1809 registers in order to output a leaf function. The following macros
1813 @findex LEAF_REGISTERS
1814 @item LEAF_REGISTERS
1815 Name of a char vector, indexed by hard register number, which
1816 contains 1 for a register that is allowable in a candidate for leaf
1819 If leaf function treatment involves renumbering the registers, then the
1820 registers marked here should be the ones before renumbering---those that
1821 GCC would ordinarily allocate. The registers which will actually be
1822 used in the assembler code, after renumbering, should not be marked with 1
1825 Define this macro only if the target machine offers a way to optimize
1826 the treatment of leaf functions.
1828 @findex LEAF_REG_REMAP
1829 @item LEAF_REG_REMAP (@var{regno})
1830 A C expression whose value is the register number to which @var{regno}
1831 should be renumbered, when a function is treated as a leaf function.
1833 If @var{regno} is a register number which should not appear in a leaf
1834 function before renumbering, then the expression should yield -1, which
1835 will cause the compiler to abort.
1837 Define this macro only if the target machine offers a way to optimize the
1838 treatment of leaf functions, and registers need to be renumbered to do
1842 @findex current_function_is_leaf
1843 @findex current_function_uses_only_leaf_regs
1844 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
1845 treat leaf functions specially. They can test the C variable
1846 @code{current_function_is_leaf} which is nonzero for leaf functions.
1847 @code{current_function_is_leaf} is set prior to local register allocation
1848 and is valid for the remaining compiler passes. They can also test the C
1849 variable @code{current_function_uses_only_leaf_regs} which is nonzero for
1850 leaf functions which only use leaf registers.
1851 @code{current_function_uses_only_leaf_regs} is valid after reload and is
1852 only useful if @code{LEAF_REGISTERS} is defined.
1853 @c changed this to fix overfull. ALSO: why the "it" at the beginning
1854 @c of the next paragraph?! --mew 2feb93
1856 @node Stack Registers
1857 @subsection Registers That Form a Stack
1859 There are special features to handle computers where some of the
1860 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
1861 Stack registers are normally written by pushing onto the stack, and are
1862 numbered relative to the top of the stack.
1864 Currently, GCC can only handle one group of stack-like registers, and
1865 they must be consecutively numbered.
1870 Define this if the machine has any stack-like registers.
1872 @findex FIRST_STACK_REG
1873 @item FIRST_STACK_REG
1874 The number of the first stack-like register. This one is the top
1877 @findex LAST_STACK_REG
1878 @item LAST_STACK_REG
1879 The number of the last stack-like register. This one is the bottom of
1883 @node Register Classes
1884 @section Register Classes
1885 @cindex register class definitions
1886 @cindex class definitions, register
1888 On many machines, the numbered registers are not all equivalent.
1889 For example, certain registers may not be allowed for indexed addressing;
1890 certain registers may not be allowed in some instructions. These machine
1891 restrictions are described to the compiler using @dfn{register classes}.
1893 You define a number of register classes, giving each one a name and saying
1894 which of the registers belong to it. Then you can specify register classes
1895 that are allowed as operands to particular instruction patterns.
1899 In general, each register will belong to several classes. In fact, one
1900 class must be named @code{ALL_REGS} and contain all the registers. Another
1901 class must be named @code{NO_REGS} and contain no registers. Often the
1902 union of two classes will be another class; however, this is not required.
1904 @findex GENERAL_REGS
1905 One of the classes must be named @code{GENERAL_REGS}. There is nothing
1906 terribly special about the name, but the operand constraint letters
1907 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
1908 the same as @code{ALL_REGS}, just define it as a macro which expands
1911 Order the classes so that if class @var{x} is contained in class @var{y}
1912 then @var{x} has a lower class number than @var{y}.
1914 The way classes other than @code{GENERAL_REGS} are specified in operand
1915 constraints is through machine-dependent operand constraint letters.
1916 You can define such letters to correspond to various classes, then use
1917 them in operand constraints.
1919 You should define a class for the union of two classes whenever some
1920 instruction allows both classes. For example, if an instruction allows
1921 either a floating point (coprocessor) register or a general register for a
1922 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
1923 which includes both of them. Otherwise you will get suboptimal code.
1925 You must also specify certain redundant information about the register
1926 classes: for each class, which classes contain it and which ones are
1927 contained in it; for each pair of classes, the largest class contained
1930 When a value occupying several consecutive registers is expected in a
1931 certain class, all the registers used must belong to that class.
1932 Therefore, register classes cannot be used to enforce a requirement for
1933 a register pair to start with an even-numbered register. The way to
1934 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
1936 Register classes used for input-operands of bitwise-and or shift
1937 instructions have a special requirement: each such class must have, for
1938 each fixed-point machine mode, a subclass whose registers can transfer that
1939 mode to or from memory. For example, on some machines, the operations for
1940 single-byte values (@code{QImode}) are limited to certain registers. When
1941 this is so, each register class that is used in a bitwise-and or shift
1942 instruction must have a subclass consisting of registers from which
1943 single-byte values can be loaded or stored. This is so that
1944 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
1947 @findex enum reg_class
1948 @item enum reg_class
1949 An enumeral type that must be defined with all the register class names
1950 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
1951 must be the last register class, followed by one more enumeral value,
1952 @code{LIM_REG_CLASSES}, which is not a register class but rather
1953 tells how many classes there are.
1955 Each register class has a number, which is the value of casting
1956 the class name to type @code{int}. The number serves as an index
1957 in many of the tables described below.
1959 @findex N_REG_CLASSES
1961 The number of distinct register classes, defined as follows:
1964 #define N_REG_CLASSES (int) LIM_REG_CLASSES
1967 @findex REG_CLASS_NAMES
1968 @item REG_CLASS_NAMES
1969 An initializer containing the names of the register classes as C string
1970 constants. These names are used in writing some of the debugging dumps.
1972 @findex REG_CLASS_CONTENTS
1973 @item REG_CLASS_CONTENTS
1974 An initializer containing the contents of the register classes, as integers
1975 which are bit masks. The @var{n}th integer specifies the contents of class
1976 @var{n}. The way the integer @var{mask} is interpreted is that
1977 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
1979 When the machine has more than 32 registers, an integer does not suffice.
1980 Then the integers are replaced by sub-initializers, braced groupings containing
1981 several integers. Each sub-initializer must be suitable as an initializer
1982 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
1983 In this situation, the first integer in each sub-initializer corresponds to
1984 registers 0 through 31, the second integer to registers 32 through 63, and
1987 @findex REGNO_REG_CLASS
1988 @item REGNO_REG_CLASS (@var{regno})
1989 A C expression whose value is a register class containing hard register
1990 @var{regno}. In general there is more than one such class; choose a class
1991 which is @dfn{minimal}, meaning that no smaller class also contains the
1994 @findex BASE_REG_CLASS
1995 @item BASE_REG_CLASS
1996 A macro whose definition is the name of the class to which a valid
1997 base register must belong. A base register is one used in an address
1998 which is the register value plus a displacement.
2000 @findex INDEX_REG_CLASS
2001 @item INDEX_REG_CLASS
2002 A macro whose definition is the name of the class to which a valid
2003 index register must belong. An index register is one used in an
2004 address where its value is either multiplied by a scale factor or
2005 added to another register (as well as added to a displacement).
2007 @findex REG_CLASS_FROM_LETTER
2008 @item REG_CLASS_FROM_LETTER (@var{char})
2009 A C expression which defines the machine-dependent operand constraint
2010 letters for register classes. If @var{char} is such a letter, the
2011 value should be the register class corresponding to it. Otherwise,
2012 the value should be @code{NO_REGS}. The register letter @samp{r},
2013 corresponding to class @code{GENERAL_REGS}, will not be passed
2014 to this macro; you do not need to handle it.
2016 @findex REGNO_OK_FOR_BASE_P
2017 @item REGNO_OK_FOR_BASE_P (@var{num})
2018 A C expression which is nonzero if register number @var{num} is
2019 suitable for use as a base register in operand addresses. It may be
2020 either a suitable hard register or a pseudo register that has been
2021 allocated such a hard register.
2023 @findex REGNO_MODE_OK_FOR_BASE_P
2024 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2025 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2026 that expression may examine the mode of the memory reference in
2027 @var{mode}. You should define this macro if the mode of the memory
2028 reference affects whether a register may be used as a base register. If
2029 you define this macro, the compiler will use it instead of
2030 @code{REGNO_OK_FOR_BASE_P}.
2032 @findex REGNO_OK_FOR_INDEX_P
2033 @item REGNO_OK_FOR_INDEX_P (@var{num})
2034 A C expression which is nonzero if register number @var{num} is
2035 suitable for use as an index register in operand addresses. It may be
2036 either a suitable hard register or a pseudo register that has been
2037 allocated such a hard register.
2039 The difference between an index register and a base register is that
2040 the index register may be scaled. If an address involves the sum of
2041 two registers, neither one of them scaled, then either one may be
2042 labeled the ``base'' and the other the ``index''; but whichever
2043 labeling is used must fit the machine's constraints of which registers
2044 may serve in each capacity. The compiler will try both labelings,
2045 looking for one that is valid, and will reload one or both registers
2046 only if neither labeling works.
2048 @findex PREFERRED_RELOAD_CLASS
2049 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2050 A C expression that places additional restrictions on the register class
2051 to use when it is necessary to copy value @var{x} into a register in class
2052 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2053 another, smaller class. On many machines, the following definition is
2057 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2060 Sometimes returning a more restrictive class makes better code. For
2061 example, on the 68000, when @var{x} is an integer constant that is in range
2062 for a @samp{moveq} instruction, the value of this macro is always
2063 @code{DATA_REGS} as long as @var{class} includes the data registers.
2064 Requiring a data register guarantees that a @samp{moveq} will be used.
2066 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2067 you can force @var{x} into a memory constant. This is useful on
2068 certain machines where immediate floating values cannot be loaded into
2069 certain kinds of registers.
2071 @findex PREFERRED_OUTPUT_RELOAD_CLASS
2072 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2073 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2074 input reloads. If you don't define this macro, the default is to use
2075 @var{class}, unchanged.
2077 @findex LIMIT_RELOAD_CLASS
2078 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2079 A C expression that places additional restrictions on the register class
2080 to use when it is necessary to be able to hold a value of mode
2081 @var{mode} in a reload register for which class @var{class} would
2084 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2085 there are certain modes that simply can't go in certain reload classes.
2087 The value is a register class; perhaps @var{class}, or perhaps another,
2090 Don't define this macro unless the target machine has limitations which
2091 require the macro to do something nontrivial.
2093 @findex SECONDARY_RELOAD_CLASS
2094 @findex SECONDARY_INPUT_RELOAD_CLASS
2095 @findex SECONDARY_OUTPUT_RELOAD_CLASS
2096 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2097 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2098 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2099 Many machines have some registers that cannot be copied directly to or
2100 from memory or even from other types of registers. An example is the
2101 @samp{MQ} register, which on most machines, can only be copied to or
2102 from general registers, but not memory. Some machines allow copying all
2103 registers to and from memory, but require a scratch register for stores
2104 to some memory locations (e.g., those with symbolic address on the RT,
2105 and those with certain symbolic address on the Sparc when compiling
2106 PIC). In some cases, both an intermediate and a scratch register are
2109 You should define these macros to indicate to the reload phase that it may
2110 need to allocate at least one register for a reload in addition to the
2111 register to contain the data. Specifically, if copying @var{x} to a
2112 register @var{class} in @var{mode} requires an intermediate register,
2113 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2114 largest register class all of whose registers can be used as
2115 intermediate registers or scratch registers.
2117 If copying a register @var{class} in @var{mode} to @var{x} requires an
2118 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2119 should be defined to return the largest register class required. If the
2120 requirements for input and output reloads are the same, the macro
2121 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2124 The values returned by these macros are often @code{GENERAL_REGS}.
2125 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2126 can be directly copied to or from a register of @var{class} in
2127 @var{mode} without requiring a scratch register. Do not define this
2128 macro if it would always return @code{NO_REGS}.
2130 If a scratch register is required (either with or without an
2131 intermediate register), you should define patterns for
2132 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2133 (@pxref{Standard Names}. These patterns, which will normally be
2134 implemented with a @code{define_expand}, should be similar to the
2135 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2138 Define constraints for the reload register and scratch register that
2139 contain a single register class. If the original reload register (whose
2140 class is @var{class}) can meet the constraint given in the pattern, the
2141 value returned by these macros is used for the class of the scratch
2142 register. Otherwise, two additional reload registers are required.
2143 Their classes are obtained from the constraints in the insn pattern.
2145 @var{x} might be a pseudo-register or a @code{subreg} of a
2146 pseudo-register, which could either be in a hard register or in memory.
2147 Use @code{true_regnum} to find out; it will return -1 if the pseudo is
2148 in memory and the hard register number if it is in a register.
2150 These macros should not be used in the case where a particular class of
2151 registers can only be copied to memory and not to another class of
2152 registers. In that case, secondary reload registers are not needed and
2153 would not be helpful. Instead, a stack location must be used to perform
2154 the copy and the @code{mov@var{m}} pattern should use memory as a
2155 intermediate storage. This case often occurs between floating-point and
2158 @findex SECONDARY_MEMORY_NEEDED
2159 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2160 Certain machines have the property that some registers cannot be copied
2161 to some other registers without using memory. Define this macro on
2162 those machines to be a C expression that is non-zero if objects of mode
2163 @var{m} in registers of @var{class1} can only be copied to registers of
2164 class @var{class2} by storing a register of @var{class1} into memory
2165 and loading that memory location into a register of @var{class2}.
2167 Do not define this macro if its value would always be zero.
2169 @findex SECONDARY_MEMORY_NEEDED_RTX
2170 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2171 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2172 allocates a stack slot for a memory location needed for register copies.
2173 If this macro is defined, the compiler instead uses the memory location
2174 defined by this macro.
2176 Do not define this macro if you do not define
2177 @code{SECONDARY_MEMORY_NEEDED}.
2179 @findex SECONDARY_MEMORY_NEEDED_MODE
2180 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2181 When the compiler needs a secondary memory location to copy between two
2182 registers of mode @var{mode}, it normally allocates sufficient memory to
2183 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2184 load operations in a mode that many bits wide and whose class is the
2185 same as that of @var{mode}.
2187 This is right thing to do on most machines because it ensures that all
2188 bits of the register are copied and prevents accesses to the registers
2189 in a narrower mode, which some machines prohibit for floating-point
2192 However, this default behavior is not correct on some machines, such as
2193 the DEC Alpha, that store short integers in floating-point registers
2194 differently than in integer registers. On those machines, the default
2195 widening will not work correctly and you must define this macro to
2196 suppress that widening in some cases. See the file @file{alpha.h} for
2199 Do not define this macro if you do not define
2200 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2201 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2203 @findex SMALL_REGISTER_CLASSES
2204 @item SMALL_REGISTER_CLASSES
2205 On some machines, it is risky to let hard registers live across arbitrary
2206 insns. Typically, these machines have instructions that require values
2207 to be in specific registers (like an accumulator), and reload will fail
2208 if the required hard register is used for another purpose across such an
2211 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a non-zero
2212 value on these machines. When this macro has a non-zero value, the
2213 compiler will try to minimize the lifetime of hard registers.
2215 It is always safe to define this macro with a non-zero value, but if you
2216 unnecessarily define it, you will reduce the amount of optimizations
2217 that can be performed in some cases. If you do not define this macro
2218 with a non-zero value when it is required, the compiler will run out of
2219 spill registers and print a fatal error message. For most machines, you
2220 should not define this macro at all.
2222 @findex CLASS_LIKELY_SPILLED_P
2223 @item CLASS_LIKELY_SPILLED_P (@var{class})
2224 A C expression whose value is nonzero if pseudos that have been assigned
2225 to registers of class @var{class} would likely be spilled because
2226 registers of @var{class} are needed for spill registers.
2228 The default value of this macro returns 1 if @var{class} has exactly one
2229 register and zero otherwise. On most machines, this default should be
2230 used. Only define this macro to some other expression if pseudos
2231 allocated by @file{local-alloc.c} end up in memory because their hard
2232 registers were needed for spill registers. If this macro returns nonzero
2233 for those classes, those pseudos will only be allocated by
2234 @file{global.c}, which knows how to reallocate the pseudo to another
2235 register. If there would not be another register available for
2236 reallocation, you should not change the definition of this macro since
2237 the only effect of such a definition would be to slow down register
2240 @findex CLASS_MAX_NREGS
2241 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2242 A C expression for the maximum number of consecutive registers
2243 of class @var{class} needed to hold a value of mode @var{mode}.
2245 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2246 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2247 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2248 @var{mode})} for all @var{regno} values in the class @var{class}.
2250 This macro helps control the handling of multiple-word values
2253 @item CLASS_CANNOT_CHANGE_MODE
2254 If defined, a C expression for a class that contains registers for
2255 which the compiler may not change modes arbitrarily.
2257 @item CLASS_CANNOT_CHANGE_MODE_P(@var{from}, @var{to})
2258 A C expression that is true if, for a register in
2259 @code{CLASS_CANNOT_CHANGE_MODE}, the requested mode punning is illegal.
2261 For the example, loading 32-bit integer or floating-point objects into
2262 floating-point registers on the Alpha extends them to 64-bits.
2263 Therefore loading a 64-bit object and then storing it as a 32-bit object
2264 does not store the low-order 32-bits, as would be the case for a normal
2265 register. Therefore, @file{alpha.h} defines @code{CLASS_CANNOT_CHANGE_MODE}
2266 as @code{FLOAT_REGS} and @code{CLASS_CANNOT_CHANGE_MODE_P} restricts
2267 mode changes to same-size modes.
2269 Compare this to IA-64, which extends floating-point values to 82-bits,
2270 and stores 64-bit integers in a different format than 64-bit doubles.
2271 Therefore @code{CLASS_CANNOT_CHANGE_MODE_P} is always true.
2274 Three other special macros describe which operands fit which constraint
2278 @findex CONST_OK_FOR_LETTER_P
2279 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2280 A C expression that defines the machine-dependent operand constraint
2281 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2282 particular ranges of integer values. If @var{c} is one of those
2283 letters, the expression should check that @var{value}, an integer, is in
2284 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2285 not one of those letters, the value should be 0 regardless of
2288 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2289 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2290 A C expression that defines the machine-dependent operand constraint
2291 letters that specify particular ranges of @code{const_double} values
2292 (@samp{G} or @samp{H}).
2294 If @var{c} is one of those letters, the expression should check that
2295 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2296 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2297 letters, the value should be 0 regardless of @var{value}.
2299 @code{const_double} is used for all floating-point constants and for
2300 @code{DImode} fixed-point constants. A given letter can accept either
2301 or both kinds of values. It can use @code{GET_MODE} to distinguish
2302 between these kinds.
2304 @findex EXTRA_CONSTRAINT
2305 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2306 A C expression that defines the optional machine-dependent constraint
2307 letters that can be used to segregate specific types of operands, usually
2308 memory references, for the target machine. Any letter that is not
2309 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER}
2310 may be used. Normally this macro will not be defined.
2312 If it is required for a particular target machine, it should return 1
2313 if @var{value} corresponds to the operand type represented by the
2314 constraint letter @var{c}. If @var{c} is not defined as an extra
2315 constraint, the value returned should be 0 regardless of @var{value}.
2317 For example, on the ROMP, load instructions cannot have their output
2318 in r0 if the memory reference contains a symbolic address. Constraint
2319 letter @samp{Q} is defined as representing a memory address that does
2320 @emph{not} contain a symbolic address. An alternative is specified with
2321 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2322 alternative specifies @samp{m} on the input and a register class that
2323 does not include r0 on the output.
2326 @node Stack and Calling
2327 @section Stack Layout and Calling Conventions
2328 @cindex calling conventions
2330 @c prevent bad page break with this line
2331 This describes the stack layout and calling conventions.
2339 * Register Arguments::
2341 * Aggregate Return::
2350 @subsection Basic Stack Layout
2351 @cindex stack frame layout
2352 @cindex frame layout
2354 @c prevent bad page break with this line
2355 Here is the basic stack layout.
2358 @findex STACK_GROWS_DOWNWARD
2359 @item STACK_GROWS_DOWNWARD
2360 Define this macro if pushing a word onto the stack moves the stack
2361 pointer to a smaller address.
2363 When we say, ``define this macro if @dots{},'' it means that the
2364 compiler checks this macro only with @code{#ifdef} so the precise
2365 definition used does not matter.
2367 @findex FRAME_GROWS_DOWNWARD
2368 @item FRAME_GROWS_DOWNWARD
2369 Define this macro if the addresses of local variable slots are at negative
2370 offsets from the frame pointer.
2372 @findex ARGS_GROW_DOWNWARD
2373 @item ARGS_GROW_DOWNWARD
2374 Define this macro if successive arguments to a function occupy decreasing
2375 addresses on the stack.
2377 @findex STARTING_FRAME_OFFSET
2378 @item STARTING_FRAME_OFFSET
2379 Offset from the frame pointer to the first local variable slot to be allocated.
2381 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2382 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2383 Otherwise, it is found by adding the length of the first slot to the
2384 value @code{STARTING_FRAME_OFFSET}.
2385 @c i'm not sure if the above is still correct.. had to change it to get
2386 @c rid of an overfull. --mew 2feb93
2388 @findex STACK_POINTER_OFFSET
2389 @item STACK_POINTER_OFFSET
2390 Offset from the stack pointer register to the first location at which
2391 outgoing arguments are placed. If not specified, the default value of
2392 zero is used. This is the proper value for most machines.
2394 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2395 the first location at which outgoing arguments are placed.
2397 @findex FIRST_PARM_OFFSET
2398 @item FIRST_PARM_OFFSET (@var{fundecl})
2399 Offset from the argument pointer register to the first argument's
2400 address. On some machines it may depend on the data type of the
2403 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2404 the first argument's address.
2406 @findex STACK_DYNAMIC_OFFSET
2407 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2408 Offset from the stack pointer register to an item dynamically allocated
2409 on the stack, e.g., by @code{alloca}.
2411 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2412 length of the outgoing arguments. The default is correct for most
2413 machines. See @file{function.c} for details.
2415 @findex DYNAMIC_CHAIN_ADDRESS
2416 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2417 A C expression whose value is RTL representing the address in a stack
2418 frame where the pointer to the caller's frame is stored. Assume that
2419 @var{frameaddr} is an RTL expression for the address of the stack frame
2422 If you don't define this macro, the default is to return the value
2423 of @var{frameaddr}---that is, the stack frame address is also the
2424 address of the stack word that points to the previous frame.
2426 @findex SETUP_FRAME_ADDRESSES
2427 @item SETUP_FRAME_ADDRESSES
2428 If defined, a C expression that produces the machine-specific code to
2429 setup the stack so that arbitrary frames can be accessed. For example,
2430 on the Sparc, we must flush all of the register windows to the stack
2431 before we can access arbitrary stack frames. You will seldom need to
2434 @findex BUILTIN_SETJMP_FRAME_VALUE
2435 @item BUILTIN_SETJMP_FRAME_VALUE
2436 If defined, a C expression that contains an rtx that is used to store
2437 the address of the current frame into the built in @code{setjmp} buffer.
2438 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2439 machines. One reason you may need to define this macro is if
2440 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2442 @findex RETURN_ADDR_RTX
2443 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2444 A C expression whose value is RTL representing the value of the return
2445 address for the frame @var{count} steps up from the current frame, after
2446 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2447 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2448 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2450 The value of the expression must always be the correct address when
2451 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2452 determine the return address of other frames.
2454 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2455 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2456 Define this if the return address of a particular stack frame is accessed
2457 from the frame pointer of the previous stack frame.
2459 @findex INCOMING_RETURN_ADDR_RTX
2460 @item INCOMING_RETURN_ADDR_RTX
2461 A C expression whose value is RTL representing the location of the
2462 incoming return address at the beginning of any function, before the
2463 prologue. This RTL is either a @code{REG}, indicating that the return
2464 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2467 You only need to define this macro if you want to support call frame
2468 debugging information like that provided by DWARF 2.
2470 If this RTL is a @code{REG}, you should also define
2471 DWARF_FRAME_RETURN_COLUMN to @code{DWARF_FRAME_REGNUM (REGNO)}.
2473 @findex INCOMING_FRAME_SP_OFFSET
2474 @item INCOMING_FRAME_SP_OFFSET
2475 A C expression whose value is an integer giving the offset, in bytes,
2476 from the value of the stack pointer register to the top of the stack
2477 frame at the beginning of any function, before the prologue. The top of
2478 the frame is defined to be the value of the stack pointer in the
2479 previous frame, just before the call instruction.
2481 You only need to define this macro if you want to support call frame
2482 debugging information like that provided by DWARF 2.
2484 @findex ARG_POINTER_CFA_OFFSET
2485 @item ARG_POINTER_CFA_OFFSET (@var{fundecl})
2486 A C expression whose value is an integer giving the offset, in bytes,
2487 from the argument pointer to the canonical frame address (cfa). The
2488 final value should coincide with that calculated by
2489 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2490 during virtual register instantiation.
2492 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2493 which is correct for most machines; in general, the arguments are found
2494 immediately before the stack frame. Note that this is not the case on
2495 some targets that save registers into the caller's frame, such as SPARC
2496 and rs6000, and so such targets need to define this macro.
2498 You only need to define this macro if the default is incorrect, and you
2499 want to support call frame debugging information like that provided by
2502 @findex EH_RETURN_DATA_REGNO
2503 @item EH_RETURN_DATA_REGNO (@var{N})
2504 A C expression whose value is the @var{N}th register number used for
2505 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2506 @var{N} registers are usable.
2508 The exception handling library routines communicate with the exception
2509 handlers via a set of agreed upon registers. Ideally these registers
2510 should be call-clobbered; it is possible to use call-saved registers,
2511 but may negatively impact code size. The target must support at least
2512 2 data registers, but should define 4 if there are enough free registers.
2514 You must define this macro if you want to support call frame exception
2515 handling like that provided by DWARF 2.
2517 @findex EH_RETURN_STACKADJ_RTX
2518 @item EH_RETURN_STACKADJ_RTX
2519 A C expression whose value is RTL representing a location in which
2520 to store a stack adjustment to be applied before function return.
2521 This is used to unwind the stack to an exception handler's call frame.
2522 It will be assigned zero on code paths that return normally.
2524 Typically this is a call-clobbered hard register that is otherwise
2525 untouched by the epilogue, but could also be a stack slot.
2527 You must define this macro if you want to support call frame exception
2528 handling like that provided by DWARF 2.
2530 @findex EH_RETURN_HANDLER_RTX
2531 @item EH_RETURN_HANDLER_RTX
2532 A C expression whose value is RTL representing a location in which
2533 to store the address of an exception handler to which we should
2534 return. It will not be assigned on code paths that return normally.
2536 Typically this is the location in the call frame at which the normal
2537 return address is stored. For targets that return by popping an
2538 address off the stack, this might be a memory address just below
2539 the @emph{target} call frame rather than inside the current call
2540 frame. @code{EH_RETURN_STACKADJ_RTX} will have already been assigned,
2541 so it may be used to calculate the location of the target call frame.
2543 Some targets have more complex requirements than storing to an
2544 address calculable during initial code generation. In that case
2545 the @code{eh_return} instruction pattern should be used instead.
2547 If you want to support call frame exception handling, you must
2548 define either this macro or the @code{eh_return} instruction pattern.
2552 Define this macro if the stack size for the target is very small. This
2553 has the effect of disabling gcc's builtin @samp{alloca}, though
2554 @samp{__builtin_alloca} is not affected.
2557 @node Stack Checking
2558 @subsection Specifying How Stack Checking is Done
2560 GCC will check that stack references are within the boundaries of
2561 the stack, if the @samp{-fstack-check} is specified, in one of three ways:
2565 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2566 will assume that you have arranged for stack checking to be done at
2567 appropriate places in the configuration files, e.g., in
2568 @code{FUNCTION_PROLOGUE}. GCC will do not other special processing.
2571 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
2572 called @code{check_stack} in your @file{md} file, GCC will call that
2573 pattern with one argument which is the address to compare the stack
2574 value against. You must arrange for this pattern to report an error if
2575 the stack pointer is out of range.
2578 If neither of the above are true, GCC will generate code to periodically
2579 ``probe'' the stack pointer using the values of the macros defined below.
2582 Normally, you will use the default values of these macros, so GCC
2583 will use the third approach.
2586 @findex STACK_CHECK_BUILTIN
2587 @item STACK_CHECK_BUILTIN
2588 A nonzero value if stack checking is done by the configuration files in a
2589 machine-dependent manner. You should define this macro if stack checking
2590 is require by the ABI of your machine or if you would like to have to stack
2591 checking in some more efficient way than GCC's portable approach.
2592 The default value of this macro is zero.
2594 @findex STACK_CHECK_PROBE_INTERVAL
2595 @item STACK_CHECK_PROBE_INTERVAL
2596 An integer representing the interval at which GCC must generate stack
2597 probe instructions. You will normally define this macro to be no larger
2598 than the size of the ``guard pages'' at the end of a stack area. The
2599 default value of 4096 is suitable for most systems.
2601 @findex STACK_CHECK_PROBE_LOAD
2602 @item STACK_CHECK_PROBE_LOAD
2603 A integer which is nonzero if GCC should perform the stack probe
2604 as a load instruction and zero if GCC should use a store instruction.
2605 The default is zero, which is the most efficient choice on most systems.
2607 @findex STACK_CHECK_PROTECT
2608 @item STACK_CHECK_PROTECT
2609 The number of bytes of stack needed to recover from a stack overflow,
2610 for languages where such a recovery is supported. The default value of
2611 75 words should be adequate for most machines.
2613 @findex STACK_CHECK_MAX_FRAME_SIZE
2614 @item STACK_CHECK_MAX_FRAME_SIZE
2615 The maximum size of a stack frame, in bytes. GCC will generate probe
2616 instructions in non-leaf functions to ensure at least this many bytes of
2617 stack are available. If a stack frame is larger than this size, stack
2618 checking will not be reliable and GCC will issue a warning. The
2619 default is chosen so that GCC only generates one instruction on most
2620 systems. You should normally not change the default value of this macro.
2622 @findex STACK_CHECK_FIXED_FRAME_SIZE
2623 @item STACK_CHECK_FIXED_FRAME_SIZE
2624 GCC uses this value to generate the above warning message. It
2625 represents the amount of fixed frame used by a function, not including
2626 space for any callee-saved registers, temporaries and user variables.
2627 You need only specify an upper bound for this amount and will normally
2628 use the default of four words.
2630 @findex STACK_CHECK_MAX_VAR_SIZE
2631 @item STACK_CHECK_MAX_VAR_SIZE
2632 The maximum size, in bytes, of an object that GCC will place in the
2633 fixed area of the stack frame when the user specifies
2634 @samp{-fstack-check}.
2635 GCC computed the default from the values of the above macros and you will
2636 normally not need to override that default.
2640 @node Frame Registers
2641 @subsection Registers That Address the Stack Frame
2643 @c prevent bad page break with this line
2644 This discusses registers that address the stack frame.
2647 @findex STACK_POINTER_REGNUM
2648 @item STACK_POINTER_REGNUM
2649 The register number of the stack pointer register, which must also be a
2650 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2651 the hardware determines which register this is.
2653 @findex FRAME_POINTER_REGNUM
2654 @item FRAME_POINTER_REGNUM
2655 The register number of the frame pointer register, which is used to
2656 access automatic variables in the stack frame. On some machines, the
2657 hardware determines which register this is. On other machines, you can
2658 choose any register you wish for this purpose.
