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 LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
329 @item LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
330 A nonzero value causes collect2 to remove duplicate -L<directory> search
331 directories from linking commands. Do not give it a nonzero value if
332 removing duplicate search directories changes the linker's semantics.
334 @findex MULTILIB_DEFAULTS
335 @item MULTILIB_DEFAULTS
336 Define this macro as a C expression for the initializer of an array of
337 string to tell the driver program which options are defaults for this
338 target and thus do not need to be handled specially when using
339 @code{MULTILIB_OPTIONS}.
341 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
342 the target makefile fragment or if none of the options listed in
343 @code{MULTILIB_OPTIONS} are set by default.
344 @xref{Target Fragment}.
346 @findex RELATIVE_PREFIX_NOT_LINKDIR
347 @item RELATIVE_PREFIX_NOT_LINKDIR
348 Define this macro to tell @code{gcc} that it should only translate
349 a @samp{-B} prefix into a @samp{-L} linker option if the prefix
350 indicates an absolute file name.
352 @findex STANDARD_EXEC_PREFIX
353 @item STANDARD_EXEC_PREFIX
354 Define this macro as a C string constant if you wish to override the
355 standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
356 try when searching for the executable files of the compiler.
358 @findex MD_EXEC_PREFIX
360 If defined, this macro is an additional prefix to try after
361 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
362 when the @samp{-b} option is used, or the compiler is built as a cross
363 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
364 to the list of directories used to find the assembler in @file{configure.in}.
366 @findex STANDARD_STARTFILE_PREFIX
367 @item STANDARD_STARTFILE_PREFIX
368 Define this macro as a C string constant if you wish to override the
369 standard choice of @file{/usr/local/lib/} as the default prefix to
370 try when searching for startup files such as @file{crt0.o}.
372 @findex MD_STARTFILE_PREFIX
373 @item MD_STARTFILE_PREFIX
374 If defined, this macro supplies an additional prefix to try after the
375 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
376 @samp{-b} option is used, or when the compiler is built as a cross
379 @findex MD_STARTFILE_PREFIX_1
380 @item MD_STARTFILE_PREFIX_1
381 If defined, this macro supplies yet another prefix to try after the
382 standard prefixes. It is not searched when the @samp{-b} option is
383 used, or when the compiler is built as a cross compiler.
385 @findex INIT_ENVIRONMENT
386 @item INIT_ENVIRONMENT
387 Define this macro as a C string constant if you wish to set environment
388 variables for programs called by the driver, such as the assembler and
389 loader. The driver passes the value of this macro to @code{putenv} to
390 initialize the necessary environment variables.
392 @findex LOCAL_INCLUDE_DIR
393 @item LOCAL_INCLUDE_DIR
394 Define this macro as a C string constant if you wish to override the
395 standard choice of @file{/usr/local/include} as the default prefix to
396 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
397 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
399 Cross compilers do not use this macro and do not search either
400 @file{/usr/local/include} or its replacement.
402 @findex MODIFY_TARGET_NAME
403 @item MODIFY_TARGET_NAME
404 Define this macro if you with to define command-line switches that modify the
407 For each switch, you can include a string to be appended to the first
408 part of the configuration name or a string to be deleted from the
409 configuration name, if present. The definition should be an initializer
410 for an array of structures. Each array element should have three
411 elements: the switch name (a string constant, including the initial
412 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
413 indicate whether the string should be inserted or deleted, and the string
414 to be inserted or deleted (a string constant).
416 For example, on a machine where @samp{64} at the end of the
417 configuration name denotes a 64-bit target and you want the @samp{-32}
418 and @samp{-64} switches to select between 32- and 64-bit targets, you would
422 #define MODIFY_TARGET_NAME \
423 @{ @{ "-32", DELETE, "64"@}, \
424 @{"-64", ADD, "64"@}@}
428 @findex SYSTEM_INCLUDE_DIR
429 @item SYSTEM_INCLUDE_DIR
430 Define this macro as a C string constant if you wish to specify a
431 system-specific directory to search for header files before the standard
432 directory. @code{SYSTEM_INCLUDE_DIR} comes before
433 @code{STANDARD_INCLUDE_DIR} in the search order.
435 Cross compilers do not use this macro and do not search the directory
438 @findex STANDARD_INCLUDE_DIR
439 @item STANDARD_INCLUDE_DIR
440 Define this macro as a C string constant if you wish to override the
441 standard choice of @file{/usr/include} as the default prefix to
442 try when searching for header files.
444 Cross compilers do not use this macro and do not search either
445 @file{/usr/include} or its replacement.
447 @findex STANDARD_INCLUDE_COMPONENT
448 @item STANDARD_INCLUDE_COMPONENT
449 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
450 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
451 If you do not define this macro, no component is used.
453 @findex INCLUDE_DEFAULTS
454 @item INCLUDE_DEFAULTS
455 Define this macro if you wish to override the entire default search path
456 for include files. For a native compiler, the default search path
457 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
458 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
459 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
460 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
461 and specify private search areas for GCC. The directory
462 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
464 The definition should be an initializer for an array of structures.
465 Each array element should have four elements: the directory name (a
466 string constant), the component name (also a string constant), a flag
467 for C++-only directories,
468 and a flag showing that the includes in the directory don't need to be
469 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
470 the array with a null element.
472 The component name denotes what GNU package the include file is part of,
473 if any, in all upper-case letters. For example, it might be @samp{GCC}
474 or @samp{BINUTILS}. If the package is part of a vendor-supplied
475 operating system, code the component name as @samp{0}.
477 For example, here is the definition used for VAX/VMS:
480 #define INCLUDE_DEFAULTS \
482 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
483 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
484 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
491 Here is the order of prefixes tried for exec files:
495 Any prefixes specified by the user with @samp{-B}.
498 The environment variable @code{GCC_EXEC_PREFIX}, if any.
501 The directories specified by the environment variable @code{COMPILER_PATH}.
504 The macro @code{STANDARD_EXEC_PREFIX}.
507 @file{/usr/lib/gcc/}.
510 The macro @code{MD_EXEC_PREFIX}, if any.
513 Here is the order of prefixes tried for startfiles:
517 Any prefixes specified by the user with @samp{-B}.
520 The environment variable @code{GCC_EXEC_PREFIX}, if any.
523 The directories specified by the environment variable @code{LIBRARY_PATH}
524 (or port-specific name; native only, cross compilers do not use this).
527 The macro @code{STANDARD_EXEC_PREFIX}.
530 @file{/usr/lib/gcc/}.
533 The macro @code{MD_EXEC_PREFIX}, if any.
536 The macro @code{MD_STARTFILE_PREFIX}, if any.
539 The macro @code{STANDARD_STARTFILE_PREFIX}.
548 @node Run-time Target
549 @section Run-time Target Specification
550 @cindex run-time target specification
551 @cindex predefined macros
552 @cindex target specifications
554 @c prevent bad page break with this line
555 Here are run-time target specifications.
558 @findex CPP_PREDEFINES
560 Define this to be a string constant containing @samp{-D} options to
561 define the predefined macros that identify this machine and system.
562 These macros will be predefined unless the @option{-ansi} option (or a
563 @option{-std} option for strict ISO C conformance) is specified.
565 In addition, a parallel set of macros are predefined, whose names are
566 made by appending @samp{__} at the beginning and at the end. These
567 @samp{__} macros are permitted by the ISO standard, so they are
568 predefined regardless of whether @option{-ansi} or a @option{-std} option
571 For example, on the Sun, one can use the following value:
574 "-Dmc68000 -Dsun -Dunix"
577 The result is to define the macros @code{__mc68000__}, @code{__sun__}
578 and @code{__unix__} unconditionally, and the macros @code{mc68000},
579 @code{sun} and @code{unix} provided @samp{-ansi} is not specified.
581 @findex extern int target_flags
582 @item extern int target_flags;
583 This declaration should be present.
585 @cindex optional hardware or system features
586 @cindex features, optional, in system conventions
588 This series of macros is to allow compiler command arguments to
589 enable or disable the use of optional features of the target machine.
590 For example, one machine description serves both the 68000 and
591 the 68020; a command argument tells the compiler whether it should
592 use 68020-only instructions or not. This command argument works
593 by means of a macro @code{TARGET_68020} that tests a bit in
596 Define a macro @code{TARGET_@var{featurename}} for each such option.
597 Its definition should test a bit in @code{target_flags}. It is
598 recommended that a helper macro @code{TARGET_MASK_@var{featurename}}
599 is defined for each bit-value to test, and used in
600 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
604 #define TARGET_MASK_68020 1
605 #define TARGET_68020 (target_flags & TARGET_MASK_68020)
608 One place where these macros are used is in the condition-expressions
609 of instruction patterns. Note how @code{TARGET_68020} appears
610 frequently in the 68000 machine description file, @file{m68k.md}.
611 Another place they are used is in the definitions of the other
612 macros in the @file{@var{machine}.h} file.
614 @findex TARGET_SWITCHES
615 @item TARGET_SWITCHES
616 This macro defines names of command options to set and clear
617 bits in @code{target_flags}. Its definition is an initializer
618 with a subgrouping for each command option.
620 Each subgrouping contains a string constant, that defines the option
621 name, a number, which contains the bits to set in
622 @code{target_flags}, and a second string which is the description
623 displayed by --help. If the number is negative then the bits specified
624 by the number are cleared instead of being set. If the description
625 string is present but empty, then no help information will be displayed
626 for that option, but it will not count as an undocumented option. The
627 actual option name is made by appending @samp{-m} to the specified name.
629 One of the subgroupings should have a null string. The number in
630 this grouping is the default value for @code{target_flags}. Any
631 target options act starting with that value.
633 Here is an example which defines @samp{-m68000} and @samp{-m68020}
634 with opposite meanings, and picks the latter as the default:
637 #define TARGET_SWITCHES \
638 @{ @{ "68020", TARGET_MASK_68020, "" @}, \
639 @{ "68000", -TARGET_MASK_68020, "Compile for the 68000" @}, \
640 @{ "", TARGET_MASK_68020, "" @}@}
643 @findex TARGET_OPTIONS
645 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
646 options that have values. Its definition is an initializer with a
647 subgrouping for each command option.
649 Each subgrouping contains a string constant, that defines the fixed part
650 of the option name, the address of a variable, and a description string.
651 The variable, type @code{char *}, is set to the variable part of the
652 given option if the fixed part matches. The actual option name is made
653 by appending @samp{-m} to the specified name.
655 Here is an example which defines @samp{-mshort-data-@var{number}}. If the
656 given option is @samp{-mshort-data-512}, the variable @code{m88k_short_data}
657 will be set to the string @code{"512"}.
660 extern char *m88k_short_data;
661 #define TARGET_OPTIONS \
662 @{ @{ "short-data-", &m88k_short_data, "Specify the size of the short data section" @} @}
665 @findex TARGET_VERSION
667 This macro is a C statement to print on @code{stderr} a string
668 describing the particular machine description choice. Every machine
669 description should define @code{TARGET_VERSION}. For example:
673 #define TARGET_VERSION \
674 fprintf (stderr, " (68k, Motorola syntax)");
676 #define TARGET_VERSION \
677 fprintf (stderr, " (68k, MIT syntax)");
681 @findex OVERRIDE_OPTIONS
682 @item OVERRIDE_OPTIONS
683 Sometimes certain combinations of command options do not make sense on
684 a particular target machine. You can define a macro
685 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
686 defined, is executed once just after all the command options have been
689 Don't use this macro to turn on various extra optimizations for
690 @samp{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
692 @findex OPTIMIZATION_OPTIONS
693 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
694 Some machines may desire to change what optimizations are performed for
695 various optimization levels. This macro, if defined, is executed once
696 just after the optimization level is determined and before the remainder
697 of the command options have been parsed. Values set in this macro are
698 used as the default values for the other command line options.
700 @var{level} is the optimization level specified; 2 if @samp{-O2} is
701 specified, 1 if @samp{-O} is specified, and 0 if neither is specified.
703 @var{size} is non-zero if @samp{-Os} is specified and zero otherwise.
705 You should not use this macro to change options that are not
706 machine-specific. These should uniformly selected by the same
707 optimization level on all supported machines. Use this macro to enable
708 machine-specific optimizations.
710 @strong{Do not examine @code{write_symbols} in
711 this macro!} The debugging options are not supposed to alter the
714 @findex CAN_DEBUG_WITHOUT_FP
715 @item CAN_DEBUG_WITHOUT_FP
716 Define this macro if debugging can be performed even without a frame
717 pointer. If this macro is defined, GCC will turn on the
718 @samp{-fomit-frame-pointer} option whenever @samp{-O} is specified.
721 @node Per-Function Data
722 @section Defining data structures for per-function information.
723 @cindex per-function data
724 @cindex data structures
726 If the target needs to store information on a per-function basis, GCC
727 provides a macro and a couple of variables to allow this. Note, just
728 using statics to store the information is a bad idea, since GCC supports
729 nested functions, so you can be halfway through encoding one function
730 when another one comes along.
732 GCC defines a data structure called @code{struct function} which
733 contains all of the data specific to an individual function. This
734 structure contains a field called @code{machine} whose type is
735 @code{struct machine_function *}, which can be used by targets to point
736 to their own specific data.
738 If a target needs per-function specific data it should define the type
739 @code{struct machine_function} and also the macro
740 @code{INIT_EXPANDERS}. This macro should be used to initialise some or
741 all of the function pointers @code{init_machine_status},
742 @code{free_machine_status} and @code{mark_machine_status}. These
743 pointers are explained below.
745 One typical use of per-function, target specific data is to create an
746 RTX to hold the register containing the function's return address. This
747 RTX can then be used to implement the @code{__builtin_return_address}
748 function, for level 0.
750 Note - earlier implementations of GCC used a single data area to hold
751 all of the per-function information. Thus when processing of a nested
752 function began the old per-function data had to be pushed onto a
753 stack, and when the processing was finished, it had to be popped off the
754 stack. GCC used to provide function pointers called
755 @code{save_machine_status} and @code{restore_machine_status} to handle
756 the saving and restoring of the target specific information. Since the
757 single data area approach is no longer used, these pointers are no
760 The macro and function pointers are described below.
763 @findex INIT_EXPANDERS
765 Macro called to initialise any target specific information. This macro
766 is called once per function, before generation of any RTL has begun.
767 The intention of this macro is to allow the initialisation of the
768 function pointers below.
770 @findex init_machine_status
771 @item init_machine_status
772 This is a @code{void (*)(struct function *)} function pointer. If this
773 pointer is non-NULL it will be called once per function, before function
774 compilation starts, in order to allow the target to perform any target
775 specific initialisation of the @code{struct function} structure. It is
776 intended that this would be used to initialise the @code{machine} of
779 @findex free_machine_status
780 @item free_machine_status
781 This is a @code{void (*)(struct function *)} function pointer. If this
782 pointer is non-NULL it will be called once per function, after the
783 function has been compiled, in order to allow any memory allocated
784 during the @code{init_machine_status} function call to be freed.
786 @findex mark_machine_status
787 @item mark_machine_status
788 This is a @code{void (*)(struct function *)} function pointer. If this
789 pointer is non-NULL it will be called once per function in order to mark
790 any data items in the @code{struct machine_function} structure which
791 need garbage collection.
796 @section Storage Layout
797 @cindex storage layout
799 Note that the definitions of the macros in this table which are sizes or
800 alignments measured in bits do not need to be constant. They can be C
801 expressions that refer to static variables, such as the @code{target_flags}.
802 @xref{Run-time Target}.
805 @findex BITS_BIG_ENDIAN
806 @item BITS_BIG_ENDIAN
807 Define this macro to have the value 1 if the most significant bit in a
808 byte has the lowest number; otherwise define it to have the value zero.
809 This means that bit-field instructions count from the most significant
810 bit. If the machine has no bit-field instructions, then this must still
811 be defined, but it doesn't matter which value it is defined to. This
812 macro need not be a constant.
814 This macro does not affect the way structure fields are packed into
815 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
817 @findex BYTES_BIG_ENDIAN
818 @item BYTES_BIG_ENDIAN
819 Define this macro to have the value 1 if the most significant byte in a
820 word has the lowest number. This macro need not be a constant.
822 @findex WORDS_BIG_ENDIAN
823 @item WORDS_BIG_ENDIAN
824 Define this macro to have the value 1 if, in a multiword object, the
825 most significant word has the lowest number. This applies to both
826 memory locations and registers; GCC fundamentally assumes that the
827 order of words in memory is the same as the order in registers. This
828 macro need not be a constant.
830 @findex LIBGCC2_WORDS_BIG_ENDIAN
831 @item LIBGCC2_WORDS_BIG_ENDIAN
832 Define this macro if WORDS_BIG_ENDIAN is not constant. This must be a
833 constant value with the same meaning as WORDS_BIG_ENDIAN, which will be
834 used only when compiling libgcc2.c. Typically the value will be set
835 based on preprocessor defines.
837 @findex FLOAT_WORDS_BIG_ENDIAN
838 @item FLOAT_WORDS_BIG_ENDIAN
839 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
840 @code{TFmode} floating point numbers are stored in memory with the word
841 containing the sign bit at the lowest address; otherwise define it to
842 have the value 0. This macro need not be a constant.
844 You need not define this macro if the ordering is the same as for
847 @findex BITS_PER_UNIT
849 Define this macro to be the number of bits in an addressable storage
850 unit (byte); normally 8.
852 @findex BITS_PER_WORD
854 Number of bits in a word; normally 32.
856 @findex MAX_BITS_PER_WORD
857 @item MAX_BITS_PER_WORD
858 Maximum number of bits in a word. If this is undefined, the default is
859 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
860 largest value that @code{BITS_PER_WORD} can have at run-time.
862 @findex UNITS_PER_WORD
864 Number of storage units in a word; normally 4.
866 @findex MIN_UNITS_PER_WORD
867 @item MIN_UNITS_PER_WORD
868 Minimum number of units in a word. If this is undefined, the default is
869 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
870 smallest value that @code{UNITS_PER_WORD} can have at run-time.
874 Width of a pointer, in bits. You must specify a value no wider than the
875 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
876 you must define @code{POINTERS_EXTEND_UNSIGNED}.
878 @findex POINTERS_EXTEND_UNSIGNED
879 @item POINTERS_EXTEND_UNSIGNED
880 A C expression whose value is nonzero if pointers that need to be
881 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
882 be zero-extended and zero if they are to be sign-extended.
884 You need not define this macro if the @code{POINTER_SIZE} is equal
885 to the width of @code{Pmode}.
888 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
889 A macro to update @var{m} and @var{unsignedp} when an object whose type
890 is @var{type} and which has the specified mode and signedness is to be
891 stored in a register. This macro is only called when @var{type} is a
894 On most RISC machines, which only have operations that operate on a full
895 register, define this macro to set @var{m} to @code{word_mode} if
896 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
897 cases, only integer modes should be widened because wider-precision
898 floating-point operations are usually more expensive than their narrower
901 For most machines, the macro definition does not change @var{unsignedp}.
902 However, some machines, have instructions that preferentially handle
903 either signed or unsigned quantities of certain modes. For example, on
904 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
905 sign-extend the result to 64 bits. On such machines, set
906 @var{unsignedp} according to which kind of extension is more efficient.
908 Do not define this macro if it would never modify @var{m}.
910 @findex PROMOTE_FUNCTION_ARGS
911 @item PROMOTE_FUNCTION_ARGS
912 Define this macro if the promotion described by @code{PROMOTE_MODE}
913 should also be done for outgoing function arguments.
915 @findex PROMOTE_FUNCTION_RETURN
916 @item PROMOTE_FUNCTION_RETURN
917 Define this macro if the promotion described by @code{PROMOTE_MODE}
918 should also be done for the return value of functions.
920 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
921 promotions done by @code{PROMOTE_MODE}.
923 @findex PROMOTE_FOR_CALL_ONLY
924 @item PROMOTE_FOR_CALL_ONLY
925 Define this macro if the promotion described by @code{PROMOTE_MODE}
926 should @emph{only} be performed for outgoing function arguments or
927 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
928 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
930 @findex PARM_BOUNDARY
932 Normal alignment required for function parameters on the stack, in
933 bits. All stack parameters receive at least this much alignment
934 regardless of data type. On most machines, this is the same as the
937 @findex STACK_BOUNDARY
939 Define this macro if there is a guaranteed alignment for the stack
940 pointer on this machine. The definition is a C expression
941 for the desired alignment (measured in bits). This value is used as a
942 default if PREFERRED_STACK_BOUNDARY is not defined.
944 @findex PREFERRED_STACK_BOUNDARY
945 @item PREFERRED_STACK_BOUNDARY
946 Define this macro if you wish to preserve a certain alignment for
947 the stack pointer. The definition is a C expression
948 for the desired alignment (measured in bits). If STACK_BOUNDARY is
949 also defined, this macro must evaluate to a value equal to or larger
952 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
953 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
954 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
955 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
956 be momentarily unaligned while pushing arguments.
958 @findex FUNCTION_BOUNDARY
959 @item FUNCTION_BOUNDARY
960 Alignment required for a function entry point, in bits.
962 @findex BIGGEST_ALIGNMENT
963 @item BIGGEST_ALIGNMENT
964 Biggest alignment that any data type can require on this machine, in bits.
966 @findex MINIMUM_ATOMIC_ALIGNMENT
967 @item MINIMUM_ATOMIC_ALIGNMENT
968 If defined, the smallest alignment, in bits, that can be given to an
969 object that can be referenced in one operation, without disturbing any
970 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
971 on machines that don't have byte or half-word store operations.
973 @findex BIGGEST_FIELD_ALIGNMENT
974 @item BIGGEST_FIELD_ALIGNMENT
975 Biggest alignment that any structure or union field can require on this
976 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
977 structure and union fields only, unless the field alignment has been set
978 by the @code{__attribute__ ((aligned (@var{n})))} construct.
980 @findex ADJUST_FIELD_ALIGN
981 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
982 An expression for the alignment of a structure field @var{field} if the
983 alignment computed in the usual way is @var{computed}. GCC uses
984 this value instead of the value in @code{BIGGEST_ALIGNMENT} or
985 @code{BIGGEST_FIELD_ALIGNMENT}, if defined, for structure fields only.
987 @findex MAX_OFILE_ALIGNMENT
988 @item MAX_OFILE_ALIGNMENT
989 Biggest alignment supported by the object file format of this machine.
990 Use this macro to limit the alignment which can be specified using the
991 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
992 the default value is @code{BIGGEST_ALIGNMENT}.
994 @findex DATA_ALIGNMENT
995 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
996 If defined, a C expression to compute the alignment for a variable in
997 the static store. @var{type} is the data type, and @var{basic-align} is
998 the alignment that the object would ordinarily have. The value of this
999 macro is used instead of that alignment to align the object.
1001 If this macro is not defined, then @var{basic-align} is used.
1004 One use of this macro is to increase alignment of medium-size data to
1005 make it all fit in fewer cache lines. Another is to cause character
1006 arrays to be word-aligned so that @code{strcpy} calls that copy
1007 constants to character arrays can be done inline.
1009 @findex CONSTANT_ALIGNMENT
1010 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1011 If defined, a C expression to compute the alignment given to a constant
1012 that is being placed in memory. @var{constant} is the constant and
1013 @var{basic-align} is the alignment that the object would ordinarily
1014 have. The value of this macro is used instead of that alignment to
1017 If this macro is not defined, then @var{basic-align} is used.
1019 The typical use of this macro is to increase alignment for string
1020 constants to be word aligned so that @code{strcpy} calls that copy
1021 constants can be done inline.
1023 @findex LOCAL_ALIGNMENT
1024 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1025 If defined, a C expression to compute the alignment for a variable in
1026 the local store. @var{type} is the data type, and @var{basic-align} is
1027 the alignment that the object would ordinarily have. The value of this
1028 macro is used instead of that alignment to align the object.
1030 If this macro is not defined, then @var{basic-align} is used.
1032 One use of this macro is to increase alignment of medium-size data to
1033 make it all fit in fewer cache lines.
1035 @findex EMPTY_FIELD_BOUNDARY
1036 @item EMPTY_FIELD_BOUNDARY
1037 Alignment in bits to be given to a structure bit field that follows an
1038 empty field such as @code{int : 0;}.
1040 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
1041 that results from an empty field.
1043 @findex STRUCTURE_SIZE_BOUNDARY
1044 @item STRUCTURE_SIZE_BOUNDARY
1045 Number of bits which any structure or union's size must be a multiple of.
1046 Each structure or union's size is rounded up to a multiple of this.
1048 If you do not define this macro, the default is the same as
1049 @code{BITS_PER_UNIT}.
1051 @findex STRICT_ALIGNMENT
1052 @item STRICT_ALIGNMENT
1053 Define this macro to be the value 1 if instructions will fail to work
1054 if given data not on the nominal alignment. If instructions will merely
1055 go slower in that case, define this macro as 0.
1057 @findex PCC_BITFIELD_TYPE_MATTERS
1058 @item PCC_BITFIELD_TYPE_MATTERS
1059 Define this if you wish to imitate the way many other C compilers handle
1060 alignment of bitfields and the structures that contain them.
1062 The behavior is that the type written for a bitfield (@code{int},
1063 @code{short}, or other integer type) imposes an alignment for the
1064 entire structure, as if the structure really did contain an ordinary
1065 field of that type. In addition, the bitfield is placed within the
1066 structure so that it would fit within such a field, not crossing a
1069 Thus, on most machines, a bitfield whose type is written as @code{int}
1070 would not cross a four-byte boundary, and would force four-byte
1071 alignment for the whole structure. (The alignment used may not be four
1072 bytes; it is controlled by the other alignment parameters.)
1074 If the macro is defined, its definition should be a C expression;
1075 a nonzero value for the expression enables this behavior.
1077 Note that if this macro is not defined, or its value is zero, some
1078 bitfields may cross more than one alignment boundary. The compiler can
1079 support such references if there are @samp{insv}, @samp{extv}, and
1080 @samp{extzv} insns that can directly reference memory.
1082 The other known way of making bitfields work is to define
1083 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1084 Then every structure can be accessed with fullwords.
1086 Unless the machine has bitfield instructions or you define
1087 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1088 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1090 If your aim is to make GCC use the same conventions for laying out
1091 bitfields as are used by another compiler, here is how to investigate
1092 what the other compiler does. Compile and run this program:
1111 printf ("Size of foo1 is %d\n",
1112 sizeof (struct foo1));
1113 printf ("Size of foo2 is %d\n",
1114 sizeof (struct foo2));
1119 If this prints 2 and 5, then the compiler's behavior is what you would
1120 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1122 @findex BITFIELD_NBYTES_LIMITED
1123 @item BITFIELD_NBYTES_LIMITED
1124 Like PCC_BITFIELD_TYPE_MATTERS except that its effect is limited to
1125 aligning a bitfield within the structure.
1127 @findex MEMBER_TYPE_FORCES_BLK
1128 @item MEMBER_TYPE_FORCES_BLK (@var{field})
1129 Return 1 if a structure or array containing @var{field} should be accessed using
1132 Normally, this is not needed. See the file @file{c4x.h} for an example
1133 of how to use this macro to prevent a structure having a floating point
1134 field from being accessed in an integer mode.
1136 @findex ROUND_TYPE_SIZE
1137 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
1138 Define this macro as an expression for the overall size of a type
1139 (given by @var{type} as a tree node) when the size computed in the
1140 usual way is @var{computed} and the alignment is @var{specified}.
1142 The default is to round @var{computed} up to a multiple of @var{specified}.
1144 @findex ROUND_TYPE_SIZE_UNIT
1145 @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified})
1146 Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are
1147 specified in units (bytes). If you define @code{ROUND_TYPE_SIZE},
1148 you must also define this macro and they must be defined consistently
1151 @findex ROUND_TYPE_ALIGN
1152 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1153 Define this macro as an expression for the alignment of a type (given
1154 by @var{type} as a tree node) if the alignment computed in the usual
1155 way is @var{computed} and the alignment explicitly specified was
1158 The default is to use @var{specified} if it is larger; otherwise, use
1159 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1161 @findex MAX_FIXED_MODE_SIZE
1162 @item MAX_FIXED_MODE_SIZE
1163 An integer expression for the size in bits of the largest integer
1164 machine mode that should actually be used. All integer machine modes of
1165 this size or smaller can be used for structures and unions with the
1166 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1167 (DImode)} is assumed.
1169 @findex VECTOR_MODE_SUPPORTED_P
1170 @item VECTOR_MODE_SUPPORTED_P(@var{mode})
1171 Define this macro to be nonzero if the port is prepared to handle insns
1172 involving vector mode @var{mode}. At the very least, it must have move
1173 patterns for this mode.
1175 @findex STACK_SAVEAREA_MODE
1176 @item STACK_SAVEAREA_MODE (@var{save_level})
1177 If defined, an expression of type @code{enum machine_mode} that
1178 specifies the mode of the save area operand of a
1179 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1180 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1181 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1182 having its mode specified.
1184 You need not define this macro if it always returns @code{Pmode}. You
1185 would most commonly define this macro if the
1186 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1189 @findex STACK_SIZE_MODE
1190 @item STACK_SIZE_MODE
1191 If defined, an expression of type @code{enum machine_mode} that
1192 specifies the mode of the size increment operand of an
1193 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1195 You need not define this macro if it always returns @code{word_mode}.
1196 You would most commonly define this macro if the @code{allocate_stack}
1197 pattern needs to support both a 32- and a 64-bit mode.
1199 @findex CHECK_FLOAT_VALUE
1200 @item CHECK_FLOAT_VALUE (@var{mode}, @var{value}, @var{overflow})
1201 A C statement to validate the value @var{value} (of type
1202 @code{double}) for mode @var{mode}. This means that you check whether
1203 @var{value} fits within the possible range of values for mode
1204 @var{mode} on this target machine. The mode @var{mode} is always
1205 a mode of class @code{MODE_FLOAT}. @var{overflow} is nonzero if
1206 the value is already known to be out of range.
1208 If @var{value} is not valid or if @var{overflow} is nonzero, you should
1209 set @var{overflow} to 1 and then assign some valid value to @var{value}.
1210 Allowing an invalid value to go through the compiler can produce
1211 incorrect assembler code which may even cause Unix assemblers to crash.
1213 This macro need not be defined if there is no work for it to do.
1215 @findex TARGET_FLOAT_FORMAT
1216 @item TARGET_FLOAT_FORMAT
1217 A code distinguishing the floating point format of the target machine.
1218 There are three defined values:
1221 @findex IEEE_FLOAT_FORMAT
1222 @item IEEE_FLOAT_FORMAT
1223 This code indicates IEEE floating point. It is the default; there is no
1224 need to define this macro when the format is IEEE.
1226 @findex VAX_FLOAT_FORMAT
1227 @item VAX_FLOAT_FORMAT
1228 This code indicates the peculiar format used on the Vax.
1230 @findex UNKNOWN_FLOAT_FORMAT
1231 @item UNKNOWN_FLOAT_FORMAT
1232 This code indicates any other format.
1235 The value of this macro is compared with @code{HOST_FLOAT_FORMAT}
1236 (@pxref{Config}) to determine whether the target machine has the same
1237 format as the host machine. If any other formats are actually in use on
1238 supported machines, new codes should be defined for them.
1240 The ordering of the component words of floating point values stored in
1241 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN} for the target
1242 machine and @code{HOST_FLOAT_WORDS_BIG_ENDIAN} for the host.
1244 @findex DEFAULT_VTABLE_THUNKS
1245 @item DEFAULT_VTABLE_THUNKS
1246 GCC supports two ways of implementing C++ vtables: traditional or with
1247 so-called ``thunks''. The flag @samp{-fvtable-thunk} chooses between them.
1248 Define this macro to be a C expression for the default value of that flag.
1249 If @code{DEFAULT_VTABLE_THUNKS} is 0, GCC uses the traditional
1250 implementation by default. The ``thunk'' implementation is more efficient
1251 (especially if you have provided an implementation of
1252 @code{ASM_OUTPUT_MI_THUNK}, see @ref{Function Entry}), but is not binary
1253 compatible with code compiled using the traditional implementation.
1254 If you are writing a new port, define @code{DEFAULT_VTABLE_THUNKS} to 1.
1256 If you do not define this macro, the default for @samp{-fvtable-thunk} is 0.
1260 @section Layout of Source Language Data Types
1262 These macros define the sizes and other characteristics of the standard
1263 basic data types used in programs being compiled. Unlike the macros in
1264 the previous section, these apply to specific features of C and related
1265 languages, rather than to fundamental aspects of storage layout.
1268 @findex INT_TYPE_SIZE
1270 A C expression for the size in bits of the type @code{int} on the
1271 target machine. If you don't define this, the default is one word.
1273 @findex MAX_INT_TYPE_SIZE
1274 @item MAX_INT_TYPE_SIZE
1275 Maximum number for the size in bits of the type @code{int} on the target
1276 machine. If this is undefined, the default is @code{INT_TYPE_SIZE}.
1277 Otherwise, it is the constant value that is the largest value that
1278 @code{INT_TYPE_SIZE} can have at run-time. This is used in @code{cpp}.
1280 @findex SHORT_TYPE_SIZE
1281 @item SHORT_TYPE_SIZE
1282 A C expression for the size in bits of the type @code{short} on the
1283 target machine. If you don't define this, the default is half a word.
1284 (If this would be less than one storage unit, it is rounded up to one
1287 @findex LONG_TYPE_SIZE
1288 @item LONG_TYPE_SIZE
1289 A C expression for the size in bits of the type @code{long} on the
1290 target machine. If you don't define this, the default is one word.
1292 @findex MAX_LONG_TYPE_SIZE
1293 @item MAX_LONG_TYPE_SIZE
1294 Maximum number for the size in bits of the type @code{long} on the
1295 target machine. If this is undefined, the default is
1296 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1297 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1300 @findex LONG_LONG_TYPE_SIZE
1301 @item LONG_LONG_TYPE_SIZE
1302 A C expression for the size in bits of the type @code{long long} on the
1303 target machine. If you don't define this, the default is two
1304 words. If you want to support GNU Ada on your machine, the value of this
1305 macro must be at least 64.
1307 @findex CHAR_TYPE_SIZE
1308 @item CHAR_TYPE_SIZE
1309 A C expression for the size in bits of the type @code{char} on the
1310 target machine. If you don't define this, the default is
1311 @code{BITS_PER_UNIT}.
1313 @findex MAX_CHAR_TYPE_SIZE
1314 @item MAX_CHAR_TYPE_SIZE
1315 Maximum number for the size in bits of the type @code{char} on the
1316 target machine. If this is undefined, the default is
1317 @code{CHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1318 largest value that @code{CHAR_TYPE_SIZE} can have at run-time. This is
1321 @findex FLOAT_TYPE_SIZE
1322 @item FLOAT_TYPE_SIZE
1323 A C expression for the size in bits of the type @code{float} on the
1324 target machine. If you don't define this, the default is one word.
1326 @findex DOUBLE_TYPE_SIZE
1327 @item DOUBLE_TYPE_SIZE
1328 A C expression for the size in bits of the type @code{double} on the
1329 target machine. If you don't define this, the default is two
1332 @findex LONG_DOUBLE_TYPE_SIZE
1333 @item LONG_DOUBLE_TYPE_SIZE
1334 A C expression for the size in bits of the type @code{long double} on
1335 the target machine. If you don't define this, the default is two
1338 @findex WIDEST_HARDWARE_FP_SIZE
1339 @item WIDEST_HARDWARE_FP_SIZE
1340 A C expression for the size in bits of the widest floating-point format
1341 supported by the hardware. If you define this macro, you must specify a
1342 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1343 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1346 @findex DEFAULT_SIGNED_CHAR
1347 @item DEFAULT_SIGNED_CHAR
1348 An expression whose value is 1 or 0, according to whether the type
1349 @code{char} should be signed or unsigned by default. The user can
1350 always override this default with the options @samp{-fsigned-char}
1351 and @samp{-funsigned-char}.
1353 @findex DEFAULT_SHORT_ENUMS
1354 @item DEFAULT_SHORT_ENUMS
1355 A C expression to determine whether to give an @code{enum} type
1356 only as many bytes as it takes to represent the range of possible values
1357 of that type. A nonzero value means to do that; a zero value means all
1358 @code{enum} types should be allocated like @code{int}.
1360 If you don't define the macro, the default is 0.
1364 A C expression for a string describing the name of the data type to use
1365 for size values. The typedef name @code{size_t} is defined using the
1366 contents of the string.
1368 The string can contain more than one keyword. If so, separate them with
1369 spaces, and write first any length keyword, then @code{unsigned} if
1370 appropriate, and finally @code{int}. The string must exactly match one
1371 of the data type names defined in the function
1372 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1373 omit @code{int} or change the order---that would cause the compiler to
1376 If you don't define this macro, the default is @code{"long unsigned
1379 @findex PTRDIFF_TYPE
1381 A C expression for a string describing the name of the data type to use
1382 for the result of subtracting two pointers. The typedef name
1383 @code{ptrdiff_t} is defined using the contents of the string. See
1384 @code{SIZE_TYPE} above for more information.
1386 If you don't define this macro, the default is @code{"long int"}.
1390 A C expression for a string describing the name of the data type to use
1391 for wide characters. The typedef name @code{wchar_t} is defined using
1392 the contents of the string. See @code{SIZE_TYPE} above for more
1395 If you don't define this macro, the default is @code{"int"}.
1397 @findex WCHAR_TYPE_SIZE
1398 @item WCHAR_TYPE_SIZE
1399 A C expression for the size in bits of the data type for wide
1400 characters. This is used in @code{cpp}, which cannot make use of
1403 @findex MAX_WCHAR_TYPE_SIZE
1404 @item MAX_WCHAR_TYPE_SIZE
1405 Maximum number for the size in bits of the data type for wide
1406 characters. If this is undefined, the default is
1407 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1408 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1413 A C expression for a string describing the name of the data type to
1414 use for wide characters passed to @code{printf} and returned from
1415 @code{getwc}. The typedef name @code{wint_t} is defined using the
1416 contents of the string. See @code{SIZE_TYPE} above for more
1419 If you don't define this macro, the default is @code{"unsigned int"}.
1423 A C expression for a string describing the name of the data type that
1424 can represent any value of any standard or extended signed integer type.
1425 The typedef name @code{intmax_t} is defined using the contents of the
1426 string. See @code{SIZE_TYPE} above for more information.
1428 If you don't define this macro, the default is the first of
1429 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1430 much precision as @code{long long int}.
1432 @findex UINTMAX_TYPE
1434 A C expression for a string describing the name of the data type that
1435 can represent any value of any standard or extended unsigned integer
1436 type. The typedef name @code{uintmax_t} is defined using the contents
1437 of the string. See @code{SIZE_TYPE} above for more information.
1439 If you don't define this macro, the default is the first of
1440 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1441 unsigned int"} that has as much precision as @code{long long unsigned
1444 @findex OBJC_SELECTORS_WITHOUT_LABELS
1445 @item OBJC_SELECTORS_WITHOUT_LABELS
1446 Define this macro if the compiler can group all the selectors together
1447 into a vector and use just one label at the beginning of the vector.
1448 Otherwise, the compiler must give each selector its own assembler
1451 On certain machines, it is important to have a separate label for each
1452 selector because this enables the linker to eliminate duplicate selectors.
1456 A C constant expression for the integer value for escape sequence
1461 @findex TARGET_NEWLINE
1464 @itemx TARGET_NEWLINE
1465 C constant expressions for the integer values for escape sequences
1466 @samp{\b}, @samp{\t} and @samp{\n}.
1474 C constant expressions for the integer values for escape sequences
1475 @samp{\v}, @samp{\f} and @samp{\r}.
1479 @section Register Usage
1480 @cindex register usage
1482 This section explains how to describe what registers the target machine
1483 has, and how (in general) they can be used.
1485 The description of which registers a specific instruction can use is
1486 done with register classes; see @ref{Register Classes}. For information
1487 on using registers to access a stack frame, see @ref{Frame Registers}.
1488 For passing values in registers, see @ref{Register Arguments}.
1489 For returning values in registers, see @ref{Scalar Return}.
1492 * Register Basics:: Number and kinds of registers.
1493 * Allocation Order:: Order in which registers are allocated.
1494 * Values in Registers:: What kinds of values each reg can hold.
1495 * Leaf Functions:: Renumbering registers for leaf functions.
1496 * Stack Registers:: Handling a register stack such as 80387.
1499 @node Register Basics
1500 @subsection Basic Characteristics of Registers
1502 @c prevent bad page break with this line
1503 Registers have various characteristics.
1506 @findex FIRST_PSEUDO_REGISTER
1507 @item FIRST_PSEUDO_REGISTER
1508 Number of hardware registers known to the compiler. They receive
1509 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1510 pseudo register's number really is assigned the number
1511 @code{FIRST_PSEUDO_REGISTER}.
1513 @item FIXED_REGISTERS
1514 @findex FIXED_REGISTERS
1515 @cindex fixed register
1516 An initializer that says which registers are used for fixed purposes
1517 all throughout the compiled code and are therefore not available for
1518 general allocation. These would include the stack pointer, the frame
1519 pointer (except on machines where that can be used as a general
1520 register when no frame pointer is needed), the program counter on
1521 machines where that is considered one of the addressable registers,
1522 and any other numbered register with a standard use.
1524 This information is expressed as a sequence of numbers, separated by
1525 commas and surrounded by braces. The @var{n}th number is 1 if
1526 register @var{n} is fixed, 0 otherwise.
1528 The table initialized from this macro, and the table initialized by
1529 the following one, may be overridden at run time either automatically,
1530 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1531 the user with the command options @samp{-ffixed-@var{reg}},
1532 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}}.
1534 @findex CALL_USED_REGISTERS
1535 @item CALL_USED_REGISTERS
1536 @cindex call-used register
1537 @cindex call-clobbered register
1538 @cindex call-saved register
1539 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1540 clobbered (in general) by function calls as well as for fixed
1541 registers. This macro therefore identifies the registers that are not
1542 available for general allocation of values that must live across
1545 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1546 automatically saves it on function entry and restores it on function
1547 exit, if the register is used within the function.
1549 @findex HARD_REGNO_CALL_PART_CLOBBERED
1550 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1551 @cindex call-used register
1552 @cindex call-clobbered register
1553 @cindex call-saved register
1554 A C expression that is non-zero if it is not permissible to store a
1555 value of mode @var{mode} in hard register number @var{regno} across a
1556 call without some part of it being clobbered. For most machines this
1557 macro need not be defined. It is only required for machines that do not
1558 preserve the entire contents of a register across a call.
1560 @findex CONDITIONAL_REGISTER_USAGE
1562 @findex call_used_regs
1563 @item CONDITIONAL_REGISTER_USAGE
1564 Zero or more C statements that may conditionally modify five variables
1565 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1566 (these three are of type @code{char []}), @code{reg_names} (of type
1567 @code{const char * []}) and @code{reg_class_contents} (of type
1568 @code{HARD_REG_SET}).
1569 Before the macro is called @code{fixed_regs}, @code{call_used_regs}
1570 @code{reg_class_contents} and @code{reg_names} have been initialized
1571 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1572 @code{REG_CLASS_CONTENTS} and @code{REGISTER_NAMES}, respectively,
1573 @code{global_regs} has been cleared, and any @samp{-ffixed-@var{reg}},
1574 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}} command
1575 options have been applied.
1577 This is necessary in case the fixed or call-clobbered registers depend
1580 You need not define this macro if it has no work to do.
1582 @cindex disabling certain registers
1583 @cindex controlling register usage
1584 If the usage of an entire class of registers depends on the target
1585 flags, you may indicate this to GCC by using this macro to modify
1586 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1587 registers in the classes which should not be used by GCC. Also define
1588 the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1589 is called with a letter for a class that shouldn't be used.
1591 (However, if this class is not included in @code{GENERAL_REGS} and all
1592 of the insn patterns whose constraints permit this class are
1593 controlled by target switches, then GCC will automatically avoid using
1594 these registers when the target switches are opposed to them.)
1596 @findex NON_SAVING_SETJMP
1597 @item NON_SAVING_SETJMP
1598 If this macro is defined and has a nonzero value, it means that
1599 @code{setjmp} and related functions fail to save the registers, or that
1600 @code{longjmp} fails to restore them. To compensate, the compiler
1601 avoids putting variables in registers in functions that use
1604 @findex INCOMING_REGNO
1605 @item INCOMING_REGNO (@var{out})
1606 Define this macro if the target machine has register windows. This C
1607 expression returns the register number as seen by the called function
1608 corresponding to the register number @var{out} as seen by the calling
1609 function. Return @var{out} if register number @var{out} is not an
1612 @findex OUTGOING_REGNO
1613 @item OUTGOING_REGNO (@var{in})
1614 Define this macro if the target machine has register windows. This C
1615 expression returns the register number as seen by the calling function
1616 corresponding to the register number @var{in} as seen by the called
1617 function. Return @var{in} if register number @var{in} is not an inbound
1621 @item LOCAL_REGNO (@var{regno})
1622 Define this macro if the target machine has register windows. This C
1623 expression returns true if the register is call-saved but is in the
1624 register window. Unlike most call-saved registers, such registers
1625 need not be explicitly restored on function exit or during non-local
1631 If the program counter has a register number, define this as that
1632 register number. Otherwise, do not define it.
1636 @node Allocation Order
1637 @subsection Order of Allocation of Registers
1638 @cindex order of register allocation
1639 @cindex register allocation order
1641 @c prevent bad page break with this line
1642 Registers are allocated in order.
1645 @findex REG_ALLOC_ORDER
1646 @item REG_ALLOC_ORDER
1647 If defined, an initializer for a vector of integers, containing the
1648 numbers of hard registers in the order in which GCC should prefer
1649 to use them (from most preferred to least).
1651 If this macro is not defined, registers are used lowest numbered first
1652 (all else being equal).
1654 One use of this macro is on machines where the highest numbered
1655 registers must always be saved and the save-multiple-registers
1656 instruction supports only sequences of consecutive registers. On such
1657 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1658 the highest numbered allocable register first.
1660 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1661 @item ORDER_REGS_FOR_LOCAL_ALLOC
1662 A C statement (sans semicolon) to choose the order in which to allocate
1663 hard registers for pseudo-registers local to a basic block.
1665 Store the desired register order in the array @code{reg_alloc_order}.
1666 Element 0 should be the register to allocate first; element 1, the next
1667 register; and so on.
1669 The macro body should not assume anything about the contents of
1670 @code{reg_alloc_order} before execution of the macro.
1672 On most machines, it is not necessary to define this macro.
1675 @node Values in Registers
1676 @subsection How Values Fit in Registers
1678 This section discusses the macros that describe which kinds of values
1679 (specifically, which machine modes) each register can hold, and how many
1680 consecutive registers are needed for a given mode.
1683 @findex HARD_REGNO_NREGS
1684 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1685 A C expression for the number of consecutive hard registers, starting
1686 at register number @var{regno}, required to hold a value of mode
1689 On a machine where all registers are exactly one word, a suitable
1690 definition of this macro is
1693 #define HARD_REGNO_NREGS(REGNO, MODE) \
1694 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1698 @findex HARD_REGNO_MODE_OK
1699 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1700 A C expression that is nonzero if it is permissible to store a value
1701 of mode @var{mode} in hard register number @var{regno} (or in several
1702 registers starting with that one). For a machine where all registers
1703 are equivalent, a suitable definition is
1706 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1709 You need not include code to check for the numbers of fixed registers,
1710 because the allocation mechanism considers them to be always occupied.
1712 @cindex register pairs
1713 On some machines, double-precision values must be kept in even/odd
1714 register pairs. You can implement that by defining this macro to reject
1715 odd register numbers for such modes.
1717 The minimum requirement for a mode to be OK in a register is that the
1718 @samp{mov@var{mode}} instruction pattern support moves between the
1719 register and other hard register in the same class and that moving a
1720 value into the register and back out not alter it.
1722 Since the same instruction used to move @code{word_mode} will work for
1723 all narrower integer modes, it is not necessary on any machine for
1724 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1725 you define patterns @samp{movhi}, etc., to take advantage of this. This
1726 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1727 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1730 Many machines have special registers for floating point arithmetic.
1731 Often people assume that floating point machine modes are allowed only
1732 in floating point registers. This is not true. Any registers that
1733 can hold integers can safely @emph{hold} a floating point machine
1734 mode, whether or not floating arithmetic can be done on it in those
1735 registers. Integer move instructions can be used to move the values.
1737 On some machines, though, the converse is true: fixed-point machine
1738 modes may not go in floating registers. This is true if the floating
1739 registers normalize any value stored in them, because storing a
1740 non-floating value there would garble it. In this case,
1741 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1742 floating registers. But if the floating registers do not automatically
1743 normalize, if you can store any bit pattern in one and retrieve it
1744 unchanged without a trap, then any machine mode may go in a floating
1745 register, so you can define this macro to say so.
1747 The primary significance of special floating registers is rather that
1748 they are the registers acceptable in floating point arithmetic
1749 instructions. However, this is of no concern to
1750 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1751 constraints for those instructions.
1753 On some machines, the floating registers are especially slow to access,
1754 so that it is better to store a value in a stack frame than in such a
1755 register if floating point arithmetic is not being done. As long as the
1756 floating registers are not in class @code{GENERAL_REGS}, they will not
1757 be used unless some pattern's constraint asks for one.
1759 @findex MODES_TIEABLE_P
1760 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
1761 A C expression that is nonzero if a value of mode
1762 @var{mode1} is accessible in mode @var{mode2} without copying.
1764 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
1765 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
1766 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
1767 should be nonzero. If they differ for any @var{r}, you should define
1768 this macro to return zero unless some other mechanism ensures the
1769 accessibility of the value in a narrower mode.
1771 You should define this macro to return nonzero in as many cases as
1772 possible since doing so will allow GCC to perform better register
1775 @findex AVOID_CCMODE_COPIES
1776 @item AVOID_CCMODE_COPIES
1777 Define this macro if the compiler should avoid copies to/from @code{CCmode}
1778 registers. You should only define this macro if support for copying to/from
1779 @code{CCmode} is incomplete.
1781 @findex SUBREG_REGNO_OFFSET
1782 @item SUBREG_REGNO_OFFSET
1783 Define this macro if the compiler needs to handle subregs in a non-standard
1784 way. The macro returns the correct regno offset for mode @code{YMODE} given
1785 a subreg of type @code{XMODE}.
1786 This macro takes 4 parameters:
1787 @code{XREGNO} - A regno of an inner hard subreg_reg (or what will become one).
1788 @code{XMODE} - The mode of xregno.
1789 @code{OFFSET} - The byte offset.
1790 @code{YMODE} - The mode of a top level SUBREG (or what may become one).
1791 The default function can be found in rtlanal.c, function
1792 @code{subreg_regno_offset}. Normally this does not need to be defined.
1795 @node Leaf Functions
1796 @subsection Handling Leaf Functions
1798 @cindex leaf functions
1799 @cindex functions, leaf
1800 On some machines, a leaf function (i.e., one which makes no calls) can run
1801 more efficiently if it does not make its own register window. Often this
1802 means it is required to receive its arguments in the registers where they
1803 are passed by the caller, instead of the registers where they would
1806 The special treatment for leaf functions generally applies only when
1807 other conditions are met; for example, often they may use only those
1808 registers for its own variables and temporaries. We use the term ``leaf
1809 function'' to mean a function that is suitable for this special
1810 handling, so that functions with no calls are not necessarily ``leaf
1813 GCC assigns register numbers before it knows whether the function is
1814 suitable for leaf function treatment. So it needs to renumber the
1815 registers in order to output a leaf function. The following macros
1819 @findex LEAF_REGISTERS
1820 @item LEAF_REGISTERS
1821 Name of a char vector, indexed by hard register number, which
1822 contains 1 for a register that is allowable in a candidate for leaf
1825 If leaf function treatment involves renumbering the registers, then the
1826 registers marked here should be the ones before renumbering---those that
1827 GCC would ordinarily allocate. The registers which will actually be
1828 used in the assembler code, after renumbering, should not be marked with 1
1831 Define this macro only if the target machine offers a way to optimize
1832 the treatment of leaf functions.
1834 @findex LEAF_REG_REMAP
1835 @item LEAF_REG_REMAP (@var{regno})
1836 A C expression whose value is the register number to which @var{regno}
1837 should be renumbered, when a function is treated as a leaf function.
1839 If @var{regno} is a register number which should not appear in a leaf
1840 function before renumbering, then the expression should yield -1, which
1841 will cause the compiler to abort.
1843 Define this macro only if the target machine offers a way to optimize the
1844 treatment of leaf functions, and registers need to be renumbered to do
1848 @findex current_function_is_leaf
1849 @findex current_function_uses_only_leaf_regs
1850 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
1851 treat leaf functions specially. They can test the C variable
1852 @code{current_function_is_leaf} which is nonzero for leaf functions.
1853 @code{current_function_is_leaf} is set prior to local register allocation
1854 and is valid for the remaining compiler passes. They can also test the C
1855 variable @code{current_function_uses_only_leaf_regs} which is nonzero for
1856 leaf functions which only use leaf registers.
1857 @code{current_function_uses_only_leaf_regs} is valid after reload and is
1858 only useful if @code{LEAF_REGISTERS} is defined.
1859 @c changed this to fix overfull. ALSO: why the "it" at the beginning
1860 @c of the next paragraph?! --mew 2feb93
1862 @node Stack Registers
1863 @subsection Registers That Form a Stack
1865 There are special features to handle computers where some of the
1866 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
1867 Stack registers are normally written by pushing onto the stack, and are
1868 numbered relative to the top of the stack.
1870 Currently, GCC can only handle one group of stack-like registers, and
1871 they must be consecutively numbered.
1876 Define this if the machine has any stack-like registers.
1878 @findex FIRST_STACK_REG
1879 @item FIRST_STACK_REG
1880 The number of the first stack-like register. This one is the top
1883 @findex LAST_STACK_REG
1884 @item LAST_STACK_REG
1885 The number of the last stack-like register. This one is the bottom of
1889 @node Register Classes
1890 @section Register Classes
1891 @cindex register class definitions
1892 @cindex class definitions, register
1894 On many machines, the numbered registers are not all equivalent.
1895 For example, certain registers may not be allowed for indexed addressing;
1896 certain registers may not be allowed in some instructions. These machine
1897 restrictions are described to the compiler using @dfn{register classes}.
1899 You define a number of register classes, giving each one a name and saying
1900 which of the registers belong to it. Then you can specify register classes
1901 that are allowed as operands to particular instruction patterns.
1905 In general, each register will belong to several classes. In fact, one
1906 class must be named @code{ALL_REGS} and contain all the registers. Another
1907 class must be named @code{NO_REGS} and contain no registers. Often the
1908 union of two classes will be another class; however, this is not required.
1910 @findex GENERAL_REGS
1911 One of the classes must be named @code{GENERAL_REGS}. There is nothing
1912 terribly special about the name, but the operand constraint letters
1913 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
1914 the same as @code{ALL_REGS}, just define it as a macro which expands
1917 Order the classes so that if class @var{x} is contained in class @var{y}
1918 then @var{x} has a lower class number than @var{y}.
1920 The way classes other than @code{GENERAL_REGS} are specified in operand
1921 constraints is through machine-dependent operand constraint letters.
1922 You can define such letters to correspond to various classes, then use
1923 them in operand constraints.
1925 You should define a class for the union of two classes whenever some
1926 instruction allows both classes. For example, if an instruction allows
1927 either a floating point (coprocessor) register or a general register for a
1928 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
1929 which includes both of them. Otherwise you will get suboptimal code.
1931 You must also specify certain redundant information about the register
1932 classes: for each class, which classes contain it and which ones are
1933 contained in it; for each pair of classes, the largest class contained
1936 When a value occupying several consecutive registers is expected in a
1937 certain class, all the registers used must belong to that class.
1938 Therefore, register classes cannot be used to enforce a requirement for
1939 a register pair to start with an even-numbered register. The way to
1940 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
1942 Register classes used for input-operands of bitwise-and or shift
1943 instructions have a special requirement: each such class must have, for
1944 each fixed-point machine mode, a subclass whose registers can transfer that
1945 mode to or from memory. For example, on some machines, the operations for
1946 single-byte values (@code{QImode}) are limited to certain registers. When
1947 this is so, each register class that is used in a bitwise-and or shift
1948 instruction must have a subclass consisting of registers from which
1949 single-byte values can be loaded or stored. This is so that
1950 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
1953 @findex enum reg_class
1954 @item enum reg_class
1955 An enumeral type that must be defined with all the register class names
1956 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
1957 must be the last register class, followed by one more enumeral value,
1958 @code{LIM_REG_CLASSES}, which is not a register class but rather
1959 tells how many classes there are.
1961 Each register class has a number, which is the value of casting
1962 the class name to type @code{int}. The number serves as an index
1963 in many of the tables described below.
1965 @findex N_REG_CLASSES
1967 The number of distinct register classes, defined as follows:
1970 #define N_REG_CLASSES (int) LIM_REG_CLASSES
1973 @findex REG_CLASS_NAMES
1974 @item REG_CLASS_NAMES
1975 An initializer containing the names of the register classes as C string
1976 constants. These names are used in writing some of the debugging dumps.
1978 @findex REG_CLASS_CONTENTS
1979 @item REG_CLASS_CONTENTS
1980 An initializer containing the contents of the register classes, as integers
1981 which are bit masks. The @var{n}th integer specifies the contents of class
1982 @var{n}. The way the integer @var{mask} is interpreted is that
1983 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
1985 When the machine has more than 32 registers, an integer does not suffice.
1986 Then the integers are replaced by sub-initializers, braced groupings containing
1987 several integers. Each sub-initializer must be suitable as an initializer
1988 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
1989 In this situation, the first integer in each sub-initializer corresponds to
1990 registers 0 through 31, the second integer to registers 32 through 63, and
1993 @findex REGNO_REG_CLASS
1994 @item REGNO_REG_CLASS (@var{regno})
1995 A C expression whose value is a register class containing hard register
1996 @var{regno}. In general there is more than one such class; choose a class
1997 which is @dfn{minimal}, meaning that no smaller class also contains the
2000 @findex BASE_REG_CLASS
2001 @item BASE_REG_CLASS
2002 A macro whose definition is the name of the class to which a valid
2003 base register must belong. A base register is one used in an address
2004 which is the register value plus a displacement.
2006 @findex INDEX_REG_CLASS
2007 @item INDEX_REG_CLASS
2008 A macro whose definition is the name of the class to which a valid
2009 index register must belong. An index register is one used in an
2010 address where its value is either multiplied by a scale factor or
2011 added to another register (as well as added to a displacement).
2013 @findex REG_CLASS_FROM_LETTER
2014 @item REG_CLASS_FROM_LETTER (@var{char})
2015 A C expression which defines the machine-dependent operand constraint
2016 letters for register classes. If @var{char} is such a letter, the
2017 value should be the register class corresponding to it. Otherwise,
2018 the value should be @code{NO_REGS}. The register letter @samp{r},
2019 corresponding to class @code{GENERAL_REGS}, will not be passed
2020 to this macro; you do not need to handle it.
2022 @findex REGNO_OK_FOR_BASE_P
2023 @item REGNO_OK_FOR_BASE_P (@var{num})
2024 A C expression which is nonzero if register number @var{num} is
2025 suitable for use as a base register in operand addresses. It may be
2026 either a suitable hard register or a pseudo register that has been
2027 allocated such a hard register.
2029 @findex REGNO_MODE_OK_FOR_BASE_P
2030 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2031 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2032 that expression may examine the mode of the memory reference in
2033 @var{mode}. You should define this macro if the mode of the memory
2034 reference affects whether a register may be used as a base register. If
2035 you define this macro, the compiler will use it instead of
2036 @code{REGNO_OK_FOR_BASE_P}.
2038 @findex REGNO_OK_FOR_INDEX_P
2039 @item REGNO_OK_FOR_INDEX_P (@var{num})
2040 A C expression which is nonzero if register number @var{num} is
2041 suitable for use as an index register in operand addresses. It may be
2042 either a suitable hard register or a pseudo register that has been
2043 allocated such a hard register.
2045 The difference between an index register and a base register is that
2046 the index register may be scaled. If an address involves the sum of
2047 two registers, neither one of them scaled, then either one may be
2048 labeled the ``base'' and the other the ``index''; but whichever
2049 labeling is used must fit the machine's constraints of which registers
2050 may serve in each capacity. The compiler will try both labelings,
2051 looking for one that is valid, and will reload one or both registers
2052 only if neither labeling works.
2054 @findex PREFERRED_RELOAD_CLASS
2055 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2056 A C expression that places additional restrictions on the register class
2057 to use when it is necessary to copy value @var{x} into a register in class
2058 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2059 another, smaller class. On many machines, the following definition is
2063 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2066 Sometimes returning a more restrictive class makes better code. For
2067 example, on the 68000, when @var{x} is an integer constant that is in range
2068 for a @samp{moveq} instruction, the value of this macro is always
2069 @code{DATA_REGS} as long as @var{class} includes the data registers.
2070 Requiring a data register guarantees that a @samp{moveq} will be used.
2072 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2073 you can force @var{x} into a memory constant. This is useful on
2074 certain machines where immediate floating values cannot be loaded into
2075 certain kinds of registers.
2077 @findex PREFERRED_OUTPUT_RELOAD_CLASS
2078 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2079 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2080 input reloads. If you don't define this macro, the default is to use
2081 @var{class}, unchanged.
2083 @findex LIMIT_RELOAD_CLASS
2084 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2085 A C expression that places additional restrictions on the register class
2086 to use when it is necessary to be able to hold a value of mode
2087 @var{mode} in a reload register for which class @var{class} would
2090 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2091 there are certain modes that simply can't go in certain reload classes.
2093 The value is a register class; perhaps @var{class}, or perhaps another,
2096 Don't define this macro unless the target machine has limitations which
2097 require the macro to do something nontrivial.
2099 @findex SECONDARY_RELOAD_CLASS
2100 @findex SECONDARY_INPUT_RELOAD_CLASS
2101 @findex SECONDARY_OUTPUT_RELOAD_CLASS
2102 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2103 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2104 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2105 Many machines have some registers that cannot be copied directly to or
2106 from memory or even from other types of registers. An example is the
2107 @samp{MQ} register, which on most machines, can only be copied to or
2108 from general registers, but not memory. Some machines allow copying all
2109 registers to and from memory, but require a scratch register for stores
2110 to some memory locations (e.g., those with symbolic address on the RT,
2111 and those with certain symbolic address on the Sparc when compiling
2112 PIC). In some cases, both an intermediate and a scratch register are
2115 You should define these macros to indicate to the reload phase that it may
2116 need to allocate at least one register for a reload in addition to the
2117 register to contain the data. Specifically, if copying @var{x} to a
2118 register @var{class} in @var{mode} requires an intermediate register,
2119 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2120 largest register class all of whose registers can be used as
2121 intermediate registers or scratch registers.
2123 If copying a register @var{class} in @var{mode} to @var{x} requires an
2124 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2125 should be defined to return the largest register class required. If the
2126 requirements for input and output reloads are the same, the macro
2127 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2130 The values returned by these macros are often @code{GENERAL_REGS}.
2131 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2132 can be directly copied to or from a register of @var{class} in
2133 @var{mode} without requiring a scratch register. Do not define this
2134 macro if it would always return @code{NO_REGS}.
2136 If a scratch register is required (either with or without an
2137 intermediate register), you should define patterns for
2138 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2139 (@pxref{Standard Names}. These patterns, which will normally be
2140 implemented with a @code{define_expand}, should be similar to the
2141 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2144 Define constraints for the reload register and scratch register that
2145 contain a single register class. If the original reload register (whose
2146 class is @var{class}) can meet the constraint given in the pattern, the
2147 value returned by these macros is used for the class of the scratch
2148 register. Otherwise, two additional reload registers are required.
2149 Their classes are obtained from the constraints in the insn pattern.
2151 @var{x} might be a pseudo-register or a @code{subreg} of a
2152 pseudo-register, which could either be in a hard register or in memory.
2153 Use @code{true_regnum} to find out; it will return -1 if the pseudo is
2154 in memory and the hard register number if it is in a register.
2156 These macros should not be used in the case where a particular class of
2157 registers can only be copied to memory and not to another class of
2158 registers. In that case, secondary reload registers are not needed and
2159 would not be helpful. Instead, a stack location must be used to perform
2160 the copy and the @code{mov@var{m}} pattern should use memory as a
2161 intermediate storage. This case often occurs between floating-point and
2164 @findex SECONDARY_MEMORY_NEEDED
2165 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2166 Certain machines have the property that some registers cannot be copied
2167 to some other registers without using memory. Define this macro on
2168 those machines to be a C expression that is non-zero if objects of mode
2169 @var{m} in registers of @var{class1} can only be copied to registers of
2170 class @var{class2} by storing a register of @var{class1} into memory
2171 and loading that memory location into a register of @var{class2}.
2173 Do not define this macro if its value would always be zero.
2175 @findex SECONDARY_MEMORY_NEEDED_RTX
2176 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2177 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2178 allocates a stack slot for a memory location needed for register copies.
2179 If this macro is defined, the compiler instead uses the memory location
2180 defined by this macro.
2182 Do not define this macro if you do not define
2183 @code{SECONDARY_MEMORY_NEEDED}.
2185 @findex SECONDARY_MEMORY_NEEDED_MODE
2186 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2187 When the compiler needs a secondary memory location to copy between two
2188 registers of mode @var{mode}, it normally allocates sufficient memory to
2189 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2190 load operations in a mode that many bits wide and whose class is the
2191 same as that of @var{mode}.
2193 This is right thing to do on most machines because it ensures that all
2194 bits of the register are copied and prevents accesses to the registers
2195 in a narrower mode, which some machines prohibit for floating-point
2198 However, this default behavior is not correct on some machines, such as
2199 the DEC Alpha, that store short integers in floating-point registers
2200 differently than in integer registers. On those machines, the default
2201 widening will not work correctly and you must define this macro to
2202 suppress that widening in some cases. See the file @file{alpha.h} for
2205 Do not define this macro if you do not define
2206 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2207 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2209 @findex SMALL_REGISTER_CLASSES
2210 @item SMALL_REGISTER_CLASSES
2211 On some machines, it is risky to let hard registers live across arbitrary
2212 insns. Typically, these machines have instructions that require values
2213 to be in specific registers (like an accumulator), and reload will fail
2214 if the required hard register is used for another purpose across such an
2217 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a non-zero
2218 value on these machines. When this macro has a non-zero value, the
2219 compiler will try to minimize the lifetime of hard registers.
2221 It is always safe to define this macro with a non-zero value, but if you
2222 unnecessarily define it, you will reduce the amount of optimizations
2223 that can be performed in some cases. If you do not define this macro
2224 with a non-zero value when it is required, the compiler will run out of
2225 spill registers and print a fatal error message. For most machines, you
2226 should not define this macro at all.
2228 @findex CLASS_LIKELY_SPILLED_P
2229 @item CLASS_LIKELY_SPILLED_P (@var{class})
2230 A C expression whose value is nonzero if pseudos that have been assigned
2231 to registers of class @var{class} would likely be spilled because
2232 registers of @var{class} are needed for spill registers.
2234 The default value of this macro returns 1 if @var{class} has exactly one
2235 register and zero otherwise. On most machines, this default should be
2236 used. Only define this macro to some other expression if pseudos
2237 allocated by @file{local-alloc.c} end up in memory because their hard
2238 registers were needed for spill registers. If this macro returns nonzero
2239 for those classes, those pseudos will only be allocated by
2240 @file{global.c}, which knows how to reallocate the pseudo to another
2241 register. If there would not be another register available for
2242 reallocation, you should not change the definition of this macro since
2243 the only effect of such a definition would be to slow down register
2246 @findex CLASS_MAX_NREGS
2247 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2248 A C expression for the maximum number of consecutive registers
2249 of class @var{class} needed to hold a value of mode @var{mode}.
2251 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2252 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2253 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2254 @var{mode})} for all @var{regno} values in the class @var{class}.
2256 This macro helps control the handling of multiple-word values
2259 @item CLASS_CANNOT_CHANGE_MODE
2260 If defined, a C expression for a class that contains registers for
2261 which the compiler may not change modes arbitrarily.
2263 @item CLASS_CANNOT_CHANGE_MODE_P(@var{from}, @var{to})
2264 A C expression that is true if, for a register in
2265 @code{CLASS_CANNOT_CHANGE_MODE}, the requested mode punning is illegal.
2267 For the example, loading 32-bit integer or floating-point objects into
2268 floating-point registers on the Alpha extends them to 64-bits.
2269 Therefore loading a 64-bit object and then storing it as a 32-bit object
2270 does not store the low-order 32-bits, as would be the case for a normal
2271 register. Therefore, @file{alpha.h} defines @code{CLASS_CANNOT_CHANGE_MODE}
2272 as @code{FLOAT_REGS} and @code{CLASS_CANNOT_CHANGE_MODE_P} restricts
2273 mode changes to same-size modes.
2275 Compare this to IA-64, which extends floating-point values to 82-bits,
2276 and stores 64-bit integers in a different format than 64-bit doubles.
2277 Therefore @code{CLASS_CANNOT_CHANGE_MODE_P} is always true.
2280 Three other special macros describe which operands fit which constraint
2284 @findex CONST_OK_FOR_LETTER_P
2285 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2286 A C expression that defines the machine-dependent operand constraint
2287 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2288 particular ranges of integer values. If @var{c} is one of those
2289 letters, the expression should check that @var{value}, an integer, is in
2290 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2291 not one of those letters, the value should be 0 regardless of
2294 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2295 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2296 A C expression that defines the machine-dependent operand constraint
2297 letters that specify particular ranges of @code{const_double} values
2298 (@samp{G} or @samp{H}).
2300 If @var{c} is one of those letters, the expression should check that
2301 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2302 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2303 letters, the value should be 0 regardless of @var{value}.
2305 @code{const_double} is used for all floating-point constants and for
2306 @code{DImode} fixed-point constants. A given letter can accept either
2307 or both kinds of values. It can use @code{GET_MODE} to distinguish
2308 between these kinds.
2310 @findex EXTRA_CONSTRAINT
2311 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2312 A C expression that defines the optional machine-dependent constraint
2313 letters that can be used to segregate specific types of operands, usually
2314 memory references, for the target machine. Any letter that is not
2315 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER}
2316 may be used. Normally this macro will not be defined.
2318 If it is required for a particular target machine, it should return 1
2319 if @var{value} corresponds to the operand type represented by the
2320 constraint letter @var{c}. If @var{c} is not defined as an extra
2321 constraint, the value returned should be 0 regardless of @var{value}.
2323 For example, on the ROMP, load instructions cannot have their output
2324 in r0 if the memory reference contains a symbolic address. Constraint
2325 letter @samp{Q} is defined as representing a memory address that does
2326 @emph{not} contain a symbolic address. An alternative is specified with
2327 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2328 alternative specifies @samp{m} on the input and a register class that
2329 does not include r0 on the output.
2332 @node Stack and Calling
2333 @section Stack Layout and Calling Conventions
2334 @cindex calling conventions
2336 @c prevent bad page break with this line
2337 This describes the stack layout and calling conventions.
2345 * Register Arguments::
2347 * Aggregate Return::
2356 @subsection Basic Stack Layout
2357 @cindex stack frame layout
2358 @cindex frame layout
2360 @c prevent bad page break with this line
2361 Here is the basic stack layout.
2364 @findex STACK_GROWS_DOWNWARD
2365 @item STACK_GROWS_DOWNWARD
2366 Define this macro if pushing a word onto the stack moves the stack
2367 pointer to a smaller address.