2660 @findex HARD_FRAME_POINTER_REGNUM
2661 @item HARD_FRAME_POINTER_REGNUM
2662 On some machines the offset between the frame pointer and starting
2663 offset of the automatic variables is not known until after register
2664 allocation has been done (for example, because the saved registers are
2665 between these two locations). On those machines, define
2666 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
2667 be used internally until the offset is known, and define
2668 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
2669 used for the frame pointer.
2671 You should define this macro only in the very rare circumstances when it
2672 is not possible to calculate the offset between the frame pointer and
2673 the automatic variables until after register allocation has been
2674 completed. When this macro is defined, you must also indicate in your
2675 definition of @code{ELIMINABLE_REGS} how to eliminate
2676 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
2677 or @code{STACK_POINTER_REGNUM}.
2679 Do not define this macro if it would be the same as
2680 @code{FRAME_POINTER_REGNUM}.
2682 @findex ARG_POINTER_REGNUM
2683 @item ARG_POINTER_REGNUM
2684 The register number of the arg pointer register, which is used to access
2685 the function's argument list. On some machines, this is the same as the
2686 frame pointer register. On some machines, the hardware determines which
2687 register this is. On other machines, you can choose any register you
2688 wish for this purpose. If this is not the same register as the frame
2689 pointer register, then you must mark it as a fixed register according to
2690 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2691 (@pxref{Elimination}).
2693 @findex RETURN_ADDRESS_POINTER_REGNUM
2694 @item RETURN_ADDRESS_POINTER_REGNUM
2695 The register number of the return address pointer register, which is used to
2696 access the current function's return address from the stack. On some
2697 machines, the return address is not at a fixed offset from the frame
2698 pointer or stack pointer or argument pointer. This register can be defined
2699 to point to the return address on the stack, and then be converted by
2700 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
2702 Do not define this macro unless there is no other way to get the return
2703 address from the stack.
2705 @findex STATIC_CHAIN_REGNUM
2706 @findex STATIC_CHAIN_INCOMING_REGNUM
2707 @item STATIC_CHAIN_REGNUM
2708 @itemx STATIC_CHAIN_INCOMING_REGNUM
2709 Register numbers used for passing a function's static chain pointer. If
2710 register windows are used, the register number as seen by the called
2711 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
2712 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
2713 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
2714 not be defined.@refill
2716 The static chain register need not be a fixed register.
2718 If the static chain is passed in memory, these macros should not be
2719 defined; instead, the next two macros should be defined.
2721 @findex STATIC_CHAIN
2722 @findex STATIC_CHAIN_INCOMING
2724 @itemx STATIC_CHAIN_INCOMING
2725 If the static chain is passed in memory, these macros provide rtx giving
2726 @code{mem} expressions that denote where they are stored.
2727 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
2728 as seen by the calling and called functions, respectively. Often the former
2729 will be at an offset from the stack pointer and the latter at an offset from
2730 the frame pointer.@refill
2732 @findex stack_pointer_rtx
2733 @findex frame_pointer_rtx
2734 @findex arg_pointer_rtx
2735 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
2736 @code{arg_pointer_rtx} will have been initialized prior to the use of these
2737 macros and should be used to refer to those items.
2739 If the static chain is passed in a register, the two previous macros should
2744 @subsection Eliminating Frame Pointer and Arg Pointer
2746 @c prevent bad page break with this line
2747 This is about eliminating the frame pointer and arg pointer.
2750 @findex FRAME_POINTER_REQUIRED
2751 @item FRAME_POINTER_REQUIRED
2752 A C expression which is nonzero if a function must have and use a frame
2753 pointer. This expression is evaluated in the reload pass. If its value is
2754 nonzero the function will have a frame pointer.
2756 The expression can in principle examine the current function and decide
2757 according to the facts, but on most machines the constant 0 or the
2758 constant 1 suffices. Use 0 when the machine allows code to be generated
2759 with no frame pointer, and doing so saves some time or space. Use 1
2760 when there is no possible advantage to avoiding a frame pointer.
2762 In certain cases, the compiler does not know how to produce valid code
2763 without a frame pointer. The compiler recognizes those cases and
2764 automatically gives the function a frame pointer regardless of what
2765 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
2768 In a function that does not require a frame pointer, the frame pointer
2769 register can be allocated for ordinary usage, unless you mark it as a
2770 fixed register. See @code{FIXED_REGISTERS} for more information.
2772 @findex INITIAL_FRAME_POINTER_OFFSET
2773 @findex get_frame_size
2774 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
2775 A C statement to store in the variable @var{depth-var} the difference
2776 between the frame pointer and the stack pointer values immediately after
2777 the function prologue. The value would be computed from information
2778 such as the result of @code{get_frame_size ()} and the tables of
2779 registers @code{regs_ever_live} and @code{call_used_regs}.
2781 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
2782 need not be defined. Otherwise, it must be defined even if
2783 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
2784 case, you may set @var{depth-var} to anything.
2786 @findex ELIMINABLE_REGS
2787 @item ELIMINABLE_REGS
2788 If defined, this macro specifies a table of register pairs used to
2789 eliminate unneeded registers that point into the stack frame. If it is not
2790 defined, the only elimination attempted by the compiler is to replace
2791 references to the frame pointer with references to the stack pointer.
2793 The definition of this macro is a list of structure initializations, each
2794 of which specifies an original and replacement register.
2796 On some machines, the position of the argument pointer is not known until
2797 the compilation is completed. In such a case, a separate hard register
2798 must be used for the argument pointer. This register can be eliminated by
2799 replacing it with either the frame pointer or the argument pointer,
2800 depending on whether or not the frame pointer has been eliminated.
2802 In this case, you might specify:
2804 #define ELIMINABLE_REGS \
2805 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
2806 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
2807 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
2810 Note that the elimination of the argument pointer with the stack pointer is
2811 specified first since that is the preferred elimination.
2813 @findex CAN_ELIMINATE
2814 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
2815 A C expression that returns non-zero if the compiler is allowed to try
2816 to replace register number @var{from-reg} with register number
2817 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
2818 is defined, and will usually be the constant 1, since most of the cases
2819 preventing register elimination are things that the compiler already
2822 @findex INITIAL_ELIMINATION_OFFSET
2823 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
2824 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
2825 specifies the initial difference between the specified pair of
2826 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
2829 @findex LONGJMP_RESTORE_FROM_STACK
2830 @item LONGJMP_RESTORE_FROM_STACK
2831 Define this macro if the @code{longjmp} function restores registers from
2832 the stack frames, rather than from those saved specifically by
2833 @code{setjmp}. Certain quantities must not be kept in registers across
2834 a call to @code{setjmp} on such machines.
2837 @node Stack Arguments
2838 @subsection Passing Function Arguments on the Stack
2839 @cindex arguments on stack
2840 @cindex stack arguments
2842 The macros in this section control how arguments are passed
2843 on the stack. See the following section for other macros that
2844 control passing certain arguments in registers.
2847 @findex PROMOTE_PROTOTYPES
2848 @item PROMOTE_PROTOTYPES
2849 A C expression whose value is nonzero if an argument declared in
2850 a prototype as an integral type smaller than @code{int} should
2851 actually be passed as an @code{int}. In addition to avoiding
2852 errors in certain cases of mismatch, it also makes for better
2853 code on certain machines. If the macro is not defined in target
2854 header files, it defaults to 0.
2858 A C expression. If nonzero, push insns will be used to pass
2860 If the target machine does not have a push instruction, set it to zero.
2861 That directs GCC to use an alternate strategy: to
2862 allocate the entire argument block and then store the arguments into
2863 it. When PUSH_ARGS is nonzero, PUSH_ROUNDING must be defined too.
2864 On some machines, the definition
2866 @findex PUSH_ROUNDING
2867 @item PUSH_ROUNDING (@var{npushed})
2868 A C expression that is the number of bytes actually pushed onto the
2869 stack when an instruction attempts to push @var{npushed} bytes.
2871 On some machines, the definition
2874 #define PUSH_ROUNDING(BYTES) (BYTES)
2878 will suffice. But on other machines, instructions that appear
2879 to push one byte actually push two bytes in an attempt to maintain
2880 alignment. Then the definition should be
2883 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
2886 @findex ACCUMULATE_OUTGOING_ARGS
2887 @findex current_function_outgoing_args_size
2888 @item ACCUMULATE_OUTGOING_ARGS
2889 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
2890 will be computed and placed into the variable
2891 @code{current_function_outgoing_args_size}. No space will be pushed
2892 onto the stack for each call; instead, the function prologue should
2893 increase the stack frame size by this amount.
2895 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
2898 @findex REG_PARM_STACK_SPACE
2899 @item REG_PARM_STACK_SPACE (@var{fndecl})
2900 Define this macro if functions should assume that stack space has been
2901 allocated for arguments even when their values are passed in
2904 The value of this macro is the size, in bytes, of the area reserved for
2905 arguments passed in registers for the function represented by @var{fndecl},
2906 which can be zero if GCC is calling a library function.
2908 This space can be allocated by the caller, or be a part of the
2909 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
2911 @c above is overfull. not sure what to do. --mew 5feb93 did
2912 @c something, not sure if it looks good. --mew 10feb93
2914 @findex MAYBE_REG_PARM_STACK_SPACE
2915 @findex FINAL_REG_PARM_STACK_SPACE
2916 @item MAYBE_REG_PARM_STACK_SPACE
2917 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
2918 Define these macros in addition to the one above if functions might
2919 allocate stack space for arguments even when their values are passed
2920 in registers. These should be used when the stack space allocated
2921 for arguments in registers is not a simple constant independent of the
2922 function declaration.
2924 The value of the first macro is the size, in bytes, of the area that
2925 we should initially assume would be reserved for arguments passed in registers.
2927 The value of the second macro is the actual size, in bytes, of the area
2928 that will be reserved for arguments passed in registers. This takes two
2929 arguments: an integer representing the number of bytes of fixed sized
2930 arguments on the stack, and a tree representing the number of bytes of
2931 variable sized arguments on the stack.
2933 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
2934 called for libcall functions, the current function, or for a function
2935 being called when it is known that such stack space must be allocated.
2936 In each case this value can be easily computed.
2938 When deciding whether a called function needs such stack space, and how
2939 much space to reserve, GCC uses these two macros instead of
2940 @code{REG_PARM_STACK_SPACE}.
2942 @findex OUTGOING_REG_PARM_STACK_SPACE
2943 @item OUTGOING_REG_PARM_STACK_SPACE
2944 Define this if it is the responsibility of the caller to allocate the area
2945 reserved for arguments passed in registers.
2947 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
2948 whether the space for these arguments counts in the value of
2949 @code{current_function_outgoing_args_size}.
2951 @findex STACK_PARMS_IN_REG_PARM_AREA
2952 @item STACK_PARMS_IN_REG_PARM_AREA
2953 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
2954 stack parameters don't skip the area specified by it.
2955 @c i changed this, makes more sens and it should have taken care of the
2956 @c overfull.. not as specific, tho. --mew 5feb93
2958 Normally, when a parameter is not passed in registers, it is placed on the
2959 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
2960 suppresses this behavior and causes the parameter to be passed on the
2961 stack in its natural location.
2963 @findex RETURN_POPS_ARGS
2964 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
2965 A C expression that should indicate the number of bytes of its own
2966 arguments that a function pops on returning, or 0 if the
2967 function pops no arguments and the caller must therefore pop them all
2968 after the function returns.
2970 @var{fundecl} is a C variable whose value is a tree node that describes
2971 the function in question. Normally it is a node of type
2972 @code{FUNCTION_DECL} that describes the declaration of the function.
2973 From this you can obtain the DECL_MACHINE_ATTRIBUTES of the function.
2975 @var{funtype} is a C variable whose value is a tree node that
2976 describes the function in question. Normally it is a node of type
2977 @code{FUNCTION_TYPE} that describes the data type of the function.
2978 From this it is possible to obtain the data types of the value and
2979 arguments (if known).
2981 When a call to a library function is being considered, @var{fundecl}
2982 will contain an identifier node for the library function. Thus, if
2983 you need to distinguish among various library functions, you can do so
2984 by their names. Note that ``library function'' in this context means
2985 a function used to perform arithmetic, whose name is known specially
2986 in the compiler and was not mentioned in the C code being compiled.
2988 @var{stack-size} is the number of bytes of arguments passed on the
2989 stack. If a variable number of bytes is passed, it is zero, and
2990 argument popping will always be the responsibility of the calling function.
2992 On the Vax, all functions always pop their arguments, so the definition
2993 of this macro is @var{stack-size}. On the 68000, using the standard
2994 calling convention, no functions pop their arguments, so the value of
2995 the macro is always 0 in this case. But an alternative calling
2996 convention is available in which functions that take a fixed number of
2997 arguments pop them but other functions (such as @code{printf}) pop
2998 nothing (the caller pops all). When this convention is in use,
2999 @var{funtype} is examined to determine whether a function takes a fixed
3000 number of arguments.
3003 @node Register Arguments
3004 @subsection Passing Arguments in Registers
3005 @cindex arguments in registers
3006 @cindex registers arguments
3008 This section describes the macros which let you control how various
3009 types of arguments are passed in registers or how they are arranged in
3013 @findex FUNCTION_ARG
3014 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3015 A C expression that controls whether a function argument is passed
3016 in a register, and which register.
3018 The arguments are @var{cum}, which summarizes all the previous
3019 arguments; @var{mode}, the machine mode of the argument; @var{type},
3020 the data type of the argument as a tree node or 0 if that is not known
3021 (which happens for C support library functions); and @var{named},
3022 which is 1 for an ordinary argument and 0 for nameless arguments that
3023 correspond to @samp{@dots{}} in the called function's prototype.
3024 @var{type} can be an incomplete type if a syntax error has previously
3027 The value of the expression is usually either a @code{reg} RTX for the
3028 hard register in which to pass the argument, or zero to pass the
3029 argument on the stack.
3031 For machines like the Vax and 68000, where normally all arguments are
3032 pushed, zero suffices as a definition.
3034 The value of the expression can also be a @code{parallel} RTX. This is
3035 used when an argument is passed in multiple locations. The mode of the
3036 of the @code{parallel} should be the mode of the entire argument. The
3037 @code{parallel} holds any number of @code{expr_list} pairs; each one
3038 describes where part of the argument is passed. In each
3039 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3040 register in which to pass this part of the argument, and the mode of the
3041 register RTX indicates how large this part of the argument is. The
3042 second operand of the @code{expr_list} is a @code{const_int} which gives
3043 the offset in bytes into the entire argument of where this part starts.
3044 As a special exception the first @code{expr_list} in the @code{parallel}
3045 RTX may have a first operand of zero. This indicates that the entire
3046 argument is also stored on the stack.
3048 @cindex @file{stdarg.h} and register arguments
3049 The usual way to make the ISO library @file{stdarg.h} work on a machine
3050 where some arguments are usually passed in registers, is to cause
3051 nameless arguments to be passed on the stack instead. This is done
3052 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3054 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3055 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3056 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3057 in the definition of this macro to determine if this argument is of a
3058 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3059 is not defined and @code{FUNCTION_ARG} returns non-zero for such an
3060 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3061 defined, the argument will be computed in the stack and then loaded into
3064 @findex MUST_PASS_IN_STACK
3065 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
3066 Define as a C expression that evaluates to nonzero if we do not know how
3067 to pass TYPE solely in registers. The file @file{expr.h} defines a
3068 definition that is usually appropriate, refer to @file{expr.h} for additional
3071 @findex FUNCTION_INCOMING_ARG
3072 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3073 Define this macro if the target machine has ``register windows'', so
3074 that the register in which a function sees an arguments is not
3075 necessarily the same as the one in which the caller passed the
3078 For such machines, @code{FUNCTION_ARG} computes the register in which
3079 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3080 be defined in a similar fashion to tell the function being called
3081 where the arguments will arrive.
3083 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3084 serves both purposes.@refill
3086 @findex FUNCTION_ARG_PARTIAL_NREGS
3087 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3088 A C expression for the number of words, at the beginning of an
3089 argument, that must be put in registers. The value must be zero for
3090 arguments that are passed entirely in registers or that are entirely
3091 pushed on the stack.
3093 On some machines, certain arguments must be passed partially in
3094 registers and partially in memory. On these machines, typically the
3095 first @var{n} words of arguments are passed in registers, and the rest
3096 on the stack. If a multi-word argument (a @code{double} or a
3097 structure) crosses that boundary, its first few words must be passed
3098 in registers and the rest must be pushed. This macro tells the
3099 compiler when this occurs, and how many of the words should go in
3102 @code{FUNCTION_ARG} for these arguments should return the first
3103 register to be used by the caller for this argument; likewise
3104 @code{FUNCTION_INCOMING_ARG}, for the called function.
3106 @findex FUNCTION_ARG_PASS_BY_REFERENCE
3107 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3108 A C expression that indicates when an argument must be passed by reference.
3109 If nonzero for an argument, a copy of that argument is made in memory and a
3110 pointer to the argument is passed instead of the argument itself.
3111 The pointer is passed in whatever way is appropriate for passing a pointer
3114 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3115 definition of this macro might be
3117 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3118 (CUM, MODE, TYPE, NAMED) \
3119 MUST_PASS_IN_STACK (MODE, TYPE)
3121 @c this is *still* too long. --mew 5feb93
3123 @findex FUNCTION_ARG_CALLEE_COPIES
3124 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3125 If defined, a C expression that indicates when it is the called function's
3126 responsibility to make a copy of arguments passed by invisible reference.
3127 Normally, the caller makes a copy and passes the address of the copy to the
3128 routine being called. When FUNCTION_ARG_CALLEE_COPIES is defined and is
3129 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3130 ``live'' value. The called function must not modify this value. If it can be
3131 determined that the value won't be modified, it need not make a copy;
3132 otherwise a copy must be made.
3134 @findex CUMULATIVE_ARGS
3135 @item CUMULATIVE_ARGS
3136 A C type for declaring a variable that is used as the first argument of
3137 @code{FUNCTION_ARG} and other related values. For some target machines,
3138 the type @code{int} suffices and can hold the number of bytes of
3141 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3142 arguments that have been passed on the stack. The compiler has other
3143 variables to keep track of that. For target machines on which all
3144 arguments are passed on the stack, there is no need to store anything in
3145 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3146 should not be empty, so use @code{int}.
3148 @findex INIT_CUMULATIVE_ARGS
3149 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
3150 A C statement (sans semicolon) for initializing the variable @var{cum}
3151 for the state at the beginning of the argument list. The variable has
3152 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
3153 for the data type of the function which will receive the args, or 0
3154 if the args are to a compiler support library function. The value of
3155 @var{indirect} is nonzero when processing an indirect call, for example
3156 a call through a function pointer. The value of @var{indirect} is zero
3157 for a call to an explicitly named function, a library function call, or when
3158 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3161 When processing a call to a compiler support library function,
3162 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3163 contains the name of the function, as a string. @var{libname} is 0 when
3164 an ordinary C function call is being processed. Thus, each time this
3165 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3166 never both of them at once.
3168 @findex INIT_CUMULATIVE_LIBCALL_ARGS
3169 @item INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3170 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3171 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3172 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3173 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3174 0)} is used instead.
3176 @findex INIT_CUMULATIVE_INCOMING_ARGS
3177 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3178 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3179 finding the arguments for the function being compiled. If this macro is
3180 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3182 The value passed for @var{libname} is always 0, since library routines
3183 with special calling conventions are never compiled with GCC. The
3184 argument @var{libname} exists for symmetry with
3185 @code{INIT_CUMULATIVE_ARGS}.
3186 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3187 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3189 @findex FUNCTION_ARG_ADVANCE
3190 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3191 A C statement (sans semicolon) to update the summarizer variable
3192 @var{cum} to advance past an argument in the argument list. The
3193 values @var{mode}, @var{type} and @var{named} describe that argument.
3194 Once this is done, the variable @var{cum} is suitable for analyzing
3195 the @emph{following} argument with @code{FUNCTION_ARG}, etc.@refill
3197 This macro need not do anything if the argument in question was passed
3198 on the stack. The compiler knows how to track the amount of stack space
3199 used for arguments without any special help.
3201 @findex FUNCTION_ARG_PADDING
3202 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3203 If defined, a C expression which determines whether, and in which direction,
3204 to pad out an argument with extra space. The value should be of type
3205 @code{enum direction}: either @code{upward} to pad above the argument,
3206 @code{downward} to pad below, or @code{none} to inhibit padding.
3208 The @emph{amount} of padding is always just enough to reach the next
3209 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3212 This macro has a default definition which is right for most systems.
3213 For little-endian machines, the default is to pad upward. For
3214 big-endian machines, the default is to pad downward for an argument of
3215 constant size shorter than an @code{int}, and upward otherwise.
3217 @findex PAD_VARARGS_DOWN
3218 @item PAD_VARARGS_DOWN
3219 If defined, a C expression which determines whether the default
3220 implementation of va_arg will attempt to pad down before reading the
3221 next argument, if that argument is smaller than its aligned space as
3222 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3223 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3225 @findex FUNCTION_ARG_BOUNDARY
3226 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3227 If defined, a C expression that gives the alignment boundary, in bits,
3228 of an argument with the specified mode and type. If it is not defined,
3229 @code{PARM_BOUNDARY} is used for all arguments.
3231 @findex FUNCTION_ARG_REGNO_P
3232 @item FUNCTION_ARG_REGNO_P (@var{regno})
3233 A C expression that is nonzero if @var{regno} is the number of a hard
3234 register in which function arguments are sometimes passed. This does
3235 @emph{not} include implicit arguments such as the static chain and
3236 the structure-value address. On many machines, no registers can be
3237 used for this purpose since all function arguments are pushed on the
3240 @findex LOAD_ARGS_REVERSED
3241 @item LOAD_ARGS_REVERSED
3242 If defined, the order in which arguments are loaded into their
3243 respective argument registers is reversed so that the last
3244 argument is loaded first. This macro only affects arguments
3245 passed in registers.
3250 @subsection How Scalar Function Values Are Returned
3251 @cindex return values in registers
3252 @cindex values, returned by functions
3253 @cindex scalars, returned as values
3255 This section discusses the macros that control returning scalars as
3256 values---values that can fit in registers.
3259 @findex TRADITIONAL_RETURN_FLOAT
3260 @item TRADITIONAL_RETURN_FLOAT
3261 Define this macro if @samp{-traditional} should not cause functions
3262 declared to return @code{float} to convert the value to @code{double}.
3264 @findex FUNCTION_VALUE
3265 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3266 A C expression to create an RTX representing the place where a
3267 function returns a value of data type @var{valtype}. @var{valtype} is
3268 a tree node representing a data type. Write @code{TYPE_MODE
3269 (@var{valtype})} to get the machine mode used to represent that type.
3270 On many machines, only the mode is relevant. (Actually, on most
3271 machines, scalar values are returned in the same place regardless of
3274 The value of the expression is usually a @code{reg} RTX for the hard
3275 register where the return value is stored. The value can also be a
3276 @code{parallel} RTX, if the return value is in multiple places. See
3277 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3279 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3280 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3283 If the precise function being called is known, @var{func} is a tree
3284 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3285 pointer. This makes it possible to use a different value-returning
3286 convention for specific functions when all their calls are
3289 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3290 types, because these are returned in another way. See
3291 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3293 @findex FUNCTION_OUTGOING_VALUE
3294 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3295 Define this macro if the target machine has ``register windows''
3296 so that the register in which a function returns its value is not
3297 the same as the one in which the caller sees the value.
3299 For such machines, @code{FUNCTION_VALUE} computes the register in which
3300 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3301 defined in a similar fashion to tell the function where to put the
3304 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3305 @code{FUNCTION_VALUE} serves both purposes.@refill
3307 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3308 aggregate data types, because these are returned in another way. See
3309 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3311 @findex LIBCALL_VALUE
3312 @item LIBCALL_VALUE (@var{mode})
3313 A C expression to create an RTX representing the place where a library
3314 function returns a value of mode @var{mode}. If the precise function
3315 being called is known, @var{func} is a tree node
3316 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3317 pointer. This makes it possible to use a different value-returning
3318 convention for specific functions when all their calls are
3321 Note that ``library function'' in this context means a compiler
3322 support routine, used to perform arithmetic, whose name is known
3323 specially by the compiler and was not mentioned in the C code being
3326 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3327 data types, because none of the library functions returns such types.
3329 @findex FUNCTION_VALUE_REGNO_P
3330 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3331 A C expression that is nonzero if @var{regno} is the number of a hard
3332 register in which the values of called function may come back.
3334 A register whose use for returning values is limited to serving as the
3335 second of a pair (for a value of type @code{double}, say) need not be
3336 recognized by this macro. So for most machines, this definition
3340 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3343 If the machine has register windows, so that the caller and the called
3344 function use different registers for the return value, this macro
3345 should recognize only the caller's register numbers.
3347 @findex APPLY_RESULT_SIZE
3348 @item APPLY_RESULT_SIZE
3349 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3350 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3351 saving and restoring an arbitrary return value.
3354 @node Aggregate Return
3355 @subsection How Large Values Are Returned
3356 @cindex aggregates as return values
3357 @cindex large return values
3358 @cindex returning aggregate values
3359 @cindex structure value address
3361 When a function value's mode is @code{BLKmode} (and in some other
3362 cases), the value is not returned according to @code{FUNCTION_VALUE}
3363 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3364 block of memory in which the value should be stored. This address
3365 is called the @dfn{structure value address}.
3367 This section describes how to control returning structure values in
3371 @findex RETURN_IN_MEMORY
3372 @item RETURN_IN_MEMORY (@var{type})
3373 A C expression which can inhibit the returning of certain function
3374 values in registers, based on the type of value. A nonzero value says
3375 to return the function value in memory, just as large structures are
3376 always returned. Here @var{type} will be a C expression of type
3377 @code{tree}, representing the data type of the value.
3379 Note that values of mode @code{BLKmode} must be explicitly handled
3380 by this macro. Also, the option @samp{-fpcc-struct-return}
3381 takes effect regardless of this macro. On most systems, it is
3382 possible to leave the macro undefined; this causes a default
3383 definition to be used, whose value is the constant 1 for @code{BLKmode}
3384 values, and 0 otherwise.
3386 Do not use this macro to indicate that structures and unions should always
3387 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3390 @findex DEFAULT_PCC_STRUCT_RETURN
3391 @item DEFAULT_PCC_STRUCT_RETURN
3392 Define this macro to be 1 if all structure and union return values must be
3393 in memory. Since this results in slower code, this should be defined
3394 only if needed for compatibility with other compilers or with an ABI.
3395 If you define this macro to be 0, then the conventions used for structure
3396 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3398 If not defined, this defaults to the value 1.
3400 @findex STRUCT_VALUE_REGNUM
3401 @item STRUCT_VALUE_REGNUM
3402 If the structure value address is passed in a register, then
3403 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3405 @findex STRUCT_VALUE
3407 If the structure value address is not passed in a register, define
3408 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3409 where the address is passed. If it returns 0, the address is passed as
3410 an ``invisible'' first argument.
3412 @findex STRUCT_VALUE_INCOMING_REGNUM
3413 @item STRUCT_VALUE_INCOMING_REGNUM
3414 On some architectures the place where the structure value address
3415 is found by the called function is not the same place that the
3416 caller put it. This can be due to register windows, or it could
3417 be because the function prologue moves it to a different place.
3419 If the incoming location of the structure value address is in a
3420 register, define this macro as the register number.
3422 @findex STRUCT_VALUE_INCOMING
3423 @item STRUCT_VALUE_INCOMING
3424 If the incoming location is not a register, then you should define
3425 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3426 called function should find the value. If it should find the value on
3427 the stack, define this to create a @code{mem} which refers to the frame
3428 pointer. A definition of 0 means that the address is passed as an
3429 ``invisible'' first argument.
3431 @findex PCC_STATIC_STRUCT_RETURN
3432 @item PCC_STATIC_STRUCT_RETURN
3433 Define this macro if the usual system convention on the target machine
3434 for returning structures and unions is for the called function to return
3435 the address of a static variable containing the value.
3437 Do not define this if the usual system convention is for the caller to
3438 pass an address to the subroutine.
3440 This macro has effect in @samp{-fpcc-struct-return} mode, but it does
3441 nothing when you use @samp{-freg-struct-return} mode.
3445 @subsection Caller-Saves Register Allocation
3447 If you enable it, GCC can save registers around function calls. This
3448 makes it possible to use call-clobbered registers to hold variables that
3449 must live across calls.
3452 @findex DEFAULT_CALLER_SAVES
3453 @item DEFAULT_CALLER_SAVES
3454 Define this macro if function calls on the target machine do not preserve
3455 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3456 for all registers. When defined, this macro enables @samp{-fcaller-saves}
3457 by default for all optimization levels. It has no effect for optimization
3458 levels 2 and higher, where @samp{-fcaller-saves} is the default.
3460 @findex CALLER_SAVE_PROFITABLE
3461 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3462 A C expression to determine whether it is worthwhile to consider placing
3463 a pseudo-register in a call-clobbered hard register and saving and
3464 restoring it around each function call. The expression should be 1 when
3465 this is worth doing, and 0 otherwise.
3467 If you don't define this macro, a default is used which is good on most
3468 machines: @code{4 * @var{calls} < @var{refs}}.
3470 @findex HARD_REGNO_CALLER_SAVE_MODE
3471 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3472 A C expression specifying which mode is required for saving @var{nregs}
3473 of a pseudo-register in call-clobbered hard register @var{regno}. If
3474 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3475 returned. For most machines this macro need not be defined since GCC
3476 will select the smallest suitable mode.
3479 @node Function Entry
3480 @subsection Function Entry and Exit
3481 @cindex function entry and exit
3485 This section describes the macros that output function entry
3486 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3489 @findex FUNCTION_PROLOGUE
3490 @item FUNCTION_PROLOGUE (@var{file}, @var{size})
3491 A C compound statement that outputs the assembler code for entry to a
3492 function. The prologue is responsible for setting up the stack frame,
3493 initializing the frame pointer register, saving registers that must be
3494 saved, and allocating @var{size} additional bytes of storage for the
3495 local variables. @var{size} is an integer. @var{file} is a stdio
3496 stream to which the assembler code should be output.
3498 The label for the beginning of the function need not be output by this
3499 macro. That has already been done when the macro is run.
3501 @findex regs_ever_live
3502 To determine which registers to save, the macro can refer to the array
3503 @code{regs_ever_live}: element @var{r} is nonzero if hard register
3504 @var{r} is used anywhere within the function. This implies the function
3505 prologue should save register @var{r}, provided it is not one of the
3506 call-used registers. (@code{FUNCTION_EPILOGUE} must likewise use
3507 @code{regs_ever_live}.)
3509 On machines that have ``register windows'', the function entry code does
3510 not save on the stack the registers that are in the windows, even if
3511 they are supposed to be preserved by function calls; instead it takes
3512 appropriate steps to ``push'' the register stack, if any non-call-used
3513 registers are used in the function.
3515 @findex frame_pointer_needed
3516 On machines where functions may or may not have frame-pointers, the
3517 function entry code must vary accordingly; it must set up the frame
3518 pointer if one is wanted, and not otherwise. To determine whether a
3519 frame pointer is in wanted, the macro can refer to the variable
3520 @code{frame_pointer_needed}. The variable's value will be 1 at run
3521 time in a function that needs a frame pointer. @xref{Elimination}.
3523 The function entry code is responsible for allocating any stack space
3524 required for the function. This stack space consists of the regions
3525 listed below. In most cases, these regions are allocated in the
3526 order listed, with the last listed region closest to the top of the
3527 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
3528 the highest address if it is not defined). You can use a different order
3529 for a machine if doing so is more convenient or required for
3530 compatibility reasons. Except in cases where required by standard
3531 or by a debugger, there is no reason why the stack layout used by GCC
3532 need agree with that used by other compilers for a machine.