2369 When we say, ``define this macro if @dots{},'' it means that the
2370 compiler checks this macro only with @code{#ifdef} so the precise
2371 definition used does not matter.
2373 @findex FRAME_GROWS_DOWNWARD
2374 @item FRAME_GROWS_DOWNWARD
2375 Define this macro if the addresses of local variable slots are at negative
2376 offsets from the frame pointer.
2378 @findex ARGS_GROW_DOWNWARD
2379 @item ARGS_GROW_DOWNWARD
2380 Define this macro if successive arguments to a function occupy decreasing
2381 addresses on the stack.
2383 @findex STARTING_FRAME_OFFSET
2384 @item STARTING_FRAME_OFFSET
2385 Offset from the frame pointer to the first local variable slot to be allocated.
2387 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2388 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2389 Otherwise, it is found by adding the length of the first slot to the
2390 value @code{STARTING_FRAME_OFFSET}.
2391 @c i'm not sure if the above is still correct.. had to change it to get
2392 @c rid of an overfull. --mew 2feb93
2394 @findex STACK_POINTER_OFFSET
2395 @item STACK_POINTER_OFFSET
2396 Offset from the stack pointer register to the first location at which
2397 outgoing arguments are placed. If not specified, the default value of
2398 zero is used. This is the proper value for most machines.
2400 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2401 the first location at which outgoing arguments are placed.
2403 @findex FIRST_PARM_OFFSET
2404 @item FIRST_PARM_OFFSET (@var{fundecl})
2405 Offset from the argument pointer register to the first argument's
2406 address. On some machines it may depend on the data type of the
2409 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2410 the first argument's address.
2412 @findex STACK_DYNAMIC_OFFSET
2413 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2414 Offset from the stack pointer register to an item dynamically allocated
2415 on the stack, e.g., by @code{alloca}.
2417 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2418 length of the outgoing arguments. The default is correct for most
2419 machines. See @file{function.c} for details.
2421 @findex DYNAMIC_CHAIN_ADDRESS
2422 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2423 A C expression whose value is RTL representing the address in a stack
2424 frame where the pointer to the caller's frame is stored. Assume that
2425 @var{frameaddr} is an RTL expression for the address of the stack frame
2428 If you don't define this macro, the default is to return the value
2429 of @var{frameaddr}---that is, the stack frame address is also the
2430 address of the stack word that points to the previous frame.
2432 @findex SETUP_FRAME_ADDRESSES
2433 @item SETUP_FRAME_ADDRESSES
2434 If defined, a C expression that produces the machine-specific code to
2435 setup the stack so that arbitrary frames can be accessed. For example,
2436 on the Sparc, we must flush all of the register windows to the stack
2437 before we can access arbitrary stack frames. You will seldom need to
2440 @findex BUILTIN_SETJMP_FRAME_VALUE
2441 @item BUILTIN_SETJMP_FRAME_VALUE
2442 If defined, a C expression that contains an rtx that is used to store
2443 the address of the current frame into the built in @code{setjmp} buffer.
2444 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2445 machines. One reason you may need to define this macro is if
2446 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2448 @findex RETURN_ADDR_RTX
2449 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2450 A C expression whose value is RTL representing the value of the return
2451 address for the frame @var{count} steps up from the current frame, after
2452 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2453 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2454 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2456 The value of the expression must always be the correct address when
2457 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2458 determine the return address of other frames.
2460 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2461 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2462 Define this if the return address of a particular stack frame is accessed
2463 from the frame pointer of the previous stack frame.
2465 @findex INCOMING_RETURN_ADDR_RTX
2466 @item INCOMING_RETURN_ADDR_RTX
2467 A C expression whose value is RTL representing the location of the
2468 incoming return address at the beginning of any function, before the
2469 prologue. This RTL is either a @code{REG}, indicating that the return
2470 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2473 You only need to define this macro if you want to support call frame
2474 debugging information like that provided by DWARF 2.
2476 If this RTL is a @code{REG}, you should also define
2477 DWARF_FRAME_RETURN_COLUMN to @code{DWARF_FRAME_REGNUM (REGNO)}.
2479 @findex INCOMING_FRAME_SP_OFFSET
2480 @item INCOMING_FRAME_SP_OFFSET
2481 A C expression whose value is an integer giving the offset, in bytes,
2482 from the value of the stack pointer register to the top of the stack
2483 frame at the beginning of any function, before the prologue. The top of
2484 the frame is defined to be the value of the stack pointer in the
2485 previous frame, just before the call instruction.
2487 You only need to define this macro if you want to support call frame
2488 debugging information like that provided by DWARF 2.
2490 @findex ARG_POINTER_CFA_OFFSET
2491 @item ARG_POINTER_CFA_OFFSET (@var{fundecl})
2492 A C expression whose value is an integer giving the offset, in bytes,
2493 from the argument pointer to the canonical frame address (cfa). The
2494 final value should coincide with that calculated by
2495 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2496 during virtual register instantiation.
2498 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2499 which is correct for most machines; in general, the arguments are found
2500 immediately before the stack frame. Note that this is not the case on
2501 some targets that save registers into the caller's frame, such as SPARC
2502 and rs6000, and so such targets need to define this macro.
2504 You only need to define this macro if the default is incorrect, and you
2505 want to support call frame debugging information like that provided by
2508 @findex EH_RETURN_DATA_REGNO
2509 @item EH_RETURN_DATA_REGNO (@var{N})
2510 A C expression whose value is the @var{N}th register number used for
2511 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2512 @var{N} registers are usable.
2514 The exception handling library routines communicate with the exception
2515 handlers via a set of agreed upon registers. Ideally these registers
2516 should be call-clobbered; it is possible to use call-saved registers,
2517 but may negatively impact code size. The target must support at least
2518 2 data registers, but should define 4 if there are enough free registers.
2520 You must define this macro if you want to support call frame exception
2521 handling like that provided by DWARF 2.
2523 @findex EH_RETURN_STACKADJ_RTX
2524 @item EH_RETURN_STACKADJ_RTX
2525 A C expression whose value is RTL representing a location in which
2526 to store a stack adjustment to be applied before function return.
2527 This is used to unwind the stack to an exception handler's call frame.
2528 It will be assigned zero on code paths that return normally.
2530 Typically this is a call-clobbered hard register that is otherwise
2531 untouched by the epilogue, but could also be a stack slot.
2533 You must define this macro if you want to support call frame exception
2534 handling like that provided by DWARF 2.
2536 @findex EH_RETURN_HANDLER_RTX
2537 @item EH_RETURN_HANDLER_RTX
2538 A C expression whose value is RTL representing a location in which
2539 to store the address of an exception handler to which we should
2540 return. It will not be assigned on code paths that return normally.
2542 Typically this is the location in the call frame at which the normal
2543 return address is stored. For targets that return by popping an
2544 address off the stack, this might be a memory address just below
2545 the @emph{target} call frame rather than inside the current call
2546 frame. @code{EH_RETURN_STACKADJ_RTX} will have already been assigned,
2547 so it may be used to calculate the location of the target call frame.
2549 Some targets have more complex requirements than storing to an
2550 address calculable during initial code generation. In that case
2551 the @code{eh_return} instruction pattern should be used instead.
2553 If you want to support call frame exception handling, you must
2554 define either this macro or the @code{eh_return} instruction pattern.
2558 Define this macro if the stack size for the target is very small. This
2559 has the effect of disabling gcc's builtin @samp{alloca}, though
2560 @samp{__builtin_alloca} is not affected.
2563 @node Stack Checking
2564 @subsection Specifying How Stack Checking is Done
2566 GCC will check that stack references are within the boundaries of
2567 the stack, if the @samp{-fstack-check} is specified, in one of three ways:
2571 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2572 will assume that you have arranged for stack checking to be done at
2573 appropriate places in the configuration files, e.g., in
2574 @code{FUNCTION_PROLOGUE}. GCC will do not other special processing.
2577 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
2578 called @code{check_stack} in your @file{md} file, GCC will call that
2579 pattern with one argument which is the address to compare the stack
2580 value against. You must arrange for this pattern to report an error if
2581 the stack pointer is out of range.
2584 If neither of the above are true, GCC will generate code to periodically
2585 ``probe'' the stack pointer using the values of the macros defined below.
2588 Normally, you will use the default values of these macros, so GCC
2589 will use the third approach.
2592 @findex STACK_CHECK_BUILTIN
2593 @item STACK_CHECK_BUILTIN
2594 A nonzero value if stack checking is done by the configuration files in a
2595 machine-dependent manner. You should define this macro if stack checking
2596 is require by the ABI of your machine or if you would like to have to stack
2597 checking in some more efficient way than GCC's portable approach.
2598 The default value of this macro is zero.
2600 @findex STACK_CHECK_PROBE_INTERVAL
2601 @item STACK_CHECK_PROBE_INTERVAL
2602 An integer representing the interval at which GCC must generate stack
2603 probe instructions. You will normally define this macro to be no larger
2604 than the size of the ``guard pages'' at the end of a stack area. The
2605 default value of 4096 is suitable for most systems.
2607 @findex STACK_CHECK_PROBE_LOAD
2608 @item STACK_CHECK_PROBE_LOAD
2609 A integer which is nonzero if GCC should perform the stack probe
2610 as a load instruction and zero if GCC should use a store instruction.
2611 The default is zero, which is the most efficient choice on most systems.
2613 @findex STACK_CHECK_PROTECT
2614 @item STACK_CHECK_PROTECT
2615 The number of bytes of stack needed to recover from a stack overflow,
2616 for languages where such a recovery is supported. The default value of
2617 75 words should be adequate for most machines.
2619 @findex STACK_CHECK_MAX_FRAME_SIZE
2620 @item STACK_CHECK_MAX_FRAME_SIZE
2621 The maximum size of a stack frame, in bytes. GCC will generate probe
2622 instructions in non-leaf functions to ensure at least this many bytes of
2623 stack are available. If a stack frame is larger than this size, stack
2624 checking will not be reliable and GCC will issue a warning. The
2625 default is chosen so that GCC only generates one instruction on most
2626 systems. You should normally not change the default value of this macro.
2628 @findex STACK_CHECK_FIXED_FRAME_SIZE
2629 @item STACK_CHECK_FIXED_FRAME_SIZE
2630 GCC uses this value to generate the above warning message. It
2631 represents the amount of fixed frame used by a function, not including
2632 space for any callee-saved registers, temporaries and user variables.
2633 You need only specify an upper bound for this amount and will normally
2634 use the default of four words.
2636 @findex STACK_CHECK_MAX_VAR_SIZE
2637 @item STACK_CHECK_MAX_VAR_SIZE
2638 The maximum size, in bytes, of an object that GCC will place in the
2639 fixed area of the stack frame when the user specifies
2640 @samp{-fstack-check}.
2641 GCC computed the default from the values of the above macros and you will
2642 normally not need to override that default.
2646 @node Frame Registers
2647 @subsection Registers That Address the Stack Frame
2649 @c prevent bad page break with this line
2650 This discusses registers that address the stack frame.
2653 @findex STACK_POINTER_REGNUM
2654 @item STACK_POINTER_REGNUM
2655 The register number of the stack pointer register, which must also be a
2656 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2657 the hardware determines which register this is.
2659 @findex FRAME_POINTER_REGNUM
2660 @item FRAME_POINTER_REGNUM
2661 The register number of the frame pointer register, which is used to
2662 access automatic variables in the stack frame. On some machines, the
2663 hardware determines which register this is. On other machines, you can
2664 choose any register you wish for this purpose.
2666 @findex HARD_FRAME_POINTER_REGNUM
2667 @item HARD_FRAME_POINTER_REGNUM
2668 On some machines the offset between the frame pointer and starting
2669 offset of the automatic variables is not known until after register
2670 allocation has been done (for example, because the saved registers are
2671 between these two locations). On those machines, define
2672 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
2673 be used internally until the offset is known, and define
2674 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
2675 used for the frame pointer.
2677 You should define this macro only in the very rare circumstances when it
2678 is not possible to calculate the offset between the frame pointer and
2679 the automatic variables until after register allocation has been
2680 completed. When this macro is defined, you must also indicate in your
2681 definition of @code{ELIMINABLE_REGS} how to eliminate
2682 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
2683 or @code{STACK_POINTER_REGNUM}.
2685 Do not define this macro if it would be the same as
2686 @code{FRAME_POINTER_REGNUM}.
2688 @findex ARG_POINTER_REGNUM
2689 @item ARG_POINTER_REGNUM
2690 The register number of the arg pointer register, which is used to access
2691 the function's argument list. On some machines, this is the same as the
2692 frame pointer register. On some machines, the hardware determines which
2693 register this is. On other machines, you can choose any register you
2694 wish for this purpose. If this is not the same register as the frame
2695 pointer register, then you must mark it as a fixed register according to
2696 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2697 (@pxref{Elimination}).
2699 @findex RETURN_ADDRESS_POINTER_REGNUM
2700 @item RETURN_ADDRESS_POINTER_REGNUM
2701 The register number of the return address pointer register, which is used to
2702 access the current function's return address from the stack. On some
2703 machines, the return address is not at a fixed offset from the frame
2704 pointer or stack pointer or argument pointer. This register can be defined
2705 to point to the return address on the stack, and then be converted by
2706 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
2708 Do not define this macro unless there is no other way to get the return
2709 address from the stack.
2711 @findex STATIC_CHAIN_REGNUM
2712 @findex STATIC_CHAIN_INCOMING_REGNUM
2713 @item STATIC_CHAIN_REGNUM
2714 @itemx STATIC_CHAIN_INCOMING_REGNUM
2715 Register numbers used for passing a function's static chain pointer. If
2716 register windows are used, the register number as seen by the called
2717 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
2718 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
2719 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
2720 not be defined.@refill
2722 The static chain register need not be a fixed register.
2724 If the static chain is passed in memory, these macros should not be
2725 defined; instead, the next two macros should be defined.
2727 @findex STATIC_CHAIN
2728 @findex STATIC_CHAIN_INCOMING
2730 @itemx STATIC_CHAIN_INCOMING
2731 If the static chain is passed in memory, these macros provide rtx giving
2732 @code{mem} expressions that denote where they are stored.
2733 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
2734 as seen by the calling and called functions, respectively. Often the former
2735 will be at an offset from the stack pointer and the latter at an offset from
2736 the frame pointer.@refill
2738 @findex stack_pointer_rtx
2739 @findex frame_pointer_rtx
2740 @findex arg_pointer_rtx
2741 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
2742 @code{arg_pointer_rtx} will have been initialized prior to the use of these
2743 macros and should be used to refer to those items.
2745 If the static chain is passed in a register, the two previous macros should
2750 @subsection Eliminating Frame Pointer and Arg Pointer
2752 @c prevent bad page break with this line
2753 This is about eliminating the frame pointer and arg pointer.
2756 @findex FRAME_POINTER_REQUIRED
2757 @item FRAME_POINTER_REQUIRED
2758 A C expression which is nonzero if a function must have and use a frame
2759 pointer. This expression is evaluated in the reload pass. If its value is
2760 nonzero the function will have a frame pointer.
2762 The expression can in principle examine the current function and decide
2763 according to the facts, but on most machines the constant 0 or the
2764 constant 1 suffices. Use 0 when the machine allows code to be generated
2765 with no frame pointer, and doing so saves some time or space. Use 1
2766 when there is no possible advantage to avoiding a frame pointer.
2768 In certain cases, the compiler does not know how to produce valid code
2769 without a frame pointer. The compiler recognizes those cases and
2770 automatically gives the function a frame pointer regardless of what
2771 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
2774 In a function that does not require a frame pointer, the frame pointer
2775 register can be allocated for ordinary usage, unless you mark it as a
2776 fixed register. See @code{FIXED_REGISTERS} for more information.
2778 @findex INITIAL_FRAME_POINTER_OFFSET
2779 @findex get_frame_size
2780 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
2781 A C statement to store in the variable @var{depth-var} the difference
2782 between the frame pointer and the stack pointer values immediately after
2783 the function prologue. The value would be computed from information
2784 such as the result of @code{get_frame_size ()} and the tables of
2785 registers @code{regs_ever_live} and @code{call_used_regs}.
2787 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
2788 need not be defined. Otherwise, it must be defined even if
2789 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
2790 case, you may set @var{depth-var} to anything.
2792 @findex ELIMINABLE_REGS
2793 @item ELIMINABLE_REGS
2794 If defined, this macro specifies a table of register pairs used to
2795 eliminate unneeded registers that point into the stack frame. If it is not
2796 defined, the only elimination attempted by the compiler is to replace
2797 references to the frame pointer with references to the stack pointer.
2799 The definition of this macro is a list of structure initializations, each
2800 of which specifies an original and replacement register.
2802 On some machines, the position of the argument pointer is not known until
2803 the compilation is completed. In such a case, a separate hard register
2804 must be used for the argument pointer. This register can be eliminated by
2805 replacing it with either the frame pointer or the argument pointer,
2806 depending on whether or not the frame pointer has been eliminated.
2808 In this case, you might specify:
2810 #define ELIMINABLE_REGS \
2811 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
2812 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
2813 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
2816 Note that the elimination of the argument pointer with the stack pointer is
2817 specified first since that is the preferred elimination.
2819 @findex CAN_ELIMINATE
2820 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
2821 A C expression that returns non-zero if the compiler is allowed to try
2822 to replace register number @var{from-reg} with register number
2823 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
2824 is defined, and will usually be the constant 1, since most of the cases
2825 preventing register elimination are things that the compiler already
2828 @findex INITIAL_ELIMINATION_OFFSET
2829 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
2830 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
2831 specifies the initial difference between the specified pair of
2832 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
2835 @findex LONGJMP_RESTORE_FROM_STACK
2836 @item LONGJMP_RESTORE_FROM_STACK
2837 Define this macro if the @code{longjmp} function restores registers from
2838 the stack frames, rather than from those saved specifically by
2839 @code{setjmp}. Certain quantities must not be kept in registers across
2840 a call to @code{setjmp} on such machines.
2843 @node Stack Arguments
2844 @subsection Passing Function Arguments on the Stack
2845 @cindex arguments on stack
2846 @cindex stack arguments
2848 The macros in this section control how arguments are passed
2849 on the stack. See the following section for other macros that
2850 control passing certain arguments in registers.
2853 @findex PROMOTE_PROTOTYPES
2854 @item PROMOTE_PROTOTYPES
2855 A C expression whose value is nonzero if an argument declared in
2856 a prototype as an integral type smaller than @code{int} should
2857 actually be passed as an @code{int}. In addition to avoiding
2858 errors in certain cases of mismatch, it also makes for better
2859 code on certain machines. If the macro is not defined in target
2860 header files, it defaults to 0.
2864 A C expression. If nonzero, push insns will be used to pass
2866 If the target machine does not have a push instruction, set it to zero.
2867 That directs GCC to use an alternate strategy: to
2868 allocate the entire argument block and then store the arguments into
2869 it. When PUSH_ARGS is nonzero, PUSH_ROUNDING must be defined too.
2870 On some machines, the definition
2872 @findex PUSH_ROUNDING
2873 @item PUSH_ROUNDING (@var{npushed})
2874 A C expression that is the number of bytes actually pushed onto the
2875 stack when an instruction attempts to push @var{npushed} bytes.
2877 On some machines, the definition
2880 #define PUSH_ROUNDING(BYTES) (BYTES)
2884 will suffice. But on other machines, instructions that appear
2885 to push one byte actually push two bytes in an attempt to maintain
2886 alignment. Then the definition should be
2889 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
2892 @findex ACCUMULATE_OUTGOING_ARGS
2893 @findex current_function_outgoing_args_size
2894 @item ACCUMULATE_OUTGOING_ARGS
2895 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
2896 will be computed and placed into the variable
2897 @code{current_function_outgoing_args_size}. No space will be pushed
2898 onto the stack for each call; instead, the function prologue should
2899 increase the stack frame size by this amount.
2901 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
2904 @findex REG_PARM_STACK_SPACE
2905 @item REG_PARM_STACK_SPACE (@var{fndecl})
2906 Define this macro if functions should assume that stack space has been
2907 allocated for arguments even when their values are passed in
2910 The value of this macro is the size, in bytes, of the area reserved for
2911 arguments passed in registers for the function represented by @var{fndecl},
2912 which can be zero if GCC is calling a library function.
2914 This space can be allocated by the caller, or be a part of the
2915 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
2917 @c above is overfull. not sure what to do. --mew 5feb93 did
2918 @c something, not sure if it looks good. --mew 10feb93
2920 @findex MAYBE_REG_PARM_STACK_SPACE
2921 @findex FINAL_REG_PARM_STACK_SPACE
2922 @item MAYBE_REG_PARM_STACK_SPACE
2923 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
2924 Define these macros in addition to the one above if functions might
2925 allocate stack space for arguments even when their values are passed
2926 in registers. These should be used when the stack space allocated
2927 for arguments in registers is not a simple constant independent of the
2928 function declaration.
2930 The value of the first macro is the size, in bytes, of the area that
2931 we should initially assume would be reserved for arguments passed in registers.
2933 The value of the second macro is the actual size, in bytes, of the area
2934 that will be reserved for arguments passed in registers. This takes two
2935 arguments: an integer representing the number of bytes of fixed sized
2936 arguments on the stack, and a tree representing the number of bytes of
2937 variable sized arguments on the stack.
2939 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
2940 called for libcall functions, the current function, or for a function
2941 being called when it is known that such stack space must be allocated.
2942 In each case this value can be easily computed.
2944 When deciding whether a called function needs such stack space, and how
2945 much space to reserve, GCC uses these two macros instead of
2946 @code{REG_PARM_STACK_SPACE}.
2948 @findex OUTGOING_REG_PARM_STACK_SPACE
2949 @item OUTGOING_REG_PARM_STACK_SPACE
2950 Define this if it is the responsibility of the caller to allocate the area
2951 reserved for arguments passed in registers.
2953 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
2954 whether the space for these arguments counts in the value of
2955 @code{current_function_outgoing_args_size}.
2957 @findex STACK_PARMS_IN_REG_PARM_AREA
2958 @item STACK_PARMS_IN_REG_PARM_AREA
2959 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
2960 stack parameters don't skip the area specified by it.
2961 @c i changed this, makes more sens and it should have taken care of the
2962 @c overfull.. not as specific, tho. --mew 5feb93
2964 Normally, when a parameter is not passed in registers, it is placed on the
2965 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
2966 suppresses this behavior and causes the parameter to be passed on the
2967 stack in its natural location.
2969 @findex RETURN_POPS_ARGS
2970 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
2971 A C expression that should indicate the number of bytes of its own
2972 arguments that a function pops on returning, or 0 if the
2973 function pops no arguments and the caller must therefore pop them all
2974 after the function returns.
2976 @var{fundecl} is a C variable whose value is a tree node that describes
2977 the function in question. Normally it is a node of type
2978 @code{FUNCTION_DECL} that describes the declaration of the function.
2979 From this you can obtain the DECL_MACHINE_ATTRIBUTES of the function.
2981 @var{funtype} is a C variable whose value is a tree node that
2982 describes the function in question. Normally it is a node of type
2983 @code{FUNCTION_TYPE} that describes the data type of the function.
2984 From this it is possible to obtain the data types of the value and
2985 arguments (if known).
2987 When a call to a library function is being considered, @var{fundecl}
2988 will contain an identifier node for the library function. Thus, if
2989 you need to distinguish among various library functions, you can do so
2990 by their names. Note that ``library function'' in this context means
2991 a function used to perform arithmetic, whose name is known specially
2992 in the compiler and was not mentioned in the C code being compiled.
2994 @var{stack-size} is the number of bytes of arguments passed on the
2995 stack. If a variable number of bytes is passed, it is zero, and
2996 argument popping will always be the responsibility of the calling function.
2998 On the Vax, all functions always pop their arguments, so the definition
2999 of this macro is @var{stack-size}. On the 68000, using the standard
3000 calling convention, no functions pop their arguments, so the value of
3001 the macro is always 0 in this case. But an alternative calling
3002 convention is available in which functions that take a fixed number of
3003 arguments pop them but other functions (such as @code{printf}) pop
3004 nothing (the caller pops all). When this convention is in use,
3005 @var{funtype} is examined to determine whether a function takes a fixed
3006 number of arguments.
3009 @node Register Arguments
3010 @subsection Passing Arguments in Registers
3011 @cindex arguments in registers
3012 @cindex registers arguments
3014 This section describes the macros which let you control how various
3015 types of arguments are passed in registers or how they are arranged in
3019 @findex FUNCTION_ARG
3020 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3021 A C expression that controls whether a function argument is passed
3022 in a register, and which register.
3024 The arguments are @var{cum}, which summarizes all the previous
3025 arguments; @var{mode}, the machine mode of the argument; @var{type},
3026 the data type of the argument as a tree node or 0 if that is not known
3027 (which happens for C support library functions); and @var{named},
3028 which is 1 for an ordinary argument and 0 for nameless arguments that
3029 correspond to @samp{@dots{}} in the called function's prototype.
3030 @var{type} can be an incomplete type if a syntax error has previously
3033 The value of the expression is usually either a @code{reg} RTX for the
3034 hard register in which to pass the argument, or zero to pass the
3035 argument on the stack.
3037 For machines like the Vax and 68000, where normally all arguments are
3038 pushed, zero suffices as a definition.
3040 The value of the expression can also be a @code{parallel} RTX. This is
3041 used when an argument is passed in multiple locations. The mode of the
3042 of the @code{parallel} should be the mode of the entire argument. The
3043 @code{parallel} holds any number of @code{expr_list} pairs; each one
3044 describes where part of the argument is passed. In each
3045 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3046 register in which to pass this part of the argument, and the mode of the
3047 register RTX indicates how large this part of the argument is. The
3048 second operand of the @code{expr_list} is a @code{const_int} which gives
3049 the offset in bytes into the entire argument of where this part starts.
3050 As a special exception the first @code{expr_list} in the @code{parallel}
3051 RTX may have a first operand of zero. This indicates that the entire
3052 argument is also stored on the stack.
3054 @cindex @file{stdarg.h} and register arguments
3055 The usual way to make the ISO library @file{stdarg.h} work on a machine
3056 where some arguments are usually passed in registers, is to cause
3057 nameless arguments to be passed on the stack instead. This is done
3058 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3060 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3061 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3062 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3063 in the definition of this macro to determine if this argument is of a
3064 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3065 is not defined and @code{FUNCTION_ARG} returns non-zero for such an
3066 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3067 defined, the argument will be computed in the stack and then loaded into
3070 @findex MUST_PASS_IN_STACK
3071 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
3072 Define as a C expression that evaluates to nonzero if we do not know how
3073 to pass TYPE solely in registers. The file @file{expr.h} defines a
3074 definition that is usually appropriate, refer to @file{expr.h} for additional
3077 @findex FUNCTION_INCOMING_ARG
3078 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3079 Define this macro if the target machine has ``register windows'', so
3080 that the register in which a function sees an arguments is not
3081 necessarily the same as the one in which the caller passed the
3084 For such machines, @code{FUNCTION_ARG} computes the register in which
3085 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3086 be defined in a similar fashion to tell the function being called
3087 where the arguments will arrive.
3089 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3090 serves both purposes.@refill
3092 @findex FUNCTION_ARG_PARTIAL_NREGS
3093 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3094 A C expression for the number of words, at the beginning of an
3095 argument, that must be put in registers. The value must be zero for
3096 arguments that are passed entirely in registers or that are entirely
3097 pushed on the stack.
3099 On some machines, certain arguments must be passed partially in
3100 registers and partially in memory. On these machines, typically the
3101 first @var{n} words of arguments are passed in registers, and the rest
3102 on the stack. If a multi-word argument (a @code{double} or a
3103 structure) crosses that boundary, its first few words must be passed
3104 in registers and the rest must be pushed. This macro tells the
3105 compiler when this occurs, and how many of the words should go in
3108 @code{FUNCTION_ARG} for these arguments should return the first
3109 register to be used by the caller for this argument; likewise
3110 @code{FUNCTION_INCOMING_ARG}, for the called function.
3112 @findex FUNCTION_ARG_PASS_BY_REFERENCE
3113 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3114 A C expression that indicates when an argument must be passed by reference.
3115 If nonzero for an argument, a copy of that argument is made in memory and a
3116 pointer to the argument is passed instead of the argument itself.
3117 The pointer is passed in whatever way is appropriate for passing a pointer
3120 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3121 definition of this macro might be
3123 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3124 (CUM, MODE, TYPE, NAMED) \
3125 MUST_PASS_IN_STACK (MODE, TYPE)
3127 @c this is *still* too long. --mew 5feb93
3129 @findex FUNCTION_ARG_CALLEE_COPIES
3130 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3131 If defined, a C expression that indicates when it is the called function's
3132 responsibility to make a copy of arguments passed by invisible reference.
3133 Normally, the caller makes a copy and passes the address of the copy to the
3134 routine being called. When FUNCTION_ARG_CALLEE_COPIES is defined and is
3135 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3136 ``live'' value. The called function must not modify this value. If it can be
3137 determined that the value won't be modified, it need not make a copy;
3138 otherwise a copy must be made.
3140 @findex CUMULATIVE_ARGS
3141 @item CUMULATIVE_ARGS
3142 A C type for declaring a variable that is used as the first argument of
3143 @code{FUNCTION_ARG} and other related values. For some target machines,
3144 the type @code{int} suffices and can hold the number of bytes of
3147 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3148 arguments that have been passed on the stack. The compiler has other
3149 variables to keep track of that. For target machines on which all
3150 arguments are passed on the stack, there is no need to store anything in
3151 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3152 should not be empty, so use @code{int}.
3154 @findex INIT_CUMULATIVE_ARGS
3155 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
3156 A C statement (sans semicolon) for initializing the variable @var{cum}
3157 for the state at the beginning of the argument list. The variable has
3158 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
3159 for the data type of the function which will receive the args, or 0
3160 if the args are to a compiler support library function. The value of
3161 @var{indirect} is nonzero when processing an indirect call, for example
3162 a call through a function pointer. The value of @var{indirect} is zero
3163 for a call to an explicitly named function, a library function call, or when
3164 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3167 When processing a call to a compiler support library function,
3168 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3169 contains the name of the function, as a string. @var{libname} is 0 when
3170 an ordinary C function call is being processed. Thus, each time this
3171 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3172 never both of them at once.
3174 @findex INIT_CUMULATIVE_LIBCALL_ARGS
3175 @item INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3176 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3177 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3178 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3179 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3180 0)} is used instead.
3182 @findex INIT_CUMULATIVE_INCOMING_ARGS
3183 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3184 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3185 finding the arguments for the function being compiled. If this macro is
3186 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3188 The value passed for @var{libname} is always 0, since library routines
3189 with special calling conventions are never compiled with GCC. The
3190 argument @var{libname} exists for symmetry with
3191 @code{INIT_CUMULATIVE_ARGS}.
3192 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3193 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3195 @findex FUNCTION_ARG_ADVANCE
3196 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3197 A C statement (sans semicolon) to update the summarizer variable
3198 @var{cum} to advance past an argument in the argument list. The
3199 values @var{mode}, @var{type} and @var{named} describe that argument.
3200 Once this is done, the variable @var{cum} is suitable for analyzing
3201 the @emph{following} argument with @code{FUNCTION_ARG}, etc.@refill
3203 This macro need not do anything if the argument in question was passed
3204 on the stack. The compiler knows how to track the amount of stack space
3205 used for arguments without any special help.
3207 @findex FUNCTION_ARG_PADDING
3208 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3209 If defined, a C expression which determines whether, and in which direction,
3210 to pad out an argument with extra space. The value should be of type
3211 @code{enum direction}: either @code{upward} to pad above the argument,
3212 @code{downward} to pad below, or @code{none} to inhibit padding.
3214 The @emph{amount} of padding is always just enough to reach the next
3215 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3218 This macro has a default definition which is right for most systems.
3219 For little-endian machines, the default is to pad upward. For
3220 big-endian machines, the default is to pad downward for an argument of
3221 constant size shorter than an @code{int}, and upward otherwise.
3223 @findex PAD_VARARGS_DOWN
3224 @item PAD_VARARGS_DOWN
3225 If defined, a C expression which determines whether the default
3226 implementation of va_arg will attempt to pad down before reading the
3227 next argument, if that argument is smaller than its aligned space as
3228 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3229 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3231 @findex FUNCTION_ARG_BOUNDARY
3232 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3233 If defined, a C expression that gives the alignment boundary, in bits,
3234 of an argument with the specified mode and type. If it is not defined,
3235 @code{PARM_BOUNDARY} is used for all arguments.
3237 @findex FUNCTION_ARG_REGNO_P
3238 @item FUNCTION_ARG_REGNO_P (@var{regno})
3239 A C expression that is nonzero if @var{regno} is the number of a hard
3240 register in which function arguments are sometimes passed. This does
3241 @emph{not} include implicit arguments such as the static chain and
3242 the structure-value address. On many machines, no registers can be
3243 used for this purpose since all function arguments are pushed on the
3246 @findex LOAD_ARGS_REVERSED
3247 @item LOAD_ARGS_REVERSED
3248 If defined, the order in which arguments are loaded into their
3249 respective argument registers is reversed so that the last
3250 argument is loaded first. This macro only affects arguments
3251 passed in registers.
3256 @subsection How Scalar Function Values Are Returned
3257 @cindex return values in registers
3258 @cindex values, returned by functions
3259 @cindex scalars, returned as values
3261 This section discusses the macros that control returning scalars as
3262 values---values that can fit in registers.
3265 @findex TRADITIONAL_RETURN_FLOAT
3266 @item TRADITIONAL_RETURN_FLOAT
3267 Define this macro if @samp{-traditional} should not cause functions
3268 declared to return @code{float} to convert the value to @code{double}.
3270 @findex FUNCTION_VALUE
3271 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3272 A C expression to create an RTX representing the place where a
3273 function returns a value of data type @var{valtype}. @var{valtype} is
3274 a tree node representing a data type. Write @code{TYPE_MODE
3275 (@var{valtype})} to get the machine mode used to represent that type.
3276 On many machines, only the mode is relevant. (Actually, on most
3277 machines, scalar values are returned in the same place regardless of
3280 The value of the expression is usually a @code{reg} RTX for the hard
3281 register where the return value is stored. The value can also be a
3282 @code{parallel} RTX, if the return value is in multiple places. See
3283 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3285 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3286 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3289 If the precise function being called is known, @var{func} is a tree
3290 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3291 pointer. This makes it possible to use a different value-returning
3292 convention for specific functions when all their calls are
3295 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3296 types, because these are returned in another way. See
3297 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3299 @findex FUNCTION_OUTGOING_VALUE
3300 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3301 Define this macro if the target machine has ``register windows''
3302 so that the register in which a function returns its value is not
3303 the same as the one in which the caller sees the value.
3305 For such machines, @code{FUNCTION_VALUE} computes the register in which
3306 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3307 defined in a similar fashion to tell the function where to put the
3310 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3311 @code{FUNCTION_VALUE} serves both purposes.@refill
3313 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3314 aggregate data types, because these are returned in another way. See
3315 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3317 @findex LIBCALL_VALUE
3318 @item LIBCALL_VALUE (@var{mode})
3319 A C expression to create an RTX representing the place where a library
3320 function returns a value of mode @var{mode}. If the precise function
3321 being called is known, @var{func} is a tree node
3322 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3323 pointer. This makes it possible to use a different value-returning
3324 convention for specific functions when all their calls are
3327 Note that ``library function'' in this context means a compiler
3328 support routine, used to perform arithmetic, whose name is known
3329 specially by the compiler and was not mentioned in the C code being
3332 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3333 data types, because none of the library functions returns such types.
3335 @findex FUNCTION_VALUE_REGNO_P
3336 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3337 A C expression that is nonzero if @var{regno} is the number of a hard
3338 register in which the values of called function may come back.
3340 A register whose use for returning values is limited to serving as the
3341 second of a pair (for a value of type @code{double}, say) need not be
3342 recognized by this macro. So for most machines, this definition
3346 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3349 If the machine has register windows, so that the caller and the called
3350 function use different registers for the return value, this macro
3351 should recognize only the caller's register numbers.
3353 @findex APPLY_RESULT_SIZE
3354 @item APPLY_RESULT_SIZE
3355 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3356 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3357 saving and restoring an arbitrary return value.
3360 @node Aggregate Return
3361 @subsection How Large Values Are Returned
3362 @cindex aggregates as return values
3363 @cindex large return values
3364 @cindex returning aggregate values
3365 @cindex structure value address
3367 When a function value's mode is @code{BLKmode} (and in some other
3368 cases), the value is not returned according to @code{FUNCTION_VALUE}
3369 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3370 block of memory in which the value should be stored. This address
3371 is called the @dfn{structure value address}.
3373 This section describes how to control returning structure values in
3377 @findex RETURN_IN_MEMORY
3378 @item RETURN_IN_MEMORY (@var{type})
3379 A C expression which can inhibit the returning of certain function
3380 values in registers, based on the type of value. A nonzero value says
3381 to return the function value in memory, just as large structures are
3382 always returned. Here @var{type} will be a C expression of type
3383 @code{tree}, representing the data type of the value.
3385 Note that values of mode @code{BLKmode} must be explicitly handled
3386 by this macro. Also, the option @samp{-fpcc-struct-return}
3387 takes effect regardless of this macro. On most systems, it is
3388 possible to leave the macro undefined; this causes a default
3389 definition to be used, whose value is the constant 1 for @code{BLKmode}
3390 values, and 0 otherwise.
3392 Do not use this macro to indicate that structures and unions should always
3393 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3396 @findex DEFAULT_PCC_STRUCT_RETURN
3397 @item DEFAULT_PCC_STRUCT_RETURN
3398 Define this macro to be 1 if all structure and union return values must be
3399 in memory. Since this results in slower code, this should be defined
3400 only if needed for compatibility with other compilers or with an ABI.
3401 If you define this macro to be 0, then the conventions used for structure
3402 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3404 If not defined, this defaults to the value 1.
3406 @findex STRUCT_VALUE_REGNUM
3407 @item STRUCT_VALUE_REGNUM
3408 If the structure value address is passed in a register, then
3409 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3411 @findex STRUCT_VALUE
3413 If the structure value address is not passed in a register, define
3414 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3415 where the address is passed. If it returns 0, the address is passed as
3416 an ``invisible'' first argument.
3418 @findex STRUCT_VALUE_INCOMING_REGNUM
3419 @item STRUCT_VALUE_INCOMING_REGNUM
3420 On some architectures the place where the structure value address
3421 is found by the called function is not the same place that the
3422 caller put it. This can be due to register windows, or it could
3423 be because the function prologue moves it to a different place.
3425 If the incoming location of the structure value address is in a
3426 register, define this macro as the register number.
3428 @findex STRUCT_VALUE_INCOMING
3429 @item STRUCT_VALUE_INCOMING
3430 If the incoming location is not a register, then you should define
3431 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3432 called function should find the value. If it should find the value on
3433 the stack, define this to create a @code{mem} which refers to the frame
3434 pointer. A definition of 0 means that the address is passed as an
3435 ``invisible'' first argument.
3437 @findex PCC_STATIC_STRUCT_RETURN
3438 @item PCC_STATIC_STRUCT_RETURN
3439 Define this macro if the usual system convention on the target machine
3440 for returning structures and unions is for the called function to return
3441 the address of a static variable containing the value.
3443 Do not define this if the usual system convention is for the caller to
3444 pass an address to the subroutine.
3446 This macro has effect in @samp{-fpcc-struct-return} mode, but it does
3447 nothing when you use @samp{-freg-struct-return} mode.
3451 @subsection Caller-Saves Register Allocation
3453 If you enable it, GCC can save registers around function calls. This
3454 makes it possible to use call-clobbered registers to hold variables that
3455 must live across calls.
3458 @findex DEFAULT_CALLER_SAVES
3459 @item DEFAULT_CALLER_SAVES
3460 Define this macro if function calls on the target machine do not preserve
3461 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3462 for all registers. When defined, this macro enables @samp{-fcaller-saves}
3463 by default for all optimization levels. It has no effect for optimization
3464 levels 2 and higher, where @samp{-fcaller-saves} is the default.
3466 @findex CALLER_SAVE_PROFITABLE
3467 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3468 A C expression to determine whether it is worthwhile to consider placing
3469 a pseudo-register in a call-clobbered hard register and saving and
3470 restoring it around each function call. The expression should be 1 when
3471 this is worth doing, and 0 otherwise.
3473 If you don't define this macro, a default is used which is good on most
3474 machines: @code{4 * @var{calls} < @var{refs}}.
3476 @findex HARD_REGNO_CALLER_SAVE_MODE
3477 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3478 A C expression specifying which mode is required for saving @var{nregs}
3479 of a pseudo-register in call-clobbered hard register @var{regno}. If
3480 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3481 returned. For most machines this macro need not be defined since GCC
3482 will select the smallest suitable mode.
3485 @node Function Entry
3486 @subsection Function Entry and Exit
3487 @cindex function entry and exit
3491 This section describes the macros that output function entry
3492 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3495 @findex FUNCTION_PROLOGUE
3496 @item FUNCTION_PROLOGUE (@var{file}, @var{size})
3497 A C compound statement that outputs the assembler code for entry to a
3498 function. The prologue is responsible for setting up the stack frame,
3499 initializing the frame pointer register, saving registers that must be
3500 saved, and allocating @var{size} additional bytes of storage for the
3501 local variables. @var{size} is an integer. @var{file} is a stdio
3502 stream to which the assembler code should be output.
3504 The label for the beginning of the function need not be output by this
3505 macro. That has already been done when the macro is run.
3507 @findex regs_ever_live
3508 To determine which registers to save, the macro can refer to the array
3509 @code{regs_ever_live}: element @var{r} is nonzero if hard register
3510 @var{r} is used anywhere within the function. This implies the function
3511 prologue should save register @var{r}, provided it is not one of the
3512 call-used registers. (@code{FUNCTION_EPILOGUE} must likewise use
3513 @code{regs_ever_live}.)
3515 On machines that have ``register windows'', the function entry code does
3516 not save on the stack the registers that are in the windows, even if
3517 they are supposed to be preserved by function calls; instead it takes
3518 appropriate steps to ``push'' the register stack, if any non-call-used
3519 registers are used in the function.
3521 @findex frame_pointer_needed
3522 On machines where functions may or may not have frame-pointers, the
3523 function entry code must vary accordingly; it must set up the frame
3524 pointer if one is wanted, and not otherwise. To determine whether a
3525 frame pointer is in wanted, the macro can refer to the variable
3526 @code{frame_pointer_needed}. The variable's value will be 1 at run
3527 time in a function that needs a frame pointer. @xref{Elimination}.
3529 The function entry code is responsible for allocating any stack space
3530 required for the function. This stack space consists of the regions
3531 listed below. In most cases, these regions are allocated in the
3532 order listed, with the last listed region closest to the top of the
3533 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
3534 the highest address if it is not defined). You can use a different order
3535 for a machine if doing so is more convenient or required for
3536 compatibility reasons. Except in cases where required by standard
3537 or by a debugger, there is no reason why the stack layout used by GCC
3538 need agree with that used by other compilers for a machine.
3542 @findex current_function_pretend_args_size
3543 A region of @code{current_function_pretend_args_size} bytes of
3544 uninitialized space just underneath the first argument arriving on the
3545 stack. (This may not be at the very start of the allocated stack region
3546 if the calling sequence has pushed anything else since pushing the stack
3547 arguments. But usually, on such machines, nothing else has been pushed
3548 yet, because the function prologue itself does all the pushing.) This
3549 region is used on machines where an argument may be passed partly in
3550 registers and partly in memory, and, in some cases to support the
3551 features in @file{varargs.h} and @file{stdargs.h}.
3554 An area of memory used to save certain registers used by the function.
3555 The size of this area, which may also include space for such things as
3556 the return address and pointers to previous stack frames, is
3557 machine-specific and usually depends on which registers have been used
3558 in the function. Machines with register windows often do not require
3562 A region of at least @var{size} bytes, possibly rounded up to an allocation
3563 boundary, to contain the local variables of the function. On some machines,
3564 this region and the save area may occur in the opposite order, with the
3565 save area closer to the top of the stack.
3568 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3569 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3570 @code{current_function_outgoing_args_size} bytes to be used for outgoing
3571 argument lists of the function. @xref{Stack Arguments}.
3574 Normally, it is necessary for the macros @code{FUNCTION_PROLOGUE} and
3575 @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C
3576 variable @code{current_function_is_leaf} is nonzero for such a function.
3578 @findex EXIT_IGNORE_STACK
3579 @item EXIT_IGNORE_STACK
3580 Define this macro as a C expression that is nonzero if the return
3581 instruction or the function epilogue ignores the value of the stack
3582 pointer; in other words, if it is safe to delete an instruction to
3583 adjust the stack pointer before a return from the function.
3585 Note that this macro's value is relevant only for functions for which
3586 frame pointers are maintained. It is never safe to delete a final
3587 stack adjustment in a function that has no frame pointer, and the
3588 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3590 @findex EPILOGUE_USES
3591 @item EPILOGUE_USES (@var{regno})
3592 Define this macro as a C expression that is nonzero for registers that are
3593 used by the epilogue or the @samp{return} pattern. The stack and frame
3594 pointer registers are already be assumed to be used as needed.
3596 @findex FUNCTION_EPILOGUE
3597 @item FUNCTION_EPILOGUE (@var{file}, @var{size})
3598 A C compound statement that outputs the assembler code for exit from a
3599 function. The epilogue is responsible for restoring the saved
3600 registers and stack pointer to their values when the function was
3601 called, and returning control to the caller. This macro takes the
3602 same arguments as the macro @code{FUNCTION_PROLOGUE}, and the
3603 registers to restore are determined from @code{regs_ever_live} and
3604 @code{CALL_USED_REGISTERS} in the same way.
3606 On some machines, there is a single instruction that does all the work
3607 of returning from the function. On these machines, give that
3608 instruction the name @samp{return} and do not define the macro
3609 @code{FUNCTION_EPILOGUE} at all.
3611 Do not define a pattern named @samp{return} if you want the
3612 @code{FUNCTION_EPILOGUE} to be used. If you want the target switches
3613 to control whether return instructions or epilogues are used, define a
3614 @samp{return} pattern with a validity condition that tests the target
3615 switches appropriately. If the @samp{return} pattern's validity
3616 condition is false, epilogues will be used.
3618 On machines where functions may or may not have frame-pointers, the
3619 function exit code must vary accordingly. Sometimes the code for these
3620 two cases is completely different. To determine whether a frame pointer
3621 is wanted, the macro can refer to the variable
3622 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
3623 a function that needs a frame pointer.
3625 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
3626 treat leaf functions specially. The C variable @code{current_function_is_leaf}
3627 is nonzero for such a function. @xref{Leaf Functions}.
3629 On some machines, some functions pop their arguments on exit while
3630 others leave that for the caller to do. For example, the 68020 when
3631 given @samp{-mrtd} pops arguments in functions that take a fixed
3632 number of arguments.
3634 @findex current_function_pops_args
3635 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
3636 functions pop their own arguments. @code{FUNCTION_EPILOGUE} needs to
3637 know what was decided. The variable that is called
3638 @code{current_function_pops_args} is the number of bytes of its
3639 arguments that a function should pop. @xref{Scalar Return}.
3640 @c what is the "its arguments" in the above sentence referring to, pray
3641 @c tell? --mew 5feb93
3643 @findex DELAY_SLOTS_FOR_EPILOGUE
3644 @item DELAY_SLOTS_FOR_EPILOGUE
3645 Define this macro if the function epilogue contains delay slots to which
3646 instructions from the rest of the function can be ``moved''. The
3647 definition should be a C expression whose value is an integer
3648 representing the number of delay slots there.
3650 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
3651 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
3652 A C expression that returns 1 if @var{insn} can be placed in delay
3653 slot number @var{n} of the epilogue.
3655 The argument @var{n} is an integer which identifies the delay slot now
3656 being considered (since different slots may have different rules of
3657 eligibility). It is never negative and is always less than the number
3658 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
3659 If you reject a particular insn for a given delay slot, in principle, it
3660 may be reconsidered for a subsequent delay slot. Also, other insns may
3661 (at least in principle) be considered for the so far unfilled delay
3664 @findex current_function_epilogue_delay_list
3665 @findex final_scan_insn
3666 The insns accepted to fill the epilogue delay slots are put in an RTL
3667 list made with @code{insn_list} objects, stored in the variable
3668 @code{current_function_epilogue_delay_list}. The insn for the first
3669 delay slot comes first in the list. Your definition of the macro
3670 @code{FUNCTION_EPILOGUE} should fill the delay slots by outputting the
3671 insns in this list, usually by calling @code{final_scan_insn}.
3673 You need not define this macro if you did not define
3674 @code{DELAY_SLOTS_FOR_EPILOGUE}.
3676 @findex ASM_OUTPUT_MI_THUNK
3677 @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function})
3678 A C compound statement that outputs the assembler code for a thunk
3679 function, used to implement C++ virtual function calls with multiple
3680 inheritance. The thunk acts as a wrapper around a virtual function,
3681 adjusting the implicit object parameter before handing control off to
3684 First, emit code to add the integer @var{delta} to the location that
3685 contains the incoming first argument. Assume that this argument
3686 contains a pointer, and is the one used to pass the @code{this} pointer
3687 in C++. This is the incoming argument @emph{before} the function prologue,
3688 e.g. @samp{%o0} on a sparc. The addition must preserve the values of
3689 all other incoming arguments.
3691 After the addition, emit code to jump to @var{function}, which is a
3692 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
3693 not touch the return address. Hence returning from @var{FUNCTION} will
3694 return to whoever called the current @samp{thunk}.
3696 The effect must be as if @var{function} had been called directly with
3697 the adjusted first argument. This macro is responsible for emitting all
3698 of the code for a thunk function; @code{FUNCTION_PROLOGUE} and
3699 @code{FUNCTION_EPILOGUE} are not invoked.
3701 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
3702 have already been extracted from it.) It might possibly be useful on
3703 some targets, but probably not.
3705 If you do not define this macro, the target-independent code in the C++
3706 frontend will generate a less efficient heavyweight thunk that calls
3707 @var{function} instead of jumping to it. The generic approach does
3708 not support varargs.
3712 @subsection Generating Code for Profiling
3713 @cindex profiling, code generation
3715 These macros will help you generate code for profiling.
3718 @findex FUNCTION_PROFILER
3719 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
3720 A C statement or compound statement to output to @var{file} some
3721 assembler code to call the profiling subroutine @code{mcount}.
3724 The details of how @code{mcount} expects to be called are determined by
3725 your operating system environment, not by GCC. To figure them out,
3726 compile a small program for profiling using the system's installed C
3727 compiler and look at the assembler code that results.
3729 Older implementations of @code{mcount} expect the address of a counter
3730 variable to be loaded into some register. The name of this variable is
3731 @samp{LP} followed by the number @var{labelno}, so you would generate
3732 the name using @samp{LP%d} in a @code{fprintf}.
3734 @findex PROFILE_HOOK
3736 A C statement or compound statement to output to @var{file} some assembly
3737 code to call the profiling subroutine @code{mcount} even the target does
3738 not support profiling.
3740 @findex NO_PROFILE_COUNTERS
3741 @item NO_PROFILE_COUNTERS
3742 Define this macro if the @code{mcount} subroutine on your system does
3743 not need a counter variable allocated for each function. This is true
3744 for almost all modern implementations. If you define this macro, you
3745 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
3747 @findex PROFILE_BEFORE_PROLOGUE
3748 @item PROFILE_BEFORE_PROLOGUE
3749 Define this macro if the code for function profiling should come before
3750 the function prologue. Normally, the profiling code comes after.
3752 @findex FUNCTION_BLOCK_PROFILER
3753 @vindex profile_block_flag
3754 @item FUNCTION_BLOCK_PROFILER (@var{file}, @var{labelno})
3755 A C statement or compound statement to output to @var{file} some
3756 assembler code to initialize basic-block profiling for the current
3757 object module. The global compile flag @code{profile_block_flag}
3758 distinguishes two profile modes.
3761 @findex __bb_init_func
3762 @item profile_block_flag != 2
3763 Output code to call the subroutine @code{__bb_init_func} once per
3764 object module, passing it as its sole argument the address of a block
3765 allocated in the object module.
3767 The name of the block is a local symbol made with this statement:
3770 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3773 Of course, since you are writing the definition of
3774 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3775 can take a short cut in the definition of this macro and use the name
3776 that you know will result.
3778 The first word of this block is a flag which will be nonzero if the
3779 object module has already been initialized. So test this word first,
3780 and do not call @code{__bb_init_func} if the flag is
3781 nonzero. BLOCK_OR_LABEL contains a unique number which may be used to
3782 generate a label as a branch destination when @code{__bb_init_func}
3785 Described in assembler language, the code to be output looks like:
3795 @findex __bb_init_trace_func
3796 @item profile_block_flag == 2
3797 Output code to call the subroutine @code{__bb_init_trace_func}
3798 and pass two parameters to it. The first parameter is the same as
3799 for @code{__bb_init_func}. The second parameter is the number of the
3800 first basic block of the function as given by BLOCK_OR_LABEL. Note
3801 that @code{__bb_init_trace_func} has to be called, even if the object
3802 module has been initialized already.
3804 Described in assembler language, the code to be output looks like:
3807 parameter2 <- BLOCK_OR_LABEL
3808 call __bb_init_trace_func
3812 @findex BLOCK_PROFILER
3813 @vindex profile_block_flag
3814 @item BLOCK_PROFILER (@var{file}, @var{blockno})
3815 A C statement or compound statement to output to @var{file} some
3816 assembler code to increment the count associated with the basic
3817 block number @var{blockno}. The global compile flag
3818 @code{profile_block_flag} distinguishes two profile modes.
3821 @item profile_block_flag != 2
3822 Output code to increment the counter directly. Basic blocks are
3823 numbered separately from zero within each compilation. The count
3824 associated with block number @var{blockno} is at index
3825 @var{blockno} in a vector of words; the name of this array is a local
3826 symbol made with this statement:
3829 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 2);
3832 @c This paragraph is the same as one a few paragraphs up.
3833 @c That is not an error.
3834 Of course, since you are writing the definition of
3835 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3836 can take a short cut in the definition of this macro and use the name
3837 that you know will result.
3839 Described in assembler language, the code to be output looks like:
3842 inc (LPBX2+4*BLOCKNO)
3846 @findex __bb_trace_func
3847 @item profile_block_flag == 2
3848 Output code to initialize the global structure @code{__bb} and
3849 call the function @code{__bb_trace_func}, which will increment the
3852 @code{__bb} consists of two words. In the first word, the current
3853 basic block number, as given by BLOCKNO, has to be stored. In
3854 the second word, the address of a block allocated in the object
3855 module has to be stored. The address is given by the label created
3856 with this statement:
3859 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3862 Described in assembler language, the code to be output looks like:
3864 move BLOCKNO -> (__bb)
3865 move LPBX0 -> (__bb+4)
3866 call __bb_trace_func
3870 @findex FUNCTION_BLOCK_PROFILER_EXIT
3871 @findex __bb_trace_ret
3872 @vindex profile_block_flag
3873 @item FUNCTION_BLOCK_PROFILER_EXIT (@var{file})
3874 A C statement or compound statement to output to @var{file}
3875 assembler code to call function @code{__bb_trace_ret}. The
3876 assembler code should only be output
3877 if the global compile flag @code{profile_block_flag} == 2. This
3878 macro has to be used at every place where code for returning from
3879 a function is generated (e.g. @code{FUNCTION_EPILOGUE}). Although
3880 you have to write the definition of @code{FUNCTION_EPILOGUE}
3881 as well, you have to define this macro to tell the compiler, that
3882 the proper call to @code{__bb_trace_ret} is produced.
3884 @findex MACHINE_STATE_SAVE
3885 @findex __bb_init_trace_func
3886 @findex __bb_trace_func
3887 @findex __bb_trace_ret
3888 @item MACHINE_STATE_SAVE (@var{id})
3889 A C statement or compound statement to save all registers, which may
3890 be clobbered by a function call, including condition codes. The
3891 @code{asm} statement will be mostly likely needed to handle this
3892 task. Local labels in the assembler code can be concatenated with the
3893 string @var{id}, to obtain a unique label name.
3895 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3896 @code{FUNCTION_EPILOGUE} must be saved in the macros
3897 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3898 @code{BLOCK_PROFILER} prior calling @code{__bb_init_trace_func},
3899 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3901 @findex MACHINE_STATE_RESTORE
3902 @findex __bb_init_trace_func
3903 @findex __bb_trace_func
3904 @findex __bb_trace_ret
3905 @item MACHINE_STATE_RESTORE (@var{id})
3906 A C statement or compound statement to restore all registers, including
3907 condition codes, saved by @code{MACHINE_STATE_SAVE}.
3909 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3910 @code{FUNCTION_EPILOGUE} must be restored in the macros
3911 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3912 @code{BLOCK_PROFILER} after calling @code{__bb_init_trace_func},
3913 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3915 @findex BLOCK_PROFILER_CODE
3916 @item BLOCK_PROFILER_CODE
3917 A C function or functions which are needed in the library to
3918 support block profiling.
3922 @subsection Permitting inlining of functions with attributes
3925 By default if a function has a target specific attribute attached to it,
3926 it will not be inlined. This behaviour can be overridden if the target
3927 defines the @samp{FUNCTION_ATTRIBUTE_INLINABLE_P} macro. This macro
3928 takes one argument, a @samp{DECL} describing the function. It should
3929 return non-zero if the function can be inlined, otherwise it should
3933 @subsection Permitting tail calls to functions
3935 @cindex sibling calls
3938 @findex FUNCTION_OK_FOR_SIBCALL
3939 @item FUNCTION_OK_FOR_SIBCALL (@var{decl})
3940 A C expression that evaluates to true if it is ok to perform a sibling
3943 It is not uncommon for limitations of calling conventions to prevent
3944 tail calls to functions outside the current unit of translation, or
3945 during PIC compilation. Use this macro to enforce these restrictions,
3946 as the @code{sibcall} md pattern can not fail, or fall over to a
3951 @section Implementing the Varargs Macros
3952 @cindex varargs implementation
3954 GCC comes with an implementation of @file{varargs.h} and
3955 @file{stdarg.h} that work without change on machines that pass arguments
3956 on the stack. Other machines require their own implementations of
3957 varargs, and the two machine independent header files must have
3958 conditionals to include it.
3960 ISO @file{stdarg.h} differs from traditional @file{varargs.h} mainly in
3961 the calling convention for @code{va_start}. The traditional
3962 implementation takes just one argument, which is the variable in which
3963 to store the argument pointer. The ISO implementation of
3964 @code{va_start} takes an additional second argument. The user is
3965 supposed to write the last named argument of the function here.
3967 However, @code{va_start} should not use this argument. The way to find
3968 the end of the named arguments is with the built-in functions described
3972 @findex __builtin_saveregs
3973 @item __builtin_saveregs ()
3974 Use this built-in function to save the argument registers in memory so
3975 that the varargs mechanism can access them. Both ISO and traditional
3976 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3977 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
3979 On some machines, @code{__builtin_saveregs} is open-coded under the
3980 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
3981 it calls a routine written in assembler language, found in
3984 Code generated for the call to @code{__builtin_saveregs} appears at the
3985 beginning of the function, as opposed to where the call to
3986 @code{__builtin_saveregs} is written, regardless of what the code is.
3987 This is because the registers must be saved before the function starts
3988 to use them for its own purposes.
3989 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3992 @findex __builtin_args_info
3993 @item __builtin_args_info (@var{category})
3994 Use this built-in function to find the first anonymous arguments in
3997 In general, a machine may have several categories of registers used for
3998 arguments, each for a particular category of data types. (For example,
3999 on some machines, floating-point registers are used for floating-point
4000 arguments while other arguments are passed in the general registers.)
4001 To make non-varargs functions use the proper calling convention, you
4002 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4003 registers in each category have been used so far
4005 @code{__builtin_args_info} accesses the same data structure of type
4006 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4007 with it, with @var{category} specifying which word to access. Thus, the
4008 value indicates the first unused register in a given category.
4010 Normally, you would use @code{__builtin_args_info} in the implementation
4011 of @code{va_start}, accessing each category just once and storing the
4012 value in the @code{va_list} object. This is because @code{va_list} will
4013 have to update the values, and there is no way to alter the
4014 values accessed by @code{__builtin_args_info}.
4016 @findex __builtin_next_arg
4017 @item __builtin_next_arg (@var{lastarg})
4018 This is the equivalent of @code{__builtin_args_info}, for stack
4019 arguments. It returns the address of the first anonymous stack
4020 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4021 returns the address of the location above the first anonymous stack
4022 argument. Use it in @code{va_start} to initialize the pointer for
4023 fetching arguments from the stack. Also use it in @code{va_start} to
4024 verify that the second parameter @var{lastarg} is the last named argument
4025 of the current function.
4027 @findex __builtin_classify_type
4028 @item __builtin_classify_type (@var{object})
4029 Since each machine has its own conventions for which data types are
4030 passed in which kind of register, your implementation of @code{va_arg}
4031 has to embody these conventions. The easiest way to categorize the
4032 specified data type is to use @code{__builtin_classify_type} together
4033 with @code{sizeof} and @code{__alignof__}.
4035 @code{__builtin_classify_type} ignores the value of @var{object},
4036 considering only its data type. It returns an integer describing what
4037 kind of type that is---integer, floating, pointer, structure, and so on.
4039 The file @file{typeclass.h} defines an enumeration that you can use to
4040 interpret the values of @code{__builtin_classify_type}.
4043 These machine description macros help implement varargs:
4046 @findex EXPAND_BUILTIN_SAVEREGS
4047 @item EXPAND_BUILTIN_SAVEREGS ()
4048 If defined, is a C expression that produces the machine-specific code
4049 for a call to @code{__builtin_saveregs}. This code will be moved to the
4050 very beginning of the function, before any parameter access are made.
4051 The return value of this function should be an RTX that contains the
4052 value to use as the return of @code{__builtin_saveregs}.
4054 @findex SETUP_INCOMING_VARARGS
4055 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
4056 This macro offers an alternative to using @code{__builtin_saveregs} and
4057 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
4058 anonymous register arguments into the stack so that all the arguments
4059 appear to have been passed consecutively on the stack. Once this is
4060 done, you can use the standard implementation of varargs that works for
4061 machines that pass all their arguments on the stack.
4063 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
4064 structure, containing the values that are obtained after processing the
4065 named arguments. The arguments @var{mode} and @var{type} describe the
4066 last named argument---its machine mode and its data type as a tree node.
4068 The macro implementation should do two things: first, push onto the
4069 stack all the argument registers @emph{not} used for the named
4070 arguments, and second, store the size of the data thus pushed into the
4071 @code{int}-valued variable whose name is supplied as the argument
4072 @var{pretend_args_size}. The value that you store here will serve as
4073 additional offset for setting up the stack frame.
4075 Because you must generate code to push the anonymous arguments at
4076 compile time without knowing their data types,
4077 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
4078 a single category of argument register and use it uniformly for all data
4081 If the argument @var{second_time} is nonzero, it means that the
4082 arguments of the function are being analyzed for the second time. This
4083 happens for an inline function, which is not actually compiled until the
4084 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
4085 not generate any instructions in this case.
4087 @findex STRICT_ARGUMENT_NAMING
4088 @item STRICT_ARGUMENT_NAMING
4089 Define this macro to be a nonzero value if the location where a function
4090 argument is passed depends on whether or not it is a named argument.
4092 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
4093 is set for varargs and stdarg functions. If this macro returns a
4094 nonzero value, the @var{named} argument is always true for named
4095 arguments, and false for unnamed arguments. If it returns a value of
4096 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
4097 are treated as named. Otherwise, all named arguments except the last
4098 are treated as named.
4100 You need not define this macro if it always returns zero.
4102 @findex PRETEND_OUTGOING_VARARGS_NAMED
4103 @item PRETEND_OUTGOING_VARARGS_NAMED
4104 If you need to conditionally change ABIs so that one works with
4105 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
4106 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
4107 defined, then define this macro to return nonzero if
4108 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
4109 Otherwise, you should not define this macro.
4113 @section Trampolines for Nested Functions
4114 @cindex trampolines for nested functions
4115 @cindex nested functions, trampolines for
4117 A @dfn{trampoline} is a small piece of code that is created at run time
4118 when the address of a nested function is taken. It normally resides on
4119 the stack, in the stack frame of the containing function. These macros
4120 tell GCC how to generate code to allocate and initialize a
4123 The instructions in the trampoline must do two things: load a constant
4124 address into the static chain register, and jump to the real address of
4125 the nested function. On CISC machines such as the m68k, this requires
4126 two instructions, a move immediate and a jump. Then the two addresses
4127 exist in the trampoline as word-long immediate operands. On RISC
4128 machines, it is often necessary to load each address into a register in
4129 two parts. Then pieces of each address form separate immediate
4132 The code generated to initialize the trampoline must store the variable
4133 parts---the static chain value and the function address---into the
4134 immediate operands of the instructions. On a CISC machine, this is
4135 simply a matter of copying each address to a memory reference at the
4136 proper offset from the start of the trampoline. On a RISC machine, it
4137 may be necessary to take out pieces of the address and store them
4141 @findex TRAMPOLINE_TEMPLATE
4142 @item TRAMPOLINE_TEMPLATE (@var{file})
4143 A C statement to output, on the stream @var{file}, assembler code for a
4144 block of data that contains the constant parts of a trampoline. This
4145 code should not include a label---the label is taken care of
4148 If you do not define this macro, it means no template is needed
4149 for the target. Do not define this macro on systems where the block move
4150 code to copy the trampoline into place would be larger than the code
4151 to generate it on the spot.
4153 @findex TRAMPOLINE_SECTION
4154 @item TRAMPOLINE_SECTION
4155 The name of a subroutine to switch to the section in which the
4156 trampoline template is to be placed (@pxref{Sections}). The default is
4157 a value of @samp{readonly_data_section}, which places the trampoline in
4158 the section containing read-only data.
4160 @findex TRAMPOLINE_SIZE
4161 @item TRAMPOLINE_SIZE
4162 A C expression for the size in bytes of the trampoline, as an integer.
4164 @findex TRAMPOLINE_ALIGNMENT
4165 @item TRAMPOLINE_ALIGNMENT
4166 Alignment required for trampolines, in bits.
4168 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4169 is used for aligning trampolines.
4171 @findex INITIALIZE_TRAMPOLINE
4172 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4173 A C statement to initialize the variable parts of a trampoline.
4174 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4175 an RTX for the address of the nested function; @var{static_chain} is an
4176 RTX for the static chain value that should be passed to the function
4179 @findex TRAMPOLINE_ADJUST_ADDRESS
4180 @item TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4181 A C statement that should perform any machine-specific adjustment in
4182 the address of the trampoline. Its argument contains the address that
4183 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4184 used for a function call should be different from the address in which
4185 the template was stored, the different address should be assigned to
4186 @var{addr}. If this macro is not defined, @var{addr} will be used for
4189 @findex ALLOCATE_TRAMPOLINE
4190 @item ALLOCATE_TRAMPOLINE (@var{fp})
4191 A C expression to allocate run-time space for a trampoline. The
4192 expression value should be an RTX representing a memory reference to the
4193 space for the trampoline.
4195 @cindex @code{FUNCTION_EPILOGUE} and trampolines
4196 @cindex @code{FUNCTION_PROLOGUE} and trampolines
4197 If this macro is not defined, by default the trampoline is allocated as
4198 a stack slot. This default is right for most machines. The exceptions
4199 are machines where it is impossible to execute instructions in the stack
4200 area. On such machines, you may have to implement a separate stack,
4201 using this macro in conjunction with @code{FUNCTION_PROLOGUE} and
4202 @code{FUNCTION_EPILOGUE}.
4204 @var{fp} points to a data structure, a @code{struct function}, which
4205 describes the compilation status of the immediate containing function of
4206 the function which the trampoline is for. Normally (when
4207 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
4208 trampoline is in the stack frame of this containing function. Other
4209 allocation strategies probably must do something analogous with this
4213 Implementing trampolines is difficult on many machines because they have
4214 separate instruction and data caches. Writing into a stack location
4215 fails to clear the memory in the instruction cache, so when the program
4216 jumps to that location, it executes the old contents.
4218 Here are two possible solutions. One is to clear the relevant parts of
4219 the instruction cache whenever a trampoline is set up. The other is to
4220 make all trampolines identical, by having them jump to a standard
4221 subroutine. The former technique makes trampoline execution faster; the
4222 latter makes initialization faster.
4224 To clear the instruction cache when a trampoline is initialized, define
4225 the following macros which describe the shape of the cache.
4228 @findex INSN_CACHE_SIZE
4229 @item INSN_CACHE_SIZE
4230 The total size in bytes of the cache.
4232 @findex INSN_CACHE_LINE_WIDTH
4233 @item INSN_CACHE_LINE_WIDTH
4234 The length in bytes of each cache line. The cache is divided into cache
4235 lines which are disjoint slots, each holding a contiguous chunk of data
4236 fetched from memory. Each time data is brought into the cache, an
4237 entire line is read at once. The data loaded into a cache line is
4238 always aligned on a boundary equal to the line size.
4240 @findex INSN_CACHE_DEPTH
4241 @item INSN_CACHE_DEPTH
4242 The number of alternative cache lines that can hold any particular memory
4246 Alternatively, if the machine has system calls or instructions to clear
4247 the instruction cache directly, you can define the following macro.
4250 @findex CLEAR_INSN_CACHE
4251 @item CLEAR_INSN_CACHE (@var{BEG}, @var{END})
4252 If defined, expands to a C expression clearing the @emph{instruction
4253 cache} in the specified interval. If it is not defined, and the macro
4254 INSN_CACHE_SIZE is defined, some generic code is generated to clear the
4255 cache. The definition of this macro would typically be a series of
4256 @code{asm} statements. Both @var{BEG} and @var{END} are both pointer
4260 To use a standard subroutine, define the following macro. In addition,
4261 you must make sure that the instructions in a trampoline fill an entire
4262 cache line with identical instructions, or else ensure that the
4263 beginning of the trampoline code is always aligned at the same point in
4264 its cache line. Look in @file{m68k.h} as a guide.
4267 @findex TRANSFER_FROM_TRAMPOLINE
4268 @item TRANSFER_FROM_TRAMPOLINE
4269 Define this macro if trampolines need a special subroutine to do their
4270 work. The macro should expand to a series of @code{asm} statements
4271 which will be compiled with GCC. They go in a library function named
4272 @code{__transfer_from_trampoline}.
4274 If you need to avoid executing the ordinary prologue code of a compiled
4275 C function when you jump to the subroutine, you can do so by placing a
4276 special label of your own in the assembler code. Use one @code{asm}
4277 statement to generate an assembler label, and another to make the label
4278 global. Then trampolines can use that label to jump directly to your
4279 special assembler code.
4283 @section Implicit Calls to Library Routines
4284 @cindex library subroutine names
4285 @cindex @file{libgcc.a}
4287 @c prevent bad page break with this line
4288 Here is an explanation of implicit calls to library routines.
4291 @findex MULSI3_LIBCALL
4292 @item MULSI3_LIBCALL
4293 A C string constant giving the name of the function to call for
4294 multiplication of one signed full-word by another. If you do not
4295 define this macro, the default name is used, which is @code{__mulsi3},
4296 a function defined in @file{libgcc.a}.
4298 @findex DIVSI3_LIBCALL
4299 @item DIVSI3_LIBCALL
4300 A C string constant giving the name of the function to call for
4301 division of one signed full-word by another. If you do not define
4302 this macro, the default name is used, which is @code{__divsi3}, a
4303 function defined in @file{libgcc.a}.