3536 @findex current_function_pretend_args_size
3537 A region of @code{current_function_pretend_args_size} bytes of
3538 uninitialized space just underneath the first argument arriving on the
3539 stack. (This may not be at the very start of the allocated stack region
3540 if the calling sequence has pushed anything else since pushing the stack
3541 arguments. But usually, on such machines, nothing else has been pushed
3542 yet, because the function prologue itself does all the pushing.) This
3543 region is used on machines where an argument may be passed partly in
3544 registers and partly in memory, and, in some cases to support the
3545 features in @file{varargs.h} and @file{stdargs.h}.
3548 An area of memory used to save certain registers used by the function.
3549 The size of this area, which may also include space for such things as
3550 the return address and pointers to previous stack frames, is
3551 machine-specific and usually depends on which registers have been used
3552 in the function. Machines with register windows often do not require
3556 A region of at least @var{size} bytes, possibly rounded up to an allocation
3557 boundary, to contain the local variables of the function. On some machines,
3558 this region and the save area may occur in the opposite order, with the
3559 save area closer to the top of the stack.
3562 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3563 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3564 @code{current_function_outgoing_args_size} bytes to be used for outgoing
3565 argument lists of the function. @xref{Stack Arguments}.
3568 Normally, it is necessary for the macros @code{FUNCTION_PROLOGUE} and
3569 @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C
3570 variable @code{current_function_is_leaf} is nonzero for such a function.
3572 @findex EXIT_IGNORE_STACK
3573 @item EXIT_IGNORE_STACK
3574 Define this macro as a C expression that is nonzero if the return
3575 instruction or the function epilogue ignores the value of the stack
3576 pointer; in other words, if it is safe to delete an instruction to
3577 adjust the stack pointer before a return from the function.
3579 Note that this macro's value is relevant only for functions for which
3580 frame pointers are maintained. It is never safe to delete a final
3581 stack adjustment in a function that has no frame pointer, and the
3582 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3584 @findex EPILOGUE_USES
3585 @item EPILOGUE_USES (@var{regno})
3586 Define this macro as a C expression that is nonzero for registers that are
3587 used by the epilogue or the @samp{return} pattern. The stack and frame
3588 pointer registers are already be assumed to be used as needed.
3590 @findex FUNCTION_EPILOGUE
3591 @item FUNCTION_EPILOGUE (@var{file}, @var{size})
3592 A C compound statement that outputs the assembler code for exit from a
3593 function. The epilogue is responsible for restoring the saved
3594 registers and stack pointer to their values when the function was
3595 called, and returning control to the caller. This macro takes the
3596 same arguments as the macro @code{FUNCTION_PROLOGUE}, and the
3597 registers to restore are determined from @code{regs_ever_live} and
3598 @code{CALL_USED_REGISTERS} in the same way.
3600 On some machines, there is a single instruction that does all the work
3601 of returning from the function. On these machines, give that
3602 instruction the name @samp{return} and do not define the macro
3603 @code{FUNCTION_EPILOGUE} at all.
3605 Do not define a pattern named @samp{return} if you want the
3606 @code{FUNCTION_EPILOGUE} to be used. If you want the target switches
3607 to control whether return instructions or epilogues are used, define a
3608 @samp{return} pattern with a validity condition that tests the target
3609 switches appropriately. If the @samp{return} pattern's validity
3610 condition is false, epilogues will be used.
3612 On machines where functions may or may not have frame-pointers, the
3613 function exit code must vary accordingly. Sometimes the code for these
3614 two cases is completely different. To determine whether a frame pointer
3615 is wanted, the macro can refer to the variable
3616 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
3617 a function that needs a frame pointer.
3619 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
3620 treat leaf functions specially. The C variable @code{current_function_is_leaf}
3621 is nonzero for such a function. @xref{Leaf Functions}.
3623 On some machines, some functions pop their arguments on exit while
3624 others leave that for the caller to do. For example, the 68020 when
3625 given @samp{-mrtd} pops arguments in functions that take a fixed
3626 number of arguments.
3628 @findex current_function_pops_args
3629 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
3630 functions pop their own arguments. @code{FUNCTION_EPILOGUE} needs to
3631 know what was decided. The variable that is called
3632 @code{current_function_pops_args} is the number of bytes of its
3633 arguments that a function should pop. @xref{Scalar Return}.
3634 @c what is the "its arguments" in the above sentence referring to, pray
3635 @c tell? --mew 5feb93
3637 @findex DELAY_SLOTS_FOR_EPILOGUE
3638 @item DELAY_SLOTS_FOR_EPILOGUE
3639 Define this macro if the function epilogue contains delay slots to which
3640 instructions from the rest of the function can be ``moved''. The
3641 definition should be a C expression whose value is an integer
3642 representing the number of delay slots there.
3644 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
3645 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
3646 A C expression that returns 1 if @var{insn} can be placed in delay
3647 slot number @var{n} of the epilogue.
3649 The argument @var{n} is an integer which identifies the delay slot now
3650 being considered (since different slots may have different rules of
3651 eligibility). It is never negative and is always less than the number
3652 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
3653 If you reject a particular insn for a given delay slot, in principle, it
3654 may be reconsidered for a subsequent delay slot. Also, other insns may
3655 (at least in principle) be considered for the so far unfilled delay
3658 @findex current_function_epilogue_delay_list
3659 @findex final_scan_insn
3660 The insns accepted to fill the epilogue delay slots are put in an RTL
3661 list made with @code{insn_list} objects, stored in the variable
3662 @code{current_function_epilogue_delay_list}. The insn for the first
3663 delay slot comes first in the list. Your definition of the macro
3664 @code{FUNCTION_EPILOGUE} should fill the delay slots by outputting the
3665 insns in this list, usually by calling @code{final_scan_insn}.
3667 You need not define this macro if you did not define
3668 @code{DELAY_SLOTS_FOR_EPILOGUE}.
3670 @findex ASM_OUTPUT_MI_THUNK
3671 @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function})
3672 A C compound statement that outputs the assembler code for a thunk
3673 function, used to implement C++ virtual function calls with multiple
3674 inheritance. The thunk acts as a wrapper around a virtual function,
3675 adjusting the implicit object parameter before handing control off to
3678 First, emit code to add the integer @var{delta} to the location that
3679 contains the incoming first argument. Assume that this argument
3680 contains a pointer, and is the one used to pass the @code{this} pointer
3681 in C++. This is the incoming argument @emph{before} the function prologue,
3682 e.g. @samp{%o0} on a sparc. The addition must preserve the values of
3683 all other incoming arguments.
3685 After the addition, emit code to jump to @var{function}, which is a
3686 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
3687 not touch the return address. Hence returning from @var{FUNCTION} will
3688 return to whoever called the current @samp{thunk}.
3690 The effect must be as if @var{function} had been called directly with
3691 the adjusted first argument. This macro is responsible for emitting all
3692 of the code for a thunk function; @code{FUNCTION_PROLOGUE} and
3693 @code{FUNCTION_EPILOGUE} are not invoked.
3695 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
3696 have already been extracted from it.) It might possibly be useful on
3697 some targets, but probably not.
3699 If you do not define this macro, the target-independent code in the C++
3700 frontend will generate a less efficient heavyweight thunk that calls
3701 @var{function} instead of jumping to it. The generic approach does
3702 not support varargs.
3706 @subsection Generating Code for Profiling
3707 @cindex profiling, code generation
3709 These macros will help you generate code for profiling.
3712 @findex FUNCTION_PROFILER
3713 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
3714 A C statement or compound statement to output to @var{file} some
3715 assembler code to call the profiling subroutine @code{mcount}.
3718 The details of how @code{mcount} expects to be called are determined by
3719 your operating system environment, not by GCC. To figure them out,
3720 compile a small program for profiling using the system's installed C
3721 compiler and look at the assembler code that results.
3723 Older implementations of @code{mcount} expect the address of a counter
3724 variable to be loaded into some register. The name of this variable is
3725 @samp{LP} followed by the number @var{labelno}, so you would generate
3726 the name using @samp{LP%d} in a @code{fprintf}.
3728 @findex PROFILE_HOOK
3730 A C statement or compound statement to output to @var{file} some assembly
3731 code to call the profiling subroutine @code{mcount} even the target does
3732 not support profiling.
3734 @findex NO_PROFILE_COUNTERS
3735 @item NO_PROFILE_COUNTERS
3736 Define this macro if the @code{mcount} subroutine on your system does
3737 not need a counter variable allocated for each function. This is true
3738 for almost all modern implementations. If you define this macro, you
3739 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
3741 @findex PROFILE_BEFORE_PROLOGUE
3742 @item PROFILE_BEFORE_PROLOGUE
3743 Define this macro if the code for function profiling should come before
3744 the function prologue. Normally, the profiling code comes after.
3746 @findex FUNCTION_BLOCK_PROFILER
3747 @vindex profile_block_flag
3748 @item FUNCTION_BLOCK_PROFILER (@var{file}, @var{labelno})
3749 A C statement or compound statement to output to @var{file} some
3750 assembler code to initialize basic-block profiling for the current
3751 object module. The global compile flag @code{profile_block_flag}
3752 distinguishes two profile modes.
3755 @findex __bb_init_func
3756 @item profile_block_flag != 2
3757 Output code to call the subroutine @code{__bb_init_func} once per
3758 object module, passing it as its sole argument the address of a block
3759 allocated in the object module.
3761 The name of the block is a local symbol made with this statement:
3764 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3767 Of course, since you are writing the definition of
3768 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3769 can take a short cut in the definition of this macro and use the name
3770 that you know will result.
3772 The first word of this block is a flag which will be nonzero if the
3773 object module has already been initialized. So test this word first,
3774 and do not call @code{__bb_init_func} if the flag is
3775 nonzero. BLOCK_OR_LABEL contains a unique number which may be used to
3776 generate a label as a branch destination when @code{__bb_init_func}
3779 Described in assembler language, the code to be output looks like:
3789 @findex __bb_init_trace_func
3790 @item profile_block_flag == 2
3791 Output code to call the subroutine @code{__bb_init_trace_func}
3792 and pass two parameters to it. The first parameter is the same as
3793 for @code{__bb_init_func}. The second parameter is the number of the
3794 first basic block of the function as given by BLOCK_OR_LABEL. Note
3795 that @code{__bb_init_trace_func} has to be called, even if the object
3796 module has been initialized already.
3798 Described in assembler language, the code to be output looks like:
3801 parameter2 <- BLOCK_OR_LABEL
3802 call __bb_init_trace_func
3806 @findex BLOCK_PROFILER
3807 @vindex profile_block_flag
3808 @item BLOCK_PROFILER (@var{file}, @var{blockno})
3809 A C statement or compound statement to output to @var{file} some
3810 assembler code to increment the count associated with the basic
3811 block number @var{blockno}. The global compile flag
3812 @code{profile_block_flag} distinguishes two profile modes.
3815 @item profile_block_flag != 2
3816 Output code to increment the counter directly. Basic blocks are
3817 numbered separately from zero within each compilation. The count
3818 associated with block number @var{blockno} is at index
3819 @var{blockno} in a vector of words; the name of this array is a local
3820 symbol made with this statement:
3823 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 2);
3826 @c This paragraph is the same as one a few paragraphs up.
3827 @c That is not an error.
3828 Of course, since you are writing the definition of
3829 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3830 can take a short cut in the definition of this macro and use the name
3831 that you know will result.
3833 Described in assembler language, the code to be output looks like:
3836 inc (LPBX2+4*BLOCKNO)
3840 @findex __bb_trace_func
3841 @item profile_block_flag == 2
3842 Output code to initialize the global structure @code{__bb} and
3843 call the function @code{__bb_trace_func}, which will increment the
3846 @code{__bb} consists of two words. In the first word, the current
3847 basic block number, as given by BLOCKNO, has to be stored. In
3848 the second word, the address of a block allocated in the object
3849 module has to be stored. The address is given by the label created
3850 with this statement:
3853 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3856 Described in assembler language, the code to be output looks like:
3858 move BLOCKNO -> (__bb)
3859 move LPBX0 -> (__bb+4)
3860 call __bb_trace_func
3864 @findex FUNCTION_BLOCK_PROFILER_EXIT
3865 @findex __bb_trace_ret
3866 @vindex profile_block_flag
3867 @item FUNCTION_BLOCK_PROFILER_EXIT (@var{file})
3868 A C statement or compound statement to output to @var{file}
3869 assembler code to call function @code{__bb_trace_ret}. The
3870 assembler code should only be output
3871 if the global compile flag @code{profile_block_flag} == 2. This
3872 macro has to be used at every place where code for returning from
3873 a function is generated (e.g. @code{FUNCTION_EPILOGUE}). Although
3874 you have to write the definition of @code{FUNCTION_EPILOGUE}
3875 as well, you have to define this macro to tell the compiler, that
3876 the proper call to @code{__bb_trace_ret} is produced.
3878 @findex MACHINE_STATE_SAVE
3879 @findex __bb_init_trace_func
3880 @findex __bb_trace_func
3881 @findex __bb_trace_ret
3882 @item MACHINE_STATE_SAVE (@var{id})
3883 A C statement or compound statement to save all registers, which may
3884 be clobbered by a function call, including condition codes. The
3885 @code{asm} statement will be mostly likely needed to handle this
3886 task. Local labels in the assembler code can be concatenated with the
3887 string @var{id}, to obtain a unique label name.
3889 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3890 @code{FUNCTION_EPILOGUE} must be saved in the macros
3891 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3892 @code{BLOCK_PROFILER} prior calling @code{__bb_init_trace_func},
3893 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3895 @findex MACHINE_STATE_RESTORE
3896 @findex __bb_init_trace_func
3897 @findex __bb_trace_func
3898 @findex __bb_trace_ret
3899 @item MACHINE_STATE_RESTORE (@var{id})
3900 A C statement or compound statement to restore all registers, including
3901 condition codes, saved by @code{MACHINE_STATE_SAVE}.
3903 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3904 @code{FUNCTION_EPILOGUE} must be restored in the macros
3905 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3906 @code{BLOCK_PROFILER} after calling @code{__bb_init_trace_func},
3907 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3909 @findex BLOCK_PROFILER_CODE
3910 @item BLOCK_PROFILER_CODE
3911 A C function or functions which are needed in the library to
3912 support block profiling.
3916 @subsection Permitting inlining of functions with attributes
3919 By default if a function has a target specific attribute attached to it,
3920 it will not be inlined. This behaviour can be overridden if the target
3921 defines the @samp{FUNCTION_ATTRIBUTE_INLINABLE_P} macro. This macro
3922 takes one argument, a @samp{DECL} describing the function. It should
3923 return non-zero if the function can be inlined, otherwise it should
3927 @subsection Permitting tail calls to functions
3929 @cindex sibling calls
3932 @findex FUNCTION_OK_FOR_SIBCALL
3933 @item FUNCTION_OK_FOR_SIBCALL (@var{decl})
3934 A C expression that evaluates to true if it is ok to perform a sibling
3937 It is not uncommon for limitations of calling conventions to prevent
3938 tail calls to functions outside the current unit of translation, or
3939 during PIC compilation. Use this macro to enforce these restrictions,
3940 as the @code{sibcall} md pattern can not fail, or fall over to a
3945 @section Implementing the Varargs Macros
3946 @cindex varargs implementation
3948 GCC comes with an implementation of @file{varargs.h} and
3949 @file{stdarg.h} that work without change on machines that pass arguments
3950 on the stack. Other machines require their own implementations of
3951 varargs, and the two machine independent header files must have
3952 conditionals to include it.
3954 ISO @file{stdarg.h} differs from traditional @file{varargs.h} mainly in
3955 the calling convention for @code{va_start}. The traditional
3956 implementation takes just one argument, which is the variable in which
3957 to store the argument pointer. The ISO implementation of
3958 @code{va_start} takes an additional second argument. The user is
3959 supposed to write the last named argument of the function here.
3961 However, @code{va_start} should not use this argument. The way to find
3962 the end of the named arguments is with the built-in functions described
3966 @findex __builtin_saveregs
3967 @item __builtin_saveregs ()
3968 Use this built-in function to save the argument registers in memory so
3969 that the varargs mechanism can access them. Both ISO and traditional
3970 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3971 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
3973 On some machines, @code{__builtin_saveregs} is open-coded under the
3974 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
3975 it calls a routine written in assembler language, found in
3978 Code generated for the call to @code{__builtin_saveregs} appears at the
3979 beginning of the function, as opposed to where the call to
3980 @code{__builtin_saveregs} is written, regardless of what the code is.
3981 This is because the registers must be saved before the function starts
3982 to use them for its own purposes.
3983 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3986 @findex __builtin_args_info
3987 @item __builtin_args_info (@var{category})
3988 Use this built-in function to find the first anonymous arguments in
3991 In general, a machine may have several categories of registers used for
3992 arguments, each for a particular category of data types. (For example,
3993 on some machines, floating-point registers are used for floating-point
3994 arguments while other arguments are passed in the general registers.)
3995 To make non-varargs functions use the proper calling convention, you
3996 have defined the @code{CUMULATIVE_ARGS} data type to record how many
3997 registers in each category have been used so far
3999 @code{__builtin_args_info} accesses the same data structure of type
4000 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4001 with it, with @var{category} specifying which word to access. Thus, the
4002 value indicates the first unused register in a given category.
4004 Normally, you would use @code{__builtin_args_info} in the implementation
4005 of @code{va_start}, accessing each category just once and storing the
4006 value in the @code{va_list} object. This is because @code{va_list} will
4007 have to update the values, and there is no way to alter the
4008 values accessed by @code{__builtin_args_info}.
4010 @findex __builtin_next_arg
4011 @item __builtin_next_arg (@var{lastarg})
4012 This is the equivalent of @code{__builtin_args_info}, for stack
4013 arguments. It returns the address of the first anonymous stack
4014 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4015 returns the address of the location above the first anonymous stack
4016 argument. Use it in @code{va_start} to initialize the pointer for
4017 fetching arguments from the stack. Also use it in @code{va_start} to
4018 verify that the second parameter @var{lastarg} is the last named argument
4019 of the current function.
4021 @findex __builtin_classify_type
4022 @item __builtin_classify_type (@var{object})
4023 Since each machine has its own conventions for which data types are
4024 passed in which kind of register, your implementation of @code{va_arg}
4025 has to embody these conventions. The easiest way to categorize the
4026 specified data type is to use @code{__builtin_classify_type} together
4027 with @code{sizeof} and @code{__alignof__}.
4029 @code{__builtin_classify_type} ignores the value of @var{object},
4030 considering only its data type. It returns an integer describing what
4031 kind of type that is---integer, floating, pointer, structure, and so on.
4033 The file @file{typeclass.h} defines an enumeration that you can use to
4034 interpret the values of @code{__builtin_classify_type}.
4037 These machine description macros help implement varargs:
4040 @findex EXPAND_BUILTIN_SAVEREGS
4041 @item EXPAND_BUILTIN_SAVEREGS ()
4042 If defined, is a C expression that produces the machine-specific code
4043 for a call to @code{__builtin_saveregs}. This code will be moved to the
4044 very beginning of the function, before any parameter access are made.
4045 The return value of this function should be an RTX that contains the
4046 value to use as the return of @code{__builtin_saveregs}.
4048 @findex SETUP_INCOMING_VARARGS
4049 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
4050 This macro offers an alternative to using @code{__builtin_saveregs} and
4051 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
4052 anonymous register arguments into the stack so that all the arguments
4053 appear to have been passed consecutively on the stack. Once this is
4054 done, you can use the standard implementation of varargs that works for
4055 machines that pass all their arguments on the stack.
4057 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
4058 structure, containing the values that are obtained after processing the
4059 named arguments. The arguments @var{mode} and @var{type} describe the
4060 last named argument---its machine mode and its data type as a tree node.
4062 The macro implementation should do two things: first, push onto the
4063 stack all the argument registers @emph{not} used for the named
4064 arguments, and second, store the size of the data thus pushed into the
4065 @code{int}-valued variable whose name is supplied as the argument
4066 @var{pretend_args_size}. The value that you store here will serve as
4067 additional offset for setting up the stack frame.
4069 Because you must generate code to push the anonymous arguments at
4070 compile time without knowing their data types,
4071 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
4072 a single category of argument register and use it uniformly for all data
4075 If the argument @var{second_time} is nonzero, it means that the
4076 arguments of the function are being analyzed for the second time. This
4077 happens for an inline function, which is not actually compiled until the
4078 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
4079 not generate any instructions in this case.
4081 @findex STRICT_ARGUMENT_NAMING
4082 @item STRICT_ARGUMENT_NAMING
4083 Define this macro to be a nonzero value if the location where a function
4084 argument is passed depends on whether or not it is a named argument.
4086 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
4087 is set for varargs and stdarg functions. If this macro returns a
4088 nonzero value, the @var{named} argument is always true for named
4089 arguments, and false for unnamed arguments. If it returns a value of
4090 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
4091 are treated as named. Otherwise, all named arguments except the last
4092 are treated as named.
4094 You need not define this macro if it always returns zero.
4096 @findex PRETEND_OUTGOING_VARARGS_NAMED
4097 @item PRETEND_OUTGOING_VARARGS_NAMED
4098 If you need to conditionally change ABIs so that one works with
4099 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
4100 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
4101 defined, then define this macro to return nonzero if
4102 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
4103 Otherwise, you should not define this macro.
4107 @section Trampolines for Nested Functions
4108 @cindex trampolines for nested functions
4109 @cindex nested functions, trampolines for
4111 A @dfn{trampoline} is a small piece of code that is created at run time
4112 when the address of a nested function is taken. It normally resides on
4113 the stack, in the stack frame of the containing function. These macros
4114 tell GCC how to generate code to allocate and initialize a
4117 The instructions in the trampoline must do two things: load a constant
4118 address into the static chain register, and jump to the real address of
4119 the nested function. On CISC machines such as the m68k, this requires
4120 two instructions, a move immediate and a jump. Then the two addresses
4121 exist in the trampoline as word-long immediate operands. On RISC
4122 machines, it is often necessary to load each address into a register in
4123 two parts. Then pieces of each address form separate immediate
4126 The code generated to initialize the trampoline must store the variable
4127 parts---the static chain value and the function address---into the
4128 immediate operands of the instructions. On a CISC machine, this is
4129 simply a matter of copying each address to a memory reference at the
4130 proper offset from the start of the trampoline. On a RISC machine, it
4131 may be necessary to take out pieces of the address and store them
4135 @findex TRAMPOLINE_TEMPLATE
4136 @item TRAMPOLINE_TEMPLATE (@var{file})
4137 A C statement to output, on the stream @var{file}, assembler code for a
4138 block of data that contains the constant parts of a trampoline. This
4139 code should not include a label---the label is taken care of
4142 If you do not define this macro, it means no template is needed
4143 for the target. Do not define this macro on systems where the block move
4144 code to copy the trampoline into place would be larger than the code
4145 to generate it on the spot.
4147 @findex TRAMPOLINE_SECTION
4148 @item TRAMPOLINE_SECTION
4149 The name of a subroutine to switch to the section in which the
4150 trampoline template is to be placed (@pxref{Sections}). The default is
4151 a value of @samp{readonly_data_section}, which places the trampoline in
4152 the section containing read-only data.
4154 @findex TRAMPOLINE_SIZE
4155 @item TRAMPOLINE_SIZE
4156 A C expression for the size in bytes of the trampoline, as an integer.
4158 @findex TRAMPOLINE_ALIGNMENT
4159 @item TRAMPOLINE_ALIGNMENT
4160 Alignment required for trampolines, in bits.
4162 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4163 is used for aligning trampolines.
4165 @findex INITIALIZE_TRAMPOLINE
4166 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4167 A C statement to initialize the variable parts of a trampoline.
4168 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4169 an RTX for the address of the nested function; @var{static_chain} is an
4170 RTX for the static chain value that should be passed to the function
4173 @findex TRAMPOLINE_ADJUST_ADDRESS
4174 @item TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4175 A C statement that should perform any machine-specific adjustment in
4176 the address of the trampoline. Its argument contains the address that
4177 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4178 used for a function call should be different from the address in which
4179 the template was stored, the different address should be assigned to
4180 @var{addr}. If this macro is not defined, @var{addr} will be used for
4183 @findex ALLOCATE_TRAMPOLINE
4184 @item ALLOCATE_TRAMPOLINE (@var{fp})
4185 A C expression to allocate run-time space for a trampoline. The
4186 expression value should be an RTX representing a memory reference to the
4187 space for the trampoline.
4189 @cindex @code{FUNCTION_EPILOGUE} and trampolines
4190 @cindex @code{FUNCTION_PROLOGUE} and trampolines
4191 If this macro is not defined, by default the trampoline is allocated as
4192 a stack slot. This default is right for most machines. The exceptions
4193 are machines where it is impossible to execute instructions in the stack
4194 area. On such machines, you may have to implement a separate stack,
4195 using this macro in conjunction with @code{FUNCTION_PROLOGUE} and
4196 @code{FUNCTION_EPILOGUE}.
4198 @var{fp} points to a data structure, a @code{struct function}, which
4199 describes the compilation status of the immediate containing function of
4200 the function which the trampoline is for. Normally (when
4201 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
4202 trampoline is in the stack frame of this containing function. Other
4203 allocation strategies probably must do something analogous with this
4207 Implementing trampolines is difficult on many machines because they have
4208 separate instruction and data caches. Writing into a stack location
4209 fails to clear the memory in the instruction cache, so when the program
4210 jumps to that location, it executes the old contents.
4212 Here are two possible solutions. One is to clear the relevant parts of
4213 the instruction cache whenever a trampoline is set up. The other is to
4214 make all trampolines identical, by having them jump to a standard
4215 subroutine. The former technique makes trampoline execution faster; the
4216 latter makes initialization faster.
4218 To clear the instruction cache when a trampoline is initialized, define
4219 the following macros which describe the shape of the cache.
4222 @findex INSN_CACHE_SIZE
4223 @item INSN_CACHE_SIZE
4224 The total size in bytes of the cache.
4226 @findex INSN_CACHE_LINE_WIDTH
4227 @item INSN_CACHE_LINE_WIDTH
4228 The length in bytes of each cache line. The cache is divided into cache
4229 lines which are disjoint slots, each holding a contiguous chunk of data
4230 fetched from memory. Each time data is brought into the cache, an
4231 entire line is read at once. The data loaded into a cache line is
4232 always aligned on a boundary equal to the line size.
4234 @findex INSN_CACHE_DEPTH
4235 @item INSN_CACHE_DEPTH
4236 The number of alternative cache lines that can hold any particular memory
4240 Alternatively, if the machine has system calls or instructions to clear
4241 the instruction cache directly, you can define the following macro.
4244 @findex CLEAR_INSN_CACHE
4245 @item CLEAR_INSN_CACHE (@var{BEG}, @var{END})
4246 If defined, expands to a C expression clearing the @emph{instruction
4247 cache} in the specified interval. If it is not defined, and the macro
4248 INSN_CACHE_SIZE is defined, some generic code is generated to clear the
4249 cache. The definition of this macro would typically be a series of
4250 @code{asm} statements. Both @var{BEG} and @var{END} are both pointer
4254 To use a standard subroutine, define the following macro. In addition,
4255 you must make sure that the instructions in a trampoline fill an entire
4256 cache line with identical instructions, or else ensure that the
4257 beginning of the trampoline code is always aligned at the same point in
4258 its cache line. Look in @file{m68k.h} as a guide.
4261 @findex TRANSFER_FROM_TRAMPOLINE
4262 @item TRANSFER_FROM_TRAMPOLINE
4263 Define this macro if trampolines need a special subroutine to do their
4264 work. The macro should expand to a series of @code{asm} statements
4265 which will be compiled with GCC. They go in a library function named
4266 @code{__transfer_from_trampoline}.
4268 If you need to avoid executing the ordinary prologue code of a compiled
4269 C function when you jump to the subroutine, you can do so by placing a
4270 special label of your own in the assembler code. Use one @code{asm}
4271 statement to generate an assembler label, and another to make the label
4272 global. Then trampolines can use that label to jump directly to your
4273 special assembler code.
4277 @section Implicit Calls to Library Routines
4278 @cindex library subroutine names
4279 @cindex @file{libgcc.a}
4281 @c prevent bad page break with this line
4282 Here is an explanation of implicit calls to library routines.
4285 @findex MULSI3_LIBCALL
4286 @item MULSI3_LIBCALL
4287 A C string constant giving the name of the function to call for
4288 multiplication of one signed full-word by another. If you do not
4289 define this macro, the default name is used, which is @code{__mulsi3},
4290 a function defined in @file{libgcc.a}.
4292 @findex DIVSI3_LIBCALL
4293 @item DIVSI3_LIBCALL
4294 A C string constant giving the name of the function to call for
4295 division of one signed full-word by another. If you do not define
4296 this macro, the default name is used, which is @code{__divsi3}, a
4297 function defined in @file{libgcc.a}.
4299 @findex UDIVSI3_LIBCALL
4300 @item UDIVSI3_LIBCALL
4301 A C string constant giving the name of the function to call for
4302 division of one unsigned full-word by another. If you do not define
4303 this macro, the default name is used, which is @code{__udivsi3}, a
4304 function defined in @file{libgcc.a}.
4306 @findex MODSI3_LIBCALL
4307 @item MODSI3_LIBCALL
4308 A C string constant giving the name of the function to call for the
4309 remainder in division of one signed full-word by another. If you do
4310 not define this macro, the default name is used, which is
4311 @code{__modsi3}, a function defined in @file{libgcc.a}.
4313 @findex UMODSI3_LIBCALL
4314 @item UMODSI3_LIBCALL
4315 A C string constant giving the name of the function to call for the
4316 remainder in division of one unsigned full-word by another. If you do
4317 not define this macro, the default name is used, which is
4318 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4320 @findex MULDI3_LIBCALL
4321 @item MULDI3_LIBCALL
4322 A C string constant giving the name of the function to call for
4323 multiplication of one signed double-word by another. If you do not
4324 define this macro, the default name is used, which is @code{__muldi3},
4325 a function defined in @file{libgcc.a}.
4327 @findex DIVDI3_LIBCALL
4328 @item DIVDI3_LIBCALL
4329 A C string constant giving the name of the function to call for
4330 division of one signed double-word by another. If you do not define
4331 this macro, the default name is used, which is @code{__divdi3}, a
4332 function defined in @file{libgcc.a}.
4334 @findex UDIVDI3_LIBCALL
4335 @item UDIVDI3_LIBCALL
4336 A C string constant giving the name of the function to call for
4337 division of one unsigned full-word by another. If you do not define
4338 this macro, the default name is used, which is @code{__udivdi3}, a
4339 function defined in @file{libgcc.a}.
4341 @findex MODDI3_LIBCALL
4342 @item MODDI3_LIBCALL
4343 A C string constant giving the name of the function to call for the
4344 remainder in division of one signed double-word by another. If you do
4345 not define this macro, the default name is used, which is
4346 @code{__moddi3}, a function defined in @file{libgcc.a}.
4348 @findex UMODDI3_LIBCALL
4349 @item UMODDI3_LIBCALL
4350 A C string constant giving the name of the function to call for the
4351 remainder in division of one unsigned full-word by another. If you do
4352 not define this macro, the default name is used, which is
4353 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4355 @findex INIT_TARGET_OPTABS
4356 @item INIT_TARGET_OPTABS
4357 Define this macro as a C statement that declares additional library
4358 routines renames existing ones. @code{init_optabs} calls this macro after
4359 initializing all the normal library routines.