4305 @findex UDIVSI3_LIBCALL
4306 @item UDIVSI3_LIBCALL
4307 A C string constant giving the name of the function to call for
4308 division of one unsigned full-word by another. If you do not define
4309 this macro, the default name is used, which is @code{__udivsi3}, a
4310 function defined in @file{libgcc.a}.
4312 @findex MODSI3_LIBCALL
4313 @item MODSI3_LIBCALL
4314 A C string constant giving the name of the function to call for the
4315 remainder in division of one signed full-word by another. If you do
4316 not define this macro, the default name is used, which is
4317 @code{__modsi3}, a function defined in @file{libgcc.a}.
4319 @findex UMODSI3_LIBCALL
4320 @item UMODSI3_LIBCALL
4321 A C string constant giving the name of the function to call for the
4322 remainder in division of one unsigned full-word by another. If you do
4323 not define this macro, the default name is used, which is
4324 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4326 @findex MULDI3_LIBCALL
4327 @item MULDI3_LIBCALL
4328 A C string constant giving the name of the function to call for
4329 multiplication of one signed double-word by another. If you do not
4330 define this macro, the default name is used, which is @code{__muldi3},
4331 a function defined in @file{libgcc.a}.
4333 @findex DIVDI3_LIBCALL
4334 @item DIVDI3_LIBCALL
4335 A C string constant giving the name of the function to call for
4336 division of one signed double-word by another. If you do not define
4337 this macro, the default name is used, which is @code{__divdi3}, a
4338 function defined in @file{libgcc.a}.
4340 @findex UDIVDI3_LIBCALL
4341 @item UDIVDI3_LIBCALL
4342 A C string constant giving the name of the function to call for
4343 division of one unsigned full-word by another. If you do not define
4344 this macro, the default name is used, which is @code{__udivdi3}, a
4345 function defined in @file{libgcc.a}.
4347 @findex MODDI3_LIBCALL
4348 @item MODDI3_LIBCALL
4349 A C string constant giving the name of the function to call for the
4350 remainder in division of one signed double-word by another. If you do
4351 not define this macro, the default name is used, which is
4352 @code{__moddi3}, a function defined in @file{libgcc.a}.
4354 @findex UMODDI3_LIBCALL
4355 @item UMODDI3_LIBCALL
4356 A C string constant giving the name of the function to call for the
4357 remainder in division of one unsigned full-word by another. If you do
4358 not define this macro, the default name is used, which is
4359 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4361 @findex INIT_TARGET_OPTABS
4362 @item INIT_TARGET_OPTABS
4363 Define this macro as a C statement that declares additional library
4364 routines renames existing ones. @code{init_optabs} calls this macro after
4365 initializing all the normal library routines.
4367 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4368 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4369 Define this macro as a C statement that returns nonzero if a call to
4370 the floating point comparison library function will return a boolean
4371 value that indicates the result of the comparison. It should return
4372 zero if one of gcc's own libgcc functions is called.
4374 Most ports don't need to define this macro.
4377 @cindex @code{EDOM}, implicit usage
4379 The value of @code{EDOM} on the target machine, as a C integer constant
4380 expression. If you don't define this macro, GCC does not attempt to
4381 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4382 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4385 If you do not define @code{TARGET_EDOM}, then compiled code reports
4386 domain errors by calling the library function and letting it report the
4387 error. If mathematical functions on your system use @code{matherr} when
4388 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4389 that @code{matherr} is used normally.
4391 @findex GEN_ERRNO_RTX
4392 @cindex @code{errno}, implicit usage
4394 Define this macro as a C expression to create an rtl expression that
4395 refers to the global ``variable'' @code{errno}. (On certain systems,
4396 @code{errno} may not actually be a variable.) If you don't define this
4397 macro, a reasonable default is used.
4399 @findex TARGET_MEM_FUNCTIONS
4400 @cindex @code{bcopy}, implicit usage
4401 @cindex @code{memcpy}, implicit usage
4402 @cindex @code{memmove}, implicit usage
4403 @cindex @code{bzero}, implicit usage
4404 @cindex @code{memset}, implicit usage
4405 @item TARGET_MEM_FUNCTIONS
4406 Define this macro if GCC should generate calls to the ISO C
4407 (and System V) library functions @code{memcpy}, @code{memmove} and
4408 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4410 @findex LIBGCC_NEEDS_DOUBLE
4411 @item LIBGCC_NEEDS_DOUBLE
4412 Define this macro if only @code{float} arguments cannot be passed to
4413 library routines (so they must be converted to @code{double}). This
4414 macro affects both how library calls are generated and how the library
4415 routines in @file{libgcc1.c} accept their arguments. It is useful on
4416 machines where floating and fixed point arguments are passed
4417 differently, such as the i860.
4419 @findex FLOAT_ARG_TYPE
4420 @item FLOAT_ARG_TYPE
4421 Define this macro to override the type used by the library routines to
4422 pick up arguments of type @code{float}. (By default, they use a union
4423 of @code{float} and @code{int}.)
4425 The obvious choice would be @code{float}---but that won't work with
4426 traditional C compilers that expect all arguments declared as @code{float}
4427 to arrive as @code{double}. To avoid this conversion, the library routines
4428 ask for the value as some other type and then treat it as a @code{float}.
4430 On some systems, no other type will work for this. For these systems,
4431 you must use @code{LIBGCC_NEEDS_DOUBLE} instead, to force conversion of
4432 the values @code{double} before they are passed.
4435 @item FLOATIFY (@var{passed-value})
4436 Define this macro to override the way library routines redesignate a
4437 @code{float} argument as a @code{float} instead of the type it was
4438 passed as. The default is an expression which takes the @code{float}
4441 @findex FLOAT_VALUE_TYPE
4442 @item FLOAT_VALUE_TYPE
4443 Define this macro to override the type used by the library routines to
4444 return values that ought to have type @code{float}. (By default, they
4447 The obvious choice would be @code{float}---but that won't work with
4448 traditional C compilers gratuitously convert values declared as
4449 @code{float} into @code{double}.
4452 @item INTIFY (@var{float-value})
4453 Define this macro to override the way the value of a
4454 @code{float}-returning library routine should be packaged in order to
4455 return it. These functions are actually declared to return type
4456 @code{FLOAT_VALUE_TYPE} (normally @code{int}).
4458 These values can't be returned as type @code{float} because traditional
4459 C compilers would gratuitously convert the value to a @code{double}.
4461 A local variable named @code{intify} is always available when the macro
4462 @code{INTIFY} is used. It is a union of a @code{float} field named
4463 @code{f} and a field named @code{i} whose type is
4464 @code{FLOAT_VALUE_TYPE} or @code{int}.
4466 If you don't define this macro, the default definition works by copying
4467 the value through that union.
4469 @findex nongcc_SI_type
4470 @item nongcc_SI_type
4471 Define this macro as the name of the data type corresponding to
4472 @code{SImode} in the system's own C compiler.
4474 You need not define this macro if that type is @code{long int}, as it usually
4477 @findex nongcc_word_type
4478 @item nongcc_word_type
4479 Define this macro as the name of the data type corresponding to the
4480 word_mode in the system's own C compiler.
4482 You need not define this macro if that type is @code{long int}, as it usually
4485 @findex perform_@dots{}
4486 @item perform_@dots{}
4487 Define these macros to supply explicit C statements to carry out various
4488 arithmetic operations on types @code{float} and @code{double} in the
4489 library routines in @file{libgcc1.c}. See that file for a full list
4490 of these macros and their arguments.
4492 On most machines, you don't need to define any of these macros, because
4493 the C compiler that comes with the system takes care of doing them.
4495 @findex NEXT_OBJC_RUNTIME
4496 @item NEXT_OBJC_RUNTIME
4497 Define this macro to generate code for Objective C message sending using
4498 the calling convention of the NeXT system. This calling convention
4499 involves passing the object, the selector and the method arguments all
4500 at once to the method-lookup library function.
4502 The default calling convention passes just the object and the selector
4503 to the lookup function, which returns a pointer to the method.
4506 @node Addressing Modes
4507 @section Addressing Modes
4508 @cindex addressing modes
4510 @c prevent bad page break with this line
4511 This is about addressing modes.
4514 @findex HAVE_PRE_INCREMENT
4515 @findex HAVE_PRE_DECREMENT
4516 @findex HAVE_POST_INCREMENT
4517 @findex HAVE_POST_DECREMENT
4518 @item HAVE_PRE_INCREMENT
4519 @itemx HAVE_PRE_DECREMENT
4520 @itemx HAVE_POST_INCREMENT
4521 @itemx HAVE_POST_DECREMENT
4522 A C expression that is non-zero if the machine supports pre-increment,
4523 pre-decrement, post-increment, or post-decrement addressing respectively.
4525 @findex HAVE_POST_MODIFY_DISP
4526 @findex HAVE_PRE_MODIFY_DISP
4527 @item HAVE_PRE_MODIFY_DISP
4528 @itemx HAVE_POST_MODIFY_DISP
4529 A C expression that is non-zero if the machine supports pre- or
4530 post-address side-effect generation involving constants other than
4531 the size of the memory operand.
4533 @findex HAVE_POST_MODIFY_REG
4534 @findex HAVE_PRE_MODIFY_REG
4535 @item HAVE_PRE_MODIFY_REG
4536 @itemx HAVE_POST_MODIFY_REG
4537 A C expression that is non-zero if the machine supports pre- or
4538 post-address side-effect generation involving a register displacement.
4540 @findex CONSTANT_ADDRESS_P
4541 @item CONSTANT_ADDRESS_P (@var{x})
4542 A C expression that is 1 if the RTX @var{x} is a constant which
4543 is a valid address. On most machines, this can be defined as
4544 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4545 in which constant addresses are supported.
4548 @code{CONSTANT_P} accepts integer-values expressions whose values are
4549 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4550 @code{high} expressions and @code{const} arithmetic expressions, in
4551 addition to @code{const_int} and @code{const_double} expressions.
4553 @findex MAX_REGS_PER_ADDRESS
4554 @item MAX_REGS_PER_ADDRESS
4555 A number, the maximum number of registers that can appear in a valid
4556 memory address. Note that it is up to you to specify a value equal to
4557 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4560 @findex GO_IF_LEGITIMATE_ADDRESS
4561 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4562 A C compound statement with a conditional @code{goto @var{label};}
4563 executed if @var{x} (an RTX) is a legitimate memory address on the
4564 target machine for a memory operand of mode @var{mode}.
4566 It usually pays to define several simpler macros to serve as
4567 subroutines for this one. Otherwise it may be too complicated to
4570 This macro must exist in two variants: a strict variant and a
4571 non-strict one. The strict variant is used in the reload pass. It
4572 must be defined so that any pseudo-register that has not been
4573 allocated a hard register is considered a memory reference. In
4574 contexts where some kind of register is required, a pseudo-register
4575 with no hard register must be rejected.
4577 The non-strict variant is used in other passes. It must be defined to
4578 accept all pseudo-registers in every context where some kind of
4579 register is required.
4581 @findex REG_OK_STRICT
4582 Compiler source files that want to use the strict variant of this
4583 macro define the macro @code{REG_OK_STRICT}. You should use an
4584 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4585 in that case and the non-strict variant otherwise.
4587 Subroutines to check for acceptable registers for various purposes (one
4588 for base registers, one for index registers, and so on) are typically
4589 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4590 Then only these subroutine macros need have two variants; the higher
4591 levels of macros may be the same whether strict or not.@refill
4593 Normally, constant addresses which are the sum of a @code{symbol_ref}
4594 and an integer are stored inside a @code{const} RTX to mark them as
4595 constant. Therefore, there is no need to recognize such sums
4596 specifically as legitimate addresses. Normally you would simply
4597 recognize any @code{const} as legitimate.
4599 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4600 sums that are not marked with @code{const}. It assumes that a naked
4601 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4602 naked constant sums as illegitimate addresses, so that none of them will
4603 be given to @code{PRINT_OPERAND_ADDRESS}.
4605 @cindex @code{ENCODE_SECTION_INFO} and address validation
4606 On some machines, whether a symbolic address is legitimate depends on
4607 the section that the address refers to. On these machines, define the
4608 macro @code{ENCODE_SECTION_INFO} to store the information into the
4609 @code{symbol_ref}, and then check for it here. When you see a
4610 @code{const}, you will have to look inside it to find the
4611 @code{symbol_ref} in order to determine the section. @xref{Assembler
4614 @findex saveable_obstack
4615 The best way to modify the name string is by adding text to the
4616 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4617 the new name in @code{saveable_obstack}. You will have to modify
4618 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4619 output the name accordingly, and define @code{STRIP_NAME_ENCODING} to
4620 access the original name string.
4622 You can check the information stored here into the @code{symbol_ref} in
4623 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4624 @code{PRINT_OPERAND_ADDRESS}.
4626 @findex REG_OK_FOR_BASE_P
4627 @item REG_OK_FOR_BASE_P (@var{x})
4628 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4629 RTX) is valid for use as a base register. For hard registers, it
4630 should always accept those which the hardware permits and reject the
4631 others. Whether the macro accepts or rejects pseudo registers must be
4632 controlled by @code{REG_OK_STRICT} as described above. This usually
4633 requires two variant definitions, of which @code{REG_OK_STRICT}
4634 controls the one actually used.
4636 @findex REG_MODE_OK_FOR_BASE_P
4637 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4638 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4639 that expression may examine the mode of the memory reference in
4640 @var{mode}. You should define this macro if the mode of the memory
4641 reference affects whether a register may be used as a base register. If
4642 you define this macro, the compiler will use it instead of
4643 @code{REG_OK_FOR_BASE_P}.
4645 @findex REG_OK_FOR_INDEX_P
4646 @item REG_OK_FOR_INDEX_P (@var{x})
4647 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4648 RTX) is valid for use as an index register.
4650 The difference between an index register and a base register is that
4651 the index register may be scaled. If an address involves the sum of
4652 two registers, neither one of them scaled, then either one may be
4653 labeled the ``base'' and the other the ``index''; but whichever
4654 labeling is used must fit the machine's constraints of which registers
4655 may serve in each capacity. The compiler will try both labelings,
4656 looking for one that is valid, and will reload one or both registers
4657 only if neither labeling works.
4659 @findex FIND_BASE_TERM
4660 @item FIND_BASE_TERM (@var{x})
4661 A C expression to determine the base term of address @var{x}.
4662 This macro is used in only one place: `find_base_term' in alias.c.
4664 It is always safe for this macro to not be defined. It exists so
4665 that alias analysis can understand machine-dependent addresses.
4667 The typical use of this macro is to handle addresses containing
4668 a label_ref or symbol_ref within an UNSPEC.
4670 @findex LEGITIMIZE_ADDRESS
4671 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4672 A C compound statement that attempts to replace @var{x} with a valid
4673 memory address for an operand of mode @var{mode}. @var{win} will be a
4674 C statement label elsewhere in the code; the macro definition may use
4677 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4681 to avoid further processing if the address has become legitimate.
4683 @findex break_out_memory_refs
4684 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4685 and @var{oldx} will be the operand that was given to that function to produce
4688 The code generated by this macro should not alter the substructure of
4689 @var{x}. If it transforms @var{x} into a more legitimate form, it
4690 should assign @var{x} (which will always be a C variable) a new value.
4692 It is not necessary for this macro to come up with a legitimate
4693 address. The compiler has standard ways of doing so in all cases. In
4694 fact, it is safe for this macro to do nothing. But often a
4695 machine-dependent strategy can generate better code.
4697 @findex LEGITIMIZE_RELOAD_ADDRESS
4698 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4699 A C compound statement that attempts to replace @var{x}, which is an address
4700 that needs reloading, with a valid memory address for an operand of mode
4701 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4702 It is not necessary to define this macro, but it might be useful for
4703 performance reasons.
4705 For example, on the i386, it is sometimes possible to use a single
4706 reload register instead of two by reloading a sum of two pseudo
4707 registers into a register. On the other hand, for number of RISC
4708 processors offsets are limited so that often an intermediate address
4709 needs to be generated in order to address a stack slot. By defining
4710 LEGITIMIZE_RELOAD_ADDRESS appropriately, the intermediate addresses
4711 generated for adjacent some stack slots can be made identical, and thus
4714 @emph{Note}: This macro should be used with caution. It is necessary
4715 to know something of how reload works in order to effectively use this,
4716 and it is quite easy to produce macros that build in too much knowledge
4717 of reload internals.
4719 @emph{Note}: This macro must be able to reload an address created by a
4720 previous invocation of this macro. If it fails to handle such addresses
4721 then the compiler may generate incorrect code or abort.
4724 The macro definition should use @code{push_reload} to indicate parts that
4725 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4726 suitable to be passed unaltered to @code{push_reload}.
4728 The code generated by this macro must not alter the substructure of
4729 @var{x}. If it transforms @var{x} into a more legitimate form, it
4730 should assign @var{x} (which will always be a C variable) a new value.
4731 This also applies to parts that you change indirectly by calling
4734 @findex strict_memory_address_p
4735 The macro definition may use @code{strict_memory_address_p} to test if
4736 the address has become legitimate.
4739 If you want to change only a part of @var{x}, one standard way of doing
4740 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4741 single level of rtl. Thus, if the part to be changed is not at the
4742 top level, you'll need to replace first the top leve
4743 It is not necessary for this macro to come up with a legitimate
4744 address; but often a machine-dependent strategy can generate better code.
4746 @findex GO_IF_MODE_DEPENDENT_ADDRESS
4747 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4748 A C statement or compound statement with a conditional @code{goto
4749 @var{label};} executed if memory address @var{x} (an RTX) can have
4750 different meanings depending on the machine mode of the memory
4751 reference it is used for or if the address is valid for some modes
4754 Autoincrement and autodecrement addresses typically have mode-dependent
4755 effects because the amount of the increment or decrement is the size
4756 of the operand being addressed. Some machines have other mode-dependent
4757 addresses. Many RISC machines have no mode-dependent addresses.
4759 You may assume that @var{addr} is a valid address for the machine.
4761 @findex LEGITIMATE_CONSTANT_P
4762 @item LEGITIMATE_CONSTANT_P (@var{x})
4763 A C expression that is nonzero if @var{x} is a legitimate constant for
4764 an immediate operand on the target machine. You can assume that
4765 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4766 @samp{1} is a suitable definition for this macro on machines where
4767 anything @code{CONSTANT_P} is valid.@refill
4770 @node Condition Code
4771 @section Condition Code Status
4772 @cindex condition code status
4774 @c prevent bad page break with this line
4775 This describes the condition code status.
4778 The file @file{conditions.h} defines a variable @code{cc_status} to
4779 describe how the condition code was computed (in case the interpretation of
4780 the condition code depends on the instruction that it was set by). This
4781 variable contains the RTL expressions on which the condition code is
4782 currently based, and several standard flags.
4784 Sometimes additional machine-specific flags must be defined in the machine
4785 description header file. It can also add additional machine-specific
4786 information by defining @code{CC_STATUS_MDEP}.
4789 @findex CC_STATUS_MDEP
4790 @item CC_STATUS_MDEP
4791 C code for a data type which is used for declaring the @code{mdep}
4792 component of @code{cc_status}. It defaults to @code{int}.
4794 This macro is not used on machines that do not use @code{cc0}.
4796 @findex CC_STATUS_MDEP_INIT
4797 @item CC_STATUS_MDEP_INIT
4798 A C expression to initialize the @code{mdep} field to ``empty''.
4799 The default definition does nothing, since most machines don't use
4800 the field anyway. If you want to use the field, you should probably
4801 define this macro to initialize it.
4803 This macro is not used on machines that do not use @code{cc0}.
4805 @findex NOTICE_UPDATE_CC
4806 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4807 A C compound statement to set the components of @code{cc_status}
4808 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4809 this macro's responsibility to recognize insns that set the condition
4810 code as a byproduct of other activity as well as those that explicitly
4813 This macro is not used on machines that do not use @code{cc0}.
4815 If there are insns that do not set the condition code but do alter
4816 other machine registers, this macro must check to see whether they
4817 invalidate the expressions that the condition code is recorded as
4818 reflecting. For example, on the 68000, insns that store in address
4819 registers do not set the condition code, which means that usually
4820 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4821 insns. But suppose that the previous insn set the condition code
4822 based on location @samp{a4@@(102)} and the current insn stores a new
4823 value in @samp{a4}. Although the condition code is not changed by
4824 this, it will no longer be true that it reflects the contents of
4825 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4826 @code{cc_status} in this case to say that nothing is known about the
4827 condition code value.
4829 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4830 with the results of peephole optimization: insns whose patterns are
4831 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4832 constants which are just the operands. The RTL structure of these
4833 insns is not sufficient to indicate what the insns actually do. What
4834 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4835 @code{CC_STATUS_INIT}.
4837 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4838 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4839 @samp{cc}. This avoids having detailed information about patterns in
4840 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4842 @findex EXTRA_CC_MODES
4843 @item EXTRA_CC_MODES
4844 A list of additional modes for condition code values in registers
4845 (@pxref{Jump Patterns}). This macro should expand to a sequence of
4846 calls of the macro @code{CC} separated by white space. @code{CC} takes
4847 two arguments. The first is the enumeration name of the mode, which
4848 should begin with @samp{CC} and end with @samp{mode}. The second is a C
4849 string giving the printable name of the mode; it should be the same as
4850 the first argument, but with the trailing @samp{mode} removed.
4852 You should only define this macro if additional modes are required.
4854 A sample definition of @code{EXTRA_CC_MODES} is:
4856 #define EXTRA_CC_MODES \
4857 CC(CC_NOOVmode, "CC_NOOV") \
4858 CC(CCFPmode, "CCFP") \
4859 CC(CCFPEmode, "CCFPE")
4862 @findex SELECT_CC_MODE
4863 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4864 Returns a mode from class @code{MODE_CC} to be used when comparison
4865 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
4866 example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
4867 @pxref{Jump Patterns} for a description of the reason for this
4871 #define SELECT_CC_MODE(OP,X,Y) \
4872 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4873 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
4874 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4875 || GET_CODE (X) == NEG) \
4876 ? CC_NOOVmode : CCmode))
4879 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
4881 @findex CANONICALIZE_COMPARISON
4882 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
4883 On some machines not all possible comparisons are defined, but you can
4884 convert an invalid comparison into a valid one. For example, the Alpha
4885 does not have a @code{GT} comparison, but you can use an @code{LT}
4886 comparison instead and swap the order of the operands.
4888 On such machines, define this macro to be a C statement to do any
4889 required conversions. @var{code} is the initial comparison code
4890 and @var{op0} and @var{op1} are the left and right operands of the
4891 comparison, respectively. You should modify @var{code}, @var{op0}, and
4892 @var{op1} as required.
4894 GCC will not assume that the comparison resulting from this macro is
4895 valid but will see if the resulting insn matches a pattern in the
4898 You need not define this macro if it would never change the comparison
4901 @findex REVERSIBLE_CC_MODE
4902 @item REVERSIBLE_CC_MODE (@var{mode})
4903 A C expression whose value is one if it is always safe to reverse a
4904 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4905 can ever return @var{mode} for a floating-point inequality comparison,
4906 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4908 You need not define this macro if it would always returns zero or if the
4909 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4910 For example, here is the definition used on the Sparc, where floating-point
4911 inequality comparisons are always given @code{CCFPEmode}:
4914 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
4917 @findex REVERSE_CONDITION (@var{code}, @var{mode})
4918 A C expression whose value is reversed condition code of the @var{code} for
4919 comparison done in CC_MODE @var{mode}. The macro is used only in case
4920 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
4921 machine has some non-standard way how to reverse certain conditionals. For
4922 instance in case all floating point conditions are non-trapping, compiler may
4923 freely convert unordered compares to ordered one. Then definition may look
4927 #define REVERSE_CONDITION(CODE, MODE) \
4928 ((MODE) != CCFPmode ? reverse_condtion (CODE) \
4929 : reverse_condition_maybe_unordered (CODE))
4932 @findex REVERSE_CONDEXEC_PREDICATES_P
4933 @item REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
4934 A C expression that returns true if the conditional execution predicate
4935 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
4936 return 0 if the target has conditional execution predicates that cannot be
4937 reversed safely. If no expansion is specified, this macro is defined as
4941 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) ((x) == reverse_condition (y))
4947 @section Describing Relative Costs of Operations
4948 @cindex costs of instructions
4949 @cindex relative costs
4950 @cindex speed of instructions
4952 These macros let you describe the relative speed of various operations
4953 on the target machine.
4957 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
4958 A part of a C @code{switch} statement that describes the relative costs
4959 of constant RTL expressions. It must contain @code{case} labels for
4960 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
4961 @code{label_ref} and @code{const_double}. Each case must ultimately
4962 reach a @code{return} statement to return the relative cost of the use
4963 of that kind of constant value in an expression. The cost may depend on
4964 the precise value of the constant, which is available for examination in
4965 @var{x}, and the rtx code of the expression in which it is contained,
4966 found in @var{outer_code}.
4968 @var{code} is the expression code---redundant, since it can be
4969 obtained with @code{GET_CODE (@var{x})}.
4972 @findex COSTS_N_INSNS
4973 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4974 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
4975 This can be used, for example, to indicate how costly a multiply
4976 instruction is. In writing this macro, you can use the construct
4977 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
4978 instructions. @var{outer_code} is the code of the expression in which
4979 @var{x} is contained.
4981 This macro is optional; do not define it if the default cost assumptions
4982 are adequate for the target machine.
4984 @findex DEFAULT_RTX_COSTS
4985 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4986 This macro, if defined, is called for any case not handled by the
4987 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
4988 to put case labels into the macro, but the code, or any functions it
4989 calls, must assume that the RTL in @var{x} could be of any type that has
4990 not already been handled. The arguments are the same as for
4991 @code{RTX_COSTS}, and the macro should execute a return statement giving
4992 the cost of any RTL expressions that it can handle. The default cost
4993 calculation is used for any RTL for which this macro does not return a
4996 This macro is optional; do not define it if the default cost assumptions
4997 are adequate for the target machine.
4999 @findex ADDRESS_COST
5000 @item ADDRESS_COST (@var{address})
5001 An expression giving the cost of an addressing mode that contains
5002 @var{address}. If not defined, the cost is computed from
5003 the @var{address} expression and the @code{CONST_COSTS} values.
5005 For most CISC machines, the default cost is a good approximation of the
5006 true cost of the addressing mode. However, on RISC machines, all
5007 instructions normally have the same length and execution time. Hence
5008 all addresses will have equal costs.
5010 In cases where more than one form of an address is known, the form with
5011 the lowest cost will be used. If multiple forms have the same, lowest,
5012 cost, the one that is the most complex will be used.
5014 For example, suppose an address that is equal to the sum of a register
5015 and a constant is used twice in the same basic block. When this macro
5016 is not defined, the address will be computed in a register and memory
5017 references will be indirect through that register. On machines where
5018 the cost of the addressing mode containing the sum is no higher than
5019 that of a simple indirect reference, this will produce an additional
5020 instruction and possibly require an additional register. Proper
5021 specification of this macro eliminates this overhead for such machines.
5023 Similar use of this macro is made in strength reduction of loops.
5025 @var{address} need not be valid as an address. In such a case, the cost
5026 is not relevant and can be any value; invalid addresses need not be
5027 assigned a different cost.
5029 On machines where an address involving more than one register is as
5030 cheap as an address computation involving only one register, defining
5031 @code{ADDRESS_COST} to reflect this can cause two registers to be live
5032 over a region of code where only one would have been if
5033 @code{ADDRESS_COST} were not defined in that manner. This effect should
5034 be considered in the definition of this macro. Equivalent costs should
5035 probably only be given to addresses with different numbers of registers
5036 on machines with lots of registers.
5038 This macro will normally either not be defined or be defined as a
5041 @findex REGISTER_MOVE_COST
5042 @item REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5043 A C expression for the cost of moving data of mode @var{mode} from a
5044 register in class @var{from} to one in class @var{to}. The classes are
5045 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5046 value of 2 is the default; other values are interpreted relative to
5049 It is not required that the cost always equal 2 when @var{from} is the
5050 same as @var{to}; on some machines it is expensive to move between
5051 registers if they are not general registers.
5053 If reload sees an insn consisting of a single @code{set} between two
5054 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5055 classes returns a value of 2, reload does not check to ensure that the
5056 constraints of the insn are met. Setting a cost of other than 2 will
5057 allow reload to verify that the constraints are met. You should do this
5058 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5060 @findex MEMORY_MOVE_COST
5061 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5062 A C expression for the cost of moving data of mode @var{mode} between a
5063 register of class @var{class} and memory; @var{in} is zero if the value
5064 is to be written to memory, non-zero if it is to be read in. This cost
5065 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5066 registers and memory is more expensive than between two registers, you
5067 should define this macro to express the relative cost.
5069 If you do not define this macro, GCC uses a default cost of 4 plus
5070 the cost of copying via a secondary reload register, if one is
5071 needed. If your machine requires a secondary reload register to copy
5072 between memory and a register of @var{class} but the reload mechanism is
5073 more complex than copying via an intermediate, define this macro to
5074 reflect the actual cost of the move.
5076 GCC defines the function @code{memory_move_secondary_cost} if
5077 secondary reloads are needed. It computes the costs due to copying via
5078 a secondary register. If your machine copies from memory using a
5079 secondary register in the conventional way but the default base value of
5080 4 is not correct for your machine, define this macro to add some other
5081 value to the result of that function. The arguments to that function
5082 are the same as to this macro.
5086 A C expression for the cost of a branch instruction. A value of 1 is
5087 the default; other values are interpreted relative to that.
5090 Here are additional macros which do not specify precise relative costs,
5091 but only that certain actions are more expensive than GCC would
5095 @findex SLOW_BYTE_ACCESS
5096 @item SLOW_BYTE_ACCESS
5097 Define this macro as a C expression which is nonzero if accessing less
5098 than a word of memory (i.e. a @code{char} or a @code{short}) is no
5099 faster than accessing a word of memory, i.e., if such access
5100 require more than one instruction or if there is no difference in cost
5101 between byte and (aligned) word loads.
5103 When this macro is not defined, the compiler will access a field by
5104 finding the smallest containing object; when it is defined, a fullword
5105 load will be used if alignment permits. Unless bytes accesses are
5106 faster than word accesses, using word accesses is preferable since it
5107 may eliminate subsequent memory access if subsequent accesses occur to
5108 other fields in the same word of the structure, but to different bytes.
5110 @findex SLOW_ZERO_EXTEND
5111 @item SLOW_ZERO_EXTEND
5112 Define this macro if zero-extension (of a @code{char} or @code{short}
5113 to an @code{int}) can be done faster if the destination is a register
5114 that is known to be zero.
5116 If you define this macro, you must have instruction patterns that
5117 recognize RTL structures like this:
5120 (set (strict_low_part (subreg:QI (reg:SI @dots{}) 0)) @dots{})
5124 and likewise for @code{HImode}.
5126 @findex SLOW_UNALIGNED_ACCESS
5127 @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5128 Define this macro to be the value 1 if memory accesses described by the
5129 @var{mode} and @var{alignment} parameters have a cost many times greater
5130 than aligned accesses, for example if they are emulated in a trap
5133 When this macro is non-zero, the compiler will act as if
5134 @code{STRICT_ALIGNMENT} were non-zero when generating code for block
5135 moves. This can cause significantly more instructions to be produced.
5136 Therefore, do not set this macro non-zero if unaligned accesses only add a
5137 cycle or two to the time for a memory access.
5139 If the value of this macro is always zero, it need not be defined. If
5140 this macro is defined, it should produce a non-zero value when
5141 @code{STRICT_ALIGNMENT} is non-zero.
5143 @findex DONT_REDUCE_ADDR
5144 @item DONT_REDUCE_ADDR
5145 Define this macro to inhibit strength reduction of memory addresses.
5146 (On some machines, such strength reduction seems to do harm rather
5151 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5152 which a sequence of insns should be generated instead of a
5153 string move insn or a library call. Increasing the value will always
5154 make code faster, but eventually incurs high cost in increased code size.
5156 Note that on machines where the corresponding move insn is a
5157 @code{define_expand} that emits a sequence of insns, this macro counts
5158 the number of such sequences.
5160 If you don't define this, a reasonable default is used.
5162 @findex MOVE_BY_PIECES_P
5163 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5164 A C expression used to determine whether @code{move_by_pieces} will be used to
5165 copy a chunk of memory, or whether some other block move mechanism
5166 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5167 than @code{MOVE_RATIO}.
5169 @findex MOVE_MAX_PIECES
5170 @item MOVE_MAX_PIECES
5171 A C expression used by @code{move_by_pieces} to determine the largest unit
5172 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5174 @findex USE_LOAD_POST_INCREMENT
5175 @item USE_LOAD_POST_INCREMENT (@var{mode})
5176 A C expression used to determine whether a load postincrement is a good
5177 thing to use for a given mode. Defaults to the value of
5178 @code{HAVE_POST_INCREMENT}.
5180 @findex USE_LOAD_POST_DECREMENT
5181 @item USE_LOAD_POST_DECREMENT (@var{mode})
5182 A C expression used to determine whether a load postdecrement is a good
5183 thing to use for a given mode. Defaults to the value of
5184 @code{HAVE_POST_DECREMENT}.
5186 @findex USE_LOAD_PRE_INCREMENT
5187 @item USE_LOAD_PRE_INCREMENT (@var{mode})
5188 A C expression used to determine whether a load preincrement is a good
5189 thing to use for a given mode. Defaults to the value of
5190 @code{HAVE_PRE_INCREMENT}.
5192 @findex USE_LOAD_PRE_DECREMENT
5193 @item USE_LOAD_PRE_DECREMENT (@var{mode})
5194 A C expression used to determine whether a load predecrement is a good
5195 thing to use for a given mode. Defaults to the value of
5196 @code{HAVE_PRE_DECREMENT}.
5198 @findex USE_STORE_POST_INCREMENT
5199 @item USE_STORE_POST_INCREMENT (@var{mode})
5200 A C expression used to determine whether a store postincrement is a good
5201 thing to use for a given mode. Defaults to the value of
5202 @code{HAVE_POST_INCREMENT}.
5204 @findex USE_STORE_POST_DECREMENT
5205 @item USE_STORE_POST_DECREMENT (@var{mode})
5206 A C expression used to determine whether a store postdeccrement is a good
5207 thing to use for a given mode. Defaults to the value of
5208 @code{HAVE_POST_DECREMENT}.
5210 @findex USE_STORE_PRE_INCREMENT
5211 @item USE_STORE_PRE_INCREMENT (@var{mode})
5212 This macro is used to determine whether a store preincrement is a good
5213 thing to use for a given mode. Defaults to the value of
5214 @code{HAVE_PRE_INCREMENT}.
5216 @findex USE_STORE_PRE_DECREMENT
5217 @item USE_STORE_PRE_DECREMENT (@var{mode})
5218 This macro is used to determine whether a store predecrement is a good
5219 thing to use for a given mode. Defaults to the value of
5220 @code{HAVE_PRE_DECREMENT}.
5222 @findex NO_FUNCTION_CSE
5223 @item NO_FUNCTION_CSE
5224 Define this macro if it is as good or better to call a constant
5225 function address than to call an address kept in a register.
5227 @findex NO_RECURSIVE_FUNCTION_CSE
5228 @item NO_RECURSIVE_FUNCTION_CSE
5229 Define this macro if it is as good or better for a function to call
5230 itself with an explicit address than to call an address kept in a
5234 @item ADJUST_COST (@var{insn}, @var{link}, @var{dep_insn}, @var{cost})
5235 A C statement (sans semicolon) to update the integer variable @var{cost}
5236 based on the relationship between @var{insn} that is dependent on
5237 @var{dep_insn} through the dependence @var{link}. The default is to
5238 make no adjustment to @var{cost}. This can be used for example to
5239 specify to the scheduler that an output- or anti-dependence does not
5240 incur the same cost as a data-dependence.
5242 @findex ADJUST_PRIORITY
5243 @item ADJUST_PRIORITY (@var{insn})
5244 A C statement (sans semicolon) to update the integer scheduling
5245 priority @code{INSN_PRIORITY(@var{insn})}. Reduce the priority
5246 to execute the @var{insn} earlier, increase the priority to execute
5247 @var{insn} later. Do not define this macro if you do not need to
5248 adjust the scheduling priorities of insns.
5252 @section Dividing the Output into Sections (Texts, Data, @dots{})
5253 @c the above section title is WAY too long. maybe cut the part between
5254 @c the (...)? --mew 10feb93
5256 An object file is divided into sections containing different types of
5257 data. In the most common case, there are three sections: the @dfn{text
5258 section}, which holds instructions and read-only data; the @dfn{data
5259 section}, which holds initialized writable data; and the @dfn{bss
5260 section}, which holds uninitialized data. Some systems have other kinds
5263 The compiler must tell the assembler when to switch sections. These
5264 macros control what commands to output to tell the assembler this. You
5265 can also define additional sections.
5268 @findex TEXT_SECTION_ASM_OP
5269 @item TEXT_SECTION_ASM_OP
5270 A C expression whose value is a string, including spacing, containing the
5271 assembler operation that should precede instructions and read-only data.
5272 Normally @code{"\t.text"} is right.
5274 @findex DATA_SECTION_ASM_OP
5275 @item DATA_SECTION_ASM_OP
5276 A C expression whose value is a string, including spacing, containing the
5277 assembler operation to identify the following data as writable initialized
5278 data. Normally @code{"\t.data"} is right.
5280 @findex SHARED_SECTION_ASM_OP
5281 @item SHARED_SECTION_ASM_OP
5282 If defined, a C expression whose value is a string, including spacing,
5283 containing the assembler operation to identify the following data as
5284 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5286 @findex BSS_SECTION_ASM_OP
5287 @item BSS_SECTION_ASM_OP
5288 If defined, a C expression whose value is a string, including spacing,
5289 containing the assembler operation to identify the following data as
5290 uninitialized global data. If not defined, and neither
5291 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5292 uninitialized global data will be output in the data section if
5293 @samp{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5296 @findex SHARED_BSS_SECTION_ASM_OP
5297 @item SHARED_BSS_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 uninitialized global shared data. If not defined, and
5301 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5303 @findex INIT_SECTION_ASM_OP
5304 @item INIT_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 initialization code. If not defined, GCC will assume such a section does
5310 @findex FINI_SECTION_ASM_OP
5311 @item FINI_SECTION_ASM_OP
5312 If defined, a C expression whose value is a string, including spacing,
5313 containing the assembler operation to identify the following data as
5314 finalization code. If not defined, GCC will assume such a section does
5317 @findex CRT_CALL_STATIC_FUNCTION
5318 @item CRT_CALL_STATIC_FUNCTION
5319 If defined, a C statement that calls the function named as the sole
5320 argument of this macro. This is used in @file{crtstuff.c} if
5321 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls to
5322 initialization and finalization functions from the init and fini
5323 sections. By default, this macro is a simple function call. Some
5324 ports need hand-crafted assembly code to avoid dependencies on
5325 registers initialized in the function prologue or to ensure that
5326 constant pools don't end up too far way in the text section.
5328 @findex EXTRA_SECTIONS
5331 @item EXTRA_SECTIONS
5332 A list of names for sections other than the standard two, which are
5333 @code{in_text} and @code{in_data}. You need not define this macro
5334 on a system with no other sections (that GCC needs to use).
5336 @findex EXTRA_SECTION_FUNCTIONS
5337 @findex text_section
5338 @findex data_section
5339 @item EXTRA_SECTION_FUNCTIONS
5340 One or more functions to be defined in @file{varasm.c}. These
5341 functions should do jobs analogous to those of @code{text_section} and
5342 @code{data_section}, for your additional sections. Do not define this
5343 macro if you do not define @code{EXTRA_SECTIONS}.
5345 @findex READONLY_DATA_SECTION
5346 @item READONLY_DATA_SECTION
5347 On most machines, read-only variables, constants, and jump tables are
5348 placed in the text section. If this is not the case on your machine,
5349 this macro should be defined to be the name of a function (either
5350 @code{data_section} or a function defined in @code{EXTRA_SECTIONS}) that
5351 switches to the section to be used for read-only items.
5353 If these items should be placed in the text section, this macro should
5356 @findex SELECT_SECTION
5357 @item SELECT_SECTION (@var{exp}, @var{reloc})
5358 A C statement or statements to switch to the appropriate section for
5359 output of @var{exp}. You can assume that @var{exp} is either a
5360 @code{VAR_DECL} node or a constant of some sort. @var{reloc}
5361 indicates whether the initial value of @var{exp} requires link-time
5362 relocations. Select the section by calling @code{text_section} or one
5363 of the alternatives for other sections.
5365 Do not define this macro if you put all read-only variables and
5366 constants in the read-only data section (usually the text section).
5368 @findex SELECT_RTX_SECTION
5369 @item SELECT_RTX_SECTION (@var{mode}, @var{rtx})
5370 A C statement or statements to switch to the appropriate section for
5371 output of @var{rtx} in mode @var{mode}. You can assume that @var{rtx}
5372 is some kind of constant in RTL. The argument @var{mode} is redundant
5373 except in the case of a @code{const_int} rtx. Select the section by
5374 calling @code{text_section} or one of the alternatives for other
5377 Do not define this macro if you put all constants in the read-only
5380 @findex JUMP_TABLES_IN_TEXT_SECTION
5381 @item JUMP_TABLES_IN_TEXT_SECTION
5382 Define this macro to be an expression with a non-zero value if jump
5383 tables (for @code{tablejump} insns) should be output in the text
5384 section, along with the assembler instructions. Otherwise, the
5385 readonly data section is used.
5387 This macro is irrelevant if there is no separate readonly data section.
5389 @findex ENCODE_SECTION_INFO
5390 @item ENCODE_SECTION_INFO (@var{decl})
5391 Define this macro if references to a symbol must be treated differently
5392 depending on something about the variable or function named by the
5393 symbol (such as what section it is in).
5395 The macro definition, if any, is executed immediately after the rtl for
5396 @var{decl} has been created and stored in @code{DECL_RTL (@var{decl})}.
5397 The value of the rtl will be a @code{mem} whose address is a
5400 @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO}
5401 The usual thing for this macro to do is to record a flag in the
5402 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
5403 modified name string in the @code{symbol_ref} (if one bit is not enough
5406 @findex STRIP_NAME_ENCODING
5407 @item STRIP_NAME_ENCODING (@var{var}, @var{sym_name})
5408 Decode @var{sym_name} and store the real name part in @var{var}, sans
5409 the characters that encode section info. Define this macro if
5410 @code{ENCODE_SECTION_INFO} alters the symbol's name string.
5412 @findex UNIQUE_SECTION_P
5413 @item UNIQUE_SECTION_P (@var{decl})
5414 A C expression which evaluates to true if @var{decl} should be placed
5415 into a unique section for some target-specific reason. If you do not
5416 define this macro, the default is @samp{0}. Note that the flag
5417 @samp{-ffunction-sections} will also cause functions to be placed into
5420 @findex UNIQUE_SECTION
5421 @item UNIQUE_SECTION (@var{decl}, @var{reloc})
5422 A C statement to build up a unique section name, expressed as a
5423 STRING_CST node, and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5424 @var{reloc} indicates whether the initial value of @var{exp} requires
5425 link-time relocations. If you do not define this macro, GCC will use
5426 the symbol name prefixed by @samp{.} as the section name. Note - this
5427 macro can now be called for unitialised data items as well as
5428 initialised data and functions.
5432 @section Position Independent Code
5433 @cindex position independent code
5436 This section describes macros that help implement generation of position
5437 independent code. Simply defining these macros is not enough to
5438 generate valid PIC; you must also add support to the macros
5439 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5440 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5441 @samp{movsi} to do something appropriate when the source operand
5442 contains a symbolic address. You may also need to alter the handling of
5443 switch statements so that they use relative addresses.
5444 @c i rearranged the order of the macros above to try to force one of
5445 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5448 @findex PIC_OFFSET_TABLE_REGNUM
5449 @item PIC_OFFSET_TABLE_REGNUM
5450 The register number of the register used to address a table of static
5451 data addresses in memory. In some cases this register is defined by a
5452 processor's ``application binary interface'' (ABI). When this macro
5453 is defined, RTL is generated for this register once, as with the stack
5454 pointer and frame pointer registers. If this macro is not defined, it
5455 is up to the machine-dependent files to allocate such a register (if
5458 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5459 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5460 Define this macro if the register defined by
5461 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
5462 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5464 @findex FINALIZE_PIC
5466 By generating position-independent code, when two different programs (A
5467 and B) share a common library (libC.a), the text of the library can be
5468 shared whether or not the library is linked at the same address for both
5469 programs. In some of these environments, position-independent code
5470 requires not only the use of different addressing modes, but also
5471 special code to enable the use of these addressing modes.
5473 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
5474 codes once the function is being compiled into assembly code, but not
5475 before. (It is not done before, because in the case of compiling an
5476 inline function, it would lead to multiple PIC prologues being
5477 included in functions which used inline functions and were compiled to
5480 @findex LEGITIMATE_PIC_OPERAND_P
5481 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
5482 A C expression that is nonzero if @var{x} is a legitimate immediate
5483 operand on the target machine when generating position independent code.
5484 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5485 check this. You can also assume @var{flag_pic} is true, so you need not
5486 check it either. You need not define this macro if all constants
5487 (including @code{SYMBOL_REF}) can be immediate operands when generating
5488 position independent code.
5491 @node Assembler Format
5492 @section Defining the Output Assembler Language
5494 This section describes macros whose principal purpose is to describe how
5495 to write instructions in assembler language--rather than what the
5499 * File Framework:: Structural information for the assembler file.
5500 * Data Output:: Output of constants (numbers, strings, addresses).
5501 * Uninitialized Data:: Output of uninitialized variables.
5502 * Label Output:: Output and generation of labels.
5503 * Initialization:: General principles of initialization
5504 and termination routines.
5505 * Macros for Initialization::
5506 Specific macros that control the handling of
5507 initialization and termination routines.
5508 * Instruction Output:: Output of actual instructions.
5509 * Dispatch Tables:: Output of jump tables.
5510 * Exception Region Output:: Output of exception region code.
5511 * Alignment Output:: Pseudo ops for alignment and skipping data.
5514 @node File Framework
5515 @subsection The Overall Framework of an Assembler File
5516 @cindex assembler format
5517 @cindex output of assembler code
5519 @c prevent bad page break with this line
5520 This describes the overall framework of an assembler file.
5523 @findex ASM_FILE_START
5524 @item ASM_FILE_START (@var{stream})
5525 A C expression which outputs to the stdio stream @var{stream}
5526 some appropriate text to go at the start of an assembler file.
5528 Normally this macro is defined to output a line containing
5529 @samp{#NO_APP}, which is a comment that has no effect on most
5530 assemblers but tells the GNU assembler that it can save time by not
5531 checking for certain assembler constructs.
5533 On systems that use SDB, it is necessary to output certain commands;
5534 see @file{attasm.h}.
5536 @findex ASM_FILE_END
5537 @item ASM_FILE_END (@var{stream})
5538 A C expression which outputs to the stdio stream @var{stream}
5539 some appropriate text to go at the end of an assembler file.
5541 If this macro is not defined, the default is to output nothing
5542 special at the end of the file. Most systems don't require any
5545 On systems that use SDB, it is necessary to output certain commands;
5546 see @file{attasm.h}.
5548 @findex ASM_COMMENT_START
5549 @item ASM_COMMENT_START
5550 A C string constant describing how to begin a comment in the target
5551 assembler language. The compiler assumes that the comment will end at
5552 the end of the line.
5556 A C string constant for text to be output before each @code{asm}
5557 statement or group of consecutive ones. Normally this is
5558 @code{"#APP"}, which is a comment that has no effect on most
5559 assemblers but tells the GNU assembler that it must check the lines
5560 that follow for all valid assembler constructs.
5564 A C string constant for text to be output after each @code{asm}
5565 statement or group of consecutive ones. Normally this is
5566 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5567 time-saving assumptions that are valid for ordinary compiler output.
5569 @findex ASM_OUTPUT_SOURCE_FILENAME
5570 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5571 A C statement to output COFF information or DWARF debugging information
5572 which indicates that filename @var{name} is the current source file to
5573 the stdio stream @var{stream}.
5575 This macro need not be defined if the standard form of output
5576 for the file format in use is appropriate.
5578 @findex OUTPUT_QUOTED_STRING
5579 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5580 A C statement to output the string @var{string} to the stdio stream
5581 @var{stream}. If you do not call the function @code{output_quoted_string}
5582 in your config files, GCC will only call it to output filenames to
5583 the assembler source. So you can use it to canonicalize the format
5584 of the filename using this macro.
5586 @findex ASM_OUTPUT_SOURCE_LINE
5587 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
5588 A C statement to output DBX or SDB debugging information before code
5589 for line number @var{line} of the current source file to the
5590 stdio stream @var{stream}.
5592 This macro need not be defined if the standard form of debugging
5593 information for the debugger in use is appropriate.
5595 @findex ASM_OUTPUT_IDENT
5596 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
5597 A C statement to output something to the assembler file to handle a
5598 @samp{#ident} directive containing the text @var{string}. If this
5599 macro is not defined, nothing is output for a @samp{#ident} directive.
5601 @findex ASM_OUTPUT_SECTION_NAME
5602 @item ASM_OUTPUT_SECTION_NAME (@var{stream}, @var{decl}, @var{name}, @var{reloc})
5603 A C statement to output something to the assembler file to switch to section
5604 @var{name} for object @var{decl} which is either a @code{FUNCTION_DECL}, a
5605 @code{VAR_DECL} or @code{NULL_TREE}. @var{reloc}
5606 indicates whether the initial value of @var{exp} requires link-time
5607 relocations. The string given by @var{name} will always be the
5608 canonical version stored in the global stringpool.
5610 Some target formats do not support arbitrary sections. Do not define
5611 this macro in such cases.
5613 At present this macro is only used to support section attributes.
5614 When this macro is undefined, section attributes are disabled.
5616 @findex OBJC_PROLOGUE
5618 A C statement to output any assembler statements which are required to
5619 precede any Objective C object definitions or message sending. The
5620 statement is executed only when compiling an Objective C program.
5625 @subsection Output of Data
5627 @c prevent bad page break with this line
5628 This describes data output.
5631 @findex ASM_OUTPUT_LONG_DOUBLE
5632 @findex ASM_OUTPUT_DOUBLE
5633 @findex ASM_OUTPUT_FLOAT
5634 @item ASM_OUTPUT_LONG_DOUBLE (@var{stream}, @var{value})
5635 @itemx ASM_OUTPUT_DOUBLE (@var{stream}, @var{value})
5636 @itemx ASM_OUTPUT_FLOAT (@var{stream}, @var{value})
5637 @itemx ASM_OUTPUT_THREE_QUARTER_FLOAT (@var{stream}, @var{value})
5638 @itemx ASM_OUTPUT_SHORT_FLOAT (@var{stream}, @var{value})
5639 @itemx ASM_OUTPUT_BYTE_FLOAT (@var{stream}, @var{value})
5640 A C statement to output to the stdio stream @var{stream} an assembler
5641 instruction to assemble a floating-point constant of @code{TFmode},
5642 @code{DFmode}, @code{SFmode}, @code{TQFmode}, @code{HFmode}, or
5643 @code{QFmode}, respectively, whose value is @var{value}. @var{value}
5644 will be a C expression of type @code{REAL_VALUE_TYPE}. Macros such as
5645 @code{REAL_VALUE_TO_TARGET_DOUBLE} are useful for writing these
5648 @findex ASM_OUTPUT_QUADRUPLE_INT
5649 @findex ASM_OUTPUT_DOUBLE_INT
5650 @findex ASM_OUTPUT_INT
5651 @findex ASM_OUTPUT_SHORT
5652 @findex ASM_OUTPUT_CHAR
5653 @findex output_addr_const
5654 @item ASM_OUTPUT_QUADRUPLE_INT (@var{stream}, @var{exp})
5655 @itemx ASM_OUTPUT_DOUBLE_INT (@var{stream}, @var{exp})
5656 @itemx ASM_OUTPUT_INT (@var{stream}, @var{exp})
5657 @itemx ASM_OUTPUT_SHORT (@var{stream}, @var{exp})
5658 @itemx ASM_OUTPUT_CHAR (@var{stream}, @var{exp})
5659 A C statement to output to the stdio stream @var{stream} an assembler
5660 instruction to assemble an integer of 16, 8, 4, 2 or 1 bytes,
5661 respectively, whose value is @var{value}. The argument @var{exp} will
5662 be an RTL expression which represents a constant value. Use
5663 @samp{output_addr_const (@var{stream}, @var{exp})} to output this value
5664 as an assembler expression.@refill
5666 For sizes larger than @code{UNITS_PER_WORD}, if the action of a macro
5667 would be identical to repeatedly calling the macro corresponding to
5668 a size of @code{UNITS_PER_WORD}, once for each word, you need not define
5671 @findex OUTPUT_ADDR_CONST_EXTRA
5672 @item OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
5673 A C statement to recognize @var{rtx} patterns that
5674 @code{output_addr_const} can't deal with, and output assembly code to
5675 @var{stream} corresponding to the pattern @var{x}. This may be used to
5676 allow machine-dependent @code{UNSPEC}s to appear within constants.
5678 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
5679 @code{goto fail}, so that a standard error message is printed. If it
5680 prints an error message itself, by calling, for example,
5681 @code{output_operand_lossage}, it may just complete normally.
5683 @findex ASM_OUTPUT_BYTE
5684 @item ASM_OUTPUT_BYTE (@var{stream}, @var{value})
5685 A C statement to output to the stdio stream @var{stream} an assembler
5686 instruction to assemble a single byte containing the number @var{value}.
5690 A C string constant, including spacing, giving the pseudo-op to use for a
5691 sequence of single-byte constants. If this macro is not defined, the
5692 default is @code{"\t.byte\t"}.
5694 @findex UNALIGNED_SHORT_ASM_OP
5695 @findex UNALIGNED_INT_ASM_OP
5696 @findex UNALIGNED_DOUBLE_INT_ASM_OP
5697 @item UNALIGNED_SHORT_ASM_OP
5698 @itemx UNALIGNED_INT_ASM_OP
5699 @itemx UNALIGNED_DOUBLE_INT_ASM_OP
5700 A C string constant, including spacing, giving the pseudo-op to use
5701 to assemble 16-, 32-, and 64-bit integers respectively @emph{without}
5702 adding implicit padding or alignment. These macros are required if
5703 DWARF 2 frame unwind is used. On ELF systems, these will default
5704 to @code{.2byte}, @code{.4byte}, and @code{.8byte}.@refill
5706 @findex ASM_OUTPUT_ASCII
5707 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5708 A C statement to output to the stdio stream @var{stream} an assembler
5709 instruction to assemble a string constant containing the @var{len}
5710 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5711 @code{char *} and @var{len} a C expression of type @code{int}.
5713 If the assembler has a @code{.ascii} pseudo-op as found in the
5714 Berkeley Unix assembler, do not define the macro
5715 @code{ASM_OUTPUT_ASCII}.
5717 @findex CONSTANT_POOL_BEFORE_FUNCTION
5718 @item CONSTANT_POOL_BEFORE_FUNCTION
5719 You may define this macro as a C expression. You should define the
5720 expression to have a non-zero value if GCC should output the constant
5721 pool for a function before the code for the function, or a zero value if
5722 GCC should output the constant pool after the function. If you do
5723 not define this macro, the usual case, GCC will output the constant
5724 pool before the function.
5726 @findex ASM_OUTPUT_POOL_PROLOGUE
5727 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
5728 A C statement to output assembler commands to define the start of the
5729 constant pool for a function. @var{funname} is a string giving
5730 the name of the function. Should the return type of the function
5731 be required, it can be obtained via @var{fundecl}. @var{size}
5732 is the size, in bytes, of the constant pool that will be written
5733 immediately after this call.
5735 If no constant-pool prefix is required, the usual case, this macro need
5738 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
5739 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5740 A C statement (with or without semicolon) to output a constant in the
5741 constant pool, if it needs special treatment. (This macro need not do
5742 anything for RTL expressions that can be output normally.)
5744 The argument @var{file} is the standard I/O stream to output the
5745 assembler code on. @var{x} is the RTL expression for the constant to
5746 output, and @var{mode} is the machine mode (in case @var{x} is a
5747 @samp{const_int}). @var{align} is the required alignment for the value
5748 @var{x}; you should output an assembler directive to force this much
5751 The argument @var{labelno} is a number to use in an internal label for
5752 the address of this pool entry. The definition of this macro is
5753 responsible for outputting the label definition at the proper place.
5754 Here is how to do this:
5757 ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
5760 When you output a pool entry specially, you should end with a
5761 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5762 entry from being output a second time in the usual manner.
5764 You need not define this macro if it would do nothing.
5766 @findex CONSTANT_AFTER_FUNCTION_P
5767 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
5768 Define this macro as a C expression which is nonzero if the constant
5769 @var{exp}, of type @code{tree}, should be output after the code for a
5770 function. The compiler will normally output all constants before the
5771 function; you need not define this macro if this is OK.
5773 @findex ASM_OUTPUT_POOL_EPILOGUE
5774 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5775 A C statement to output assembler commands to at the end of the constant
5776 pool for a function. @var{funname} is a string giving the name of the
5777 function. Should the return type of the function be required, you can
5778 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5779 constant pool that GCC wrote immediately before this call.
5781 If no constant-pool epilogue is required, the usual case, you need not
5784 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
5785 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
5786 Define this macro as a C expression which is nonzero if @var{C} is
5787 used as a logical line separator by the assembler.
5789 If you do not define this macro, the default is that only
5790 the character @samp{;} is treated as a logical line separator.
5793 @findex ASM_OPEN_PAREN
5794 @findex ASM_CLOSE_PAREN
5795 @item ASM_OPEN_PAREN
5796 @itemx ASM_CLOSE_PAREN
5797 These macros are defined as C string constants, describing the syntax
5798 in the assembler for grouping arithmetic expressions. The following
5799 definitions are correct for most assemblers:
5802 #define ASM_OPEN_PAREN "("
5803 #define ASM_CLOSE_PAREN ")"
5807 These macros are provided by @file{real.h} for writing the definitions
5808 of @code{ASM_OUTPUT_DOUBLE} and the like:
5811 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5812 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5813 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5814 @findex REAL_VALUE_TO_TARGET_SINGLE
5815 @findex REAL_VALUE_TO_TARGET_DOUBLE
5816 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
5817 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
5818 floating point representation, and store its bit pattern in the array of
5819 @code{long int} whose address is @var{l}. The number of elements in the
5820 output array is determined by the size of the desired target floating
5821 point data type: 32 bits of it go in each @code{long int} array
5822 element. Each array element holds 32 bits of the result, even if
5823 @code{long int} is wider than 32 bits on the host machine.
5825 The array element values are designed so that you can print them out
5826 using @code{fprintf} in the order they should appear in the target
5829 @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
5830 @findex REAL_VALUE_TO_DECIMAL
5831 This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
5832 decimal number and stores it as a string into @var{string}.
5833 You must pass, as @var{string}, the address of a long enough block
5834 of space to hold the result.
5836 The argument @var{format} is a @code{printf}-specification that serves
5837 as a suggestion for how to format the output string.
5840 @node Uninitialized Data
5841 @subsection Output of Uninitialized Variables
5843 Each of the macros in this section is used to do the whole job of
5844 outputting a single uninitialized variable.
5847 @findex ASM_OUTPUT_COMMON
5848 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5849 A C statement (sans semicolon) to output to the stdio stream
5850 @var{stream} the assembler definition of a common-label named
5851 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5852 is the size rounded up to whatever alignment the caller wants.
5854 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5855 output the name itself; before and after that, output the additional
5856 assembler syntax for defining the name, and a newline.
5858 This macro controls how the assembler definitions of uninitialized
5859 common global variables are output.
5861 @findex ASM_OUTPUT_ALIGNED_COMMON
5862 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5863 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5864 separate, explicit argument. If you define this macro, it is used in
5865 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5866 handling the required alignment of the variable. The alignment is specified
5867 as the number of bits.
5869 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
5870 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5871 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5872 variable to be output, if there is one, or @code{NULL_TREE} if there
5873 is no corresponding variable. If you define this macro, GCC will use it
5874 in place of both @code{ASM_OUTPUT_COMMON} and
5875 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5876 the variable's decl in order to chose what to output.
5878 @findex ASM_OUTPUT_SHARED_COMMON
5879 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5880 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
5881 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
5884 @findex ASM_OUTPUT_BSS
5885 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5886 A C statement (sans semicolon) to output to the stdio stream
5887 @var{stream} the assembler definition of uninitialized global @var{decl} named
5888 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5889 is the size rounded up to whatever alignment the caller wants.
5891 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
5892 defining this macro. If unable, use the expression
5893 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5894 before and after that, output the additional assembler syntax for defining
5895 the name, and a newline.
5897 This macro controls how the assembler definitions of uninitialized global
5898 variables are output. This macro exists to properly support languages like
5899 @code{c++} which do not have @code{common} data. However, this macro currently
5900 is not defined for all targets. If this macro and
5901 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
5902 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
5903 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
5905 @findex ASM_OUTPUT_ALIGNED_BSS
5906 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5907 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
5908 separate, explicit argument. If you define this macro, it is used in
5909 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
5910 handling the required alignment of the variable. The alignment is specified
5911 as the number of bits.
5913 Try to use function @code{asm_output_aligned_bss} defined in file
5914 @file{varasm.c} when defining this macro.
5916 @findex ASM_OUTPUT_SHARED_BSS
5917 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5918 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
5919 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
5922 @findex ASM_OUTPUT_LOCAL
5923 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5924 A C statement (sans semicolon) to output to the stdio stream
5925 @var{stream} the assembler definition of a local-common-label named
5926 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5927 is the size rounded up to whatever alignment the caller wants.
5929 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5930 output the name itself; before and after that, output the additional
5931 assembler syntax for defining the name, and a newline.
5933 This macro controls how the assembler definitions of uninitialized
5934 static variables are output.
5936 @findex ASM_OUTPUT_ALIGNED_LOCAL
5937 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5938 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5939 separate, explicit argument. If you define this macro, it is used in
5940 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5941 handling the required alignment of the variable. The alignment is specified
5942 as the number of bits.
5944 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
5945 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5946 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5947 variable to be output, if there is one, or @code{NULL_TREE} if there
5948 is no corresponding variable. If you define this macro, GCC will use it
5949 in place of both @code{ASM_OUTPUT_DECL} and
5950 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5951 the variable's decl in order to chose what to output.
5953 @findex ASM_OUTPUT_SHARED_LOCAL
5954 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5955 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
5956 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
5961 @subsection Output and Generation of Labels
5963 @c prevent bad page break with this line
5964 This is about outputting labels.
5967 @findex ASM_OUTPUT_LABEL
5968 @findex assemble_name
5969 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5970 A C statement (sans semicolon) to output to the stdio stream
5971 @var{stream} the assembler definition of a label named @var{name}.
5972 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5973 output the name itself; before and after that, output the additional
5974 assembler syntax for defining the name, and a newline.
5976 @findex ASM_DECLARE_FUNCTION_NAME
5977 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5978 A C statement (sans semicolon) to output to the stdio stream
5979 @var{stream} any text necessary for declaring the name @var{name} of a
5980 function which is being defined. This macro is responsible for
5981 outputting the label definition (perhaps using
5982 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
5983 @code{FUNCTION_DECL} tree node representing the function.
5985 If this macro is not defined, then the function name is defined in the
5986 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5988 @findex ASM_DECLARE_FUNCTION_SIZE
5989 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5990 A C statement (sans semicolon) to output to the stdio stream
5991 @var{stream} any text necessary for declaring the size of a function
5992 which is being defined. The argument @var{name} is the name of the
5993 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5994 representing the function.
5996 If this macro is not defined, then the function size is not defined.
5998 @findex ASM_DECLARE_OBJECT_NAME
5999 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6000 A C statement (sans semicolon) to output to the stdio stream
6001 @var{stream} any text necessary for declaring the name @var{name} of an
6002 initialized variable which is being defined. This macro must output the
6003 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6004 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6006 If this macro is not defined, then the variable name is defined in the
6007 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6009 @findex ASM_DECLARE_REGISTER_GLOBAL
6010 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6011 A C statement (sans semicolon) to output to the stdio stream
6012 @var{stream} any text necessary for claiming a register @var{regno}
6013 for a global variable @var{decl} with name @var{name}.
6015 If you don't define this macro, that is equivalent to defining it to do
6018 @findex ASM_FINISH_DECLARE_OBJECT
6019 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6020 A C statement (sans semicolon) to finish up declaring a variable name
6021 once the compiler has processed its initializer fully and thus has had a
6022 chance to determine the size of an array when controlled by an
6023 initializer. This is used on systems where it's necessary to declare
6024 something about the size of the object.
6026 If you don't define this macro, that is equivalent to defining it to do
6029 @findex ASM_GLOBALIZE_LABEL
6030 @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
6031 A C statement (sans semicolon) to output to the stdio stream
6032 @var{stream} some commands that will make the label @var{name} global;
6033 that is, available for reference from other files. Use the expression
6034 @code{assemble_name (@var{stream}, @var{name})} to output the name
6035 itself; before and after that, output the additional assembler syntax
6036 for making that name global, and a newline.
6038 @findex ASM_WEAKEN_LABEL
6039 @item ASM_WEAKEN_LABEL
6040 A C statement (sans semicolon) to output to the stdio stream
6041 @var{stream} some commands that will make the label @var{name} weak;
6042 that is, available for reference from other files but only used if
6043 no other definition is available. Use the expression
6044 @code{assemble_name (@var{stream}, @var{name})} to output the name
6045 itself; before and after that, output the additional assembler syntax
6046 for making that name weak, and a newline.
6048 If you don't define this macro, GCC will not support weak
6049 symbols and you should not define the @code{SUPPORTS_WEAK} macro.
6051 @findex SUPPORTS_WEAK
6053 A C expression which evaluates to true if the target supports weak symbols.
6055 If you don't define this macro, @file{defaults.h} provides a default
6056 definition. If @code{ASM_WEAKEN_LABEL} is defined, the default
6057 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6058 you want to control weak symbol support with a compiler flag such as
6061 @findex MAKE_DECL_ONE_ONLY (@var{decl})
6062 @item MAKE_DECL_ONE_ONLY
6063 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6064 public symbol such that extra copies in multiple translation units will
6065 be discarded by the linker. Define this macro if your object file
6066 format provides support for this concept, such as the @samp{COMDAT}
6067 section flags in the Microsoft Windows PE/COFF format, and this support
6068 requires changes to @var{decl}, such as putting it in a separate section.
6070 @findex SUPPORTS_ONE_ONLY
6071 @item SUPPORTS_ONE_ONLY
6072 A C expression which evaluates to true if the target supports one-only
6075 If you don't define this macro, @file{varasm.c} provides a default
6076 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6077 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6078 you want to control one-only symbol support with a compiler flag, or if
6079 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6080 be emitted as one-only.
6082 @findex ASM_OUTPUT_EXTERNAL
6083 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6084 A C statement (sans semicolon) to output to the stdio stream
6085 @var{stream} any text necessary for declaring the name of an external
6086 symbol named @var{name} which is referenced in this compilation but
6087 not defined. The value of @var{decl} is the tree node for the
6090 This macro need not be defined if it does not need to output anything.
6091 The GNU assembler and most Unix assemblers don't require anything.
6093 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
6094 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
6095 A C statement (sans semicolon) to output on @var{stream} an assembler
6096 pseudo-op to declare a library function name external. The name of the
6097 library function is given by @var{symref}, which has type @code{rtx} and
6098 is a @code{symbol_ref}.
6100 This macro need not be defined if it does not need to output anything.
6101 The GNU assembler and most Unix assemblers don't require anything.
6103 @findex ASM_OUTPUT_LABELREF
6104 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6105 A C statement (sans semicolon) to output to the stdio stream
6106 @var{stream} a reference in assembler syntax to a label named
6107 @var{name}. This should add @samp{_} to the front of the name, if that
6108 is customary on your operating system, as it is in most Berkeley Unix
6109 systems. This macro is used in @code{assemble_name}.
6111 @ignore @c Seems not to exist anymore.
6112 @findex ASM_OUTPUT_LABELREF_AS_INT
6113 @item ASM_OUTPUT_LABELREF_AS_INT (@var{file}, @var{label})
6114 Define this macro for systems that use the program @code{collect2}.
6115 The definition should be a C statement to output a word containing
6116 a reference to the label @var{label}.
6119 @findex ASM_OUTPUT_SYMBOL_REF
6120 @item ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6121 A C statement (sans semicolon) to output a reference to
6122 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_output}
6123 will be used to output the name of the symbol. This macro may be used
6124 to modify the way a symbol is referenced depending on information
6125 encoded by @code{ENCODE_SECTION_INFO}.
6127 @findex ASM_OUTPUT_INTERNAL_LABEL
6128 @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
6129 A C statement to output to the stdio stream @var{stream} a label whose
6130 name is made from the string @var{prefix} and the number @var{num}.
6132 It is absolutely essential that these labels be distinct from the labels
6133 used for user-level functions and variables. Otherwise, certain programs
6134 will have name conflicts with internal labels.
6136 It is desirable to exclude internal labels from the symbol table of the
6137 object file. Most assemblers have a naming convention for labels that
6138 should be excluded; on many systems, the letter @samp{L} at the
6139 beginning of a label has this effect. You should find out what
6140 convention your system uses, and follow it.
6142 The usual definition of this macro is as follows:
6145 fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
6148 @findex ASM_OUTPUT_DEBUG_LABEL
6149 @item ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6150 A C statement to output to the stdio stream @var{stream} a debug info
6151 label whose name is made from the string @var{prefix} and the number
6152 @var{num}. This is useful for VLIW targets, where debug info labels
6153 may need to be treated differently than branch target labels. On some
6154 systems, branch target labels must be at the beginning of instruction
6155 bundles, but debug info labels can occur in the middle of instruction
6158 If this macro is not defined, then @code{ASM_OUTPUT_INTERNAL_LABEL} will be
6161 @findex ASM_OUTPUT_ALTERNATE_LABEL_NAME
6162 @item ASM_OUTPUT_ALTERNATE_LABEL_NAME (@var{stream}, @var{string})
6163 A C statement to output to the stdio stream @var{stream} the string
6166 The default definition of this macro is as follows:
6169 fprintf (@var{stream}, "%s:\n", LABEL_ALTERNATE_NAME (INSN))
6172 @findex ASM_GENERATE_INTERNAL_LABEL
6173 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6174 A C statement to store into the string @var{string} a label whose name
6175 is made from the string @var{prefix} and the number @var{num}.
6177 This string, when output subsequently by @code{assemble_name}, should
6178 produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
6179 with the same @var{prefix} and @var{num}.
6181 If the string begins with @samp{*}, then @code{assemble_name} will
6182 output the rest of the string unchanged. It is often convenient for
6183 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6184 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6185 to output the string, and may change it. (Of course,
6186 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6187 you should know what it does on your machine.)
6189 @findex ASM_FORMAT_PRIVATE_NAME
6190 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6191 A C expression to assign to @var{outvar} (which is a variable of type
6192 @code{char *}) a newly allocated string made from the string
6193 @var{name} and the number @var{number}, with some suitable punctuation
6194 added. Use @code{alloca} to get space for the string.
6196 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6197 produce an assembler label for an internal static variable whose name is
6198 @var{name}. Therefore, the string must be such as to result in valid
6199 assembler code. The argument @var{number} is different each time this
6200 macro is executed; it prevents conflicts between similarly-named
6201 internal static variables in different scopes.