4361 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4362 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4363 Define this macro as a C statement that returns nonzero if a call to
4364 the floating point comparison library function will return a boolean
4365 value that indicates the result of the comparison. It should return
4366 zero if one of gcc's own libgcc functions is called.
4368 Most ports don't need to define this macro.
4371 @cindex @code{EDOM}, implicit usage
4373 The value of @code{EDOM} on the target machine, as a C integer constant
4374 expression. If you don't define this macro, GCC does not attempt to
4375 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4376 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4379 If you do not define @code{TARGET_EDOM}, then compiled code reports
4380 domain errors by calling the library function and letting it report the
4381 error. If mathematical functions on your system use @code{matherr} when
4382 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4383 that @code{matherr} is used normally.
4385 @findex GEN_ERRNO_RTX
4386 @cindex @code{errno}, implicit usage
4388 Define this macro as a C expression to create an rtl expression that
4389 refers to the global ``variable'' @code{errno}. (On certain systems,
4390 @code{errno} may not actually be a variable.) If you don't define this
4391 macro, a reasonable default is used.
4393 @findex TARGET_MEM_FUNCTIONS
4394 @cindex @code{bcopy}, implicit usage
4395 @cindex @code{memcpy}, implicit usage
4396 @cindex @code{bzero}, implicit usage
4397 @cindex @code{memset}, implicit usage
4398 @item TARGET_MEM_FUNCTIONS
4399 Define this macro if GCC should generate calls to the ISO C
4400 (and System V) library functions @code{memcpy} and @code{memset}
4401 rather than the BSD functions @code{bcopy} and @code{bzero}.
4403 @findex LIBGCC_NEEDS_DOUBLE
4404 @item LIBGCC_NEEDS_DOUBLE
4405 Define this macro if only @code{float} arguments cannot be passed to
4406 library routines (so they must be converted to @code{double}). This
4407 macro affects both how library calls are generated and how the library
4408 routines in @file{libgcc1.c} accept their arguments. It is useful on
4409 machines where floating and fixed point arguments are passed
4410 differently, such as the i860.
4412 @findex FLOAT_ARG_TYPE
4413 @item FLOAT_ARG_TYPE
4414 Define this macro to override the type used by the library routines to
4415 pick up arguments of type @code{float}. (By default, they use a union
4416 of @code{float} and @code{int}.)
4418 The obvious choice would be @code{float}---but that won't work with
4419 traditional C compilers that expect all arguments declared as @code{float}
4420 to arrive as @code{double}. To avoid this conversion, the library routines
4421 ask for the value as some other type and then treat it as a @code{float}.
4423 On some systems, no other type will work for this. For these systems,
4424 you must use @code{LIBGCC_NEEDS_DOUBLE} instead, to force conversion of
4425 the values @code{double} before they are passed.
4428 @item FLOATIFY (@var{passed-value})
4429 Define this macro to override the way library routines redesignate a
4430 @code{float} argument as a @code{float} instead of the type it was
4431 passed as. The default is an expression which takes the @code{float}
4434 @findex FLOAT_VALUE_TYPE
4435 @item FLOAT_VALUE_TYPE
4436 Define this macro to override the type used by the library routines to
4437 return values that ought to have type @code{float}. (By default, they
4440 The obvious choice would be @code{float}---but that won't work with
4441 traditional C compilers gratuitously convert values declared as
4442 @code{float} into @code{double}.
4445 @item INTIFY (@var{float-value})
4446 Define this macro to override the way the value of a
4447 @code{float}-returning library routine should be packaged in order to
4448 return it. These functions are actually declared to return type
4449 @code{FLOAT_VALUE_TYPE} (normally @code{int}).
4451 These values can't be returned as type @code{float} because traditional
4452 C compilers would gratuitously convert the value to a @code{double}.
4454 A local variable named @code{intify} is always available when the macro
4455 @code{INTIFY} is used. It is a union of a @code{float} field named
4456 @code{f} and a field named @code{i} whose type is
4457 @code{FLOAT_VALUE_TYPE} or @code{int}.
4459 If you don't define this macro, the default definition works by copying
4460 the value through that union.
4462 @findex nongcc_SI_type
4463 @item nongcc_SI_type
4464 Define this macro as the name of the data type corresponding to
4465 @code{SImode} in the system's own C compiler.
4467 You need not define this macro if that type is @code{long int}, as it usually
4470 @findex nongcc_word_type
4471 @item nongcc_word_type
4472 Define this macro as the name of the data type corresponding to the
4473 word_mode in the system's own C compiler.
4475 You need not define this macro if that type is @code{long int}, as it usually
4478 @findex perform_@dots{}
4479 @item perform_@dots{}
4480 Define these macros to supply explicit C statements to carry out various
4481 arithmetic operations on types @code{float} and @code{double} in the
4482 library routines in @file{libgcc1.c}. See that file for a full list
4483 of these macros and their arguments.
4485 On most machines, you don't need to define any of these macros, because
4486 the C compiler that comes with the system takes care of doing them.
4488 @findex NEXT_OBJC_RUNTIME
4489 @item NEXT_OBJC_RUNTIME
4490 Define this macro to generate code for Objective C message sending using
4491 the calling convention of the NeXT system. This calling convention
4492 involves passing the object, the selector and the method arguments all
4493 at once to the method-lookup library function.
4495 The default calling convention passes just the object and the selector
4496 to the lookup function, which returns a pointer to the method.
4499 @node Addressing Modes
4500 @section Addressing Modes
4501 @cindex addressing modes
4503 @c prevent bad page break with this line
4504 This is about addressing modes.
4507 @findex HAVE_PRE_INCREMENT
4508 @findex HAVE_PRE_DECREMENT
4509 @findex HAVE_POST_INCREMENT
4510 @findex HAVE_POST_DECREMENT
4511 @item HAVE_PRE_INCREMENT
4512 @itemx HAVE_PRE_DECREMENT
4513 @itemx HAVE_POST_INCREMENT
4514 @itemx HAVE_POST_DECREMENT
4515 A C expression that is non-zero if the machine supports pre-increment,
4516 pre-decrement, post-increment, or post-decrement addressing respectively.
4518 @findex HAVE_POST_MODIFY_DISP
4519 @findex HAVE_PRE_MODIFY_DISP
4520 @item HAVE_PRE_MODIFY_DISP
4521 @itemx HAVE_POST_MODIFY_DISP
4522 A C expression that is non-zero if the machine supports pre- or
4523 post-address side-effect generation involving constants other than
4524 the size of the memory operand.
4526 @findex HAVE_POST_MODIFY_REG
4527 @findex HAVE_PRE_MODIFY_REG
4528 @item HAVE_PRE_MODIFY_REG
4529 @itemx HAVE_POST_MODIFY_REG
4530 A C expression that is non-zero if the machine supports pre- or
4531 post-address side-effect generation involving a register displacement.
4533 @findex CONSTANT_ADDRESS_P
4534 @item CONSTANT_ADDRESS_P (@var{x})
4535 A C expression that is 1 if the RTX @var{x} is a constant which
4536 is a valid address. On most machines, this can be defined as
4537 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4538 in which constant addresses are supported.
4541 @code{CONSTANT_P} accepts integer-values expressions whose values are
4542 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4543 @code{high} expressions and @code{const} arithmetic expressions, in
4544 addition to @code{const_int} and @code{const_double} expressions.
4546 @findex MAX_REGS_PER_ADDRESS
4547 @item MAX_REGS_PER_ADDRESS
4548 A number, the maximum number of registers that can appear in a valid
4549 memory address. Note that it is up to you to specify a value equal to
4550 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4553 @findex GO_IF_LEGITIMATE_ADDRESS
4554 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4555 A C compound statement with a conditional @code{goto @var{label};}
4556 executed if @var{x} (an RTX) is a legitimate memory address on the
4557 target machine for a memory operand of mode @var{mode}.
4559 It usually pays to define several simpler macros to serve as
4560 subroutines for this one. Otherwise it may be too complicated to
4563 This macro must exist in two variants: a strict variant and a
4564 non-strict one. The strict variant is used in the reload pass. It
4565 must be defined so that any pseudo-register that has not been
4566 allocated a hard register is considered a memory reference. In
4567 contexts where some kind of register is required, a pseudo-register
4568 with no hard register must be rejected.
4570 The non-strict variant is used in other passes. It must be defined to
4571 accept all pseudo-registers in every context where some kind of
4572 register is required.
4574 @findex REG_OK_STRICT
4575 Compiler source files that want to use the strict variant of this
4576 macro define the macro @code{REG_OK_STRICT}. You should use an
4577 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4578 in that case and the non-strict variant otherwise.
4580 Subroutines to check for acceptable registers for various purposes (one
4581 for base registers, one for index registers, and so on) are typically
4582 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4583 Then only these subroutine macros need have two variants; the higher
4584 levels of macros may be the same whether strict or not.@refill
4586 Normally, constant addresses which are the sum of a @code{symbol_ref}
4587 and an integer are stored inside a @code{const} RTX to mark them as
4588 constant. Therefore, there is no need to recognize such sums
4589 specifically as legitimate addresses. Normally you would simply
4590 recognize any @code{const} as legitimate.
4592 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4593 sums that are not marked with @code{const}. It assumes that a naked
4594 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4595 naked constant sums as illegitimate addresses, so that none of them will
4596 be given to @code{PRINT_OPERAND_ADDRESS}.
4598 @cindex @code{ENCODE_SECTION_INFO} and address validation
4599 On some machines, whether a symbolic address is legitimate depends on
4600 the section that the address refers to. On these machines, define the
4601 macro @code{ENCODE_SECTION_INFO} to store the information into the
4602 @code{symbol_ref}, and then check for it here. When you see a
4603 @code{const}, you will have to look inside it to find the
4604 @code{symbol_ref} in order to determine the section. @xref{Assembler
4607 @findex saveable_obstack
4608 The best way to modify the name string is by adding text to the
4609 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4610 the new name in @code{saveable_obstack}. You will have to modify
4611 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4612 output the name accordingly, and define @code{STRIP_NAME_ENCODING} to
4613 access the original name string.
4615 You can check the information stored here into the @code{symbol_ref} in
4616 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4617 @code{PRINT_OPERAND_ADDRESS}.
4619 @findex REG_OK_FOR_BASE_P
4620 @item REG_OK_FOR_BASE_P (@var{x})
4621 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4622 RTX) is valid for use as a base register. For hard registers, it
4623 should always accept those which the hardware permits and reject the
4624 others. Whether the macro accepts or rejects pseudo registers must be
4625 controlled by @code{REG_OK_STRICT} as described above. This usually
4626 requires two variant definitions, of which @code{REG_OK_STRICT}
4627 controls the one actually used.
4629 @findex REG_MODE_OK_FOR_BASE_P
4630 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4631 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4632 that expression may examine the mode of the memory reference in
4633 @var{mode}. You should define this macro if the mode of the memory
4634 reference affects whether a register may be used as a base register. If
4635 you define this macro, the compiler will use it instead of
4636 @code{REG_OK_FOR_BASE_P}.
4638 @findex REG_OK_FOR_INDEX_P
4639 @item REG_OK_FOR_INDEX_P (@var{x})
4640 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4641 RTX) is valid for use as an index register.
4643 The difference between an index register and a base register is that
4644 the index register may be scaled. If an address involves the sum of
4645 two registers, neither one of them scaled, then either one may be
4646 labeled the ``base'' and the other the ``index''; but whichever
4647 labeling is used must fit the machine's constraints of which registers
4648 may serve in each capacity. The compiler will try both labelings,
4649 looking for one that is valid, and will reload one or both registers
4650 only if neither labeling works.
4652 @findex FIND_BASE_TERM
4653 @item FIND_BASE_TERM (@var{x})
4654 A C expression to determine the base term of address @var{x}.
4655 This macro is used in only one place: `find_base_term' in alias.c.
4657 It is always safe for this macro to not be defined. It exists so
4658 that alias analysis can understand machine-dependent addresses.
4660 The typical use of this macro is to handle addresses containing
4661 a label_ref or symbol_ref within an UNSPEC.
4663 @findex LEGITIMIZE_ADDRESS
4664 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4665 A C compound statement that attempts to replace @var{x} with a valid
4666 memory address for an operand of mode @var{mode}. @var{win} will be a
4667 C statement label elsewhere in the code; the macro definition may use
4670 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4674 to avoid further processing if the address has become legitimate.
4676 @findex break_out_memory_refs
4677 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4678 and @var{oldx} will be the operand that was given to that function to produce
4681 The code generated by this macro should not alter the substructure of
4682 @var{x}. If it transforms @var{x} into a more legitimate form, it
4683 should assign @var{x} (which will always be a C variable) a new value.
4685 It is not necessary for this macro to come up with a legitimate
4686 address. The compiler has standard ways of doing so in all cases. In
4687 fact, it is safe for this macro to do nothing. But often a
4688 machine-dependent strategy can generate better code.
4690 @findex LEGITIMIZE_RELOAD_ADDRESS
4691 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4692 A C compound statement that attempts to replace @var{x}, which is an address
4693 that needs reloading, with a valid memory address for an operand of mode
4694 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4695 It is not necessary to define this macro, but it might be useful for
4696 performance reasons.
4698 For example, on the i386, it is sometimes possible to use a single
4699 reload register instead of two by reloading a sum of two pseudo
4700 registers into a register. On the other hand, for number of RISC
4701 processors offsets are limited so that often an intermediate address
4702 needs to be generated in order to address a stack slot. By defining
4703 LEGITIMIZE_RELOAD_ADDRESS appropriately, the intermediate addresses
4704 generated for adjacent some stack slots can be made identical, and thus
4707 @emph{Note}: This macro should be used with caution. It is necessary
4708 to know something of how reload works in order to effectively use this,
4709 and it is quite easy to produce macros that build in too much knowledge
4710 of reload internals.
4712 @emph{Note}: This macro must be able to reload an address created by a
4713 previous invocation of this macro. If it fails to handle such addresses
4714 then the compiler may generate incorrect code or abort.
4717 The macro definition should use @code{push_reload} to indicate parts that
4718 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4719 suitable to be passed unaltered to @code{push_reload}.
4721 The code generated by this macro must not alter the substructure of
4722 @var{x}. If it transforms @var{x} into a more legitimate form, it
4723 should assign @var{x} (which will always be a C variable) a new value.
4724 This also applies to parts that you change indirectly by calling
4727 @findex strict_memory_address_p
4728 The macro definition may use @code{strict_memory_address_p} to test if
4729 the address has become legitimate.
4732 If you want to change only a part of @var{x}, one standard way of doing
4733 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4734 single level of rtl. Thus, if the part to be changed is not at the
4735 top level, you'll need to replace first the top leve
4736 It is not necessary for this macro to come up with a legitimate
4737 address; but often a machine-dependent strategy can generate better code.
4739 @findex GO_IF_MODE_DEPENDENT_ADDRESS
4740 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4741 A C statement or compound statement with a conditional @code{goto
4742 @var{label};} executed if memory address @var{x} (an RTX) can have
4743 different meanings depending on the machine mode of the memory
4744 reference it is used for or if the address is valid for some modes
4747 Autoincrement and autodecrement addresses typically have mode-dependent
4748 effects because the amount of the increment or decrement is the size
4749 of the operand being addressed. Some machines have other mode-dependent
4750 addresses. Many RISC machines have no mode-dependent addresses.
4752 You may assume that @var{addr} is a valid address for the machine.
4754 @findex LEGITIMATE_CONSTANT_P
4755 @item LEGITIMATE_CONSTANT_P (@var{x})
4756 A C expression that is nonzero if @var{x} is a legitimate constant for
4757 an immediate operand on the target machine. You can assume that
4758 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4759 @samp{1} is a suitable definition for this macro on machines where
4760 anything @code{CONSTANT_P} is valid.@refill
4763 @node Condition Code
4764 @section Condition Code Status
4765 @cindex condition code status
4767 @c prevent bad page break with this line
4768 This describes the condition code status.
4771 The file @file{conditions.h} defines a variable @code{cc_status} to
4772 describe how the condition code was computed (in case the interpretation of
4773 the condition code depends on the instruction that it was set by). This
4774 variable contains the RTL expressions on which the condition code is
4775 currently based, and several standard flags.
4777 Sometimes additional machine-specific flags must be defined in the machine
4778 description header file. It can also add additional machine-specific
4779 information by defining @code{CC_STATUS_MDEP}.
4782 @findex CC_STATUS_MDEP
4783 @item CC_STATUS_MDEP
4784 C code for a data type which is used for declaring the @code{mdep}
4785 component of @code{cc_status}. It defaults to @code{int}.
4787 This macro is not used on machines that do not use @code{cc0}.
4789 @findex CC_STATUS_MDEP_INIT
4790 @item CC_STATUS_MDEP_INIT
4791 A C expression to initialize the @code{mdep} field to ``empty''.
4792 The default definition does nothing, since most machines don't use
4793 the field anyway. If you want to use the field, you should probably
4794 define this macro to initialize it.
4796 This macro is not used on machines that do not use @code{cc0}.
4798 @findex NOTICE_UPDATE_CC
4799 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4800 A C compound statement to set the components of @code{cc_status}
4801 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4802 this macro's responsibility to recognize insns that set the condition
4803 code as a byproduct of other activity as well as those that explicitly
4806 This macro is not used on machines that do not use @code{cc0}.
4808 If there are insns that do not set the condition code but do alter
4809 other machine registers, this macro must check to see whether they
4810 invalidate the expressions that the condition code is recorded as
4811 reflecting. For example, on the 68000, insns that store in address
4812 registers do not set the condition code, which means that usually
4813 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4814 insns. But suppose that the previous insn set the condition code
4815 based on location @samp{a4@@(102)} and the current insn stores a new
4816 value in @samp{a4}. Although the condition code is not changed by
4817 this, it will no longer be true that it reflects the contents of
4818 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4819 @code{cc_status} in this case to say that nothing is known about the
4820 condition code value.
4822 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4823 with the results of peephole optimization: insns whose patterns are
4824 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4825 constants which are just the operands. The RTL structure of these
4826 insns is not sufficient to indicate what the insns actually do. What
4827 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4828 @code{CC_STATUS_INIT}.
4830 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4831 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4832 @samp{cc}. This avoids having detailed information about patterns in
4833 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4835 @findex EXTRA_CC_MODES
4836 @item EXTRA_CC_MODES
4837 A list of additional modes for condition code values in registers
4838 (@pxref{Jump Patterns}). This macro should expand to a sequence of
4839 calls of the macro @code{CC} separated by white space. @code{CC} takes
4840 two arguments. The first is the enumeration name of the mode, which
4841 should begin with @samp{CC} and end with @samp{mode}. The second is a C
4842 string giving the printable name of the mode; it should be the same as
4843 the first argument, but with the trailing @samp{mode} removed.
4845 You should only define this macro if additional modes are required.
4847 A sample definition of @code{EXTRA_CC_MODES} is:
4849 #define EXTRA_CC_MODES \
4850 CC(CC_NOOVmode, "CC_NOOV") \
4851 CC(CCFPmode, "CCFP") \
4852 CC(CCFPEmode, "CCFPE")
4855 @findex SELECT_CC_MODE
4856 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4857 Returns a mode from class @code{MODE_CC} to be used when comparison
4858 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
4859 example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
4860 @pxref{Jump Patterns} for a description of the reason for this
4864 #define SELECT_CC_MODE(OP,X,Y) \
4865 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4866 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
4867 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4868 || GET_CODE (X) == NEG) \
4869 ? CC_NOOVmode : CCmode))
4872 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
4874 @findex CANONICALIZE_COMPARISON
4875 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
4876 On some machines not all possible comparisons are defined, but you can
4877 convert an invalid comparison into a valid one. For example, the Alpha
4878 does not have a @code{GT} comparison, but you can use an @code{LT}
4879 comparison instead and swap the order of the operands.
4881 On such machines, define this macro to be a C statement to do any
4882 required conversions. @var{code} is the initial comparison code
4883 and @var{op0} and @var{op1} are the left and right operands of the
4884 comparison, respectively. You should modify @var{code}, @var{op0}, and
4885 @var{op1} as required.
4887 GCC will not assume that the comparison resulting from this macro is
4888 valid but will see if the resulting insn matches a pattern in the
4891 You need not define this macro if it would never change the comparison
4894 @findex REVERSIBLE_CC_MODE
4895 @item REVERSIBLE_CC_MODE (@var{mode})
4896 A C expression whose value is one if it is always safe to reverse a
4897 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4898 can ever return @var{mode} for a floating-point inequality comparison,
4899 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4901 You need not define this macro if it would always returns zero or if the
4902 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4903 For example, here is the definition used on the Sparc, where floating-point
4904 inequality comparisons are always given @code{CCFPEmode}:
4907 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
4910 @findex REVERSE_CONDITION (@var{code}, @var{mode})
4911 A C expression whose value is reversed condition code of the @var{code} for
4912 comparison done in CC_MODE @var{mode}. The macro is used only in case
4913 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
4914 machine has some non-standard way how to reverse certain conditionals. For
4915 instance in case all floating point conditions are non-trapping, compiler may
4916 freely convert unordered compares to ordered one. Then definition may look
4920 #define REVERSE_CONDITION(CODE, MODE) \
4921 ((MODE) != CCFPmode ? reverse_condtion (CODE) \
4922 : reverse_condition_maybe_unordered (CODE))
4925 @findex REVERSE_CONDEXEC_PREDICATES_P
4926 @item REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
4927 A C expression that returns true if the conditional execution predicate
4928 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
4929 return 0 if the target has conditional execution predicates that cannot be
4930 reversed safely. If no expansion is specified, this macro is defined as
4934 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) ((x) == reverse_condition (y))
4940 @section Describing Relative Costs of Operations
4941 @cindex costs of instructions
4942 @cindex relative costs
4943 @cindex speed of instructions
4945 These macros let you describe the relative speed of various operations
4946 on the target machine.
4950 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
4951 A part of a C @code{switch} statement that describes the relative costs
4952 of constant RTL expressions. It must contain @code{case} labels for
4953 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
4954 @code{label_ref} and @code{const_double}. Each case must ultimately
4955 reach a @code{return} statement to return the relative cost of the use
4956 of that kind of constant value in an expression. The cost may depend on
4957 the precise value of the constant, which is available for examination in
4958 @var{x}, and the rtx code of the expression in which it is contained,
4959 found in @var{outer_code}.
4961 @var{code} is the expression code---redundant, since it can be
4962 obtained with @code{GET_CODE (@var{x})}.
4965 @findex COSTS_N_INSNS
4966 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4967 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
4968 This can be used, for example, to indicate how costly a multiply
4969 instruction is. In writing this macro, you can use the construct
4970 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
4971 instructions. @var{outer_code} is the code of the expression in which
4972 @var{x} is contained.
4974 This macro is optional; do not define it if the default cost assumptions
4975 are adequate for the target machine.
4977 @findex DEFAULT_RTX_COSTS
4978 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4979 This macro, if defined, is called for any case not handled by the
4980 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
4981 to put case labels into the macro, but the code, or any functions it
4982 calls, must assume that the RTL in @var{x} could be of any type that has
4983 not already been handled. The arguments are the same as for
4984 @code{RTX_COSTS}, and the macro should execute a return statement giving
4985 the cost of any RTL expressions that it can handle. The default cost
4986 calculation is used for any RTL for which this macro does not return a
4989 This macro is optional; do not define it if the default cost assumptions
4990 are adequate for the target machine.
4992 @findex ADDRESS_COST
4993 @item ADDRESS_COST (@var{address})
4994 An expression giving the cost of an addressing mode that contains
4995 @var{address}. If not defined, the cost is computed from
4996 the @var{address} expression and the @code{CONST_COSTS} values.
4998 For most CISC machines, the default cost is a good approximation of the
4999 true cost of the addressing mode. However, on RISC machines, all
5000 instructions normally have the same length and execution time. Hence
5001 all addresses will have equal costs.
5003 In cases where more than one form of an address is known, the form with
5004 the lowest cost will be used. If multiple forms have the same, lowest,
5005 cost, the one that is the most complex will be used.
5007 For example, suppose an address that is equal to the sum of a register
5008 and a constant is used twice in the same basic block. When this macro
5009 is not defined, the address will be computed in a register and memory
5010 references will be indirect through that register. On machines where
5011 the cost of the addressing mode containing the sum is no higher than
5012 that of a simple indirect reference, this will produce an additional
5013 instruction and possibly require an additional register. Proper
5014 specification of this macro eliminates this overhead for such machines.
5016 Similar use of this macro is made in strength reduction of loops.
5018 @var{address} need not be valid as an address. In such a case, the cost
5019 is not relevant and can be any value; invalid addresses need not be
5020 assigned a different cost.
5022 On machines where an address involving more than one register is as
5023 cheap as an address computation involving only one register, defining
5024 @code{ADDRESS_COST} to reflect this can cause two registers to be live
5025 over a region of code where only one would have been if
5026 @code{ADDRESS_COST} were not defined in that manner. This effect should
5027 be considered in the definition of this macro. Equivalent costs should
5028 probably only be given to addresses with different numbers of registers
5029 on machines with lots of registers.
5031 This macro will normally either not be defined or be defined as a
5034 @findex REGISTER_MOVE_COST
5035 @item REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5036 A C expression for the cost of moving data of mode @var{mode} from a
5037 register in class @var{from} to one in class @var{to}. The classes are
5038 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5039 value of 2 is the default; other values are interpreted relative to
5042 It is not required that the cost always equal 2 when @var{from} is the
5043 same as @var{to}; on some machines it is expensive to move between
5044 registers if they are not general registers.
5046 If reload sees an insn consisting of a single @code{set} between two
5047 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5048 classes returns a value of 2, reload does not check to ensure that the
5049 constraints of the insn are met. Setting a cost of other than 2 will
5050 allow reload to verify that the constraints are met. You should do this
5051 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5053 @findex MEMORY_MOVE_COST
5054 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5055 A C expression for the cost of moving data of mode @var{mode} between a
5056 register of class @var{class} and memory; @var{in} is zero if the value
5057 is to be written to memory, non-zero if it is to be read in. This cost
5058 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5059 registers and memory is more expensive than between two registers, you
5060 should define this macro to express the relative cost.
5062 If you do not define this macro, GCC uses a default cost of 4 plus
5063 the cost of copying via a secondary reload register, if one is
5064 needed. If your machine requires a secondary reload register to copy
5065 between memory and a register of @var{class} but the reload mechanism is
5066 more complex than copying via an intermediate, define this macro to
5067 reflect the actual cost of the move.
5069 GCC defines the function @code{memory_move_secondary_cost} if
5070 secondary reloads are needed. It computes the costs due to copying via
5071 a secondary register. If your machine copies from memory using a
5072 secondary register in the conventional way but the default base value of
5073 4 is not correct for your machine, define this macro to add some other
5074 value to the result of that function. The arguments to that function
5075 are the same as to this macro.
5079 A C expression for the cost of a branch instruction. A value of 1 is
5080 the default; other values are interpreted relative to that.
5083 Here are additional macros which do not specify precise relative costs,
5084 but only that certain actions are more expensive than GCC would
5088 @findex SLOW_BYTE_ACCESS
5089 @item SLOW_BYTE_ACCESS
5090 Define this macro as a C expression which is nonzero if accessing less
5091 than a word of memory (i.e. a @code{char} or a @code{short}) is no
5092 faster than accessing a word of memory, i.e., if such access
5093 require more than one instruction or if there is no difference in cost
5094 between byte and (aligned) word loads.
5096 When this macro is not defined, the compiler will access a field by
5097 finding the smallest containing object; when it is defined, a fullword
5098 load will be used if alignment permits. Unless bytes accesses are
5099 faster than word accesses, using word accesses is preferable since it
5100 may eliminate subsequent memory access if subsequent accesses occur to
5101 other fields in the same word of the structure, but to different bytes.
5103 @findex SLOW_ZERO_EXTEND
5104 @item SLOW_ZERO_EXTEND
5105 Define this macro if zero-extension (of a @code{char} or @code{short}
5106 to an @code{int}) can be done faster if the destination is a register
5107 that is known to be zero.
5109 If you define this macro, you must have instruction patterns that
5110 recognize RTL structures like this:
5113 (set (strict_low_part (subreg:QI (reg:SI @dots{}) 0)) @dots{})
5117 and likewise for @code{HImode}.
5119 @findex SLOW_UNALIGNED_ACCESS
5120 @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5121 Define this macro to be the value 1 if memory accesses described by the
5122 @var{mode} and @var{alignment} parameters have a cost many times greater
5123 than aligned accesses, for example if they are emulated in a trap
5126 When this macro is non-zero, the compiler will act as if
5127 @code{STRICT_ALIGNMENT} were non-zero when generating code for block
5128 moves. This can cause significantly more instructions to be produced.
5129 Therefore, do not set this macro non-zero if unaligned accesses only add a
5130 cycle or two to the time for a memory access.
5132 If the value of this macro is always zero, it need not be defined. If
5133 this macro is defined, it should produce a non-zero value when
5134 @code{STRICT_ALIGNMENT} is non-zero.
5136 @findex DONT_REDUCE_ADDR
5137 @item DONT_REDUCE_ADDR
5138 Define this macro to inhibit strength reduction of memory addresses.
5139 (On some machines, such strength reduction seems to do harm rather
5144 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5145 which a sequence of insns should be generated instead of a
5146 string move insn or a library call. Increasing the value will always
5147 make code faster, but eventually incurs high cost in increased code size.
5149 Note that on machines where the corresponding move insn is a
5150 @code{define_expand} that emits a sequence of insns, this macro counts
5151 the number of such sequences.
5153 If you don't define this, a reasonable default is used.
5155 @findex MOVE_BY_PIECES_P
5156 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5157 A C expression used to determine whether @code{move_by_pieces} will be used to
5158 copy a chunk of memory, or whether some other block move mechanism
5159 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5160 than @code{MOVE_RATIO}.
5162 @findex MOVE_MAX_PIECES
5163 @item MOVE_MAX_PIECES
5164 A C expression used by @code{move_by_pieces} to determine the largest unit
5165 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5167 @findex USE_LOAD_POST_INCREMENT
5168 @item USE_LOAD_POST_INCREMENT (@var{mode})
5169 A C expression used to determine whether a load postincrement is a good
5170 thing to use for a given mode. Defaults to the value of
5171 @code{HAVE_POST_INCREMENT}.
5173 @findex USE_LOAD_POST_DECREMENT
5174 @item USE_LOAD_POST_DECREMENT (@var{mode})
5175 A C expression used to determine whether a load postdecrement is a good
5176 thing to use for a given mode. Defaults to the value of
5177 @code{HAVE_POST_DECREMENT}.
5179 @findex USE_LOAD_PRE_INCREMENT
5180 @item USE_LOAD_PRE_INCREMENT (@var{mode})
5181 A C expression used to determine whether a load preincrement is a good
5182 thing to use for a given mode. Defaults to the value of
5183 @code{HAVE_PRE_INCREMENT}.
5185 @findex USE_LOAD_PRE_DECREMENT
5186 @item USE_LOAD_PRE_DECREMENT (@var{mode})
5187 A C expression used to determine whether a load predecrement is a good
5188 thing to use for a given mode. Defaults to the value of
5189 @code{HAVE_PRE_DECREMENT}.
5191 @findex USE_STORE_POST_INCREMENT
5192 @item USE_STORE_POST_INCREMENT (@var{mode})
5193 A C expression used to determine whether a store postincrement is a good
5194 thing to use for a given mode. Defaults to the value of
5195 @code{HAVE_POST_INCREMENT}.