6203 Ideally this string should not be a valid C identifier, to prevent any
6204 conflict with the user's own symbols. Most assemblers allow periods
6205 or percent signs in assembler symbols; putting at least one of these
6206 between the name and the number will suffice.
6208 @findex ASM_OUTPUT_DEF
6209 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6210 A C statement to output to the stdio stream @var{stream} assembler code
6211 which defines (equates) the symbol @var{name} to have the value @var{value}.
6214 If SET_ASM_OP is defined, a default definition is provided which is
6215 correct for most systems.
6217 @findex ASM_OUTPUT_DEF_FROM_DECLS
6218 @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6219 A C statement to output to the stdio stream @var{stream} assembler code
6220 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6221 to have the value of the tree node @var{decl_of_value}. This macro will
6222 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6223 the tree nodes are available.
6225 @findex ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL
6226 @item ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL (@var{stream}, @var{symbol}, @var{high}, @var{low})
6227 A C statement to output to the stdio stream @var{stream} assembler code
6228 which defines (equates) the symbol @var{symbol} to have a value equal to
6229 the difference of the two symbols @var{high} and @var{low}, i.e.
6230 @var{high} minus @var{low}. GCC guarantees that the symbols @var{high}
6231 and @var{low} are already known by the assembler so that the difference
6232 resolves into a constant.
6235 If SET_ASM_OP is defined, a default definition is provided which is
6236 correct for most systems.
6238 @findex ASM_OUTPUT_WEAK_ALIAS
6239 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6240 A C statement to output to the stdio stream @var{stream} assembler code
6241 which defines (equates) the weak symbol @var{name} to have the value
6244 Define this macro if the target only supports weak aliases; define
6245 ASM_OUTPUT_DEF instead if possible.
6247 @findex OBJC_GEN_METHOD_LABEL
6248 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6249 Define this macro to override the default assembler names used for
6250 Objective C methods.
6252 The default name is a unique method number followed by the name of the
6253 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6254 the category is also included in the assembler name (e.g.@:
6257 These names are safe on most systems, but make debugging difficult since
6258 the method's selector is not present in the name. Therefore, particular
6259 systems define other ways of computing names.
6261 @var{buf} is an expression of type @code{char *} which gives you a
6262 buffer in which to store the name; its length is as long as
6263 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6264 50 characters extra.
6266 The argument @var{is_inst} specifies whether the method is an instance
6267 method or a class method; @var{class_name} is the name of the class;
6268 @var{cat_name} is the name of the category (or NULL if the method is not
6269 in a category); and @var{sel_name} is the name of the selector.
6271 On systems where the assembler can handle quoted names, you can use this
6272 macro to provide more human-readable names.
6274 @findex ASM_DECLARE_UNRESOLVED_REFERENCE
6275 @item ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6276 A C statement (sans semicolon) to output to the stdio stream
6277 @var{stream} commands to declare that the label @var{name} is an
6278 unresolved Objective-C class reference. This is only needed for targets
6279 whose linkers have special support for NeXT-style runtimes.
6282 @node Initialization
6283 @subsection How Initialization Functions Are Handled
6284 @cindex initialization routines
6285 @cindex termination routines
6286 @cindex constructors, output of
6287 @cindex destructors, output of
6289 The compiled code for certain languages includes @dfn{constructors}
6290 (also called @dfn{initialization routines})---functions to initialize
6291 data in the program when the program is started. These functions need
6292 to be called before the program is ``started''---that is to say, before
6293 @code{main} is called.
6295 Compiling some languages generates @dfn{destructors} (also called
6296 @dfn{termination routines}) that should be called when the program
6299 To make the initialization and termination functions work, the compiler
6300 must output something in the assembler code to cause those functions to
6301 be called at the appropriate time. When you port the compiler to a new
6302 system, you need to specify how to do this.
6304 There are two major ways that GCC currently supports the execution of
6305 initialization and termination functions. Each way has two variants.
6306 Much of the structure is common to all four variations.
6308 @findex __CTOR_LIST__
6309 @findex __DTOR_LIST__
6310 The linker must build two lists of these functions---a list of
6311 initialization functions, called @code{__CTOR_LIST__}, and a list of
6312 termination functions, called @code{__DTOR_LIST__}.
6314 Each list always begins with an ignored function pointer (which may hold
6315 0, @minus{}1, or a count of the function pointers after it, depending on
6316 the environment). This is followed by a series of zero or more function
6317 pointers to constructors (or destructors), followed by a function
6318 pointer containing zero.
6320 Depending on the operating system and its executable file format, either
6321 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6322 time and exit time. Constructors are called in reverse order of the
6323 list; destructors in forward order.
6325 The best way to handle static constructors works only for object file
6326 formats which provide arbitrarily-named sections. A section is set
6327 aside for a list of constructors, and another for a list of destructors.
6328 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6329 object file that defines an initialization function also puts a word in
6330 the constructor section to point to that function. The linker
6331 accumulates all these words into one contiguous @samp{.ctors} section.
6332 Termination functions are handled similarly.
6334 To use this method, you need appropriate definitions of the macros
6335 @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}. Usually
6336 you can get them by including @file{svr4.h}.
6338 When arbitrary sections are available, there are two variants, depending
6339 upon how the code in @file{crtstuff.c} is called. On systems that
6340 support an @dfn{init} section which is executed at program startup,
6341 parts of @file{crtstuff.c} are compiled into that section. The
6342 program is linked by the @code{gcc} driver like this:
6345 ld -o @var{output_file} crtbegin.o @dots{} crtend.o -lgcc
6348 The head of a function (@code{__do_global_ctors}) appears in the init
6349 section of @file{crtbegin.o}; the remainder of the function appears in
6350 the init section of @file{crtend.o}. The linker will pull these two
6351 parts of the section together, making a whole function. If any of the
6352 user's object files linked into the middle of it contribute code, then that
6353 code will be executed as part of the body of @code{__do_global_ctors}.
6355 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
6358 If no init section is available, do not define
6359 @code{INIT_SECTION_ASM_OP}. Then @code{__do_global_ctors} is built into
6360 the text section like all other functions, and resides in
6361 @file{libgcc.a}. When GCC compiles any function called @code{main}, it
6362 inserts a procedure call to @code{__main} as the first executable code
6363 after the function prologue. The @code{__main} function, also defined
6364 in @file{libgcc2.c}, simply calls @file{__do_global_ctors}.
6366 In file formats that don't support arbitrary sections, there are again
6367 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
6368 and an `a.out' format must be used. In this case,
6369 @code{ASM_OUTPUT_CONSTRUCTOR} is defined to produce a @code{.stabs}
6370 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
6371 and with the address of the void function containing the initialization
6372 code as its value. The GNU linker recognizes this as a request to add
6373 the value to a ``set''; the values are accumulated, and are eventually
6374 placed in the executable as a vector in the format described above, with
6375 a leading (ignored) count and a trailing zero element.
6376 @code{ASM_OUTPUT_DESTRUCTOR} is handled similarly. Since no init
6377 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
6378 the compilation of @code{main} to call @code{__main} as above, starting
6379 the initialization process.
6381 The last variant uses neither arbitrary sections nor the GNU linker.
6382 This is preferable when you want to do dynamic linking and when using
6383 file formats which the GNU linker does not support, such as `ECOFF'. In
6384 this case, @code{ASM_OUTPUT_CONSTRUCTOR} does not produce an
6385 @code{N_SETT} symbol; initialization and termination functions are
6386 recognized simply by their names. This requires an extra program in the
6387 linkage step, called @code{collect2}. This program pretends to be the
6388 linker, for use with GCC; it does its job by running the ordinary
6389 linker, but also arranges to include the vectors of initialization and
6390 termination functions. These functions are called via @code{__main} as
6393 Choosing among these configuration options has been simplified by a set
6394 of operating-system-dependent files in the @file{config} subdirectory.
6395 These files define all of the relevant parameters. Usually it is
6396 sufficient to include one into your specific machine-dependent
6397 configuration file. These files are:
6401 For operating systems using the `a.out' format.
6404 For operating systems using the `MachO' format.
6407 For System V Release 3 and similar systems using `COFF' format.
6410 For System V Release 4 and similar systems using `ELF' format.
6413 For the VMS operating system.
6417 The following section describes the specific macros that control and
6418 customize the handling of initialization and termination functions.
6421 @node Macros for Initialization
6422 @subsection Macros Controlling Initialization Routines
6424 Here are the macros that control how the compiler handles initialization
6425 and termination functions:
6428 @findex INIT_SECTION_ASM_OP
6429 @item INIT_SECTION_ASM_OP
6430 If defined, a C string constant, including spacing, for the assembler
6431 operation to identify the following data as initialization code. If not
6432 defined, GCC will assume such a section does not exist. When you are
6433 using special sections for initialization and termination functions, this
6434 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6435 run the initialization functions.
6437 @item HAS_INIT_SECTION
6438 @findex HAS_INIT_SECTION
6439 If defined, @code{main} will not call @code{__main} as described above.
6440 This macro should be defined for systems that control the contents of the
6441 init section on a symbol-by-symbol basis, such as OSF/1, and should not
6442 be defined explicitly for systems that support
6443 @code{INIT_SECTION_ASM_OP}.
6445 @item LD_INIT_SWITCH
6446 @findex LD_INIT_SWITCH
6447 If defined, a C string constant for a switch that tells the linker that
6448 the following symbol is an initialization routine.
6450 @item LD_FINI_SWITCH
6451 @findex LD_FINI_SWITCH
6452 If defined, a C string constant for a switch that tells the linker that
6453 the following symbol is a finalization routine.
6456 @findex INVOKE__main
6457 If defined, @code{main} will call @code{__main} despite the presence of
6458 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6459 where the init section is not actually run automatically, but is still
6460 useful for collecting the lists of constructors and destructors.
6462 @item SUPPORTS_INIT_PRIORITY
6463 @findex SUPPORTS_INIT_PRIORITY
6464 If nonzero, the C++ @code{init_priority} attribute is supported and the
6465 compiler should emit instructions to control the order of initialization
6466 of objects. If zero, the compiler will issue an error message upon
6467 encountering an @code{init_priority} attribute.
6469 @item ASM_OUTPUT_CONSTRUCTOR (@var{stream}, @var{name})
6470 @findex ASM_OUTPUT_CONSTRUCTOR
6471 Define this macro as a C statement to output on the stream @var{stream}
6472 the assembler code to arrange to call the function named @var{name} at
6473 initialization time.
6475 Assume that @var{name} is the name of a C function generated
6476 automatically by the compiler. This function takes no arguments. Use
6477 the function @code{assemble_name} to output the name @var{name}; this
6478 performs any system-specific syntactic transformations such as adding an
6481 If you don't define this macro, nothing special is output to arrange to
6482 call the function. This is correct when the function will be called in
6483 some other manner---for example, by means of the @code{collect2} program,
6484 which looks through the symbol table to find these functions by their
6487 @item ASM_OUTPUT_DESTRUCTOR (@var{stream}, @var{name})
6488 @findex ASM_OUTPUT_DESTRUCTOR
6489 This is like @code{ASM_OUTPUT_CONSTRUCTOR} but used for termination
6490 functions rather than initialization functions.
6492 When @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR} are
6493 defined, the initialization routine generated for the generated object
6494 file will have static linkage.
6497 If your system uses @code{collect2} as the means of processing
6498 constructors, then that program normally uses @code{nm} to scan an
6499 object file for constructor functions to be called. On such systems you
6500 must not define @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}
6501 as the object file's initialization routine must have global scope.
6503 On certain kinds of systems, you can define these macros to make
6504 @code{collect2} work faster (and, in some cases, make it work at all):
6507 @findex OBJECT_FORMAT_COFF
6508 @item OBJECT_FORMAT_COFF
6509 Define this macro if the system uses COFF (Common Object File Format)
6510 object files, so that @code{collect2} can assume this format and scan
6511 object files directly for dynamic constructor/destructor functions.
6513 @findex OBJECT_FORMAT_ROSE
6514 @item OBJECT_FORMAT_ROSE
6515 Define this macro if the system uses ROSE format object files, so that
6516 @code{collect2} can assume this format and scan object files directly
6517 for dynamic constructor/destructor functions.
6519 These macros are effective only in a native compiler; @code{collect2} as
6520 part of a cross compiler always uses @code{nm} for the target machine.
6522 @findex REAL_NM_FILE_NAME
6523 @item REAL_NM_FILE_NAME
6524 Define this macro as a C string constant containing the file name to use
6525 to execute @code{nm}. The default is to search the path normally for
6528 If your system supports shared libraries and has a program to list the
6529 dynamic dependencies of a given library or executable, you can define
6530 these macros to enable support for running initialization and
6531 termination functions in shared libraries:
6535 Define this macro to a C string constant containing the name of the
6536 program which lists dynamic dependencies, like @code{"ldd"} under SunOS 4.
6538 @findex PARSE_LDD_OUTPUT
6539 @item PARSE_LDD_OUTPUT (@var{PTR})
6540 Define this macro to be C code that extracts filenames from the output
6541 of the program denoted by @code{LDD_SUFFIX}. @var{PTR} is a variable
6542 of type @code{char *} that points to the beginning of a line of output
6543 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6544 code must advance @var{PTR} to the beginning of the filename on that
6545 line. Otherwise, it must set @var{PTR} to @code{NULL}.
6549 @node Instruction Output
6550 @subsection Output of Assembler Instructions
6552 @c prevent bad page break with this line
6553 This describes assembler instruction output.
6556 @findex REGISTER_NAMES
6557 @item REGISTER_NAMES
6558 A C initializer containing the assembler's names for the machine
6559 registers, each one as a C string constant. This is what translates
6560 register numbers in the compiler into assembler language.
6562 @findex ADDITIONAL_REGISTER_NAMES
6563 @item ADDITIONAL_REGISTER_NAMES
6564 If defined, a C initializer for an array of structures containing a name
6565 and a register number. This macro defines additional names for hard
6566 registers, thus allowing the @code{asm} option in declarations to refer
6567 to registers using alternate names.
6569 @findex ASM_OUTPUT_OPCODE
6570 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6571 Define this macro if you are using an unusual assembler that
6572 requires different names for the machine instructions.
6574 The definition is a C statement or statements which output an
6575 assembler instruction opcode to the stdio stream @var{stream}. The
6576 macro-operand @var{ptr} is a variable of type @code{char *} which
6577 points to the opcode name in its ``internal'' form---the form that is
6578 written in the machine description. The definition should output the
6579 opcode name to @var{stream}, performing any translation you desire, and
6580 increment the variable @var{ptr} to point at the end of the opcode
6581 so that it will not be output twice.
6583 In fact, your macro definition may process less than the entire opcode
6584 name, or more than the opcode name; but if you want to process text
6585 that includes @samp{%}-sequences to substitute operands, you must take
6586 care of the substitution yourself. Just be sure to increment
6587 @var{ptr} over whatever text should not be output normally.
6589 @findex recog_operand
6590 If you need to look at the operand values, they can be found as the
6591 elements of @code{recog_operand}.
6593 If the macro definition does nothing, the instruction is output
6596 @findex FINAL_PRESCAN_INSN
6597 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6598 If defined, a C statement to be executed just prior to the output of
6599 assembler code for @var{insn}, to modify the extracted operands so
6600 they will be output differently.
6602 Here the argument @var{opvec} is the vector containing the operands
6603 extracted from @var{insn}, and @var{noperands} is the number of
6604 elements of the vector which contain meaningful data for this insn.
6605 The contents of this vector are what will be used to convert the insn
6606 template into assembler code, so you can change the assembler output
6607 by changing the contents of the vector.
6609 This macro is useful when various assembler syntaxes share a single
6610 file of instruction patterns; by defining this macro differently, you
6611 can cause a large class of instructions to be output differently (such
6612 as with rearranged operands). Naturally, variations in assembler
6613 syntax affecting individual insn patterns ought to be handled by
6614 writing conditional output routines in those patterns.
6616 If this macro is not defined, it is equivalent to a null statement.
6618 @findex FINAL_PRESCAN_LABEL
6619 @item FINAL_PRESCAN_LABEL
6620 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
6621 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
6622 @var{noperands} will be zero.
6624 @findex PRINT_OPERAND
6625 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6626 A C compound statement to output to stdio stream @var{stream} the
6627 assembler syntax for an instruction operand @var{x}. @var{x} is an
6630 @var{code} is a value that can be used to specify one of several ways
6631 of printing the operand. It is used when identical operands must be
6632 printed differently depending on the context. @var{code} comes from
6633 the @samp{%} specification that was used to request printing of the
6634 operand. If the specification was just @samp{%@var{digit}} then
6635 @var{code} is 0; if the specification was @samp{%@var{ltr}
6636 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6639 If @var{x} is a register, this macro should print the register's name.
6640 The names can be found in an array @code{reg_names} whose type is
6641 @code{char *[]}. @code{reg_names} is initialized from
6642 @code{REGISTER_NAMES}.
6644 When the machine description has a specification @samp{%@var{punct}}
6645 (a @samp{%} followed by a punctuation character), this macro is called
6646 with a null pointer for @var{x} and the punctuation character for
6649 @findex PRINT_OPERAND_PUNCT_VALID_P
6650 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6651 A C expression which evaluates to true if @var{code} is a valid
6652 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6653 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6654 punctuation characters (except for the standard one, @samp{%}) are used
6657 @findex PRINT_OPERAND_ADDRESS
6658 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6659 A C compound statement to output to stdio stream @var{stream} the
6660 assembler syntax for an instruction operand that is a memory reference
6661 whose address is @var{x}. @var{x} is an RTL expression.
6663 @cindex @code{ENCODE_SECTION_INFO} usage
6664 On some machines, the syntax for a symbolic address depends on the
6665 section that the address refers to. On these machines, define the macro
6666 @code{ENCODE_SECTION_INFO} to store the information into the
6667 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
6669 @findex DBR_OUTPUT_SEQEND
6670 @findex dbr_sequence_length
6671 @item DBR_OUTPUT_SEQEND(@var{file})
6672 A C statement, to be executed after all slot-filler instructions have
6673 been output. If necessary, call @code{dbr_sequence_length} to
6674 determine the number of slots filled in a sequence (zero if not
6675 currently outputting a sequence), to decide how many no-ops to output,
6678 Don't define this macro if it has nothing to do, but it is helpful in
6679 reading assembly output if the extent of the delay sequence is made
6680 explicit (e.g. with white space).
6682 @findex final_sequence
6683 Note that output routines for instructions with delay slots must be
6684 prepared to deal with not being output as part of a sequence (i.e.
6685 when the scheduling pass is not run, or when no slot fillers could be
6686 found.) The variable @code{final_sequence} is null when not
6687 processing a sequence, otherwise it contains the @code{sequence} rtx
6690 @findex REGISTER_PREFIX
6691 @findex LOCAL_LABEL_PREFIX
6692 @findex USER_LABEL_PREFIX
6693 @findex IMMEDIATE_PREFIX
6695 @item REGISTER_PREFIX
6696 @itemx LOCAL_LABEL_PREFIX
6697 @itemx USER_LABEL_PREFIX
6698 @itemx IMMEDIATE_PREFIX
6699 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6700 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6701 @file{final.c}). These are useful when a single @file{md} file must
6702 support multiple assembler formats. In that case, the various @file{tm.h}
6703 files can define these macros differently.
6705 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
6706 @findex ASM_FPRINTF_EXTENSIONS
6707 If defined this macro should expand to a series of @code{case}
6708 statements which will be parsed inside the @code{switch} statement of
6709 the @code{asm_fprintf} function. This allows targets to define extra
6710 printf formats which may useful when generating their assembler
6711 statements. Note that upper case letters are reserved for future
6712 generic extensions to asm_fprintf, and so are not available to target
6713 specific code. The output file is given by the parameter @var{file}.
6714 The varargs input pointer is @var{argptr} and the rest of the format
6715 string, starting the character after the one that is being switched
6716 upon, is pointed to by @var{format}.
6718 @findex ASSEMBLER_DIALECT
6719 @item ASSEMBLER_DIALECT
6720 If your target supports multiple dialects of assembler language (such as
6721 different opcodes), define this macro as a C expression that gives the
6722 numeric index of the assembler language dialect to use, with zero as the
6725 If this macro is defined, you may use constructs of the form
6726 @samp{@{option0|option1|option2@dots{}@}} in the output
6727 templates of patterns (@pxref{Output Template}) or in the first argument
6728 of @code{asm_fprintf}. This construct outputs @samp{option0},
6729 @samp{option1} or @samp{option2}, etc., if the value of
6730 @code{ASSEMBLER_DIALECT} is zero, one or two, etc. Any special
6731 characters within these strings retain their usual meaning.
6733 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6734 @samp{@}} do not have any special meaning when used in templates or
6735 operands to @code{asm_fprintf}.
6737 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6738 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6739 the variations in assembler language syntax with that mechanism. Define
6740 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6741 if the syntax variant are larger and involve such things as different
6742 opcodes or operand order.
6744 @findex ASM_OUTPUT_REG_PUSH
6745 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6746 A C expression to output to @var{stream} some assembler code
6747 which will push hard register number @var{regno} onto the stack.
6748 The code need not be optimal, since this macro is used only when
6751 @findex ASM_OUTPUT_REG_POP
6752 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6753 A C expression to output to @var{stream} some assembler code
6754 which will pop hard register number @var{regno} off of the stack.
6755 The code need not be optimal, since this macro is used only when
6759 @node Dispatch Tables
6760 @subsection Output of Dispatch Tables
6762 @c prevent bad page break with this line
6763 This concerns dispatch tables.
6766 @cindex dispatch table
6767 @findex ASM_OUTPUT_ADDR_DIFF_ELT
6768 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6769 A C statement to output to the stdio stream @var{stream} an assembler
6770 pseudo-instruction to generate a difference between two labels.
6771 @var{value} and @var{rel} are the numbers of two internal labels. The
6772 definitions of these labels are output using
6773 @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
6774 way here. For example,
6777 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6778 @var{value}, @var{rel})
6781 You must provide this macro on machines where the addresses in a
6782 dispatch table are relative to the table's own address. If defined, GNU
6783 CC will also use this macro on all machines when producing PIC.
6784 @var{body} is the body of the ADDR_DIFF_VEC; it is provided so that the
6785 mode and flags can be read.
6787 @findex ASM_OUTPUT_ADDR_VEC_ELT
6788 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6789 This macro should be provided on machines where the addresses
6790 in a dispatch table are absolute.
6792 The definition should be a C statement to output to the stdio stream
6793 @var{stream} an assembler pseudo-instruction to generate a reference to
6794 a label. @var{value} is the number of an internal label whose
6795 definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
6799 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6802 @findex ASM_OUTPUT_CASE_LABEL
6803 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6804 Define this if the label before a jump-table needs to be output
6805 specially. The first three arguments are the same as for
6806 @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
6807 jump-table which follows (a @code{jump_insn} containing an
6808 @code{addr_vec} or @code{addr_diff_vec}).
6810 This feature is used on system V to output a @code{swbeg} statement
6813 If this macro is not defined, these labels are output with
6814 @code{ASM_OUTPUT_INTERNAL_LABEL}.
6816 @findex ASM_OUTPUT_CASE_END
6817 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6818 Define this if something special must be output at the end of a
6819 jump-table. The definition should be a C statement to be executed
6820 after the assembler code for the table is written. It should write
6821 the appropriate code to stdio stream @var{stream}. The argument
6822 @var{table} is the jump-table insn, and @var{num} is the label-number
6823 of the preceding label.
6825 If this macro is not defined, nothing special is output at the end of
6829 @node Exception Region Output
6830 @subsection Assembler Commands for Exception Regions
6832 @c prevent bad page break with this line
6834 This describes commands marking the start and the end of an exception
6838 @findex ASM_OUTPUT_EH_REGION_BEG
6839 @item ASM_OUTPUT_EH_REGION_BEG ()
6840 A C expression to output text to mark the start of an exception region.
6842 This macro need not be defined on most platforms.
6844 @findex ASM_OUTPUT_EH_REGION_END
6845 @item ASM_OUTPUT_EH_REGION_END ()
6846 A C expression to output text to mark the end of an exception region.
6848 This macro need not be defined on most platforms.
6850 @findex EXCEPTION_SECTION
6851 @item EXCEPTION_SECTION ()
6852 A C expression to switch to the section in which the main
6853 exception table is to be placed (@pxref{Sections}). The default is a
6854 section named @code{.gcc_except_table} on machines that support named
6855 sections via @code{ASM_OUTPUT_SECTION_NAME}, otherwise if @samp{-fpic}
6856 or @samp{-fPIC} is in effect, the @code{data_section}, otherwise the
6857 @code{readonly_data_section}.
6859 @findex EH_FRAME_SECTION_ASM_OP
6860 @item EH_FRAME_SECTION_ASM_OP
6861 If defined, a C string constant, including spacing, for the assembler
6862 operation to switch to the section for exception handling frame unwind
6863 information. If not defined, GCC will provide a default definition if the
6864 target supports named sections. @file{crtstuff.c} uses this macro to
6865 switch to the appropriate section.
6867 You should define this symbol if your target supports DWARF 2 frame
6868 unwind information and the default definition does not work.
6870 @findex OMIT_EH_TABLE
6871 @item OMIT_EH_TABLE ()
6872 A C expression that is nonzero if the normal exception table output
6875 This macro need not be defined on most platforms.
6877 @findex EH_TABLE_LOOKUP
6878 @item EH_TABLE_LOOKUP ()
6879 Alternate runtime support for looking up an exception at runtime and
6880 finding the associated handler, if the default method won't work.
6882 This macro need not be defined on most platforms.
6884 @findex DOESNT_NEED_UNWINDER
6885 @item DOESNT_NEED_UNWINDER
6886 A C expression that decides whether or not the current function needs to
6887 have a function unwinder generated for it. See the file @code{except.c}
6888 for details on when to define this, and how.
6890 @findex MASK_RETURN_ADDR
6891 @item MASK_RETURN_ADDR
6892 An rtx used to mask the return address found via RETURN_ADDR_RTX, so
6893 that it does not contain any extraneous set bits in it.
6895 @findex DWARF2_UNWIND_INFO
6896 @item DWARF2_UNWIND_INFO
6897 Define this macro to 0 if your target supports DWARF 2 frame unwind
6898 information, but it does not yet work with exception handling.
6899 Otherwise, if your target supports this information (if it defines
6900 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
6901 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
6904 If this macro is defined to 1, the DWARF 2 unwinder will be the default
6905 exception handling mechanism; otherwise, setjmp/longjmp will be used by
6908 If this macro is defined to anything, the DWARF 2 unwinder will be used
6909 instead of inline unwinders and __unwind_function in the non-setjmp case.
6911 @findex DWARF_CIE_DATA_ALIGNMENT
6912 @item DWARF_CIE_DATA_ALIGNMENT
6913 This macro need only be defined if the target might save registers in the
6914 function prologue at an offset to the stack pointer that is not aligned to
6915 @code{UNITS_PER_WORD}. The definition should be the negative minimum
6916 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
6917 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
6918 the target supports DWARF 2 frame unwind information.
6922 @node Alignment Output
6923 @subsection Assembler Commands for Alignment
6925 @c prevent bad page break with this line
6926 This describes commands for alignment.
6929 @findex LABEL_ALIGN_AFTER_BARRIER
6930 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
6931 The alignment (log base 2) to put in front of @var{label}, which follows
6934 This macro need not be defined if you don't want any special alignment
6935 to be done at such a time. Most machine descriptions do not currently
6938 Unless it's necessary to inspect the @var{label} parameter, it is better
6939 to set the variable @var{align_jumps} in the target's
6940 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6941 selection in @var{align_jumps} in a @code{LABEL_ALIGN_AFTER_BARRIER}
6944 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6945 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6946 The maximum number of bytes to skip when applying
6947 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
6948 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6951 @item LOOP_ALIGN (@var{label})
6952 The alignment (log base 2) to put in front of @var{label}, which follows
6953 a NOTE_INSN_LOOP_BEG note.
6955 This macro need not be defined if you don't want any special alignment
6956 to be done at such a time. Most machine descriptions do not currently
6959 Unless it's necessary to inspect the @var{label} parameter, it is better
6960 to set the variable @var{align_loops} in the target's
6961 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6962 selection in @var{align_loops} in a @code{LOOP_ALIGN} implementation.
6964 @findex LOOP_ALIGN_MAX_SKIP
6965 @item LOOP_ALIGN_MAX_SKIP
6966 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
6967 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6970 @item LABEL_ALIGN (@var{label})
6971 The alignment (log base 2) to put in front of @var{label}.
6972 If LABEL_ALIGN_AFTER_BARRIER / LOOP_ALIGN specify a different alignment,
6973 the maximum of the specified values is used.
6975 Unless it's necessary to inspect the @var{label} parameter, it is better
6976 to set the variable @var{align_labels} in the target's
6977 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6978 selection in @var{align_labels} in a @code{LABEL_ALIGN} implementation.
6980 @findex LABEL_ALIGN_MAX_SKIP
6981 @item LABEL_ALIGN_MAX_SKIP
6982 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
6983 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6985 @findex ASM_OUTPUT_SKIP
6986 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6987 A C statement to output to the stdio stream @var{stream} an assembler
6988 instruction to advance the location counter by @var{nbytes} bytes.
6989 Those bytes should be zero when loaded. @var{nbytes} will be a C
6990 expression of type @code{int}.
6992 @findex ASM_NO_SKIP_IN_TEXT
6993 @item ASM_NO_SKIP_IN_TEXT
6994 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6995 text section because it fails to put zeros in the bytes that are skipped.
6996 This is true on many Unix systems, where the pseudo--op to skip bytes
6997 produces no-op instructions rather than zeros when used in the text
7000 @findex ASM_OUTPUT_ALIGN
7001 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7002 A C statement to output to the stdio stream @var{stream} an assembler
7003 command to advance the location counter to a multiple of 2 to the
7004 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7006 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
7007 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7008 A C statement to output to the stdio stream @var{stream} an assembler
7009 command to advance the location counter to a multiple of 2 to the
7010 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7011 satisfy the alignment request. @var{power} and @var{max_skip} will be
7012 a C expression of type @code{int}.
7016 @node Debugging Info
7017 @section Controlling Debugging Information Format
7019 @c prevent bad page break with this line
7020 This describes how to specify debugging information.
7023 * All Debuggers:: Macros that affect all debugging formats uniformly.
7024 * DBX Options:: Macros enabling specific options in DBX format.
7025 * DBX Hooks:: Hook macros for varying DBX format.
7026 * File Names and DBX:: Macros controlling output of file names in DBX format.
7027 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7031 @subsection Macros Affecting All Debugging Formats
7033 @c prevent bad page break with this line
7034 These macros affect all debugging formats.
7037 @findex DBX_REGISTER_NUMBER
7038 @item DBX_REGISTER_NUMBER (@var{regno})
7039 A C expression that returns the DBX register number for the compiler
7040 register number @var{regno}. In simple cases, the value of this
7041 expression may be @var{regno} itself. But sometimes there are some
7042 registers that the compiler knows about and DBX does not, or vice
7043 versa. In such cases, some register may need to have one number in
7044 the compiler and another for DBX.
7046 If two registers have consecutive numbers inside GCC, and they can be
7047 used as a pair to hold a multiword value, then they @emph{must} have
7048 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7049 Otherwise, debuggers will be unable to access such a pair, because they
7050 expect register pairs to be consecutive in their own numbering scheme.
7052 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7053 does not preserve register pairs, then what you must do instead is
7054 redefine the actual register numbering scheme.
7056 @findex DEBUGGER_AUTO_OFFSET
7057 @item DEBUGGER_AUTO_OFFSET (@var{x})
7058 A C expression that returns the integer offset value for an automatic
7059 variable having address @var{x} (an RTL expression). The default
7060 computation assumes that @var{x} is based on the frame-pointer and
7061 gives the offset from the frame-pointer. This is required for targets
7062 that produce debugging output for DBX or COFF-style debugging output
7063 for SDB and allow the frame-pointer to be eliminated when the
7064 @samp{-g} options is used.
7066 @findex DEBUGGER_ARG_OFFSET
7067 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7068 A C expression that returns the integer offset value for an argument
7069 having address @var{x} (an RTL expression). The nominal offset is
7072 @findex PREFERRED_DEBUGGING_TYPE
7073 @item PREFERRED_DEBUGGING_TYPE
7074 A C expression that returns the type of debugging output GCC should
7075 produce when the user specifies just @samp{-g}. Define
7076 this if you have arranged for GCC to support more than one format of
7077 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7078 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG}, and
7081 When the user specifies @samp{-ggdb}, GCC normally also uses the
7082 value of this macro to select the debugging output format, but with two
7083 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7084 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7085 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7086 defined, GCC uses @code{DBX_DEBUG}.
7088 The value of this macro only affects the default debugging output; the
7089 user can always get a specific type of output by using @samp{-gstabs},
7090 @samp{-gcoff}, @samp{-gdwarf-1}, @samp{-gdwarf-2}, or @samp{-gxcoff}.
7094 @subsection Specific Options for DBX Output
7096 @c prevent bad page break with this line
7097 These are specific options for DBX output.
7100 @findex DBX_DEBUGGING_INFO
7101 @item DBX_DEBUGGING_INFO
7102 Define this macro if GCC should produce debugging output for DBX
7103 in response to the @samp{-g} option.
7105 @findex XCOFF_DEBUGGING_INFO
7106 @item XCOFF_DEBUGGING_INFO
7107 Define this macro if GCC should produce XCOFF format debugging output
7108 in response to the @samp{-g} option. This is a variant of DBX format.
7110 @findex DEFAULT_GDB_EXTENSIONS
7111 @item DEFAULT_GDB_EXTENSIONS
7112 Define this macro to control whether GCC should by default generate
7113 GDB's extended version of DBX debugging information (assuming DBX-format
7114 debugging information is enabled at all). If you don't define the
7115 macro, the default is 1: always generate the extended information
7116 if there is any occasion to.
7118 @findex DEBUG_SYMS_TEXT
7119 @item DEBUG_SYMS_TEXT
7120 Define this macro if all @code{.stabs} commands should be output while
7121 in the text section.
7123 @findex ASM_STABS_OP
7125 A C string constant, including spacing, naming the assembler pseudo op to
7126 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7127 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7128 applies only to DBX debugging information format.
7130 @findex ASM_STABD_OP
7132 A C string constant, including spacing, naming the assembler pseudo op to
7133 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7134 value is the current location. If you don't define this macro,
7135 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7138 @findex ASM_STABN_OP
7140 A C string constant, including spacing, naming the assembler pseudo op to
7141 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7142 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7143 macro applies only to DBX debugging information format.
7145 @findex DBX_NO_XREFS
7147 Define this macro if DBX on your system does not support the construct
7148 @samp{xs@var{tagname}}. On some systems, this construct is used to
7149 describe a forward reference to a structure named @var{tagname}.
7150 On other systems, this construct is not supported at all.
7152 @findex DBX_CONTIN_LENGTH
7153 @item DBX_CONTIN_LENGTH
7154 A symbol name in DBX-format debugging information is normally
7155 continued (split into two separate @code{.stabs} directives) when it
7156 exceeds a certain length (by default, 80 characters). On some
7157 operating systems, DBX requires this splitting; on others, splitting
7158 must not be done. You can inhibit splitting by defining this macro
7159 with the value zero. You can override the default splitting-length by
7160 defining this macro as an expression for the length you desire.
7162 @findex DBX_CONTIN_CHAR
7163 @item DBX_CONTIN_CHAR
7164 Normally continuation is indicated by adding a @samp{\} character to
7165 the end of a @code{.stabs} string when a continuation follows. To use
7166 a different character instead, define this macro as a character
7167 constant for the character you want to use. Do not define this macro
7168 if backslash is correct for your system.
7170 @findex DBX_STATIC_STAB_DATA_SECTION
7171 @item DBX_STATIC_STAB_DATA_SECTION
7172 Define this macro if it is necessary to go to the data section before
7173 outputting the @samp{.stabs} pseudo-op for a non-global static
7176 @findex DBX_TYPE_DECL_STABS_CODE
7177 @item DBX_TYPE_DECL_STABS_CODE
7178 The value to use in the ``code'' field of the @code{.stabs} directive
7179 for a typedef. The default is @code{N_LSYM}.