5197 @findex USE_STORE_POST_DECREMENT
5198 @item USE_STORE_POST_DECREMENT (@var{mode})
5199 A C expression used to determine whether a store postdeccrement is a good
5200 thing to use for a given mode. Defaults to the value of
5201 @code{HAVE_POST_DECREMENT}.
5203 @findex USE_STORE_PRE_INCREMENT
5204 @item USE_STORE_PRE_INCREMENT (@var{mode})
5205 This macro is used to determine whether a store preincrement is a good
5206 thing to use for a given mode. Defaults to the value of
5207 @code{HAVE_PRE_INCREMENT}.
5209 @findex USE_STORE_PRE_DECREMENT
5210 @item USE_STORE_PRE_DECREMENT (@var{mode})
5211 This macro is used to determine whether a store predecrement is a good
5212 thing to use for a given mode. Defaults to the value of
5213 @code{HAVE_PRE_DECREMENT}.
5215 @findex NO_FUNCTION_CSE
5216 @item NO_FUNCTION_CSE
5217 Define this macro if it is as good or better to call a constant
5218 function address than to call an address kept in a register.
5220 @findex NO_RECURSIVE_FUNCTION_CSE
5221 @item NO_RECURSIVE_FUNCTION_CSE
5222 Define this macro if it is as good or better for a function to call
5223 itself with an explicit address than to call an address kept in a
5227 @item ADJUST_COST (@var{insn}, @var{link}, @var{dep_insn}, @var{cost})
5228 A C statement (sans semicolon) to update the integer variable @var{cost}
5229 based on the relationship between @var{insn} that is dependent on
5230 @var{dep_insn} through the dependence @var{link}. The default is to
5231 make no adjustment to @var{cost}. This can be used for example to
5232 specify to the scheduler that an output- or anti-dependence does not
5233 incur the same cost as a data-dependence.
5235 @findex ADJUST_PRIORITY
5236 @item ADJUST_PRIORITY (@var{insn})
5237 A C statement (sans semicolon) to update the integer scheduling
5238 priority @code{INSN_PRIORITY(@var{insn})}. Reduce the priority
5239 to execute the @var{insn} earlier, increase the priority to execute
5240 @var{insn} later. Do not define this macro if you do not need to
5241 adjust the scheduling priorities of insns.
5245 @section Dividing the Output into Sections (Texts, Data, @dots{})
5246 @c the above section title is WAY too long. maybe cut the part between
5247 @c the (...)? --mew 10feb93
5249 An object file is divided into sections containing different types of
5250 data. In the most common case, there are three sections: the @dfn{text
5251 section}, which holds instructions and read-only data; the @dfn{data
5252 section}, which holds initialized writable data; and the @dfn{bss
5253 section}, which holds uninitialized data. Some systems have other kinds
5256 The compiler must tell the assembler when to switch sections. These
5257 macros control what commands to output to tell the assembler this. You
5258 can also define additional sections.
5261 @findex TEXT_SECTION_ASM_OP
5262 @item TEXT_SECTION_ASM_OP
5263 A C expression whose value is a string, including spacing, containing the
5264 assembler operation that should precede instructions and read-only data.
5265 Normally @code{"\t.text"} is right.
5267 @findex DATA_SECTION_ASM_OP
5268 @item DATA_SECTION_ASM_OP
5269 A C expression whose value is a string, including spacing, containing the
5270 assembler operation to identify the following data as writable initialized
5271 data. Normally @code{"\t.data"} is right.
5273 @findex SHARED_SECTION_ASM_OP
5274 @item SHARED_SECTION_ASM_OP
5275 If defined, a C expression whose value is a string, including spacing,
5276 containing the assembler operation to identify the following data as
5277 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5279 @findex BSS_SECTION_ASM_OP
5280 @item BSS_SECTION_ASM_OP
5281 If defined, a C expression whose value is a string, including spacing,
5282 containing the assembler operation to identify the following data as
5283 uninitialized global data. If not defined, and neither
5284 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5285 uninitialized global data will be output in the data section if
5286 @samp{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5289 @findex SHARED_BSS_SECTION_ASM_OP
5290 @item SHARED_BSS_SECTION_ASM_OP
5291 If defined, a C expression whose value is a string, including spacing,
5292 containing the assembler operation to identify the following data as
5293 uninitialized global shared data. If not defined, and
5294 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5296 @findex INIT_SECTION_ASM_OP
5297 @item INIT_SECTION_ASM_OP
5298 If defined, a C expression whose value is a string, including spacing,
5299 containing the assembler operation to identify the following data as
5300 initialization code. If not defined, GCC will assume such a section does
5303 @findex FINI_SECTION_ASM_OP
5304 @item FINI_SECTION_ASM_OP
5305 If defined, a C expression whose value is a string, including spacing,
5306 containing the assembler operation to identify the following data as
5307 finalization code. If not defined, GCC will assume such a section does
5310 @findex CRT_CALL_STATIC_FUNCTION
5311 @item CRT_CALL_STATIC_FUNCTION
5312 If defined, a C statement that calls the function named as the sole
5313 argument of this macro. This is used in @file{crtstuff.c} if
5314 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls to
5315 initialization and finalization functions from the init and fini
5316 sections. By default, this macro is a simple function call. Some
5317 ports need hand-crafted assembly code to avoid dependencies on
5318 registers initialized in the function prologue or to ensure that
5319 constant pools don't end up too far way in the text section.
5321 @findex EXTRA_SECTIONS
5324 @item EXTRA_SECTIONS
5325 A list of names for sections other than the standard two, which are
5326 @code{in_text} and @code{in_data}. You need not define this macro
5327 on a system with no other sections (that GCC needs to use).
5329 @findex EXTRA_SECTION_FUNCTIONS
5330 @findex text_section
5331 @findex data_section
5332 @item EXTRA_SECTION_FUNCTIONS
5333 One or more functions to be defined in @file{varasm.c}. These
5334 functions should do jobs analogous to those of @code{text_section} and
5335 @code{data_section}, for your additional sections. Do not define this
5336 macro if you do not define @code{EXTRA_SECTIONS}.
5338 @findex READONLY_DATA_SECTION
5339 @item READONLY_DATA_SECTION
5340 On most machines, read-only variables, constants, and jump tables are
5341 placed in the text section. If this is not the case on your machine,
5342 this macro should be defined to be the name of a function (either
5343 @code{data_section} or a function defined in @code{EXTRA_SECTIONS}) that
5344 switches to the section to be used for read-only items.
5346 If these items should be placed in the text section, this macro should
5349 @findex SELECT_SECTION
5350 @item SELECT_SECTION (@var{exp}, @var{reloc})
5351 A C statement or statements to switch to the appropriate section for
5352 output of @var{exp}. You can assume that @var{exp} is either a
5353 @code{VAR_DECL} node or a constant of some sort. @var{reloc}
5354 indicates whether the initial value of @var{exp} requires link-time
5355 relocations. Select the section by calling @code{text_section} or one
5356 of the alternatives for other sections.
5358 Do not define this macro if you put all read-only variables and
5359 constants in the read-only data section (usually the text section).
5361 @findex SELECT_RTX_SECTION
5362 @item SELECT_RTX_SECTION (@var{mode}, @var{rtx})
5363 A C statement or statements to switch to the appropriate section for
5364 output of @var{rtx} in mode @var{mode}. You can assume that @var{rtx}
5365 is some kind of constant in RTL. The argument @var{mode} is redundant
5366 except in the case of a @code{const_int} rtx. Select the section by
5367 calling @code{text_section} or one of the alternatives for other
5370 Do not define this macro if you put all constants in the read-only
5373 @findex JUMP_TABLES_IN_TEXT_SECTION
5374 @item JUMP_TABLES_IN_TEXT_SECTION
5375 Define this macro to be an expression with a non-zero value if jump
5376 tables (for @code{tablejump} insns) should be output in the text
5377 section, along with the assembler instructions. Otherwise, the
5378 readonly data section is used.
5380 This macro is irrelevant if there is no separate readonly data section.
5382 @findex ENCODE_SECTION_INFO
5383 @item ENCODE_SECTION_INFO (@var{decl})
5384 Define this macro if references to a symbol must be treated differently
5385 depending on something about the variable or function named by the
5386 symbol (such as what section it is in).
5388 The macro definition, if any, is executed immediately after the rtl for
5389 @var{decl} has been created and stored in @code{DECL_RTL (@var{decl})}.
5390 The value of the rtl will be a @code{mem} whose address is a
5393 @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO}
5394 The usual thing for this macro to do is to record a flag in the
5395 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
5396 modified name string in the @code{symbol_ref} (if one bit is not enough
5399 @findex STRIP_NAME_ENCODING
5400 @item STRIP_NAME_ENCODING (@var{var}, @var{sym_name})
5401 Decode @var{sym_name} and store the real name part in @var{var}, sans
5402 the characters that encode section info. Define this macro if
5403 @code{ENCODE_SECTION_INFO} alters the symbol's name string.
5405 @findex UNIQUE_SECTION_P
5406 @item UNIQUE_SECTION_P (@var{decl})
5407 A C expression which evaluates to true if @var{decl} should be placed
5408 into a unique section for some target-specific reason. If you do not
5409 define this macro, the default is @samp{0}. Note that the flag
5410 @samp{-ffunction-sections} will also cause functions to be placed into
5413 @findex UNIQUE_SECTION
5414 @item UNIQUE_SECTION (@var{decl}, @var{reloc})
5415 A C statement to build up a unique section name, expressed as a
5416 STRING_CST node, and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5417 @var{reloc} indicates whether the initial value of @var{exp} requires
5418 link-time relocations. If you do not define this macro, GCC will use
5419 the symbol name prefixed by @samp{.} as the section name. Note - this
5420 macro can now be called for unitialised data items as well as
5421 initialised data and functions.
5425 @section Position Independent Code
5426 @cindex position independent code
5429 This section describes macros that help implement generation of position
5430 independent code. Simply defining these macros is not enough to
5431 generate valid PIC; you must also add support to the macros
5432 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5433 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5434 @samp{movsi} to do something appropriate when the source operand
5435 contains a symbolic address. You may also need to alter the handling of
5436 switch statements so that they use relative addresses.
5437 @c i rearranged the order of the macros above to try to force one of
5438 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5441 @findex PIC_OFFSET_TABLE_REGNUM
5442 @item PIC_OFFSET_TABLE_REGNUM
5443 The register number of the register used to address a table of static
5444 data addresses in memory. In some cases this register is defined by a
5445 processor's ``application binary interface'' (ABI). When this macro
5446 is defined, RTL is generated for this register once, as with the stack
5447 pointer and frame pointer registers. If this macro is not defined, it
5448 is up to the machine-dependent files to allocate such a register (if
5451 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5452 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5453 Define this macro if the register defined by
5454 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
5455 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5457 @findex FINALIZE_PIC
5459 By generating position-independent code, when two different programs (A
5460 and B) share a common library (libC.a), the text of the library can be
5461 shared whether or not the library is linked at the same address for both
5462 programs. In some of these environments, position-independent code
5463 requires not only the use of different addressing modes, but also
5464 special code to enable the use of these addressing modes.
5466 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
5467 codes once the function is being compiled into assembly code, but not
5468 before. (It is not done before, because in the case of compiling an
5469 inline function, it would lead to multiple PIC prologues being
5470 included in functions which used inline functions and were compiled to
5473 @findex LEGITIMATE_PIC_OPERAND_P
5474 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
5475 A C expression that is nonzero if @var{x} is a legitimate immediate
5476 operand on the target machine when generating position independent code.
5477 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5478 check this. You can also assume @var{flag_pic} is true, so you need not
5479 check it either. You need not define this macro if all constants
5480 (including @code{SYMBOL_REF}) can be immediate operands when generating
5481 position independent code.
5484 @node Assembler Format
5485 @section Defining the Output Assembler Language
5487 This section describes macros whose principal purpose is to describe how
5488 to write instructions in assembler language--rather than what the
5492 * File Framework:: Structural information for the assembler file.
5493 * Data Output:: Output of constants (numbers, strings, addresses).
5494 * Uninitialized Data:: Output of uninitialized variables.
5495 * Label Output:: Output and generation of labels.
5496 * Initialization:: General principles of initialization
5497 and termination routines.
5498 * Macros for Initialization::
5499 Specific macros that control the handling of
5500 initialization and termination routines.
5501 * Instruction Output:: Output of actual instructions.
5502 * Dispatch Tables:: Output of jump tables.
5503 * Exception Region Output:: Output of exception region code.
5504 * Alignment Output:: Pseudo ops for alignment and skipping data.
5507 @node File Framework
5508 @subsection The Overall Framework of an Assembler File
5509 @cindex assembler format
5510 @cindex output of assembler code
5512 @c prevent bad page break with this line
5513 This describes the overall framework of an assembler file.
5516 @findex ASM_FILE_START
5517 @item ASM_FILE_START (@var{stream})
5518 A C expression which outputs to the stdio stream @var{stream}
5519 some appropriate text to go at the start of an assembler file.
5521 Normally this macro is defined to output a line containing
5522 @samp{#NO_APP}, which is a comment that has no effect on most
5523 assemblers but tells the GNU assembler that it can save time by not
5524 checking for certain assembler constructs.
5526 On systems that use SDB, it is necessary to output certain commands;
5527 see @file{attasm.h}.
5529 @findex ASM_FILE_END
5530 @item ASM_FILE_END (@var{stream})
5531 A C expression which outputs to the stdio stream @var{stream}
5532 some appropriate text to go at the end of an assembler file.
5534 If this macro is not defined, the default is to output nothing
5535 special at the end of the file. Most systems don't require any
5538 On systems that use SDB, it is necessary to output certain commands;
5539 see @file{attasm.h}.
5541 @findex ASM_IDENTIFY_GCC
5542 @item ASM_IDENTIFY_GCC (@var{file})
5543 A C statement to output assembler commands which will identify
5544 the object file as having been compiled with GCC (or another
5547 If you don't define this macro, the string @samp{gcc_compiled.:}
5548 is output. This string is calculated to define a symbol which,
5549 on BSD systems, will never be defined for any other reason.
5550 GDB checks for the presence of this symbol when reading the
5551 symbol table of an executable.
5553 On non-BSD systems, you must arrange communication with GDB in
5554 some other fashion. If GDB is not used on your system, you can
5555 define this macro with an empty body.
5557 @findex ASM_COMMENT_START
5558 @item ASM_COMMENT_START
5559 A C string constant describing how to begin a comment in the target
5560 assembler language. The compiler assumes that the comment will end at
5561 the end of the line.
5565 A C string constant for text to be output before each @code{asm}
5566 statement or group of consecutive ones. Normally this is
5567 @code{"#APP"}, which is a comment that has no effect on most
5568 assemblers but tells the GNU assembler that it must check the lines
5569 that follow for all valid assembler constructs.
5573 A C string constant for text to be output after each @code{asm}
5574 statement or group of consecutive ones. Normally this is
5575 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5576 time-saving assumptions that are valid for ordinary compiler output.
5578 @findex ASM_OUTPUT_SOURCE_FILENAME
5579 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5580 A C statement to output COFF information or DWARF debugging information
5581 which indicates that filename @var{name} is the current source file to
5582 the stdio stream @var{stream}.
5584 This macro need not be defined if the standard form of output
5585 for the file format in use is appropriate.
5587 @findex OUTPUT_QUOTED_STRING
5588 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5589 A C statement to output the string @var{string} to the stdio stream
5590 @var{stream}. If you do not call the function @code{output_quoted_string}
5591 in your config files, GCC will only call it to output filenames to
5592 the assembler source. So you can use it to canonicalize the format
5593 of the filename using this macro.
5595 @findex ASM_OUTPUT_SOURCE_LINE
5596 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
5597 A C statement to output DBX or SDB debugging information before code
5598 for line number @var{line} of the current source file to the
5599 stdio stream @var{stream}.
5601 This macro need not be defined if the standard form of debugging
5602 information for the debugger in use is appropriate.
5604 @findex ASM_OUTPUT_IDENT
5605 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
5606 A C statement to output something to the assembler file to handle a
5607 @samp{#ident} directive containing the text @var{string}. If this
5608 macro is not defined, nothing is output for a @samp{#ident} directive.
5610 @findex ASM_OUTPUT_SECTION_NAME
5611 @item ASM_OUTPUT_SECTION_NAME (@var{stream}, @var{decl}, @var{name}, @var{reloc})
5612 A C statement to output something to the assembler file to switch to section
5613 @var{name} for object @var{decl} which is either a @code{FUNCTION_DECL}, a
5614 @code{VAR_DECL} or @code{NULL_TREE}. @var{reloc}
5615 indicates whether the initial value of @var{exp} requires link-time
5616 relocations. The string given by @var{name} will always be the
5617 canonical version stored in the global stringpool.
5619 Some target formats do not support arbitrary sections. Do not define
5620 this macro in such cases.
5622 At present this macro is only used to support section attributes.
5623 When this macro is undefined, section attributes are disabled.
5625 @findex OBJC_PROLOGUE
5627 A C statement to output any assembler statements which are required to
5628 precede any Objective C object definitions or message sending. The
5629 statement is executed only when compiling an Objective C program.
5634 @subsection Output of Data
5636 @c prevent bad page break with this line
5637 This describes data output.
5640 @findex ASM_OUTPUT_LONG_DOUBLE
5641 @findex ASM_OUTPUT_DOUBLE
5642 @findex ASM_OUTPUT_FLOAT
5643 @item ASM_OUTPUT_LONG_DOUBLE (@var{stream}, @var{value})
5644 @itemx ASM_OUTPUT_DOUBLE (@var{stream}, @var{value})
5645 @itemx ASM_OUTPUT_FLOAT (@var{stream}, @var{value})
5646 @itemx ASM_OUTPUT_THREE_QUARTER_FLOAT (@var{stream}, @var{value})
5647 @itemx ASM_OUTPUT_SHORT_FLOAT (@var{stream}, @var{value})
5648 @itemx ASM_OUTPUT_BYTE_FLOAT (@var{stream}, @var{value})
5649 A C statement to output to the stdio stream @var{stream} an assembler
5650 instruction to assemble a floating-point constant of @code{TFmode},
5651 @code{DFmode}, @code{SFmode}, @code{TQFmode}, @code{HFmode}, or
5652 @code{QFmode}, respectively, whose value is @var{value}. @var{value}
5653 will be a C expression of type @code{REAL_VALUE_TYPE}. Macros such as
5654 @code{REAL_VALUE_TO_TARGET_DOUBLE} are useful for writing these
5657 @findex ASM_OUTPUT_QUADRUPLE_INT
5658 @findex ASM_OUTPUT_DOUBLE_INT
5659 @findex ASM_OUTPUT_INT
5660 @findex ASM_OUTPUT_SHORT
5661 @findex ASM_OUTPUT_CHAR
5662 @findex output_addr_const
5663 @item ASM_OUTPUT_QUADRUPLE_INT (@var{stream}, @var{exp})
5664 @itemx ASM_OUTPUT_DOUBLE_INT (@var{stream}, @var{exp})
5665 @itemx ASM_OUTPUT_INT (@var{stream}, @var{exp})
5666 @itemx ASM_OUTPUT_SHORT (@var{stream}, @var{exp})
5667 @itemx ASM_OUTPUT_CHAR (@var{stream}, @var{exp})
5668 A C statement to output to the stdio stream @var{stream} an assembler
5669 instruction to assemble an integer of 16, 8, 4, 2 or 1 bytes,
5670 respectively, whose value is @var{value}. The argument @var{exp} will
5671 be an RTL expression which represents a constant value. Use
5672 @samp{output_addr_const (@var{stream}, @var{exp})} to output this value
5673 as an assembler expression.@refill
5675 For sizes larger than @code{UNITS_PER_WORD}, if the action of a macro
5676 would be identical to repeatedly calling the macro corresponding to
5677 a size of @code{UNITS_PER_WORD}, once for each word, you need not define
5680 @findex OUTPUT_ADDR_CONST_EXTRA
5681 @item OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
5682 A C statement to recognize @var{rtx} patterns that
5683 @code{output_addr_const} can't deal with, and output assembly code to
5684 @var{stream} corresponding to the pattern @var{x}. This may be used to
5685 allow machine-dependent @code{UNSPEC}s to appear within constants.
5687 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
5688 @code{goto fail}, so that a standard error message is printed. If it
5689 prints an error message itself, by calling, for example,
5690 @code{output_operand_lossage}, it may just complete normally.
5692 @findex ASM_OUTPUT_BYTE
5693 @item ASM_OUTPUT_BYTE (@var{stream}, @var{value})
5694 A C statement to output to the stdio stream @var{stream} an assembler
5695 instruction to assemble a single byte containing the number @var{value}.
5699 A C string constant, including spacing, giving the pseudo-op to use for a
5700 sequence of single-byte constants. If this macro is not defined, the
5701 default is @code{"\t.byte\t"}.
5703 @findex UNALIGNED_SHORT_ASM_OP
5704 @findex UNALIGNED_INT_ASM_OP
5705 @findex UNALIGNED_DOUBLE_INT_ASM_OP
5706 @item UNALIGNED_SHORT_ASM_OP
5707 @itemx UNALIGNED_INT_ASM_OP
5708 @itemx UNALIGNED_DOUBLE_INT_ASM_OP
5709 A C string constant, including spacing, giving the pseudo-op to use
5710 to assemble 16, 32, and 64 bit integers respectively @emph{without}
5711 adding implicit padding or alignment. These macros are required if
5712 DWARF 2 frame unwind is used. On ELF systems, these will default
5713 to @code{.2byte}, @code{.4byte}, and @code{.8byte}.@refill
5715 @findex ASM_OUTPUT_ASCII
5716 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5717 A C statement to output to the stdio stream @var{stream} an assembler
5718 instruction to assemble a string constant containing the @var{len}
5719 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5720 @code{char *} and @var{len} a C expression of type @code{int}.
5722 If the assembler has a @code{.ascii} pseudo-op as found in the
5723 Berkeley Unix assembler, do not define the macro
5724 @code{ASM_OUTPUT_ASCII}.
5726 @findex CONSTANT_POOL_BEFORE_FUNCTION
5727 @item CONSTANT_POOL_BEFORE_FUNCTION
5728 You may define this macro as a C expression. You should define the
5729 expression to have a non-zero value if GCC should output the constant
5730 pool for a function before the code for the function, or a zero value if
5731 GCC should output the constant pool after the function. If you do
5732 not define this macro, the usual case, GCC will output the constant
5733 pool before the function.
5735 @findex ASM_OUTPUT_POOL_PROLOGUE
5736 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
5737 A C statement to output assembler commands to define the start of the
5738 constant pool for a function. @var{funname} is a string giving
5739 the name of the function. Should the return type of the function
5740 be required, it can be obtained via @var{fundecl}. @var{size}
5741 is the size, in bytes, of the constant pool that will be written
5742 immediately after this call.
5744 If no constant-pool prefix is required, the usual case, this macro need
5747 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
5748 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5749 A C statement (with or without semicolon) to output a constant in the
5750 constant pool, if it needs special treatment. (This macro need not do
5751 anything for RTL expressions that can be output normally.)
5753 The argument @var{file} is the standard I/O stream to output the
5754 assembler code on. @var{x} is the RTL expression for the constant to
5755 output, and @var{mode} is the machine mode (in case @var{x} is a
5756 @samp{const_int}). @var{align} is the required alignment for the value
5757 @var{x}; you should output an assembler directive to force this much
5760 The argument @var{labelno} is a number to use in an internal label for
5761 the address of this pool entry. The definition of this macro is
5762 responsible for outputting the label definition at the proper place.
5763 Here is how to do this:
5766 ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
5769 When you output a pool entry specially, you should end with a
5770 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5771 entry from being output a second time in the usual manner.
5773 You need not define this macro if it would do nothing.
5775 @findex CONSTANT_AFTER_FUNCTION_P
5776 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
5777 Define this macro as a C expression which is nonzero if the constant
5778 @var{exp}, of type @code{tree}, should be output after the code for a
5779 function. The compiler will normally output all constants before the
5780 function; you need not define this macro if this is OK.
5782 @findex ASM_OUTPUT_POOL_EPILOGUE
5783 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5784 A C statement to output assembler commands to at the end of the constant
5785 pool for a function. @var{funname} is a string giving the name of the
5786 function. Should the return type of the function be required, you can
5787 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5788 constant pool that GCC wrote immediately before this call.
5790 If no constant-pool epilogue is required, the usual case, you need not
5793 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
5794 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
5795 Define this macro as a C expression which is nonzero if @var{C} is
5796 used as a logical line separator by the assembler.
5798 If you do not define this macro, the default is that only
5799 the character @samp{;} is treated as a logical line separator.
5802 @findex ASM_OPEN_PAREN
5803 @findex ASM_CLOSE_PAREN
5804 @item ASM_OPEN_PAREN
5805 @itemx ASM_CLOSE_PAREN
5806 These macros are defined as C string constants, describing the syntax
5807 in the assembler for grouping arithmetic expressions. The following
5808 definitions are correct for most assemblers:
5811 #define ASM_OPEN_PAREN "("
5812 #define ASM_CLOSE_PAREN ")"
5816 These macros are provided by @file{real.h} for writing the definitions
5817 of @code{ASM_OUTPUT_DOUBLE} and the like:
5820 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5821 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5822 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5823 @findex REAL_VALUE_TO_TARGET_SINGLE
5824 @findex REAL_VALUE_TO_TARGET_DOUBLE
5825 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
5826 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
5827 floating point representation, and store its bit pattern in the array of
5828 @code{long int} whose address is @var{l}. The number of elements in the
5829 output array is determined by the size of the desired target floating
5830 point data type: 32 bits of it go in each @code{long int} array
5831 element. Each array element holds 32 bits of the result, even if
5832 @code{long int} is wider than 32 bits on the host machine.
5834 The array element values are designed so that you can print them out
5835 using @code{fprintf} in the order they should appear in the target
5838 @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
5839 @findex REAL_VALUE_TO_DECIMAL
5840 This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
5841 decimal number and stores it as a string into @var{string}.
5842 You must pass, as @var{string}, the address of a long enough block
5843 of space to hold the result.
5845 The argument @var{format} is a @code{printf}-specification that serves
5846 as a suggestion for how to format the output string.
5849 @node Uninitialized Data
5850 @subsection Output of Uninitialized Variables
5852 Each of the macros in this section is used to do the whole job of
5853 outputting a single uninitialized variable.
5856 @findex ASM_OUTPUT_COMMON
5857 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5858 A C statement (sans semicolon) to output to the stdio stream
5859 @var{stream} the assembler definition of a common-label named
5860 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5861 is the size rounded up to whatever alignment the caller wants.
5863 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5864 output the name itself; before and after that, output the additional
5865 assembler syntax for defining the name, and a newline.
5867 This macro controls how the assembler definitions of uninitialized
5868 common global variables are output.
5870 @findex ASM_OUTPUT_ALIGNED_COMMON
5871 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5872 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5873 separate, explicit argument. If you define this macro, it is used in
5874 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5875 handling the required alignment of the variable. The alignment is specified
5876 as the number of bits.
5878 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
5879 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5880 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5881 variable to be output, if there is one, or @code{NULL_TREE} if there
5882 is no corresponding variable. If you define this macro, GCC will use it
5883 in place of both @code{ASM_OUTPUT_COMMON} and
5884 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5885 the variable's decl in order to chose what to output.
5887 @findex ASM_OUTPUT_SHARED_COMMON
5888 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5889 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
5890 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
5893 @findex ASM_OUTPUT_BSS
5894 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5895 A C statement (sans semicolon) to output to the stdio stream
5896 @var{stream} the assembler definition of uninitialized global @var{decl} named
5897 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5898 is the size rounded up to whatever alignment the caller wants.
5900 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
5901 defining this macro. If unable, use the expression
5902 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5903 before and after that, output the additional assembler syntax for defining
5904 the name, and a newline.
5906 This macro controls how the assembler definitions of uninitialized global
5907 variables are output. This macro exists to properly support languages like
5908 @code{c++} which do not have @code{common} data. However, this macro currently
5909 is not defined for all targets. If this macro and
5910 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
5911 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
5912 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
5914 @findex ASM_OUTPUT_ALIGNED_BSS
5915 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5916 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
5917 separate, explicit argument. If you define this macro, it is used in
5918 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
5919 handling the required alignment of the variable. The alignment is specified
5920 as the number of bits.
5922 Try to use function @code{asm_output_aligned_bss} defined in file
5923 @file{varasm.c} when defining this macro.
5925 @findex ASM_OUTPUT_SHARED_BSS
5926 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5927 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
5928 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
5931 @findex ASM_OUTPUT_LOCAL
5932 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5933 A C statement (sans semicolon) to output to the stdio stream
5934 @var{stream} the assembler definition of a local-common-label named
5935 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5936 is the size rounded up to whatever alignment the caller wants.
5938 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5939 output the name itself; before and after that, output the additional
5940 assembler syntax for defining the name, and a newline.
5942 This macro controls how the assembler definitions of uninitialized
5943 static variables are output.
5945 @findex ASM_OUTPUT_ALIGNED_LOCAL
5946 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5947 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5948 separate, explicit argument. If you define this macro, it is used in
5949 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5950 handling the required alignment of the variable. The alignment is specified
5951 as the number of bits.
5953 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
5954 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5955 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5956 variable to be output, if there is one, or @code{NULL_TREE} if there
5957 is no corresponding variable. If you define this macro, GCC will use it
5958 in place of both @code{ASM_OUTPUT_DECL} and
5959 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5960 the variable's decl in order to chose what to output.
5962 @findex ASM_OUTPUT_SHARED_LOCAL
5963 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5964 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
5965 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
5970 @subsection Output and Generation of Labels
5972 @c prevent bad page break with this line
5973 This is about outputting labels.
5976 @findex ASM_OUTPUT_LABEL
5977 @findex assemble_name
5978 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5979 A C statement (sans semicolon) to output to the stdio stream
5980 @var{stream} the assembler definition of a label named @var{name}.
5981 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5982 output the name itself; before and after that, output the additional
5983 assembler syntax for defining the name, and a newline.
5985 @findex ASM_DECLARE_FUNCTION_NAME
5986 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5987 A C statement (sans semicolon) to output to the stdio stream
5988 @var{stream} any text necessary for declaring the name @var{name} of a
5989 function which is being defined. This macro is responsible for
5990 outputting the label definition (perhaps using
5991 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
5992 @code{FUNCTION_DECL} tree node representing the function.
5994 If this macro is not defined, then the function name is defined in the
5995 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5997 @findex ASM_DECLARE_FUNCTION_SIZE
5998 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5999 A C statement (sans semicolon) to output to the stdio stream
6000 @var{stream} any text necessary for declaring the size of a function
6001 which is being defined. The argument @var{name} is the name of the
6002 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6003 representing the function.
6005 If this macro is not defined, then the function size is not defined.
6007 @findex ASM_DECLARE_OBJECT_NAME
6008 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6009 A C statement (sans semicolon) to output to the stdio stream
6010 @var{stream} any text necessary for declaring the name @var{name} of an
6011 initialized variable which is being defined. This macro must output the
6012 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6013 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6015 If this macro is not defined, then the variable name is defined in the
6016 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6018 @findex ASM_DECLARE_REGISTER_GLOBAL
6019 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6020 A C statement (sans semicolon) to output to the stdio stream
6021 @var{stream} any text necessary for claiming a register @var{regno}
6022 for a global variable @var{decl} with name @var{name}.
6024 If you don't define this macro, that is equivalent to defining it to do
6027 @findex ASM_FINISH_DECLARE_OBJECT
6028 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6029 A C statement (sans semicolon) to finish up declaring a variable name
6030 once the compiler has processed its initializer fully and thus has had a
6031 chance to determine the size of an array when controlled by an
6032 initializer. This is used on systems where it's necessary to declare
6033 something about the size of the object.
6035 If you don't define this macro, that is equivalent to defining it to do
6038 @findex ASM_GLOBALIZE_LABEL
6039 @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
6040 A C statement (sans semicolon) to output to the stdio stream
6041 @var{stream} some commands that will make the label @var{name} global;
6042 that is, available for reference from other files. Use the expression
6043 @code{assemble_name (@var{stream}, @var{name})} to output the name
6044 itself; before and after that, output the additional assembler syntax
6045 for making that name global, and a newline.
6047 @findex ASM_WEAKEN_LABEL
6048 @item ASM_WEAKEN_LABEL
6049 A C statement (sans semicolon) to output to the stdio stream
6050 @var{stream} some commands that will make the label @var{name} weak;
6051 that is, available for reference from other files but only used if
6052 no other definition is available. Use the expression
6053 @code{assemble_name (@var{stream}, @var{name})} to output the name
6054 itself; before and after that, output the additional assembler syntax
6055 for making that name weak, and a newline.
6057 If you don't define this macro, GCC will not support weak
6058 symbols and you should not define the @code{SUPPORTS_WEAK} macro.
6060 @findex SUPPORTS_WEAK
6062 A C expression which evaluates to true if the target supports weak symbols.
6064 If you don't define this macro, @file{defaults.h} provides a default
6065 definition. If @code{ASM_WEAKEN_LABEL} is defined, the default
6066 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6067 you want to control weak symbol support with a compiler flag such as
6070 @findex MAKE_DECL_ONE_ONLY (@var{decl})
6071 @item MAKE_DECL_ONE_ONLY
6072 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6073 public symbol such that extra copies in multiple translation units will
6074 be discarded by the linker. Define this macro if your object file
6075 format provides support for this concept, such as the @samp{COMDAT}
6076 section flags in the Microsoft Windows PE/COFF format, and this support
6077 requires changes to @var{decl}, such as putting it in a separate section.
6079 @findex SUPPORTS_ONE_ONLY
6080 @item SUPPORTS_ONE_ONLY
6081 A C expression which evaluates to true if the target supports one-only
6084 If you don't define this macro, @file{varasm.c} provides a default
6085 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6086 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6087 you want to control one-only symbol support with a compiler flag, or if
6088 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6089 be emitted as one-only.
6091 @findex ASM_OUTPUT_EXTERNAL
6092 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6093 A C statement (sans semicolon) to output to the stdio stream
6094 @var{stream} any text necessary for declaring the name of an external
6095 symbol named @var{name} which is referenced in this compilation but
6096 not defined. The value of @var{decl} is the tree node for the
6099 This macro need not be defined if it does not need to output anything.
6100 The GNU assembler and most Unix assemblers don't require anything.
6102 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
6103 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
6104 A C statement (sans semicolon) to output on @var{stream} an assembler
6105 pseudo-op to declare a library function name external. The name of the
6106 library function is given by @var{symref}, which has type @code{rtx} and
6107 is a @code{symbol_ref}.
6109 This macro need not be defined if it does not need to output anything.
6110 The GNU assembler and most Unix assemblers don't require anything.
6112 @findex ASM_OUTPUT_LABELREF
6113 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6114 A C statement (sans semicolon) to output to the stdio stream
6115 @var{stream} a reference in assembler syntax to a label named
6116 @var{name}. This should add @samp{_} to the front of the name, if that
6117 is customary on your operating system, as it is in most Berkeley Unix
6118 systems. This macro is used in @code{assemble_name}.
6120 @ignore @c Seems not to exist anymore.
6121 @findex ASM_OUTPUT_LABELREF_AS_INT
6122 @item ASM_OUTPUT_LABELREF_AS_INT (@var{file}, @var{label})
6123 Define this macro for systems that use the program @code{collect2}.
6124 The definition should be a C statement to output a word containing
6125 a reference to the label @var{label}.
6128 @findex ASM_OUTPUT_SYMBOL_REF
6129 @item ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6130 A C statement (sans semicolon) to output a reference to
6131 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_output}
6132 will be used to output the name of the symbol. This macro may be used
6133 to modify the way a symbol is referenced depending on information
6134 encoded by @code{ENCODE_SECTION_INFO}.
6136 @findex ASM_OUTPUT_INTERNAL_LABEL
6137 @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
6138 A C statement to output to the stdio stream @var{stream} a label whose
6139 name is made from the string @var{prefix} and the number @var{num}.
6141 It is absolutely essential that these labels be distinct from the labels
6142 used for user-level functions and variables. Otherwise, certain programs
6143 will have name conflicts with internal labels.
6145 It is desirable to exclude internal labels from the symbol table of the
6146 object file. Most assemblers have a naming convention for labels that
6147 should be excluded; on many systems, the letter @samp{L} at the
6148 beginning of a label has this effect. You should find out what
6149 convention your system uses, and follow it.
6151 The usual definition of this macro is as follows:
6154 fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
6157 @findex ASM_OUTPUT_DEBUG_LABEL
6158 @item ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6159 A C statement to output to the stdio stream @var{stream} a debug info
6160 label whose name is made from the string @var{prefix} and the number
6161 @var{num}. This is useful for VLIW targets, where debug info labels
6162 may need to be treated differently than branch target labels. On some
6163 systems, branch target labels must be at the beginning of instruction
6164 bundles, but debug info labels can occur in the middle of instruction
6167 If this macro is not defined, then @code{ASM_OUTPUT_INTERNAL_LABEL} will be
6170 @findex ASM_OUTPUT_ALTERNATE_LABEL_NAME
6171 @item ASM_OUTPUT_ALTERNATE_LABEL_NAME (@var{stream}, @var{string})
6172 A C statement to output to the stdio stream @var{stream} the string
6175 The default definition of this macro is as follows:
6178 fprintf (@var{stream}, "%s:\n", LABEL_ALTERNATE_NAME (INSN))
6181 @findex ASM_GENERATE_INTERNAL_LABEL
6182 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6183 A C statement to store into the string @var{string} a label whose name
6184 is made from the string @var{prefix} and the number @var{num}.
6186 This string, when output subsequently by @code{assemble_name}, should
6187 produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
6188 with the same @var{prefix} and @var{num}.
6190 If the string begins with @samp{*}, then @code{assemble_name} will
6191 output the rest of the string unchanged. It is often convenient for
6192 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6193 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6194 to output the string, and may change it. (Of course,
6195 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6196 you should know what it does on your machine.)
6198 @findex ASM_FORMAT_PRIVATE_NAME
6199 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6200 A C expression to assign to @var{outvar} (which is a variable of type
6201 @code{char *}) a newly allocated string made from the string
6202 @var{name} and the number @var{number}, with some suitable punctuation
6203 added. Use @code{alloca} to get space for the string.
6205 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6206 produce an assembler label for an internal static variable whose name is
6207 @var{name}. Therefore, the string must be such as to result in valid
6208 assembler code. The argument @var{number} is different each time this
6209 macro is executed; it prevents conflicts between similarly-named
6210 internal static variables in different scopes.
6212 Ideally this string should not be a valid C identifier, to prevent any
6213 conflict with the user's own symbols. Most assemblers allow periods
6214 or percent signs in assembler symbols; putting at least one of these
6215 between the name and the number will suffice.
6217 @findex ASM_OUTPUT_DEF
6218 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6219 A C statement to output to the stdio stream @var{stream} assembler code
6220 which defines (equates) the symbol @var{name} to have the value @var{value}.
6223 If SET_ASM_OP is defined, a default definition is provided which is
6224 correct for most systems.
6226 @findex ASM_OUTPUT_DEF_FROM_DECLS
6227 @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6228 A C statement to output to the stdio stream @var{stream} assembler code
6229 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6230 to have the value of the tree node @var{decl_of_value}. This macro will
6231 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6232 the tree nodes are available.
6234 @findex ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL
6235 @item ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL (@var{stream}, @var{symbol}, @var{high}, @var{low})
6236 A C statement to output to the stdio stream @var{stream} assembler code
6237 which defines (equates) the symbol @var{symbol} to have a value equal to
6238 the difference of the two symbols @var{high} and @var{low}, i.e.
6239 @var{high} minus @var{low}. GCC guarantees that the symbols @var{high}
6240 and @var{low} are already known by the assembler so that the difference
6241 resolves into a constant.
6244 If SET_ASM_OP is defined, a default definition is provided which is
6245 correct for most systems.
6247 @findex ASM_OUTPUT_WEAK_ALIAS
6248 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6249 A C statement to output to the stdio stream @var{stream} assembler code
6250 which defines (equates) the weak symbol @var{name} to have the value
6253 Define this macro if the target only supports weak aliases; define
6254 ASM_OUTPUT_DEF instead if possible.
6256 @findex OBJC_GEN_METHOD_LABEL
6257 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6258 Define this macro to override the default assembler names used for
6259 Objective C methods.
6261 The default name is a unique method number followed by the name of the
6262 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6263 the category is also included in the assembler name (e.g.@:
6266 These names are safe on most systems, but make debugging difficult since
6267 the method's selector is not present in the name. Therefore, particular
6268 systems define other ways of computing names.
6270 @var{buf} is an expression of type @code{char *} which gives you a
6271 buffer in which to store the name; its length is as long as
6272 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6273 50 characters extra.
6275 The argument @var{is_inst} specifies whether the method is an instance
6276 method or a class method; @var{class_name} is the name of the class;
6277 @var{cat_name} is the name of the category (or NULL if the method is not
6278 in a category); and @var{sel_name} is the name of the selector.
6280 On systems where the assembler can handle quoted names, you can use this
6281 macro to provide more human-readable names.
6284 @node Initialization
6285 @subsection How Initialization Functions Are Handled
6286 @cindex initialization routines
6287 @cindex termination routines
6288 @cindex constructors, output of
6289 @cindex destructors, output of
6291 The compiled code for certain languages includes @dfn{constructors}
6292 (also called @dfn{initialization routines})---functions to initialize
6293 data in the program when the program is started. These functions need
6294 to be called before the program is ``started''---that is to say, before
6295 @code{main} is called.
6297 Compiling some languages generates @dfn{destructors} (also called
6298 @dfn{termination routines}) that should be called when the program
6301 To make the initialization and termination functions work, the compiler
6302 must output something in the assembler code to cause those functions to
6303 be called at the appropriate time. When you port the compiler to a new
6304 system, you need to specify how to do this.
6306 There are two major ways that GCC currently supports the execution of
6307 initialization and termination functions. Each way has two variants.
6308 Much of the structure is common to all four variations.
6310 @findex __CTOR_LIST__
6311 @findex __DTOR_LIST__
6312 The linker must build two lists of these functions---a list of
6313 initialization functions, called @code{__CTOR_LIST__}, and a list of
6314 termination functions, called @code{__DTOR_LIST__}.
6316 Each list always begins with an ignored function pointer (which may hold
6317 0, @minus{}1, or a count of the function pointers after it, depending on
6318 the environment). This is followed by a series of zero or more function
6319 pointers to constructors (or destructors), followed by a function
6320 pointer containing zero.
6322 Depending on the operating system and its executable file format, either
6323 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6324 time and exit time. Constructors are called in reverse order of the
6325 list; destructors in forward order.
6327 The best way to handle static constructors works only for object file
6328 formats which provide arbitrarily-named sections. A section is set
6329 aside for a list of constructors, and another for a list of destructors.
6330 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6331 object file that defines an initialization function also puts a word in
6332 the constructor section to point to that function. The linker
6333 accumulates all these words into one contiguous @samp{.ctors} section.
6334 Termination functions are handled similarly.
6336 To use this method, you need appropriate definitions of the macros
6337 @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}. Usually
6338 you can get them by including @file{svr4.h}.
6340 When arbitrary sections are available, there are two variants, depending
6341 upon how the code in @file{crtstuff.c} is called. On systems that
6342 support an @dfn{init} section which is executed at program startup,
6343 parts of @file{crtstuff.c} are compiled into that section. The
6344 program is linked by the @code{gcc} driver like this:
6347 ld -o @var{output_file} crtbegin.o @dots{} crtend.o -lgcc
6350 The head of a function (@code{__do_global_ctors}) appears in the init
6351 section of @file{crtbegin.o}; the remainder of the function appears in
6352 the init section of @file{crtend.o}. The linker will pull these two
6353 parts of the section together, making a whole function. If any of the
6354 user's object files linked into the middle of it contribute code, then that
6355 code will be executed as part of the body of @code{__do_global_ctors}.
6357 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
6360 If no init section is available, do not define
6361 @code{INIT_SECTION_ASM_OP}. Then @code{__do_global_ctors} is built into
6362 the text section like all other functions, and resides in
6363 @file{libgcc.a}. When GCC compiles any function called @code{main}, it
6364 inserts a procedure call to @code{__main} as the first executable code
6365 after the function prologue. The @code{__main} function, also defined
6366 in @file{libgcc2.c}, simply calls @file{__do_global_ctors}.
6368 In file formats that don't support arbitrary sections, there are again
6369 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
6370 and an `a.out' format must be used. In this case,
6371 @code{ASM_OUTPUT_CONSTRUCTOR} is defined to produce a @code{.stabs}
6372 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
6373 and with the address of the void function containing the initialization
6374 code as its value. The GNU linker recognizes this as a request to add
6375 the value to a ``set''; the values are accumulated, and are eventually
6376 placed in the executable as a vector in the format described above, with
6377 a leading (ignored) count and a trailing zero element.
6378 @code{ASM_OUTPUT_DESTRUCTOR} is handled similarly. Since no init
6379 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
6380 the compilation of @code{main} to call @code{__main} as above, starting
6381 the initialization process.
6383 The last variant uses neither arbitrary sections nor the GNU linker.
6384 This is preferable when you want to do dynamic linking and when using
6385 file formats which the GNU linker does not support, such as `ECOFF'. In
6386 this case, @code{ASM_OUTPUT_CONSTRUCTOR} does not produce an
6387 @code{N_SETT} symbol; initialization and termination functions are
6388 recognized simply by their names. This requires an extra program in the
6389 linkage step, called @code{collect2}. This program pretends to be the
6390 linker, for use with GCC; it does its job by running the ordinary
6391 linker, but also arranges to include the vectors of initialization and
6392 termination functions. These functions are called via @code{__main} as
6395 Choosing among these configuration options has been simplified by a set
6396 of operating-system-dependent files in the @file{config} subdirectory.
6397 These files define all of the relevant parameters. Usually it is
6398 sufficient to include one into your specific machine-dependent
6399 configuration file. These files are:
6403 For operating systems using the `a.out' format.
6406 For operating systems using the `MachO' format.
6409 For System V Release 3 and similar systems using `COFF' format.
6412 For System V Release 4 and similar systems using `ELF' format.
6415 For the VMS operating system.
6419 The following section describes the specific macros that control and
6420 customize the handling of initialization and termination functions.
6423 @node Macros for Initialization
6424 @subsection Macros Controlling Initialization Routines
6426 Here are the macros that control how the compiler handles initialization
6427 and termination functions:
6430 @findex INIT_SECTION_ASM_OP
6431 @item INIT_SECTION_ASM_OP
6432 If defined, a C string constant, including spacing, for the assembler
6433 operation to identify the following data as initialization code. If not
6434 defined, GCC will assume such a section does not exist. When you are
6435 using special sections for initialization and termination functions, this
6436 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6437 run the initialization functions.
6439 @item HAS_INIT_SECTION
6440 @findex HAS_INIT_SECTION
6441 If defined, @code{main} will not call @code{__main} as described above.
6442 This macro should be defined for systems that control the contents of the
6443 init section on a symbol-by-symbol basis, such as OSF/1, and should not
6444 be defined explicitly for systems that support
6445 @code{INIT_SECTION_ASM_OP}.
6447 @item LD_INIT_SWITCH
6448 @findex LD_INIT_SWITCH
6449 If defined, a C string constant for a switch that tells the linker that
6450 the following symbol is an initialization routine.
6452 @item LD_FINI_SWITCH
6453 @findex LD_FINI_SWITCH
6454 If defined, a C string constant for a switch that tells the linker that
6455 the following symbol is a finalization routine.
6458 @findex INVOKE__main
6459 If defined, @code{main} will call @code{__main} despite the presence of
6460 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6461 where the init section is not actually run automatically, but is still
6462 useful for collecting the lists of constructors and destructors.
6464 @item SUPPORTS_INIT_PRIORITY
6465 @findex SUPPORTS_INIT_PRIORITY
6466 If nonzero, the C++ @code{init_priority} attribute is supported and the
6467 compiler should emit instructions to control the order of initialization
6468 of objects. If zero, the compiler will issue an error message upon
6469 encountering an @code{init_priority} attribute.
6471 @item ASM_OUTPUT_CONSTRUCTOR (@var{stream}, @var{name})
6472 @findex ASM_OUTPUT_CONSTRUCTOR
6473 Define this macro as a C statement to output on the stream @var{stream}
6474 the assembler code to arrange to call the function named @var{name} at
6475 initialization time.
6477 Assume that @var{name} is the name of a C function generated
6478 automatically by the compiler. This function takes no arguments. Use
6479 the function @code{assemble_name} to output the name @var{name}; this
6480 performs any system-specific syntactic transformations such as adding an
6483 If you don't define this macro, nothing special is output to arrange to
6484 call the function. This is correct when the function will be called in
6485 some other manner---for example, by means of the @code{collect2} program,
6486 which looks through the symbol table to find these functions by their
6489 @item ASM_OUTPUT_DESTRUCTOR (@var{stream}, @var{name})
6490 @findex ASM_OUTPUT_DESTRUCTOR
6491 This is like @code{ASM_OUTPUT_CONSTRUCTOR} but used for termination
6492 functions rather than initialization functions.
6494 When @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR} are
6495 defined, the initialization routine generated for the generated object
6496 file will have static linkage.
6499 If your system uses @code{collect2} as the means of processing
6500 constructors, then that program normally uses @code{nm} to scan an
6501 object file for constructor functions to be called. On such systems you
6502 must not define @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}
6503 as the object file's initialization routine must have global scope.
6505 On certain kinds of systems, you can define these macros to make
6506 @code{collect2} work faster (and, in some cases, make it work at all):
6509 @findex OBJECT_FORMAT_COFF
6510 @item OBJECT_FORMAT_COFF
6511 Define this macro if the system uses COFF (Common Object File Format)
6512 object files, so that @code{collect2} can assume this format and scan
6513 object files directly for dynamic constructor/destructor functions.
6515 @findex OBJECT_FORMAT_ROSE
6516 @item OBJECT_FORMAT_ROSE
6517 Define this macro if the system uses ROSE format object files, so that
6518 @code{collect2} can assume this format and scan object files directly
6519 for dynamic constructor/destructor functions.
6521 These macros are effective only in a native compiler; @code{collect2} as
6522 part of a cross compiler always uses @code{nm} for the target machine.
6524 @findex REAL_NM_FILE_NAME
6525 @item REAL_NM_FILE_NAME
6526 Define this macro as a C string constant containing the file name to use
6527 to execute @code{nm}. The default is to search the path normally for
6530 If your system supports shared libraries and has a program to list the
6531 dynamic dependencies of a given library or executable, you can define
6532 these macros to enable support for running initialization and
6533 termination functions in shared libraries:
6537 Define this macro to a C string constant containing the name of the
6538 program which lists dynamic dependencies, like @code{"ldd"} under SunOS 4.
6540 @findex PARSE_LDD_OUTPUT
6541 @item PARSE_LDD_OUTPUT (@var{PTR})
6542 Define this macro to be C code that extracts filenames from the output
6543 of the program denoted by @code{LDD_SUFFIX}. @var{PTR} is a variable
6544 of type @code{char *} that points to the beginning of a line of output
6545 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6546 code must advance @var{PTR} to the beginning of the filename on that
6547 line. Otherwise, it must set @var{PTR} to @code{NULL}.
6551 @node Instruction Output
6552 @subsection Output of Assembler Instructions
6554 @c prevent bad page break with this line
6555 This describes assembler instruction output.
6558 @findex REGISTER_NAMES
6559 @item REGISTER_NAMES
6560 A C initializer containing the assembler's names for the machine
6561 registers, each one as a C string constant. This is what translates
6562 register numbers in the compiler into assembler language.
6564 @findex ADDITIONAL_REGISTER_NAMES
6565 @item ADDITIONAL_REGISTER_NAMES
6566 If defined, a C initializer for an array of structures containing a name
6567 and a register number. This macro defines additional names for hard
6568 registers, thus allowing the @code{asm} option in declarations to refer
6569 to registers using alternate names.
6571 @findex ASM_OUTPUT_OPCODE
6572 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6573 Define this macro if you are using an unusual assembler that
6574 requires different names for the machine instructions.
6576 The definition is a C statement or statements which output an
6577 assembler instruction opcode to the stdio stream @var{stream}. The
6578 macro-operand @var{ptr} is a variable of type @code{char *} which
6579 points to the opcode name in its ``internal'' form---the form that is
6580 written in the machine description. The definition should output the
6581 opcode name to @var{stream}, performing any translation you desire, and
6582 increment the variable @var{ptr} to point at the end of the opcode
6583 so that it will not be output twice.
6585 In fact, your macro definition may process less than the entire opcode
6586 name, or more than the opcode name; but if you want to process text
6587 that includes @samp{%}-sequences to substitute operands, you must take
6588 care of the substitution yourself. Just be sure to increment
6589 @var{ptr} over whatever text should not be output normally.
6591 @findex recog_operand
6592 If you need to look at the operand values, they can be found as the
6593 elements of @code{recog_operand}.
6595 If the macro definition does nothing, the instruction is output
6598 @findex FINAL_PRESCAN_INSN
6599 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6600 If defined, a C statement to be executed just prior to the output of
6601 assembler code for @var{insn}, to modify the extracted operands so
6602 they will be output differently.
6604 Here the argument @var{opvec} is the vector containing the operands
6605 extracted from @var{insn}, and @var{noperands} is the number of
6606 elements of the vector which contain meaningful data for this insn.
6607 The contents of this vector are what will be used to convert the insn
6608 template into assembler code, so you can change the assembler output
6609 by changing the contents of the vector.
6611 This macro is useful when various assembler syntaxes share a single
6612 file of instruction patterns; by defining this macro differently, you
6613 can cause a large class of instructions to be output differently (such
6614 as with rearranged operands). Naturally, variations in assembler
6615 syntax affecting individual insn patterns ought to be handled by
6616 writing conditional output routines in those patterns.
6618 If this macro is not defined, it is equivalent to a null statement.
6620 @findex FINAL_PRESCAN_LABEL
6621 @item FINAL_PRESCAN_LABEL
6622 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
6623 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
6624 @var{noperands} will be zero.
6626 @findex PRINT_OPERAND
6627 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6628 A C compound statement to output to stdio stream @var{stream} the
6629 assembler syntax for an instruction operand @var{x}. @var{x} is an
6632 @var{code} is a value that can be used to specify one of several ways
6633 of printing the operand. It is used when identical operands must be
6634 printed differently depending on the context. @var{code} comes from
6635 the @samp{%} specification that was used to request printing of the
6636 operand. If the specification was just @samp{%@var{digit}} then
6637 @var{code} is 0; if the specification was @samp{%@var{ltr}
6638 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6641 If @var{x} is a register, this macro should print the register's name.
6642 The names can be found in an array @code{reg_names} whose type is
6643 @code{char *[]}. @code{reg_names} is initialized from
6644 @code{REGISTER_NAMES}.
6646 When the machine description has a specification @samp{%@var{punct}}
6647 (a @samp{%} followed by a punctuation character), this macro is called
6648 with a null pointer for @var{x} and the punctuation character for
6651 @findex PRINT_OPERAND_PUNCT_VALID_P
6652 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6653 A C expression which evaluates to true if @var{code} is a valid
6654 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6655 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6656 punctuation characters (except for the standard one, @samp{%}) are used
6659 @findex PRINT_OPERAND_ADDRESS
6660 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6661 A C compound statement to output to stdio stream @var{stream} the
6662 assembler syntax for an instruction operand that is a memory reference
6663 whose address is @var{x}. @var{x} is an RTL expression.
6665 @cindex @code{ENCODE_SECTION_INFO} usage
6666 On some machines, the syntax for a symbolic address depends on the
6667 section that the address refers to. On these machines, define the macro
6668 @code{ENCODE_SECTION_INFO} to store the information into the
6669 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
6671 @findex DBR_OUTPUT_SEQEND
6672 @findex dbr_sequence_length
6673 @item DBR_OUTPUT_SEQEND(@var{file})
6674 A C statement, to be executed after all slot-filler instructions have
6675 been output. If necessary, call @code{dbr_sequence_length} to
6676 determine the number of slots filled in a sequence (zero if not
6677 currently outputting a sequence), to decide how many no-ops to output,
6680 Don't define this macro if it has nothing to do, but it is helpful in
6681 reading assembly output if the extent of the delay sequence is made
6682 explicit (e.g. with white space).
6684 @findex final_sequence
6685 Note that output routines for instructions with delay slots must be
6686 prepared to deal with not being output as part of a sequence (i.e.
6687 when the scheduling pass is not run, or when no slot fillers could be
6688 found.) The variable @code{final_sequence} is null when not
6689 processing a sequence, otherwise it contains the @code{sequence} rtx
6692 @findex REGISTER_PREFIX
6693 @findex LOCAL_LABEL_PREFIX
6694 @findex USER_LABEL_PREFIX
6695 @findex IMMEDIATE_PREFIX
6697 @item REGISTER_PREFIX
6698 @itemx LOCAL_LABEL_PREFIX
6699 @itemx USER_LABEL_PREFIX
6700 @itemx IMMEDIATE_PREFIX
6701 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6702 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6703 @file{final.c}). These are useful when a single @file{md} file must
6704 support multiple assembler formats. In that case, the various @file{tm.h}
6705 files can define these macros differently.
6707 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
6708 @findex ASM_FPRINTF_EXTENSIONS
6709 If defined this macro should expand to a series of @code{case}
6710 statements which will be parsed inside the @code{switch} statement of
6711 the @code{asm_fprintf} function. This allows targets to define extra
6712 printf formats which may useful when generating their assembler
6713 statements. Note that upper case letters are reserved for future
6714 generic extensions to asm_fprintf, and so are not available to target
6715 specific code. The output file is given by the parameter @var{file}.
6716 The varargs input pointer is @var{argptr} and the rest of the format
6717 string, starting the character after the one that is being switched
6718 upon, is pointed to by @var{format}.
6720 @findex ASSEMBLER_DIALECT
6721 @item ASSEMBLER_DIALECT
6722 If your target supports multiple dialects of assembler language (such as
6723 different opcodes), define this macro as a C expression that gives the
6724 numeric index of the assembler language dialect to use, with zero as the
6727 If this macro is defined, you may use constructs of the form
6728 @samp{@{option0|option1|option2@dots{}@}} in the output
6729 templates of patterns (@pxref{Output Template}) or in the first argument
6730 of @code{asm_fprintf}. This construct outputs @samp{option0},
6731 @samp{option1} or @samp{option2}, etc., if the value of
6732 @code{ASSEMBLER_DIALECT} is zero, one or two, etc. Any special
6733 characters within these strings retain their usual meaning.
6735 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6736 @samp{@}} do not have any special meaning when used in templates or
6737 operands to @code{asm_fprintf}.
6739 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6740 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6741 the variations in assembler language syntax with that mechanism. Define
6742 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6743 if the syntax variant are larger and involve such things as different
6744 opcodes or operand order.
6746 @findex ASM_OUTPUT_REG_PUSH
6747 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6748 A C expression to output to @var{stream} some assembler code
6749 which will push hard register number @var{regno} onto the stack.
6750 The code need not be optimal, since this macro is used only when
6753 @findex ASM_OUTPUT_REG_POP
6754 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6755 A C expression to output to @var{stream} some assembler code
6756 which will pop hard register number @var{regno} off of the stack.
6757 The code need not be optimal, since this macro is used only when
6761 @node Dispatch Tables
6762 @subsection Output of Dispatch Tables
6764 @c prevent bad page break with this line
6765 This concerns dispatch tables.
6768 @cindex dispatch table
6769 @findex ASM_OUTPUT_ADDR_DIFF_ELT
6770 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6771 A C statement to output to the stdio stream @var{stream} an assembler
6772 pseudo-instruction to generate a difference between two labels.
6773 @var{value} and @var{rel} are the numbers of two internal labels. The
6774 definitions of these labels are output using
6775 @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
6776 way here. For example,
6779 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6780 @var{value}, @var{rel})
6783 You must provide this macro on machines where the addresses in a
6784 dispatch table are relative to the table's own address. If defined, GNU
6785 CC will also use this macro on all machines when producing PIC.
6786 @var{body} is the body of the ADDR_DIFF_VEC; it is provided so that the
6787 mode and flags can be read.
6789 @findex ASM_OUTPUT_ADDR_VEC_ELT
6790 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6791 This macro should be provided on machines where the addresses
6792 in a dispatch table are absolute.
6794 The definition should be a C statement to output to the stdio stream
6795 @var{stream} an assembler pseudo-instruction to generate a reference to
6796 a label. @var{value} is the number of an internal label whose
6797 definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
6801 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6804 @findex ASM_OUTPUT_CASE_LABEL
6805 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6806 Define this if the label before a jump-table needs to be output
6807 specially. The first three arguments are the same as for
6808 @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
6809 jump-table which follows (a @code{jump_insn} containing an
6810 @code{addr_vec} or @code{addr_diff_vec}).
6812 This feature is used on system V to output a @code{swbeg} statement
6815 If this macro is not defined, these labels are output with
6816 @code{ASM_OUTPUT_INTERNAL_LABEL}.
6818 @findex ASM_OUTPUT_CASE_END
6819 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6820 Define this if something special must be output at the end of a
6821 jump-table. The definition should be a C statement to be executed
6822 after the assembler code for the table is written. It should write
6823 the appropriate code to stdio stream @var{stream}. The argument
6824 @var{table} is the jump-table insn, and @var{num} is the label-number
6825 of the preceding label.
6827 If this macro is not defined, nothing special is output at the end of
6831 @node Exception Region Output
6832 @subsection Assembler Commands for Exception Regions
6834 @c prevent bad page break with this line
6836 This describes commands marking the start and the end of an exception
6840 @findex ASM_OUTPUT_EH_REGION_BEG
6841 @item ASM_OUTPUT_EH_REGION_BEG ()
6842 A C expression to output text to mark the start of an exception region.
6844 This macro need not be defined on most platforms.
6846 @findex ASM_OUTPUT_EH_REGION_END
6847 @item ASM_OUTPUT_EH_REGION_END ()
6848 A C expression to output text to mark the end of an exception region.
6850 This macro need not be defined on most platforms.
6852 @findex EXCEPTION_SECTION
6853 @item EXCEPTION_SECTION ()
6854 A C expression to switch to the section in which the main
6855 exception table is to be placed (@pxref{Sections}). The default is a
6856 section named @code{.gcc_except_table} on machines that support named
6857 sections via @code{ASM_OUTPUT_SECTION_NAME}, otherwise if @samp{-fpic}
6858 or @samp{-fPIC} is in effect, the @code{data_section}, otherwise the
6859 @code{readonly_data_section}.
6861 @findex EH_FRAME_SECTION_ASM_OP
6862 @item EH_FRAME_SECTION_ASM_OP
6863 If defined, a C string constant, including spacing, for the assembler
6864 operation to switch to the section for exception handling frame unwind
6865 information. If not defined, GCC will provide a default definition if the
6866 target supports named sections. @file{crtstuff.c} uses this macro to
6867 switch to the appropriate section.
6869 You should define this symbol if your target supports DWARF 2 frame
6870 unwind information and the default definition does not work.
6872 @findex OMIT_EH_TABLE
6873 @item OMIT_EH_TABLE ()
6874 A C expression that is nonzero if the normal exception table output
6877 This macro need not be defined on most platforms.
6879 @findex EH_TABLE_LOOKUP
6880 @item EH_TABLE_LOOKUP ()
6881 Alternate runtime support for looking up an exception at runtime and
6882 finding the associated handler, if the default method won't work.
6884 This macro need not be defined on most platforms.
6886 @findex DOESNT_NEED_UNWINDER
6887 @item DOESNT_NEED_UNWINDER
6888 A C expression that decides whether or not the current function needs to
6889 have a function unwinder generated for it. See the file @code{except.c}
6890 for details on when to define this, and how.
6892 @findex MASK_RETURN_ADDR
6893 @item MASK_RETURN_ADDR
6894 An rtx used to mask the return address found via RETURN_ADDR_RTX, so
6895 that it does not contain any extraneous set bits in it.
6897 @findex DWARF2_UNWIND_INFO
6898 @item DWARF2_UNWIND_INFO
6899 Define this macro to 0 if your target supports DWARF 2 frame unwind
6900 information, but it does not yet work with exception handling.
6901 Otherwise, if your target supports this information (if it defines
6902 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
6903 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
6906 If this macro is defined to 1, the DWARF 2 unwinder will be the default
6907 exception handling mechanism; otherwise, setjmp/longjmp will be used by
6910 If this macro is defined to anything, the DWARF 2 unwinder will be used
6911 instead of inline unwinders and __unwind_function in the non-setjmp case.
6913 @findex DWARF_CIE_DATA_ALIGNMENT
6914 @item DWARF_CIE_DATA_ALIGNMENT
6915 This macro need only be defined if the target might save registers in the
6916 function prologue at an offset to the stack pointer that is not aligned to
6917 @code{UNITS_PER_WORD}. The definition should be the negative minimum
6918 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
6919 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
6920 the target supports DWARF 2 frame unwind information.
6924 @node Alignment Output
6925 @subsection Assembler Commands for Alignment
6927 @c prevent bad page break with this line
6928 This describes commands for alignment.
6931 @findex LABEL_ALIGN_AFTER_BARRIER
6932 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
6933 The alignment (log base 2) to put in front of @var{label}, which follows
6936 This macro need not be defined if you don't want any special alignment
6937 to be done at such a time. Most machine descriptions do not currently
6940 Unless it's necessary to inspect the @var{label} parameter, it is better
6941 to set the variable @var{align_jumps} in the target's
6942 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6943 selection in @var{align_jumps} in a @code{LABEL_ALIGN_AFTER_BARRIER}
6946 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6947 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6948 The maximum number of bytes to skip when applying
6949 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
6950 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6953 @item LOOP_ALIGN (@var{label})
6954 The alignment (log base 2) to put in front of @var{label}, which follows
6955 a NOTE_INSN_LOOP_BEG note.
6957 This macro need not be defined if you don't want any special alignment
6958 to be done at such a time. Most machine descriptions do not currently
6961 Unless it's necessary to inspect the @var{label} parameter, it is better
6962 to set the variable @var{align_loops} in the target's
6963 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6964 selection in @var{align_loops} in a @code{LOOP_ALIGN} implementation.
6966 @findex LOOP_ALIGN_MAX_SKIP
6967 @item LOOP_ALIGN_MAX_SKIP
6968 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
6969 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6972 @item LABEL_ALIGN (@var{label})
6973 The alignment (log base 2) to put in front of @var{label}.
6974 If LABEL_ALIGN_AFTER_BARRIER / LOOP_ALIGN specify a different alignment,
6975 the maximum of the specified values is used.
6977 Unless it's necessary to inspect the @var{label} parameter, it is better
6978 to set the variable @var{align_labels} in the target's
6979 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6980 selection in @var{align_labels} in a @code{LABEL_ALIGN} implementation.
6982 @findex LABEL_ALIGN_MAX_SKIP
6983 @item LABEL_ALIGN_MAX_SKIP
6984 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
6985 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6987 @findex ASM_OUTPUT_SKIP
6988 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6989 A C statement to output to the stdio stream @var{stream} an assembler
6990 instruction to advance the location counter by @var{nbytes} bytes.
6991 Those bytes should be zero when loaded. @var{nbytes} will be a C
6992 expression of type @code{int}.
6994 @findex ASM_NO_SKIP_IN_TEXT
6995 @item ASM_NO_SKIP_IN_TEXT
6996 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6997 text section because it fails to put zeros in the bytes that are skipped.
6998 This is true on many Unix systems, where the pseudo--op to skip bytes
6999 produces no-op instructions rather than zeros when used in the text
7002 @findex ASM_OUTPUT_ALIGN
7003 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7004 A C statement to output to the stdio stream @var{stream} an assembler
7005 command to advance the location counter to a multiple of 2 to the
7006 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7008 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
7009 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7010 A C statement to output to the stdio stream @var{stream} an assembler
7011 command to advance the location counter to a multiple of 2 to the
7012 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7013 satisfy the alignment request. @var{power} and @var{max_skip} will be
7014 a C expression of type @code{int}.
7018 @node Debugging Info
7019 @section Controlling Debugging Information Format
7021 @c prevent bad page break with this line
7022 This describes how to specify debugging information.
7025 * All Debuggers:: Macros that affect all debugging formats uniformly.
7026 * DBX Options:: Macros enabling specific options in DBX format.
7027 * DBX Hooks:: Hook macros for varying DBX format.
7028 * File Names and DBX:: Macros controlling output of file names in DBX format.
7029 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7033 @subsection Macros Affecting All Debugging Formats
7035 @c prevent bad page break with this line
7036 These macros affect all debugging formats.
7039 @findex DBX_REGISTER_NUMBER
7040 @item DBX_REGISTER_NUMBER (@var{regno})
7041 A C expression that returns the DBX register number for the compiler
7042 register number @var{regno}. In simple cases, the value of this
7043 expression may be @var{regno} itself. But sometimes there are some
7044 registers that the compiler knows about and DBX does not, or vice
7045 versa. In such cases, some register may need to have one number in
7046 the compiler and another for DBX.
7048 If two registers have consecutive numbers inside GCC, and they can be
7049 used as a pair to hold a multiword value, then they @emph{must} have
7050 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7051 Otherwise, debuggers will be unable to access such a pair, because they
7052 expect register pairs to be consecutive in their own numbering scheme.
7054 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7055 does not preserve register pairs, then what you must do instead is
7056 redefine the actual register numbering scheme.
7058 @findex DEBUGGER_AUTO_OFFSET
7059 @item DEBUGGER_AUTO_OFFSET (@var{x})
7060 A C expression that returns the integer offset value for an automatic
7061 variable having address @var{x} (an RTL expression). The default
7062 computation assumes that @var{x} is based on the frame-pointer and
7063 gives the offset from the frame-pointer. This is required for targets
7064 that produce debugging output for DBX or COFF-style debugging output
7065 for SDB and allow the frame-pointer to be eliminated when the
7066 @samp{-g} options is used.
7068 @findex DEBUGGER_ARG_OFFSET
7069 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7070 A C expression that returns the integer offset value for an argument
7071 having address @var{x} (an RTL expression). The nominal offset is
7074 @findex PREFERRED_DEBUGGING_TYPE
7075 @item PREFERRED_DEBUGGING_TYPE
7076 A C expression that returns the type of debugging output GCC should
7077 produce when the user specifies just @samp{-g}. Define
7078 this if you have arranged for GCC to support more than one format of
7079 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7080 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG}, and
7083 When the user specifies @samp{-ggdb}, GCC normally also uses the
7084 value of this macro to select the debugging output format, but with two
7085 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7086 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7087 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7088 defined, GCC uses @code{DBX_DEBUG}.
7090 The value of this macro only affects the default debugging output; the
7091 user can always get a specific type of output by using @samp{-gstabs},
7092 @samp{-gcoff}, @samp{-gdwarf-1}, @samp{-gdwarf-2}, or @samp{-gxcoff}.
7096 @subsection Specific Options for DBX Output
7098 @c prevent bad page break with this line
7099 These are specific options for DBX output.
7102 @findex DBX_DEBUGGING_INFO
7103 @item DBX_DEBUGGING_INFO
7104 Define this macro if GCC should produce debugging output for DBX
7105 in response to the @samp{-g} option.
7107 @findex XCOFF_DEBUGGING_INFO
7108 @item XCOFF_DEBUGGING_INFO
7109 Define this macro if GCC should produce XCOFF format debugging output
7110 in response to the @samp{-g} option. This is a variant of DBX format.
7112 @findex DEFAULT_GDB_EXTENSIONS
7113 @item DEFAULT_GDB_EXTENSIONS
7114 Define this macro to control whether GCC should by default generate
7115 GDB's extended version of DBX debugging information (assuming DBX-format
7116 debugging information is enabled at all). If you don't define the
7117 macro, the default is 1: always generate the extended information
7118 if there is any occasion to.
7120 @findex DEBUG_SYMS_TEXT
7121 @item DEBUG_SYMS_TEXT
7122 Define this macro if all @code{.stabs} commands should be output while
7123 in the text section.
7125 @findex ASM_STABS_OP
7127 A C string constant, including spacing, naming the assembler pseudo op to
7128 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7129 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7130 applies only to DBX debugging information format.
7132 @findex ASM_STABD_OP
7134 A C string constant, including spacing, naming the assembler pseudo op to
7135 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7136 value is the current location. If you don't define this macro,
7137 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7140 @findex ASM_STABN_OP
7142 A C string constant, including spacing, naming the assembler pseudo op to
7143 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7144 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7145 macro applies only to DBX debugging information format.
7147 @findex DBX_NO_XREFS
7149 Define this macro if DBX on your system does not support the construct
7150 @samp{xs@var{tagname}}. On some systems, this construct is used to
7151 describe a forward reference to a structure named @var{tagname}.
7152 On other systems, this construct is not supported at all.
7154 @findex DBX_CONTIN_LENGTH
7155 @item DBX_CONTIN_LENGTH
7156 A symbol name in DBX-format debugging information is normally
7157 continued (split into two separate @code{.stabs} directives) when it
7158 exceeds a certain length (by default, 80 characters). On some
7159 operating systems, DBX requires this splitting; on others, splitting
7160 must not be done. You can inhibit splitting by defining this macro
7161 with the value zero. You can override the default splitting-length by
7162 defining this macro as an expression for the length you desire.
7164 @findex DBX_CONTIN_CHAR
7165 @item DBX_CONTIN_CHAR
7166 Normally continuation is indicated by adding a @samp{\} character to
7167 the end of a @code{.stabs} string when a continuation follows. To use
7168 a different character instead, define this macro as a character
7169 constant for the character you want to use. Do not define this macro
7170 if backslash is correct for your system.
7172 @findex DBX_STATIC_STAB_DATA_SECTION
7173 @item DBX_STATIC_STAB_DATA_SECTION
7174 Define this macro if it is necessary to go to the data section before
7175 outputting the @samp{.stabs} pseudo-op for a non-global static
7178 @findex DBX_TYPE_DECL_STABS_CODE
7179 @item DBX_TYPE_DECL_STABS_CODE
7180 The value to use in the ``code'' field of the @code{.stabs} directive
7181 for a typedef. The default is @code{N_LSYM}.
7183 @findex DBX_STATIC_CONST_VAR_CODE
7184 @item DBX_STATIC_CONST_VAR_CODE
7185 The value to use in the ``code'' field of the @code{.stabs} directive
7186 for a static variable located in the text section. DBX format does not
7187 provide any ``right'' way to do this. The default is @code{N_FUN}.
7189 @findex DBX_REGPARM_STABS_CODE
7190 @item DBX_REGPARM_STABS_CODE
7191 The value to use in the ``code'' field of the @code{.stabs} directive
7192 for a parameter passed in registers. DBX format does not provide any
7193 ``right'' way to do this. The default is @code{N_RSYM}.
7195 @findex DBX_REGPARM_STABS_LETTER
7196 @item DBX_REGPARM_STABS_LETTER
7197 The letter to use in DBX symbol data to identify a symbol as a parameter
7198 passed in registers. DBX format does not customarily provide any way to
7199 do this. The default is @code{'P'}.
7201 @findex DBX_MEMPARM_STABS_LETTER
7202 @item DBX_MEMPARM_STABS_LETTER
7203 The letter to use in DBX symbol data to identify a symbol as a stack
7204 parameter. The default is @code{'p'}.
7206 @findex DBX_FUNCTION_FIRST
7207 @item DBX_FUNCTION_FIRST
7208 Define this macro if the DBX information for a function and its
7209 arguments should precede the assembler code for the function. Normally,
7210 in DBX format, the debugging information entirely follows the assembler
7213 @findex DBX_LBRAC_FIRST
7214 @item DBX_LBRAC_FIRST
7215 Define this macro if the @code{N_LBRAC} symbol for a block should
7216 precede the debugging information for variables and functions defined in
7217 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
7220 @findex DBX_BLOCKS_FUNCTION_RELATIVE
7221 @item DBX_BLOCKS_FUNCTION_RELATIVE
7222 Define this macro if the value of a symbol describing the scope of a
7223 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7224 of the enclosing function. Normally, GNU C uses an absolute address.
7226 @findex DBX_USE_BINCL
7228 Define this macro if GNU C should generate @code{N_BINCL} and
7229 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7230 macro also directs GNU C to output a type number as a pair of a file
7231 number and a type number within the file. Normally, GNU C does not
7232 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7233 number for a type number.
7237 @subsection Open-Ended Hooks for DBX Format
7239 @c prevent bad page break with this line
7240 These are hooks for DBX format.
7243 @findex DBX_OUTPUT_LBRAC
7244 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7245 Define this macro to say how to output to @var{stream} the debugging
7246 information for the start of a scope level for variable names. The
7247 argument @var{name} is the name of an assembler symbol (for use with
7248 @code{assemble_name}) whose value is the address where the scope begins.
7250 @findex DBX_OUTPUT_RBRAC
7251 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7252 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7254 @findex DBX_OUTPUT_ENUM
7255 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
7256 Define this macro if the target machine requires special handling to
7257 output an enumeration type. The definition should be a C statement
7258 (sans semicolon) to output the appropriate information to @var{stream}
7259 for the type @var{type}.
7261 @findex DBX_OUTPUT_FUNCTION_END
7262 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7263 Define this macro if the target machine requires special output at the
7264 end of the debugging information for a function. The definition should
7265 be a C statement (sans semicolon) to output the appropriate information
7266 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7269 @findex DBX_OUTPUT_STANDARD_TYPES
7270 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7271 Define this macro if you need to control the order of output of the
7272 standard data types at the beginning of compilation. The argument
7273 @var{syms} is a @code{tree} which is a chain of all the predefined
7274 global symbols, including names of data types.
7276 Normally, DBX output starts with definitions of the types for integers
7277 and characters, followed by all the other predefined types of the
7278 particular language in no particular order.
7280 On some machines, it is necessary to output different particular types
7281 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
7282 those symbols in the necessary order. Any predefined types that you
7283 don't explicitly output will be output afterward in no particular order.
7285 Be careful not to define this macro so that it works only for C. There
7286 are no global variables to access most of the built-in types, because
7287 another language may have another set of types. The way to output a
7288 particular type is to look through @var{syms} to see if you can find it.
7294 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7295 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
7297 dbxout_symbol (decl);
7303 This does nothing if the expected type does not exist.
7305 See the function @code{init_decl_processing} in @file{c-decl.c} to find
7306 the names to use for all the built-in C types.
7308 Here is another way of finding a particular type:
7310 @c this is still overfull. --mew 10feb93
7314 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7315 if (TREE_CODE (decl) == TYPE_DECL
7316 && (TREE_CODE (TREE_TYPE (decl))
7318 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
7319 && TYPE_UNSIGNED (TREE_TYPE (decl)))
7321 /* @r{This must be @code{unsigned short}.} */
7322 dbxout_symbol (decl);
7328 @findex NO_DBX_FUNCTION_END
7329 @item NO_DBX_FUNCTION_END
7330 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7331 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extention construct.
7332 On those machines, define this macro to turn this feature off without
7333 disturbing the rest of the gdb extensions.
7337 @node File Names and DBX
7338 @subsection File Names in DBX Format
7340 @c prevent bad page break with this line
7341 This describes file names in DBX format.
7344 @findex DBX_WORKING_DIRECTORY
7345 @item DBX_WORKING_DIRECTORY
7346 Define this if DBX wants to have the current directory recorded in each
7349 Note that the working directory is always recorded if GDB extensions are
7352 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
7353 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7354 A C statement to output DBX debugging information to the stdio stream
7355 @var{stream} which indicates that file @var{name} is the main source
7356 file---the file specified as the input file for compilation.
7357 This macro is called only once, at the beginning of compilation.
7359 This macro need not be defined if the standard form of output
7360 for DBX debugging information is appropriate.
7362 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
7363 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
7364 A C statement to output DBX debugging information to the stdio stream
7365 @var{stream} which indicates that the current directory during
7366 compilation is named @var{name}.
7368 This macro need not be defined if the standard form of output
7369 for DBX debugging information is appropriate.
7371 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
7372 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7373 A C statement to output DBX debugging information at the end of
7374 compilation of the main source file @var{name}.
7376 If you don't define this macro, nothing special is output at the end
7377 of compilation, which is correct for most machines.
7379 @findex DBX_OUTPUT_SOURCE_FILENAME
7380 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7381 A C statement to output DBX debugging information to the stdio stream
7382 @var{stream} which indicates that file @var{name} is the current source
7383 file. This output is generated each time input shifts to a different
7384 source file as a result of @samp{#include}, the end of an included file,
7385 or a @samp{#line} command.
7387 This macro need not be defined if the standard form of output
7388 for DBX debugging information is appropriate.
7393 @subsection Macros for SDB and DWARF Output
7395 @c prevent bad page break with this line
7396 Here are macros for SDB and DWARF output.
7399 @findex SDB_DEBUGGING_INFO
7400 @item SDB_DEBUGGING_INFO
7401 Define this macro if GCC should produce COFF-style debugging output
7402 for SDB in response to the @samp{-g} option.
7404 @findex DWARF_DEBUGGING_INFO
7405 @item DWARF_DEBUGGING_INFO
7406 Define this macro if GCC should produce dwarf format debugging output
7407 in response to the @samp{-g} option.
7409 @findex DWARF2_DEBUGGING_INFO
7410 @item DWARF2_DEBUGGING_INFO
7411 Define this macro if GCC should produce dwarf version 2 format
7412 debugging output in response to the @samp{-g} option.
7414 To support optional call frame debugging information, you must also
7415 define @code{INCOMING_RETURN_ADDR_RTX} and either set
7416 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
7417 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
7418 as appropriate from @code{FUNCTION_PROLOGUE} if you don't.
7420 @findex DWARF2_FRAME_INFO
7421 @item DWARF2_FRAME_INFO
7422 Define this macro to a nonzero value if GCC should always output
7423 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
7424 (@pxref{Exception Region Output} is nonzero, GCC will output this
7425 information not matter how you define @code{DWARF2_FRAME_INFO}.
7427 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
7428 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
7429 Define this macro if the linker does not work with Dwarf version 2.
7430 Normally, if the user specifies only @samp{-ggdb} GCC will use Dwarf
7431 version 2 if available; this macro disables this. See the description
7432 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
7434 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
7435 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
7436 By default, the Dwarf 2 debugging information generator will generate a
7437 label to mark the beginning of the text section. If it is better simply
7438 to use the name of the text section itself, rather than an explicit label,
7439 to indicate the beginning of the text section, define this macro to zero.
7441 @findex DWARF2_ASM_LINE_DEBUG_INFO
7442 @item DWARF2_ASM_LINE_DEBUG_INFO
7443 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
7444 line debug info sections. This will result in much more compact line number
7445 tables, and hence is desirable if it works.
7447 @findex PUT_SDB_@dots{}
7448 @item PUT_SDB_@dots{}
7449 Define these macros to override the assembler syntax for the special
7450 SDB assembler directives. See @file{sdbout.c} for a list of these
7451 macros and their arguments. If the standard syntax is used, you need
7452 not define them yourself.
7456 Some assemblers do not support a semicolon as a delimiter, even between
7457 SDB assembler directives. In that case, define this macro to be the
7458 delimiter to use (usually @samp{\n}). It is not necessary to define
7459 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
7462 @findex SDB_GENERATE_FAKE
7463 @item SDB_GENERATE_FAKE
7464 Define this macro to override the usual method of constructing a dummy
7465 name for anonymous structure and union types. See @file{sdbout.c} for
7468 @findex SDB_ALLOW_UNKNOWN_REFERENCES
7469 @item SDB_ALLOW_UNKNOWN_REFERENCES
7470 Define this macro to allow references to unknown structure,
7471 union, or enumeration tags to be emitted. Standard COFF does not
7472 allow handling of unknown references, MIPS ECOFF has support for
7475 @findex SDB_ALLOW_FORWARD_REFERENCES
7476 @item SDB_ALLOW_FORWARD_REFERENCES
7477 Define this macro to allow references to structure, union, or
7478 enumeration tags that have not yet been seen to be handled. Some
7479 assemblers choke if forward tags are used, while some require it.
7482 @node Cross-compilation
7483 @section Cross Compilation and Floating Point
7484 @cindex cross compilation and floating point
7485 @cindex floating point and cross compilation
7487 While all modern machines use 2's complement representation for integers,
7488 there are a variety of representations for floating point numbers. This
7489 means that in a cross-compiler the representation of floating point numbers
7490 in the compiled program may be different from that used in the machine
7491 doing the compilation.
7494 Because different representation systems may offer different amounts of
7495 range and precision, the cross compiler cannot safely use the host
7496 machine's floating point arithmetic. Therefore, floating point constants
7497 must be represented in the target machine's format. This means that the
7498 cross compiler cannot use @code{atof} to parse a floating point constant;
7499 it must have its own special routine to use instead. Also, constant
7500 folding must emulate the target machine's arithmetic (or must not be done
7503 The macros in the following table should be defined only if you are cross
7504 compiling between different floating point formats.
7506 Otherwise, don't define them. Then default definitions will be set up which
7507 use @code{double} as the data type, @code{==} to test for equality, etc.
7509 You don't need to worry about how many times you use an operand of any
7510 of these macros. The compiler never uses operands which have side effects.
7513 @findex REAL_VALUE_TYPE
7514 @item REAL_VALUE_TYPE
7515 A macro for the C data type to be used to hold a floating point value
7516 in the target machine's format. Typically this would be a
7517 @code{struct} containing an array of @code{int}.
7519 @findex REAL_VALUES_EQUAL
7520 @item REAL_VALUES_EQUAL (@var{x}, @var{y})
7521 A macro for a C expression which compares for equality the two values,
7522 @var{x} and @var{y}, both of type @code{REAL_VALUE_TYPE}.
7524 @findex REAL_VALUES_LESS
7525 @item REAL_VALUES_LESS (@var{x}, @var{y})
7526 A macro for a C expression which tests whether @var{x} is less than
7527 @var{y}, both values being of type @code{REAL_VALUE_TYPE} and
7528 interpreted as floating point numbers in the target machine's
7531 @findex REAL_VALUE_LDEXP
7533 @item REAL_VALUE_LDEXP (@var{x}, @var{scale})
7534 A macro for a C expression which performs the standard library
7535 function @code{ldexp}, but using the target machine's floating point
7536 representation. Both @var{x} and the value of the expression have
7537 type @code{REAL_VALUE_TYPE}. The second argument, @var{scale}, is an
7540 @findex REAL_VALUE_FIX
7541 @item REAL_VALUE_FIX (@var{x})
7542 A macro whose definition is a C expression to convert the target-machine
7543 floating point value @var{x} to a signed integer. @var{x} has type
7544 @code{REAL_VALUE_TYPE}.
7546 @findex REAL_VALUE_UNSIGNED_FIX
7547 @item REAL_VALUE_UNSIGNED_FIX (@var{x})
7548 A macro whose definition is a C expression to convert the target-machine
7549 floating point value @var{x} to an unsigned integer. @var{x} has type
7550 @code{REAL_VALUE_TYPE}.
7552 @findex REAL_VALUE_RNDZINT
7553 @item REAL_VALUE_RNDZINT (@var{x})
7554 A macro whose definition is a C expression to round the target-machine
7555 floating point value @var{x} towards zero to an integer value (but still
7556 as a floating point number). @var{x} has type @code{REAL_VALUE_TYPE},
7557 and so does the value.
7559 @findex REAL_VALUE_UNSIGNED_RNDZINT
7560 @item REAL_VALUE_UNSIGNED_RNDZINT (@var{x})
7561 A macro whose definition is a C expression to round the target-machine
7562 floating point value @var{x} towards zero to an unsigned integer value
7563 (but still represented as a floating point number). @var{x} has type
7564 @code{REAL_VALUE_TYPE}, and so does the value.
7566 @findex REAL_VALUE_ATOF
7567 @item REAL_VALUE_ATOF (@var{string}, @var{mode})
7568 A macro for a C expression which converts @var{string}, an expression of
7569 type @code{char *}, into a floating point number in the target machine's
7570 representation for mode @var{mode}. The value has type
7571 @code{REAL_VALUE_TYPE}.
7573 @findex REAL_INFINITY
7575 Define this macro if infinity is a possible floating point value, and
7576 therefore division by 0 is legitimate.
7578 @findex REAL_VALUE_ISINF
7580 @item REAL_VALUE_ISINF (@var{x})
7581 A macro for a C expression which determines whether @var{x}, a floating
7582 point value, is infinity. The value has type @code{int}.
7583 By default, this is defined to call @code{isinf}.
7585 @findex REAL_VALUE_ISNAN
7587 @item REAL_VALUE_ISNAN (@var{x})
7588 A macro for a C expression which determines whether @var{x}, a floating
7589 point value, is a ``nan'' (not-a-number). The value has type
7590 @code{int}. By default, this is defined to call @code{isnan}.
7593 @cindex constant folding and floating point
7594 Define the following additional macros if you want to make floating
7595 point constant folding work while cross compiling. If you don't
7596 define them, cross compilation is still possible, but constant folding
7597 will not happen for floating point values.
7600 @findex REAL_ARITHMETIC
7601 @item REAL_ARITHMETIC (@var{output}, @var{code}, @var{x}, @var{y})
7602 A macro for a C statement which calculates an arithmetic operation of
7603 the two floating point values @var{x} and @var{y}, both of type
7604 @code{REAL_VALUE_TYPE} in the target machine's representation, to
7605 produce a result of the same type and representation which is stored
7606 in @var{output} (which will be a variable).
7608 The operation to be performed is specified by @var{code}, a tree code
7609 which will always be one of the following: @code{PLUS_EXPR},
7610 @code{MINUS_EXPR}, @code{MULT_EXPR}, @code{RDIV_EXPR},
7611 @code{MAX_EXPR}, @code{MIN_EXPR}.@refill
7613 @cindex overflow while constant folding
7614 The expansion of this macro is responsible for checking for overflow.
7615 If overflow happens, the macro expansion should execute the statement
7616 @code{return 0;}, which indicates the inability to perform the
7617 arithmetic operation requested.
7619 @findex REAL_VALUE_NEGATE
7620 @item REAL_VALUE_NEGATE (@var{x})
7621 A macro for a C expression which returns the negative of the floating
7622 point value @var{x}. Both @var{x} and the value of the expression
7623 have type @code{REAL_VALUE_TYPE} and are in the target machine's
7624 floating point representation.
7626 There is no way for this macro to report overflow, since overflow
7627 can't happen in the negation operation.
7629 @findex REAL_VALUE_TRUNCATE
7630 @item REAL_VALUE_TRUNCATE (@var{mode}, @var{x})
7631 A macro for a C expression which converts the floating point value
7632 @var{x} to mode @var{mode}.
7634 Both @var{x} and the value of the expression are in the target machine's
7635 floating point representation and have type @code{REAL_VALUE_TYPE}.
7636 However, the value should have an appropriate bit pattern to be output
7637 properly as a floating constant whose precision accords with mode
7640 There is no way for this macro to report overflow.
7642 @findex REAL_VALUE_TO_INT
7643 @item REAL_VALUE_TO_INT (@var{low}, @var{high}, @var{x})
7644 A macro for a C expression which converts a floating point value
7645 @var{x} into a double-precision integer which is then stored into
7646 @var{low} and @var{high}, two variables of type @var{int}.
7648 @item REAL_VALUE_FROM_INT (@var{x}, @var{low}, @var{high}, @var{mode})
7649 @findex REAL_VALUE_FROM_INT
7650 A macro for a C expression which converts a double-precision integer
7651 found in @var{low} and @var{high}, two variables of type @var{int},
7652 into a floating point value which is then stored into @var{x}.
7653 The value is in the target machine's representation for mode @var{mode}
7654 and has the type @code{REAL_VALUE_TYPE}.
7657 @node Mode Switching
7658 @section Mode Switching Instructions
7659 @cindex mode switching
7660 The following macros control mode switching optimizations:
7663 @findex OPTIMIZE_MODE_SWITCHING
7664 @item OPTIMIZE_MODE_SWITCHING (@var{entity})
7665 Define this macro if the port needs extra instructions inserted for mode
7666 switching in an optimizing compilation.
7668 For an example, the SH4 can perform both single and double precision
7669 floating point operations, but to perform a single precision operation,
7670 the FPSCR PR bit has to be cleared, while for a double precision
7671 operation, this bit has to be set. Changing the PR bit requires a general
7672 purpose register as a scratch register, hence these FPSCR sets have to
7673 be inserted before reload, i.e. you can't put this into instruction emitting
7674 or MACHINE_DEPENDENT_REORG.
7676 You can have multiple entities that are mode-switched, and select at run time
7677 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
7678 return non-zero for any @var{entity} that that needs mode-switching.
7679 If you define this macro, you also have to define
7680 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
7681 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
7682 @code{NORMAL_MODE} is optional.
7684 @findex NUM_MODES_FOR_MODE_SWITCHING
7685 @item NUM_MODES_FOR_MODE_SWITCHING
7686 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
7687 initializer for an array of integers. Each initializer element
7688 N refers to an entity that needs mode switching, and specifies the number
7689 of different modes that might need to be set for this entity.
7690 The position of the initializer in the initializer - starting counting at
7691 zero - determines the integer that is used to refer to the mode-switched
7693 In macros that take mode arguments / yield a mode result, modes are
7694 represented as numbers 0 .. N - 1. N is used to specify that no mode
7695 switch is needed / supplied.
7698 @item MODE_NEEDED (@var{entity}, @var{insn})
7699 @var{entity} is an integer specifying a mode-switched entity. If
7700 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
7701 return an integer value not larger than the corresponding element in
7702 NUM_MODES_FOR_MODE_SWITCHING, to denote the mode that @var{entity} must
7703 be switched into prior to the execution of INSN.
7706 @item NORMAL_MODE (@var{entity})
7707 If this macro is defined, it is evaluated for every @var{entity} that needs
7708 mode switching. It should evaluate to an integer, which is a mode that
7709 @var{entity} is assumed to be switched to at function entry and exit.
7711 @findex MODE_PRIORITY_TO_MODE
7712 @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
7713 This macro specifies the order in which modes for ENTITY are processed.
7714 0 is the highest priority, NUM_MODES_FOR_MODE_SWITCHING[ENTITY] - 1 the
7715 lowest. The value of the macro should be an integer designating a mode
7716 for ENTITY. For any fixed @var{entity}, @code{mode_priority_to_mode}
7717 (@var{entity}, @var{n}) shall be a bijection in 0 ..
7718 @code{num_modes_for_mode_switching}[@var{entity}] - 1 .
7720 @findex EMIT_MODE_SET
7721 @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
7722 Generate one or more insns to set @var{entity} to @var{mode}.
7723 @var{hard_reg_live} is the set of hard registers live at the point where
7724 the insn(s) are to be inserted.
7728 @section Miscellaneous Parameters
7729 @cindex parameters, miscellaneous
7731 @c prevent bad page break with this line
7732 Here are several miscellaneous parameters.
7735 @item PREDICATE_CODES
7736 @findex PREDICATE_CODES
7737 Define this if you have defined special-purpose predicates in the file
7738 @file{@var{machine}.c}. This macro is called within an initializer of an
7739 array of structures. The first field in the structure is the name of a
7740 predicate and the second field is an array of rtl codes. For each
7741 predicate, list all rtl codes that can be in expressions matched by the
7742 predicate. The list should have a trailing comma. Here is an example
7743 of two entries in the list for a typical RISC machine:
7746 #define PREDICATE_CODES \
7747 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
7748 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
7751 Defining this macro does not affect the generated code (however,
7752 incorrect definitions that omit an rtl code that may be matched by the
7753 predicate can cause the compiler to malfunction). Instead, it allows
7754 the table built by @file{genrecog} to be more compact and efficient,
7755 thus speeding up the compiler. The most important predicates to include
7756 in the list specified by this macro are those used in the most insn
7759 For each predicate function named in @var{PREDICATE_CODES}, a
7760 declaration will be generated in @file{insn-codes.h}.
7762 @item SPECIAL_MODE_PREDICATES
7763 @findex SPECIAL_MODE_PREDICATES
7764 Define this if you have special predicates that know special things
7765 about modes. Genrecog will warn about certain forms of
7766 @code{match_operand} without a mode; if the operand predicate is
7767 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
7770 Here is an example from the IA-32 port (@code{ext_register_operand}
7771 specially checks for @code{HImode} or @code{SImode} in preparation
7772 for a byte extraction from @code{%ah} etc.).
7775 #define SPECIAL_MODE_PREDICATES \
7776 "ext_register_operand",
7779 @findex CASE_VECTOR_MODE
7780 @item CASE_VECTOR_MODE
7781 An alias for a machine mode name. This is the machine mode that
7782 elements of a jump-table should have.
7784 @findex CASE_VECTOR_SHORTEN_MODE
7785 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7786 Optional: return the preferred mode for an @code{addr_diff_vec}
7787 when the minimum and maximum offset are known. If you define this,
7788 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7789 To make this work, you also have to define INSN_ALIGN and
7790 make the alignment for @code{addr_diff_vec} explicit.
7791 The @var{body} argument is provided so that the offset_unsigned and scale
7792 flags can be updated.
7794 @findex CASE_VECTOR_PC_RELATIVE
7795 @item CASE_VECTOR_PC_RELATIVE
7796 Define this macro to be a C expression to indicate when jump-tables
7797 should contain relative addresses. If jump-tables never contain
7798 relative addresses, then you need not define this macro.
7800 @findex CASE_DROPS_THROUGH
7801 @item CASE_DROPS_THROUGH
7802 Define this if control falls through a @code{case} insn when the index
7803 value is out of range. This means the specified default-label is
7804 actually ignored by the @code{case} insn proper.
7806 @findex CASE_VALUES_THRESHOLD
7807 @item CASE_VALUES_THRESHOLD
7808 Define this to be the smallest number of different values for which it
7809 is best to use a jump-table instead of a tree of conditional branches.
7810 The default is four for machines with a @code{casesi} instruction and
7811 five otherwise. This is best for most machines.
7813 @findex WORD_REGISTER_OPERATIONS
7814 @item WORD_REGISTER_OPERATIONS
7815 Define this macro if operations between registers with integral mode
7816 smaller than a word are always performed on the entire register.
7817 Most RISC machines have this property and most CISC machines do not.
7819 @findex LOAD_EXTEND_OP
7820 @item LOAD_EXTEND_OP (@var{mode})
7821 Define this macro to be a C expression indicating when insns that read
7822 memory in @var{mode}, an integral mode narrower than a word, set the
7823 bits outside of @var{mode} to be either the sign-extension or the
7824 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7825 of @var{mode} for which the
7826 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7827 @code{NIL} for other modes.
7829 This macro is not called with @var{mode} non-integral or with a width
7830 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7831 value in this case. Do not define this macro if it would always return
7832 @code{NIL}. On machines where this macro is defined, you will normally
7833 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7835 @findex SHORT_IMMEDIATES_SIGN_EXTEND
7836 @item SHORT_IMMEDIATES_SIGN_EXTEND
7837 Define this macro if loading short immediate values into registers sign
7840 @findex IMPLICIT_FIX_EXPR
7841 @item IMPLICIT_FIX_EXPR
7842 An alias for a tree code that should be used by default for conversion
7843 of floating point values to fixed point. Normally,
7844 @code{FIX_ROUND_EXPR} is used.@refill
7846 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
7847 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
7848 Define this macro if the same instructions that convert a floating
7849 point number to a signed fixed point number also convert validly to an
7852 @findex EASY_DIV_EXPR
7854 An alias for a tree code that is the easiest kind of division to
7855 compile code for in the general case. It may be
7856 @code{TRUNC_DIV_EXPR}, @code{FLOOR_DIV_EXPR}, @code{CEIL_DIV_EXPR} or
7857 @code{ROUND_DIV_EXPR}. These four division operators differ in how
7858 they round the result to an integer. @code{EASY_DIV_EXPR} is used
7859 when it is permissible to use any of those kinds of division and the
7860 choice should be made on the basis of efficiency.@refill
7864 The maximum number of bytes that a single instruction can move quickly
7865 between memory and registers or between two memory locations.
7867 @findex MAX_MOVE_MAX
7869 The maximum number of bytes that a single instruction can move quickly
7870 between memory and registers or between two memory locations. If this
7871 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7872 constant value that is the largest value that @code{MOVE_MAX} can have
7875 @findex SHIFT_COUNT_TRUNCATED
7876 @item SHIFT_COUNT_TRUNCATED
7877 A C expression that is nonzero if on this machine the number of bits
7878 actually used for the count of a shift operation is equal to the number
7879 of bits needed to represent the size of the object being shifted. When
7880 this macro is non-zero, the compiler will assume that it is safe to omit
7881 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7882 truncates the count of a shift operation. On machines that have
7883 instructions that act on bitfields at variable positions, which may
7884 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7885 also enables deletion of truncations of the values that serve as
7886 arguments to bitfield instructions.
7888 If both types of instructions truncate the count (for shifts) and
7889 position (for bitfield operations), or if no variable-position bitfield
7890 instructions exist, you should define this macro.
7892 However, on some machines, such as the 80386 and the 680x0, truncation
7893 only applies to shift operations and not the (real or pretended)
7894 bitfield operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7895 such machines. Instead, add patterns to the @file{md} file that include
7896 the implied truncation of the shift instructions.
7898 You need not define this macro if it would always have the value of zero.
7900 @findex TRULY_NOOP_TRUNCATION
7901 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
7902 A C expression which is nonzero if on this machine it is safe to
7903 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
7904 bits (where @var{outprec} is smaller than @var{inprec}) by merely
7905 operating on it as if it had only @var{outprec} bits.
7907 On many machines, this expression can be 1.
7909 @c rearranged this, removed the phrase "it is reported that". this was
7910 @c to fix an overfull hbox. --mew 10feb93
7911 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
7912 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
7913 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
7914 such cases may improve things.
7916 @findex STORE_FLAG_VALUE
7917 @item STORE_FLAG_VALUE
7918 A C expression describing the value returned by a comparison operator
7919 with an integral mode and stored by a store-flag instruction
7920 (@samp{s@var{cond}}) when the condition is true. This description must
7921 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
7922 comparison operators whose results have a @code{MODE_INT} mode.
7924 A value of 1 or -1 means that the instruction implementing the
7925 comparison operator returns exactly 1 or -1 when the comparison is true
7926 and 0 when the comparison is false. Otherwise, the value indicates
7927 which bits of the result are guaranteed to be 1 when the comparison is
7928 true. This value is interpreted in the mode of the comparison
7929 operation, which is given by the mode of the first operand in the
7930 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
7931 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7934 If @code{STORE_FLAG_VALUE} is neither 1 or -1, the compiler will
7935 generate code that depends only on the specified bits. It can also
7936 replace comparison operators with equivalent operations if they cause
7937 the required bits to be set, even if the remaining bits are undefined.
7938 For example, on a machine whose comparison operators return an
7939 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7940 @samp{0x80000000}, saying that just the sign bit is relevant, the
7944 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7951 (ashift:SI @var{x} (const_int @var{n}))
7955 where @var{n} is the appropriate shift count to move the bit being
7956 tested into the sign bit.
7958 There is no way to describe a machine that always sets the low-order bit
7959 for a true value, but does not guarantee the value of any other bits,
7960 but we do not know of any machine that has such an instruction. If you
7961 are trying to port GCC to such a machine, include an instruction to
7962 perform a logical-and of the result with 1 in the pattern for the
7963 comparison operators and let us know
7965 (@pxref{Bug Reporting,,How to Report Bugs}).
7968 (@pxref{Bug Reporting,,How to Report Bugs,gcc.info,Using GCC}).
7971 Often, a machine will have multiple instructions that obtain a value
7972 from a comparison (or the condition codes). Here are rules to guide the
7973 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7978 Use the shortest sequence that yields a valid definition for
7979 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7980 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7981 comparison operators to do so because there may be opportunities to
7982 combine the normalization with other operations.
7985 For equal-length sequences, use a value of 1 or -1, with -1 being
7986 slightly preferred on machines with expensive jumps and 1 preferred on
7990 As a second choice, choose a value of @samp{0x80000001} if instructions
7991 exist that set both the sign and low-order bits but do not define the
7995 Otherwise, use a value of @samp{0x80000000}.
7998 Many machines can produce both the value chosen for
7999 @code{STORE_FLAG_VALUE} and its negation in the same number of
8000 instructions. On those machines, you should also define a pattern for
8001 those cases, e.g., one matching
8004 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8007 Some machines can also perform @code{and} or @code{plus} operations on
8008 condition code values with less instructions than the corresponding
8009 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8010 machines, define the appropriate patterns. Use the names @code{incscc}
8011 and @code{decscc}, respectively, for the patterns which perform
8012 @code{plus} or @code{minus} operations on condition code values. See
8013 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8014 find such instruction sequences on other machines.
8016 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8019 @findex FLOAT_STORE_FLAG_VALUE
8020 @item FLOAT_STORE_FLAG_VALUE (@var{mode})
8021 A C expression that gives a non-zero @code{REAL_VALUE_TYPE} value that is
8022 returned when comparison operators with floating-point results are true.
8023 Define this macro on machine that have comparison operations that return
8024 floating-point values. If there are no such operations, do not define
8029 An alias for the machine mode for pointers. On most machines, define
8030 this to be the integer mode corresponding to the width of a hardware
8031 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8032 On some machines you must define this to be one of the partial integer
8033 modes, such as @code{PSImode}.
8035 The width of @code{Pmode} must be at least as large as the value of
8036 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8037 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8040 @findex FUNCTION_MODE
8042 An alias for the machine mode used for memory references to functions
8043 being called, in @code{call} RTL expressions. On most machines this
8044 should be @code{QImode}.
8046 @findex INTEGRATE_THRESHOLD
8047 @item INTEGRATE_THRESHOLD (@var{decl})
8048 A C expression for the maximum number of instructions above which the
8049 function @var{decl} should not be inlined. @var{decl} is a
8050 @code{FUNCTION_DECL} node.
8052 The default definition of this macro is 64 plus 8 times the number of
8053 arguments that the function accepts. Some people think a larger
8054 threshold should be used on RISC machines.
8056 @findex SCCS_DIRECTIVE
8057 @item SCCS_DIRECTIVE
8058 Define this if the preprocessor should ignore @code{#sccs} directives
8059 and print no error message.
8061 @findex NO_IMPLICIT_EXTERN_C
8062 @item NO_IMPLICIT_EXTERN_C
8063 Define this macro if the system header files support C++ as well as C.
8064 This macro inhibits the usual method of using system header files in
8065 C++, which is to pretend that the file's contents are enclosed in
8066 @samp{extern "C" @{@dots{}@}}.
8068 @findex HANDLE_PRAGMA
8069 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
8070 This macro is no longer supported. You must use
8071 @code{REGISTER_TARGET_PRAGMAS} instead.
8073 @findex REGISTER_TARGET_PRAGMAS
8076 @item REGISTER_TARGET_PRAGMAS (@var{pfile})
8077 Define this macro if you want to implement any target-specific pragmas.
8078 If defined, it is a C expression which makes a series of calls to the
8079 @code{cpp_register_pragma} and/or @code{cpp_register_pragma_space}
8080 functions. The @var{pfile} argument is the first argument to supply to
8081 these functions. The macro may also do setup required for the pragmas.
8083 The primary reason to define this macro is to provide compatibility with
8084 other compilers for the same target. In general, we discourage
8085 definition of target-specific pragmas for GCC.
8087 If the pragma can be implemented by attributes then the macro
8088 @samp{INSERT_ATTRIBUTES} might be a useful one to define as well.
8090 Preprocessor macros that appear on pragma lines are not expanded. All
8091 @samp{#pragma} directives that do not match any registered pragma are
8092 silently ignored, unless the user specifies @samp{-Wunknown-pragmas}.
8094 @deftypefun void cpp_register_pragma (cpp_reader *@var{pfile}, const char *@var{space}, const char *@var{name}, void (*@var{callback}) (cpp_reader *))
8096 Each call to @code{cpp_register_pragma} establishes one pragma. The
8097 @var{callback} routine will be called when the preprocessor encounters a
8101 #pragma [@var{space}] @var{name} @dots{}
8104 @var{space} must have been the subject of a previous call to
8105 @code{cpp_register_pragma_space}, or else be a null pointer. The
8106 callback routine receives @var{pfile} as its first argument, but must
8107 not use it for anything (this may change in the future). It may read
8108 any text after the @var{name} by making calls to @code{c_lex}. Text
8109 which is not read by the callback will be silently ignored.
8111 Note that both @var{space} and @var{name} are case sensitive.
8113 For an example use of this routine, see @file{c4x.h} and the callback
8114 routines defined in @file{c4x.c}.
8116 Note that the use of @code{c_lex} is specific to the C and C++
8117 compilers. It will not work in the Java or Fortran compilers, or any
8118 other language compilers for that matter. Thus if @code{c_lex} is going
8119 to be called from target-specific code, it must only be done so when
8120 building hte C and C++ compilers. This can be done by defining the
8121 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8122 target entry in the @code{config.gcc} file. These variables should name
8123 the target-specific, language-specific object file which contains the
8124 code that uses @code{c_lex}. Note it will also be necessary to add a
8125 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8126 how to build this object file.
8129 @deftypefun void cpp_register_pragma_space (cpp_reader *@var{pfile}, const char *@var{space})
8130 This routine establishes a namespace for pragmas, which will be
8131 registered by subsequent calls to @code{cpp_register_pragma}. For
8132 example, pragmas defined by the C standard are in the @samp{STDC}
8133 namespace, and pragmas specific to GCC are in the @samp{GCC} namespace.
8135 For an example use of this routine in a target header, see @file{v850.h}.
8138 @findex HANDLE_SYSV_PRAGMA
8141 @item HANDLE_SYSV_PRAGMA
8142 Define this macro (to a value of 1) if you want the System V style
8143 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8144 [=<value>]} to be supported by gcc.
8146 The pack pragma specifies the maximum alignment (in bytes) of fields
8147 within a structure, in much the same way as the @samp{__aligned__} and
8148 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8149 the behaviour to the default.
8151 The weak pragma only works if @code{SUPPORTS_WEAK} and
8152 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8153 of specifically named weak labels, optionally with a value.
8155 @findex HANDLE_PRAGMA_PACK_PUSH_POP
8158 @item HANDLE_PRAGMA_PACK_PUSH_POP
8159 Define this macro (to a value of 1) if you want to support the Win32
8160 style pragmas @samp{#pragma pack(push,<n>)} and @samp{#pragma
8161 pack(pop)}. The pack(push,<n>) pragma specifies the maximum alignment
8162 (in bytes) of fields within a structure, in much the same way as the
8163 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8164 pack value of zero resets the behaviour to the default. Successive
8165 invocations of this pragma cause the previous values to be stacked, so
8166 that invocations of @samp{#pragma pack(pop)} will return to the previous
8169 @findex VALID_MACHINE_DECL_ATTRIBUTE
8170 @item VALID_MACHINE_DECL_ATTRIBUTE (@var{decl}, @var{attributes}, @var{identifier}, @var{args})
8171 If defined, a C expression whose value is nonzero if @var{identifier} with
8172 arguments @var{args} is a valid machine specific attribute for @var{decl}.
8173 The attributes in @var{attributes} have previously been assigned to @var{decl}.
8175 @findex VALID_MACHINE_TYPE_ATTRIBUTE
8176 @item VALID_MACHINE_TYPE_ATTRIBUTE (@var{type}, @var{attributes}, @var{identifier}, @var{args})
8177 If defined, a C expression whose value is nonzero if @var{identifier} with
8178 arguments @var{args} is a valid machine specific attribute for @var{type}.
8179 The attributes in @var{attributes} have previously been assigned to @var{type}.
8181 @findex COMP_TYPE_ATTRIBUTES
8182 @item COMP_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
8183 If defined, a C expression whose value is zero if the attributes on
8184 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8185 and two if they are nearly compatible (which causes a warning to be
8188 @findex SET_DEFAULT_TYPE_ATTRIBUTES
8189 @item SET_DEFAULT_TYPE_ATTRIBUTES (@var{type})
8190 If defined, a C statement that assigns default attributes to
8191 newly defined @var{type}.
8193 @findex MERGE_MACHINE_TYPE_ATTRIBUTES
8194 @item MERGE_MACHINE_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
8195 Define this macro if the merging of type attributes needs special handling.
8196 If defined, the result is a list of the combined TYPE_ATTRIBUTES of
8197 @var{type1} and @var{type2}. It is assumed that comptypes has already been
8198 called and returned 1.
8200 @findex MERGE_MACHINE_DECL_ATTRIBUTES
8201 @item MERGE_MACHINE_DECL_ATTRIBUTES (@var{olddecl}, @var{newdecl})
8202 Define this macro if the merging of decl attributes needs special handling.
8203 If defined, the result is a list of the combined DECL_MACHINE_ATTRIBUTES of
8204 @var{olddecl} and @var{newdecl}. @var{newdecl} is a duplicate declaration
8205 of @var{olddecl}. Examples of when this is needed are when one attribute
8206 overrides another, or when an attribute is nullified by a subsequent
8209 @findex INSERT_ATTRIBUTES
8210 @item INSERT_ATTRIBUTES (@var{node}, @var{attr_ptr}, @var{prefix_ptr})
8211 Define this macro if you want to be able to add attributes to a decl
8212 when it is being created. This is normally useful for backends which
8213 wish to implement a pragma by using the attributes which correspond to
8214 the pragma's effect. The @var{node} argument is the decl which is being
8215 created. The @var{attr_ptr} argument is a pointer to the attribute list
8216 for this decl. The @var{prefix_ptr} is a pointer to the list of
8217 attributes that have appeared after the specifiers and modifiers of the
8218 declaration, but before the declaration proper.
8220 @findex SET_DEFAULT_DECL_ATTRIBUTES
8221 @item SET_DEFAULT_DECL_ATTRIBUTES (@var{decl}, @var{attributes})
8222 If defined, a C statement that assigns default attributes to
8223 newly defined @var{decl}.
8225 @findex DOLLARS_IN_IDENTIFIERS
8226 @item DOLLARS_IN_IDENTIFIERS
8227 Define this macro to control use of the character @samp{$} in identifier
8228 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
8229 1 is the default; there is no need to define this macro in that case.
8230 This macro controls the compiler proper; it does not affect the preprocessor.
8232 @findex NO_DOLLAR_IN_LABEL
8233 @item NO_DOLLAR_IN_LABEL
8234 Define this macro if the assembler does not accept the character
8235 @samp{$} in label names. By default constructors and destructors in
8236 G++ have @samp{$} in the identifiers. If this macro is defined,
8237 @samp{.} is used instead.
8239 @findex NO_DOT_IN_LABEL
8240 @item NO_DOT_IN_LABEL
8241 Define this macro if the assembler does not accept the character
8242 @samp{.} in label names. By default constructors and destructors in G++
8243 have names that use @samp{.}. If this macro is defined, these names
8244 are rewritten to avoid @samp{.}.
8246 @findex DEFAULT_MAIN_RETURN
8247 @item DEFAULT_MAIN_RETURN
8248 Define this macro if the target system expects every program's @code{main}
8249 function to return a standard ``success'' value by default (if no other
8250 value is explicitly returned).
8252 The definition should be a C statement (sans semicolon) to generate the
8253 appropriate rtl instructions. It is used only when compiling the end of
8258 Define this if the target system lacks the function @code{atexit}
8259 from the ISO C standard. If this macro is defined, a default definition
8260 will be provided to support C++. If @code{ON_EXIT} is not defined,
8261 a default @code{exit} function will also be provided.
8265 Define this macro if the target has another way to implement atexit
8266 functionality without replacing @code{exit}. For instance, SunOS 4 has
8267 a similar @code{on_exit} library function.
8269 The definition should be a functional macro which can be used just like
8270 the @code{atexit} function.
8274 Define this if your @code{exit} function needs to do something
8275 besides calling an external function @code{_cleanup} before
8276 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
8277 only needed if @code{NEED_ATEXIT} is defined and @code{ON_EXIT} is not
8280 @findex INSN_SETS_ARE_DELAYED
8281 @item INSN_SETS_ARE_DELAYED (@var{insn})
8282 Define this macro as a C expression that is nonzero if it is safe for the
8283 delay slot scheduler to place instructions in the delay slot of @var{insn},
8284 even if they appear to use a resource set or clobbered in @var{insn}.
8285 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8286 every @code{call_insn} has this behavior. On machines where some @code{insn}
8287 or @code{jump_insn} is really a function call and hence has this behavior,
8288 you should define this macro.
8290 You need not define this macro if it would always return zero.
8292 @findex INSN_REFERENCES_ARE_DELAYED
8293 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
8294 Define this macro as a C expression that is nonzero if it is safe for the
8295 delay slot scheduler to place instructions in the delay slot of @var{insn},
8296 even if they appear to set or clobber a resource referenced in @var{insn}.
8297 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8298 some @code{insn} or @code{jump_insn} is really a function call and its operands
8299 are registers whose use is actually in the subroutine it calls, you should
8300 define this macro. Doing so allows the delay slot scheduler to move
8301 instructions which copy arguments into the argument registers into the delay
8304 You need not define this macro if it would always return zero.
8306 @findex MACHINE_DEPENDENT_REORG
8307 @item MACHINE_DEPENDENT_REORG (@var{insn})
8308 In rare cases, correct code generation requires extra machine
8309 dependent processing between the second jump optimization pass and
8310 delayed branch scheduling. On those machines, define this macro as a C
8311 statement to act on the code starting at @var{insn}.
8313 @findex MULTIPLE_SYMBOL_SPACES
8314 @item MULTIPLE_SYMBOL_SPACES
8315 Define this macro if in some cases global symbols from one translation
8316 unit may not be bound to undefined symbols in another translation unit
8317 without user intervention. For instance, under Microsoft Windows
8318 symbols must be explicitly imported from shared libraries (DLLs).
8320 @findex MD_ASM_CLOBBERS
8321 @item MD_ASM_CLOBBERS
8322 A C statement that adds to @var{CLOBBERS} @code{STRING_CST} trees for
8323 any hard regs the port wishes to automatically clobber for all asms.
8327 A C expression that returns how many instructions can be issued at the
8328 same time if the machine is a superscalar machine.
8330 @findex MD_SCHED_INIT
8331 @item MD_SCHED_INIT (@var{file}, @var{verbose}, @var{max_ready})
8332 A C statement which is executed by the scheduler at the
8333 beginning of each block of instructions that are to be scheduled.
8334 @var{file} is either a null pointer, or a stdio stream to write any
8335 debug output to. @var{verbose} is the verbose level provided by
8336 @samp{-fsched-verbose-}@var{n}. @var{max_ready} is the maximum number
8337 of insns in the current scheduling region that can be live at the same
8338 time. This can be used to allocate scratch space if it is needed.
8340 @findex MD_SCHED_FINISH
8341 @item MD_SCHED_FINISH (@var{file}, @var{verbose})
8342 A C statement which is executed by the scheduler at the end of each block
8343 of instructions that are to be scheduled. It can be used to perform
8344 cleanup of any actions done by the other scheduling macros.
8345 @var{file} is either a null pointer, or a stdio stream to write any
8346 debug output to. @var{verbose} is the verbose level provided by
8347 @samp{-fsched-verbose-}@var{n}.
8349 @findex MD_SCHED_REORDER
8350 @item MD_SCHED_REORDER (@var{file}, @var{verbose}, @var{ready}, @var{n_ready}, @var{clock}, @var{can_issue_more})
8351 A C statement which is executed by the scheduler after it
8352 has scheduled the ready list to allow the machine description to reorder
8353 it (for example to combine two small instructions together on
8354 @samp{VLIW} machines). @var{file} is either a null pointer, or a stdio
8355 stream to write any debug output to. @var{verbose} is the verbose level
8356 provided by @samp{-fsched-verbose-}@var{n}. @var{ready} is a pointer to
8357 the ready list of instructions that are ready to be scheduled.
8358 @var{n_ready} is the number of elements in the ready list. The
8359 scheduler reads the ready list in reverse order, starting with
8360 @var{ready}[@var{n_ready}-1] and going to @var{ready}[0]. @var{clock}
8361 is the timer tick of the scheduler. @var{can_issue_more} is an output
8362 parameter that is set to the number of insns that can issue this clock;
8363 normally this is just @code{issue_rate}. See also @samp{MD_SCHED_REORDER2}.
8365 @findex MD_SCHED_REORDER2
8366 @item MD_SCHED_REORDER2 (@var{file}, @var{verbose}, @var{ready}, @var{n_ready}, @var{clock}, @var{can_issue_more})
8367 Like @samp{MD_SCHED_REORDER}, but called at a different time. While the
8368 @samp{MD_SCHED_REORDER} macro is called whenever the scheduler starts a
8369 new cycle, this macro is used immediately after @samp{MD_SCHED_VARIABLE_ISSUE}
8370 is called; it can reorder the ready list and set @var{can_issue_more} to
8371 determine whether there are more insns to be scheduled in the same cycle.
8372 Defining this macro can be useful if there are frequent situations where
8373 scheduling one insn causes other insns to become ready in the same cycle,
8374 these other insns can then be taken into account properly.
8376 @findex MD_SCHED_VARIABLE_ISSUE
8377 @item MD_SCHED_VARIABLE_ISSUE (@var{file}, @var{verbose}, @var{insn}, @var{more})
8378 A C statement which is executed by the scheduler after it
8379 has scheduled an insn from the ready list. @var{file} is either a null
8380 pointer, or a stdio stream to write any debug output to. @var{verbose}
8381 is the verbose level provided by @samp{-fsched-verbose-}@var{n}.
8382 @var{insn} is the instruction that was scheduled. @var{more} is the
8383 number of instructions that can be issued in the current cycle. The
8384 @samp{MD_SCHED_VARIABLE_ISSUE} macro is responsible for updating the
8385 value of @var{more} (typically by @var{more}--).
8387 @findex MAX_INTEGER_COMPUTATION_MODE
8388 @item MAX_INTEGER_COMPUTATION_MODE
8389 Define this to the largest integer machine mode which can be used for
8390 operations other than load, store and copy operations.
8392 You need only define this macro if the target holds values larger than
8393 @code{word_mode} in general purpose registers. Most targets should not define
8396 @findex MATH_LIBRARY
8398 Define this macro as a C string constant for the linker argument to link
8399 in the system math library, or @samp{""} if the target does not have a
8400 separate math library.
8402 You need only define this macro if the default of @samp{"-lm"} is wrong.
8404 @findex LIBRARY_PATH_ENV
8405 @item LIBRARY_PATH_ENV
8406 Define this macro as a C string constant for the environment variable that
8407 specifies where the linker should look for libraries.
8409 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
8412 @findex TARGET_HAS_F_SETLKW
8413 @item TARGET_HAS_F_SETLKW
8414 Define this macro if the target supports file locking with fcntl / F_SETLKW.
8415 Note that this functionality is part of POSIX.
8416 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
8417 to use file locking when exiting a program, which avoids race conditions
8418 if the program has forked.
8420 @findex MAX_CONDITIONAL_EXECUTE
8421 @item MAX_CONDITIONAL_EXECUTE
8423 A C expression for the maximum number of instructions to execute via
8424 conditional execution instructions instead of a branch. A value of
8425 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8426 1 if it does use cc0.
8428 @findex IFCVT_MODIFY_TESTS
8429 @item IFCVT_MODIFY_TESTS
8430 A C expression to modify the tests in @code{TRUE_EXPR}, and
8431 @code{FALSE_EXPPR} for use in converting insns in @code{TEST_BB},
8432 @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB} basic blocks to
8433 conditional execution. Set either @code{TRUE_EXPR} or @code{FALSE_EXPR}
8434 to a null pointer if the tests cannot be converted.
8436 @findex IFCVT_MODIFY_INSN
8437 @item IFCVT_MODIFY_INSN
8438 A C expression to modify the @code{PATTERN} of an @code{INSN} that is to
8439 be converted to conditional execution format.
8441 @findex IFCVT_MODIFY_FINAL
8442 @item IFCVT_MODIFY_FINAL
8443 A C expression to perform any final machine dependent modifications in
8444 converting code to conditional execution in the basic blocks
8445 @code{TEST_BB}, @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB}.
8447 @findex IFCVT_MODIFY_CANCEL
8448 @item IFCVT_MODIFY_CANCEL
8449 A C expression to cancel any machine dependent modifications in
8450 converting code to conditional execution in the basic blocks
8451 @code{TEST_BB}, @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB}.
8453 @findex MD_INIT_BUILTINS
8454 @item MD_INIT_BUILTINS
8455 Define this macro if you have any machine-specific builtin functions that
8456 need to be defined. It should be a C expression that performs the
8459 Machine specific builtins can be useful to expand special machine
8460 instructions that would otherwise not normally be generated because
8461 they have no equivalent in the source language (for example, SIMD vector
8462 instructions or prefetch instructions).
8464 To create a builtin function, call the function @code{builtin_function}
8465 which is defined by the language frontend. You can use any type nodes set
8466 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
8467 only language frontends that use these two functions will use
8468 @samp{MD_INIT_BUILTINS}.
8470 @findex MD_EXPAND_BUILTIN
8471 @item MD_EXPAND_BUILTIN(@var{exp}, @var{target}, @var{subtarget}, @var{mode}, @var{ignore})
8473 Expand a call to a machine specific builtin that was set up by
8474 @samp{MD_INIT_BUILTINS}. @var{exp} is the expression for the function call;
8475 the result should go to @var{target} if that is convenient, and have mode
8476 @var{mode} if that is convenient. @var{subtarget} may be used as the target
8477 for computing one of @var{exp}'s operands. @var{ignore} is nonzero if the value
8479 This macro should return the result of the call to the builtin.