7181 @findex DBX_STATIC_CONST_VAR_CODE
7182 @item DBX_STATIC_CONST_VAR_CODE
7183 The value to use in the ``code'' field of the @code{.stabs} directive
7184 for a static variable located in the text section. DBX format does not
7185 provide any ``right'' way to do this. The default is @code{N_FUN}.
7187 @findex DBX_REGPARM_STABS_CODE
7188 @item DBX_REGPARM_STABS_CODE
7189 The value to use in the ``code'' field of the @code{.stabs} directive
7190 for a parameter passed in registers. DBX format does not provide any
7191 ``right'' way to do this. The default is @code{N_RSYM}.
7193 @findex DBX_REGPARM_STABS_LETTER
7194 @item DBX_REGPARM_STABS_LETTER
7195 The letter to use in DBX symbol data to identify a symbol as a parameter
7196 passed in registers. DBX format does not customarily provide any way to
7197 do this. The default is @code{'P'}.
7199 @findex DBX_MEMPARM_STABS_LETTER
7200 @item DBX_MEMPARM_STABS_LETTER
7201 The letter to use in DBX symbol data to identify a symbol as a stack
7202 parameter. The default is @code{'p'}.
7204 @findex DBX_FUNCTION_FIRST
7205 @item DBX_FUNCTION_FIRST
7206 Define this macro if the DBX information for a function and its
7207 arguments should precede the assembler code for the function. Normally,
7208 in DBX format, the debugging information entirely follows the assembler
7211 @findex DBX_LBRAC_FIRST
7212 @item DBX_LBRAC_FIRST
7213 Define this macro if the @code{N_LBRAC} symbol for a block should
7214 precede the debugging information for variables and functions defined in
7215 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
7218 @findex DBX_BLOCKS_FUNCTION_RELATIVE
7219 @item DBX_BLOCKS_FUNCTION_RELATIVE
7220 Define this macro if the value of a symbol describing the scope of a
7221 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7222 of the enclosing function. Normally, GNU C uses an absolute address.
7224 @findex DBX_USE_BINCL
7226 Define this macro if GNU C should generate @code{N_BINCL} and
7227 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7228 macro also directs GNU C to output a type number as a pair of a file
7229 number and a type number within the file. Normally, GNU C does not
7230 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7231 number for a type number.
7235 @subsection Open-Ended Hooks for DBX Format
7237 @c prevent bad page break with this line
7238 These are hooks for DBX format.
7241 @findex DBX_OUTPUT_LBRAC
7242 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7243 Define this macro to say how to output to @var{stream} the debugging
7244 information for the start of a scope level for variable names. The
7245 argument @var{name} is the name of an assembler symbol (for use with
7246 @code{assemble_name}) whose value is the address where the scope begins.
7248 @findex DBX_OUTPUT_RBRAC
7249 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7250 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7252 @findex DBX_OUTPUT_ENUM
7253 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
7254 Define this macro if the target machine requires special handling to
7255 output an enumeration type. The definition should be a C statement
7256 (sans semicolon) to output the appropriate information to @var{stream}
7257 for the type @var{type}.
7259 @findex DBX_OUTPUT_FUNCTION_END
7260 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7261 Define this macro if the target machine requires special output at the
7262 end of the debugging information for a function. The definition should
7263 be a C statement (sans semicolon) to output the appropriate information
7264 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7267 @findex DBX_OUTPUT_STANDARD_TYPES
7268 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7269 Define this macro if you need to control the order of output of the
7270 standard data types at the beginning of compilation. The argument
7271 @var{syms} is a @code{tree} which is a chain of all the predefined
7272 global symbols, including names of data types.
7274 Normally, DBX output starts with definitions of the types for integers
7275 and characters, followed by all the other predefined types of the
7276 particular language in no particular order.
7278 On some machines, it is necessary to output different particular types
7279 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
7280 those symbols in the necessary order. Any predefined types that you
7281 don't explicitly output will be output afterward in no particular order.
7283 Be careful not to define this macro so that it works only for C. There
7284 are no global variables to access most of the built-in types, because
7285 another language may have another set of types. The way to output a
7286 particular type is to look through @var{syms} to see if you can find it.
7292 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7293 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
7295 dbxout_symbol (decl);
7301 This does nothing if the expected type does not exist.
7303 See the function @code{init_decl_processing} in @file{c-decl.c} to find
7304 the names to use for all the built-in C types.
7306 Here is another way of finding a particular type:
7308 @c this is still overfull. --mew 10feb93
7312 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7313 if (TREE_CODE (decl) == TYPE_DECL
7314 && (TREE_CODE (TREE_TYPE (decl))
7316 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
7317 && TYPE_UNSIGNED (TREE_TYPE (decl)))
7319 /* @r{This must be @code{unsigned short}.} */
7320 dbxout_symbol (decl);
7326 @findex NO_DBX_FUNCTION_END
7327 @item NO_DBX_FUNCTION_END
7328 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7329 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extention construct.
7330 On those machines, define this macro to turn this feature off without
7331 disturbing the rest of the gdb extensions.
7335 @node File Names and DBX
7336 @subsection File Names in DBX Format
7338 @c prevent bad page break with this line
7339 This describes file names in DBX format.
7342 @findex DBX_WORKING_DIRECTORY
7343 @item DBX_WORKING_DIRECTORY
7344 Define this if DBX wants to have the current directory recorded in each
7347 Note that the working directory is always recorded if GDB extensions are
7350 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
7351 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7352 A C statement to output DBX debugging information to the stdio stream
7353 @var{stream} which indicates that file @var{name} is the main source
7354 file---the file specified as the input file for compilation.
7355 This macro is called only once, at the beginning of compilation.
7357 This macro need not be defined if the standard form of output
7358 for DBX debugging information is appropriate.
7360 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
7361 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
7362 A C statement to output DBX debugging information to the stdio stream
7363 @var{stream} which indicates that the current directory during
7364 compilation is named @var{name}.
7366 This macro need not be defined if the standard form of output
7367 for DBX debugging information is appropriate.
7369 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
7370 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7371 A C statement to output DBX debugging information at the end of
7372 compilation of the main source file @var{name}.
7374 If you don't define this macro, nothing special is output at the end
7375 of compilation, which is correct for most machines.
7377 @findex DBX_OUTPUT_SOURCE_FILENAME
7378 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7379 A C statement to output DBX debugging information to the stdio stream
7380 @var{stream} which indicates that file @var{name} is the current source
7381 file. This output is generated each time input shifts to a different
7382 source file as a result of @samp{#include}, the end of an included file,
7383 or a @samp{#line} command.
7385 This macro need not be defined if the standard form of output
7386 for DBX debugging information is appropriate.
7391 @subsection Macros for SDB and DWARF Output
7393 @c prevent bad page break with this line
7394 Here are macros for SDB and DWARF output.
7397 @findex SDB_DEBUGGING_INFO
7398 @item SDB_DEBUGGING_INFO
7399 Define this macro if GCC should produce COFF-style debugging output
7400 for SDB in response to the @samp{-g} option.
7402 @findex DWARF_DEBUGGING_INFO
7403 @item DWARF_DEBUGGING_INFO
7404 Define this macro if GCC should produce dwarf format debugging output
7405 in response to the @samp{-g} option.
7407 @findex DWARF2_DEBUGGING_INFO
7408 @item DWARF2_DEBUGGING_INFO
7409 Define this macro if GCC should produce dwarf version 2 format
7410 debugging output in response to the @samp{-g} option.
7412 To support optional call frame debugging information, you must also
7413 define @code{INCOMING_RETURN_ADDR_RTX} and either set
7414 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
7415 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
7416 as appropriate from @code{FUNCTION_PROLOGUE} if you don't.
7418 @findex DWARF2_FRAME_INFO
7419 @item DWARF2_FRAME_INFO
7420 Define this macro to a nonzero value if GCC should always output
7421 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
7422 (@pxref{Exception Region Output} is nonzero, GCC will output this
7423 information not matter how you define @code{DWARF2_FRAME_INFO}.
7425 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
7426 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
7427 Define this macro if the linker does not work with Dwarf version 2.
7428 Normally, if the user specifies only @samp{-ggdb} GCC will use Dwarf
7429 version 2 if available; this macro disables this. See the description
7430 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
7432 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
7433 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
7434 By default, the Dwarf 2 debugging information generator will generate a
7435 label to mark the beginning of the text section. If it is better simply
7436 to use the name of the text section itself, rather than an explicit label,
7437 to indicate the beginning of the text section, define this macro to zero.
7439 @findex DWARF2_ASM_LINE_DEBUG_INFO
7440 @item DWARF2_ASM_LINE_DEBUG_INFO
7441 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
7442 line debug info sections. This will result in much more compact line number
7443 tables, and hence is desirable if it works.
7445 @findex PUT_SDB_@dots{}
7446 @item PUT_SDB_@dots{}
7447 Define these macros to override the assembler syntax for the special
7448 SDB assembler directives. See @file{sdbout.c} for a list of these
7449 macros and their arguments. If the standard syntax is used, you need
7450 not define them yourself.
7454 Some assemblers do not support a semicolon as a delimiter, even between
7455 SDB assembler directives. In that case, define this macro to be the
7456 delimiter to use (usually @samp{\n}). It is not necessary to define
7457 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
7460 @findex SDB_GENERATE_FAKE
7461 @item SDB_GENERATE_FAKE
7462 Define this macro to override the usual method of constructing a dummy
7463 name for anonymous structure and union types. See @file{sdbout.c} for
7466 @findex SDB_ALLOW_UNKNOWN_REFERENCES
7467 @item SDB_ALLOW_UNKNOWN_REFERENCES
7468 Define this macro to allow references to unknown structure,
7469 union, or enumeration tags to be emitted. Standard COFF does not
7470 allow handling of unknown references, MIPS ECOFF has support for
7473 @findex SDB_ALLOW_FORWARD_REFERENCES
7474 @item SDB_ALLOW_FORWARD_REFERENCES
7475 Define this macro to allow references to structure, union, or
7476 enumeration tags that have not yet been seen to be handled. Some
7477 assemblers choke if forward tags are used, while some require it.
7480 @node Cross-compilation
7481 @section Cross Compilation and Floating Point
7482 @cindex cross compilation and floating point
7483 @cindex floating point and cross compilation
7485 While all modern machines use 2's complement representation for integers,
7486 there are a variety of representations for floating point numbers. This
7487 means that in a cross-compiler the representation of floating point numbers
7488 in the compiled program may be different from that used in the machine
7489 doing the compilation.
7492 Because different representation systems may offer different amounts of
7493 range and precision, the cross compiler cannot safely use the host
7494 machine's floating point arithmetic. Therefore, floating point constants
7495 must be represented in the target machine's format. This means that the
7496 cross compiler cannot use @code{atof} to parse a floating point constant;
7497 it must have its own special routine to use instead. Also, constant
7498 folding must emulate the target machine's arithmetic (or must not be done
7501 The macros in the following table should be defined only if you are cross
7502 compiling between different floating point formats.
7504 Otherwise, don't define them. Then default definitions will be set up which
7505 use @code{double} as the data type, @code{==} to test for equality, etc.
7507 You don't need to worry about how many times you use an operand of any
7508 of these macros. The compiler never uses operands which have side effects.
7511 @findex REAL_VALUE_TYPE
7512 @item REAL_VALUE_TYPE
7513 A macro for the C data type to be used to hold a floating point value
7514 in the target machine's format. Typically this would be a
7515 @code{struct} containing an array of @code{int}.
7517 @findex REAL_VALUES_EQUAL
7518 @item REAL_VALUES_EQUAL (@var{x}, @var{y})
7519 A macro for a C expression which compares for equality the two values,
7520 @var{x} and @var{y}, both of type @code{REAL_VALUE_TYPE}.
7522 @findex REAL_VALUES_LESS
7523 @item REAL_VALUES_LESS (@var{x}, @var{y})
7524 A macro for a C expression which tests whether @var{x} is less than
7525 @var{y}, both values being of type @code{REAL_VALUE_TYPE} and
7526 interpreted as floating point numbers in the target machine's
7529 @findex REAL_VALUE_LDEXP
7531 @item REAL_VALUE_LDEXP (@var{x}, @var{scale})
7532 A macro for a C expression which performs the standard library
7533 function @code{ldexp}, but using the target machine's floating point
7534 representation. Both @var{x} and the value of the expression have
7535 type @code{REAL_VALUE_TYPE}. The second argument, @var{scale}, is an
7538 @findex REAL_VALUE_FIX
7539 @item REAL_VALUE_FIX (@var{x})
7540 A macro whose definition is a C expression to convert the target-machine
7541 floating point value @var{x} to a signed integer. @var{x} has type
7542 @code{REAL_VALUE_TYPE}.
7544 @findex REAL_VALUE_UNSIGNED_FIX
7545 @item REAL_VALUE_UNSIGNED_FIX (@var{x})
7546 A macro whose definition is a C expression to convert the target-machine
7547 floating point value @var{x} to an unsigned integer. @var{x} has type
7548 @code{REAL_VALUE_TYPE}.
7550 @findex REAL_VALUE_RNDZINT
7551 @item REAL_VALUE_RNDZINT (@var{x})
7552 A macro whose definition is a C expression to round the target-machine
7553 floating point value @var{x} towards zero to an integer value (but still
7554 as a floating point number). @var{x} has type @code{REAL_VALUE_TYPE},
7555 and so does the value.
7557 @findex REAL_VALUE_UNSIGNED_RNDZINT
7558 @item REAL_VALUE_UNSIGNED_RNDZINT (@var{x})
7559 A macro whose definition is a C expression to round the target-machine
7560 floating point value @var{x} towards zero to an unsigned integer value
7561 (but still represented as a floating point number). @var{x} has type
7562 @code{REAL_VALUE_TYPE}, and so does the value.
7564 @findex REAL_VALUE_ATOF
7565 @item REAL_VALUE_ATOF (@var{string}, @var{mode})
7566 A macro for a C expression which converts @var{string}, an expression of
7567 type @code{char *}, into a floating point number in the target machine's
7568 representation for mode @var{mode}. The value has type
7569 @code{REAL_VALUE_TYPE}.
7571 @findex REAL_INFINITY
7573 Define this macro if infinity is a possible floating point value, and
7574 therefore division by 0 is legitimate.
7576 @findex REAL_VALUE_ISINF
7578 @item REAL_VALUE_ISINF (@var{x})
7579 A macro for a C expression which determines whether @var{x}, a floating
7580 point value, is infinity. The value has type @code{int}.
7581 By default, this is defined to call @code{isinf}.
7583 @findex REAL_VALUE_ISNAN
7585 @item REAL_VALUE_ISNAN (@var{x})
7586 A macro for a C expression which determines whether @var{x}, a floating
7587 point value, is a ``nan'' (not-a-number). The value has type
7588 @code{int}. By default, this is defined to call @code{isnan}.
7591 @cindex constant folding and floating point
7592 Define the following additional macros if you want to make floating
7593 point constant folding work while cross compiling. If you don't
7594 define them, cross compilation is still possible, but constant folding
7595 will not happen for floating point values.
7598 @findex REAL_ARITHMETIC
7599 @item REAL_ARITHMETIC (@var{output}, @var{code}, @var{x}, @var{y})
7600 A macro for a C statement which calculates an arithmetic operation of
7601 the two floating point values @var{x} and @var{y}, both of type
7602 @code{REAL_VALUE_TYPE} in the target machine's representation, to
7603 produce a result of the same type and representation which is stored
7604 in @var{output} (which will be a variable).
7606 The operation to be performed is specified by @var{code}, a tree code
7607 which will always be one of the following: @code{PLUS_EXPR},
7608 @code{MINUS_EXPR}, @code{MULT_EXPR}, @code{RDIV_EXPR},
7609 @code{MAX_EXPR}, @code{MIN_EXPR}.@refill
7611 @cindex overflow while constant folding
7612 The expansion of this macro is responsible for checking for overflow.
7613 If overflow happens, the macro expansion should execute the statement
7614 @code{return 0;}, which indicates the inability to perform the
7615 arithmetic operation requested.
7617 @findex REAL_VALUE_NEGATE
7618 @item REAL_VALUE_NEGATE (@var{x})
7619 A macro for a C expression which returns the negative of the floating
7620 point value @var{x}. Both @var{x} and the value of the expression
7621 have type @code{REAL_VALUE_TYPE} and are in the target machine's
7622 floating point representation.
7624 There is no way for this macro to report overflow, since overflow
7625 can't happen in the negation operation.
7627 @findex REAL_VALUE_TRUNCATE
7628 @item REAL_VALUE_TRUNCATE (@var{mode}, @var{x})
7629 A macro for a C expression which converts the floating point value
7630 @var{x} to mode @var{mode}.
7632 Both @var{x} and the value of the expression are in the target machine's
7633 floating point representation and have type @code{REAL_VALUE_TYPE}.
7634 However, the value should have an appropriate bit pattern to be output
7635 properly as a floating constant whose precision accords with mode
7638 There is no way for this macro to report overflow.
7640 @findex REAL_VALUE_TO_INT
7641 @item REAL_VALUE_TO_INT (@var{low}, @var{high}, @var{x})
7642 A macro for a C expression which converts a floating point value
7643 @var{x} into a double-precision integer which is then stored into
7644 @var{low} and @var{high}, two variables of type @var{int}.
7646 @item REAL_VALUE_FROM_INT (@var{x}, @var{low}, @var{high}, @var{mode})
7647 @findex REAL_VALUE_FROM_INT
7648 A macro for a C expression which converts a double-precision integer
7649 found in @var{low} and @var{high}, two variables of type @var{int},
7650 into a floating point value which is then stored into @var{x}.
7651 The value is in the target machine's representation for mode @var{mode}
7652 and has the type @code{REAL_VALUE_TYPE}.
7655 @node Mode Switching
7656 @section Mode Switching Instructions
7657 @cindex mode switching
7658 The following macros control mode switching optimizations:
7661 @findex OPTIMIZE_MODE_SWITCHING
7662 @item OPTIMIZE_MODE_SWITCHING (@var{entity})
7663 Define this macro if the port needs extra instructions inserted for mode
7664 switching in an optimizing compilation.
7666 For an example, the SH4 can perform both single and double precision
7667 floating point operations, but to perform a single precision operation,
7668 the FPSCR PR bit has to be cleared, while for a double precision
7669 operation, this bit has to be set. Changing the PR bit requires a general
7670 purpose register as a scratch register, hence these FPSCR sets have to
7671 be inserted before reload, i.e. you can't put this into instruction emitting
7672 or MACHINE_DEPENDENT_REORG.
7674 You can have multiple entities that are mode-switched, and select at run time
7675 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
7676 return non-zero for any @var{entity} that that needs mode-switching.
7677 If you define this macro, you also have to define
7678 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
7679 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
7680 @code{NORMAL_MODE} is optional.
7682 @findex NUM_MODES_FOR_MODE_SWITCHING
7683 @item NUM_MODES_FOR_MODE_SWITCHING
7684 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
7685 initializer for an array of integers. Each initializer element
7686 N refers to an entity that needs mode switching, and specifies the number
7687 of different modes that might need to be set for this entity.
7688 The position of the initializer in the initializer - starting counting at
7689 zero - determines the integer that is used to refer to the mode-switched
7691 In macros that take mode arguments / yield a mode result, modes are
7692 represented as numbers 0 .. N - 1. N is used to specify that no mode
7693 switch is needed / supplied.
7696 @item MODE_NEEDED (@var{entity}, @var{insn})
7697 @var{entity} is an integer specifying a mode-switched entity. If
7698 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
7699 return an integer value not larger than the corresponding element in
7700 NUM_MODES_FOR_MODE_SWITCHING, to denote the mode that @var{entity} must
7701 be switched into prior to the execution of INSN.
7704 @item NORMAL_MODE (@var{entity})
7705 If this macro is defined, it is evaluated for every @var{entity} that needs
7706 mode switching. It should evaluate to an integer, which is a mode that
7707 @var{entity} is assumed to be switched to at function entry and exit.
7709 @findex MODE_PRIORITY_TO_MODE
7710 @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
7711 This macro specifies the order in which modes for ENTITY are processed.
7712 0 is the highest priority, NUM_MODES_FOR_MODE_SWITCHING[ENTITY] - 1 the
7713 lowest. The value of the macro should be an integer designating a mode
7714 for ENTITY. For any fixed @var{entity}, @code{mode_priority_to_mode}
7715 (@var{entity}, @var{n}) shall be a bijection in 0 ..
7716 @code{num_modes_for_mode_switching}[@var{entity}] - 1 .
7718 @findex EMIT_MODE_SET
7719 @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
7720 Generate one or more insns to set @var{entity} to @var{mode}.
7721 @var{hard_reg_live} is the set of hard registers live at the point where
7722 the insn(s) are to be inserted.
7726 @section Miscellaneous Parameters
7727 @cindex parameters, miscellaneous
7729 @c prevent bad page break with this line
7730 Here are several miscellaneous parameters.
7733 @item PREDICATE_CODES
7734 @findex PREDICATE_CODES
7735 Define this if you have defined special-purpose predicates in the file
7736 @file{@var{machine}.c}. This macro is called within an initializer of an
7737 array of structures. The first field in the structure is the name of a
7738 predicate and the second field is an array of rtl codes. For each
7739 predicate, list all rtl codes that can be in expressions matched by the
7740 predicate. The list should have a trailing comma. Here is an example
7741 of two entries in the list for a typical RISC machine:
7744 #define PREDICATE_CODES \
7745 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
7746 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
7749 Defining this macro does not affect the generated code (however,
7750 incorrect definitions that omit an rtl code that may be matched by the
7751 predicate can cause the compiler to malfunction). Instead, it allows
7752 the table built by @file{genrecog} to be more compact and efficient,
7753 thus speeding up the compiler. The most important predicates to include
7754 in the list specified by this macro are those used in the most insn
7757 For each predicate function named in @var{PREDICATE_CODES}, a
7758 declaration will be generated in @file{insn-codes.h}.
7760 @item SPECIAL_MODE_PREDICATES
7761 @findex SPECIAL_MODE_PREDICATES
7762 Define this if you have special predicates that know special things
7763 about modes. Genrecog will warn about certain forms of
7764 @code{match_operand} without a mode; if the operand predicate is
7765 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
7768 Here is an example from the IA-32 port (@code{ext_register_operand}
7769 specially checks for @code{HImode} or @code{SImode} in preparation
7770 for a byte extraction from @code{%ah} etc.).
7773 #define SPECIAL_MODE_PREDICATES \
7774 "ext_register_operand",
7777 @findex CASE_VECTOR_MODE
7778 @item CASE_VECTOR_MODE
7779 An alias for a machine mode name. This is the machine mode that
7780 elements of a jump-table should have.
7782 @findex CASE_VECTOR_SHORTEN_MODE
7783 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7784 Optional: return the preferred mode for an @code{addr_diff_vec}
7785 when the minimum and maximum offset are known. If you define this,
7786 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7787 To make this work, you also have to define INSN_ALIGN and
7788 make the alignment for @code{addr_diff_vec} explicit.
7789 The @var{body} argument is provided so that the offset_unsigned and scale
7790 flags can be updated.
7792 @findex CASE_VECTOR_PC_RELATIVE
7793 @item CASE_VECTOR_PC_RELATIVE
7794 Define this macro to be a C expression to indicate when jump-tables
7795 should contain relative addresses. If jump-tables never contain
7796 relative addresses, then you need not define this macro.
7798 @findex CASE_DROPS_THROUGH
7799 @item CASE_DROPS_THROUGH
7800 Define this if control falls through a @code{case} insn when the index
7801 value is out of range. This means the specified default-label is
7802 actually ignored by the @code{case} insn proper.
7804 @findex CASE_VALUES_THRESHOLD
7805 @item CASE_VALUES_THRESHOLD
7806 Define this to be the smallest number of different values for which it
7807 is best to use a jump-table instead of a tree of conditional branches.
7808 The default is four for machines with a @code{casesi} instruction and
7809 five otherwise. This is best for most machines.
7811 @findex WORD_REGISTER_OPERATIONS
7812 @item WORD_REGISTER_OPERATIONS
7813 Define this macro if operations between registers with integral mode
7814 smaller than a word are always performed on the entire register.
7815 Most RISC machines have this property and most CISC machines do not.
7817 @findex LOAD_EXTEND_OP
7818 @item LOAD_EXTEND_OP (@var{mode})
7819 Define this macro to be a C expression indicating when insns that read
7820 memory in @var{mode}, an integral mode narrower than a word, set the
7821 bits outside of @var{mode} to be either the sign-extension or the
7822 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7823 of @var{mode} for which the
7824 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7825 @code{NIL} for other modes.
7827 This macro is not called with @var{mode} non-integral or with a width
7828 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7829 value in this case. Do not define this macro if it would always return
7830 @code{NIL}. On machines where this macro is defined, you will normally
7831 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7833 @findex SHORT_IMMEDIATES_SIGN_EXTEND
7834 @item SHORT_IMMEDIATES_SIGN_EXTEND
7835 Define this macro if loading short immediate values into registers sign
7838 @findex IMPLICIT_FIX_EXPR
7839 @item IMPLICIT_FIX_EXPR
7840 An alias for a tree code that should be used by default for conversion
7841 of floating point values to fixed point. Normally,
7842 @code{FIX_ROUND_EXPR} is used.@refill
7844 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
7845 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
7846 Define this macro if the same instructions that convert a floating
7847 point number to a signed fixed point number also convert validly to an
7850 @findex EASY_DIV_EXPR
7852 An alias for a tree code that is the easiest kind of division to
7853 compile code for in the general case. It may be
7854 @code{TRUNC_DIV_EXPR}, @code{FLOOR_DIV_EXPR}, @code{CEIL_DIV_EXPR} or
7855 @code{ROUND_DIV_EXPR}. These four division operators differ in how
7856 they round the result to an integer. @code{EASY_DIV_EXPR} is used
7857 when it is permissible to use any of those kinds of division and the
7858 choice should be made on the basis of efficiency.@refill
7862 The maximum number of bytes that a single instruction can move quickly
7863 between memory and registers or between two memory locations.
7865 @findex MAX_MOVE_MAX
7867 The maximum number of bytes that a single instruction can move quickly
7868 between memory and registers or between two memory locations. If this
7869 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7870 constant value that is the largest value that @code{MOVE_MAX} can have
7873 @findex SHIFT_COUNT_TRUNCATED
7874 @item SHIFT_COUNT_TRUNCATED
7875 A C expression that is nonzero if on this machine the number of bits
7876 actually used for the count of a shift operation is equal to the number
7877 of bits needed to represent the size of the object being shifted. When
7878 this macro is non-zero, the compiler will assume that it is safe to omit
7879 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7880 truncates the count of a shift operation. On machines that have
7881 instructions that act on bitfields at variable positions, which may
7882 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7883 also enables deletion of truncations of the values that serve as
7884 arguments to bitfield instructions.
7886 If both types of instructions truncate the count (for shifts) and
7887 position (for bitfield operations), or if no variable-position bitfield
7888 instructions exist, you should define this macro.
7890 However, on some machines, such as the 80386 and the 680x0, truncation
7891 only applies to shift operations and not the (real or pretended)
7892 bitfield operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7893 such machines. Instead, add patterns to the @file{md} file that include
7894 the implied truncation of the shift instructions.
7896 You need not define this macro if it would always have the value of zero.
7898 @findex TRULY_NOOP_TRUNCATION
7899 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
7900 A C expression which is nonzero if on this machine it is safe to
7901 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
7902 bits (where @var{outprec} is smaller than @var{inprec}) by merely
7903 operating on it as if it had only @var{outprec} bits.
7905 On many machines, this expression can be 1.
7907 @c rearranged this, removed the phrase "it is reported that". this was
7908 @c to fix an overfull hbox. --mew 10feb93
7909 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
7910 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
7911 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
7912 such cases may improve things.
7914 @findex STORE_FLAG_VALUE
7915 @item STORE_FLAG_VALUE
7916 A C expression describing the value returned by a comparison operator
7917 with an integral mode and stored by a store-flag instruction
7918 (@samp{s@var{cond}}) when the condition is true. This description must
7919 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
7920 comparison operators whose results have a @code{MODE_INT} mode.
7922 A value of 1 or -1 means that the instruction implementing the
7923 comparison operator returns exactly 1 or -1 when the comparison is true
7924 and 0 when the comparison is false. Otherwise, the value indicates
7925 which bits of the result are guaranteed to be 1 when the comparison is
7926 true. This value is interpreted in the mode of the comparison
7927 operation, which is given by the mode of the first operand in the
7928 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
7929 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7932 If @code{STORE_FLAG_VALUE} is neither 1 or -1, the compiler will
7933 generate code that depends only on the specified bits. It can also
7934 replace comparison operators with equivalent operations if they cause
7935 the required bits to be set, even if the remaining bits are undefined.
7936 For example, on a machine whose comparison operators return an
7937 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7938 @samp{0x80000000}, saying that just the sign bit is relevant, the
7942 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7949 (ashift:SI @var{x} (const_int @var{n}))
7953 where @var{n} is the appropriate shift count to move the bit being
7954 tested into the sign bit.
7956 There is no way to describe a machine that always sets the low-order bit
7957 for a true value, but does not guarantee the value of any other bits,
7958 but we do not know of any machine that has such an instruction. If you
7959 are trying to port GCC to such a machine, include an instruction to
7960 perform a logical-and of the result with 1 in the pattern for the
7961 comparison operators and let us know
7963 (@pxref{Bug Reporting,,How to Report Bugs}).
7966 (@pxref{Bug Reporting,,How to Report Bugs,gcc.info,Using GCC}).
7969 Often, a machine will have multiple instructions that obtain a value
7970 from a comparison (or the condition codes). Here are rules to guide the
7971 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7976 Use the shortest sequence that yields a valid definition for
7977 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7978 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7979 comparison operators to do so because there may be opportunities to
7980 combine the normalization with other operations.
7983 For equal-length sequences, use a value of 1 or -1, with -1 being
7984 slightly preferred on machines with expensive jumps and 1 preferred on
7988 As a second choice, choose a value of @samp{0x80000001} if instructions
7989 exist that set both the sign and low-order bits but do not define the
7993 Otherwise, use a value of @samp{0x80000000}.
7996 Many machines can produce both the value chosen for
7997 @code{STORE_FLAG_VALUE} and its negation in the same number of
7998 instructions. On those machines, you should also define a pattern for
7999 those cases, e.g., one matching
8002 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8005 Some machines can also perform @code{and} or @code{plus} operations on
8006 condition code values with less instructions than the corresponding
8007 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8008 machines, define the appropriate patterns. Use the names @code{incscc}
8009 and @code{decscc}, respectively, for the patterns which perform
8010 @code{plus} or @code{minus} operations on condition code values. See
8011 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8012 find such instruction sequences on other machines.
8014 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8017 @findex FLOAT_STORE_FLAG_VALUE
8018 @item FLOAT_STORE_FLAG_VALUE (@var{mode})
8019 A C expression that gives a non-zero @code{REAL_VALUE_TYPE} value that is
8020 returned when comparison operators with floating-point results are true.
8021 Define this macro on machine that have comparison operations that return
8022 floating-point values. If there are no such operations, do not define
8027 An alias for the machine mode for pointers. On most machines, define
8028 this to be the integer mode corresponding to the width of a hardware
8029 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8030 On some machines you must define this to be one of the partial integer
8031 modes, such as @code{PSImode}.
8033 The width of @code{Pmode} must be at least as large as the value of
8034 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8035 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8038 @findex FUNCTION_MODE
8040 An alias for the machine mode used for memory references to functions
8041 being called, in @code{call} RTL expressions. On most machines this
8042 should be @code{QImode}.
8044 @findex INTEGRATE_THRESHOLD
8045 @item INTEGRATE_THRESHOLD (@var{decl})
8046 A C expression for the maximum number of instructions above which the
8047 function @var{decl} should not be inlined. @var{decl} is a
8048 @code{FUNCTION_DECL} node.
8050 The default definition of this macro is 64 plus 8 times the number of
8051 arguments that the function accepts. Some people think a larger
8052 threshold should be used on RISC machines.
8054 @findex SCCS_DIRECTIVE
8055 @item SCCS_DIRECTIVE
8056 Define this if the preprocessor should ignore @code{#sccs} directives
8057 and print no error message.
8059 @findex NO_IMPLICIT_EXTERN_C
8060 @item NO_IMPLICIT_EXTERN_C
8061 Define this macro if the system header files support C++ as well as C.
8062 This macro inhibits the usual method of using system header files in
8063 C++, which is to pretend that the file's contents are enclosed in
8064 @samp{extern "C" @{@dots{}@}}.
8066 @findex HANDLE_PRAGMA
8067 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
8068 This macro is no longer supported. You must use
8069 @code{REGISTER_TARGET_PRAGMAS} instead.
8071 @findex REGISTER_TARGET_PRAGMAS
8074 @item REGISTER_TARGET_PRAGMAS (@var{pfile})
8075 Define this macro if you want to implement any target-specific pragmas.
8076 If defined, it is a C expression which makes a series of calls to the
8077 @code{cpp_register_pragma} and/or @code{cpp_register_pragma_space}
8078 functions. The @var{pfile} argument is the first argument to supply to
8079 these functions. The macro may also do setup required for the pragmas.
8081 The primary reason to define this macro is to provide compatibility with
8082 other compilers for the same target. In general, we discourage
8083 definition of target-specific pragmas for GCC.
8085 If the pragma can be implemented by attributes then the macro
8086 @samp{INSERT_ATTRIBUTES} might be a useful one to define as well.
8088 Preprocessor macros that appear on pragma lines are not expanded. All
8089 @samp{#pragma} directives that do not match any registered pragma are
8090 silently ignored, unless the user specifies @samp{-Wunknown-pragmas}.
8092 @deftypefun void cpp_register_pragma (cpp_reader *@var{pfile}, const char *@var{space}, const char *@var{name}, void (*@var{callback}) (cpp_reader *))
8094 Each call to @code{cpp_register_pragma} establishes one pragma. The
8095 @var{callback} routine will be called when the preprocessor encounters a
8099 #pragma [@var{space}] @var{name} @dots{}
8102 @var{space} must have been the subject of a previous call to
8103 @code{cpp_register_pragma_space}, or else be a null pointer. The
8104 callback routine receives @var{pfile} as its first argument, but must
8105 not use it for anything (this may change in the future). It may read
8106 any text after the @var{name} by making calls to @code{c_lex}. Text
8107 which is not read by the callback will be silently ignored.
8109 Note that both @var{space} and @var{name} are case sensitive.
8111 For an example use of this routine, see @file{c4x.h} and the callback
8112 routines defined in @file{c4x.c}.
8114 Note that the use of @code{c_lex} is specific to the C and C++
8115 compilers. It will not work in the Java or Fortran compilers, or any
8116 other language compilers for that matter. Thus if @code{c_lex} is going
8117 to be called from target-specific code, it must only be done so when
8118 building hte C and C++ compilers. This can be done by defining the
8119 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8120 target entry in the @code{config.gcc} file. These variables should name
8121 the target-specific, language-specific object file which contains the
8122 code that uses @code{c_lex}. Note it will also be necessary to add a
8123 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8124 how to build this object file.
8127 @deftypefun void cpp_register_pragma_space (cpp_reader *@var{pfile}, const char *@var{space})
8128 This routine establishes a namespace for pragmas, which will be
8129 registered by subsequent calls to @code{cpp_register_pragma}. For
8130 example, pragmas defined by the C standard are in the @samp{STDC}
8131 namespace, and pragmas specific to GCC are in the @samp{GCC} namespace.
8133 For an example use of this routine in a target header, see @file{v850.h}.
8136 @findex HANDLE_SYSV_PRAGMA
8139 @item HANDLE_SYSV_PRAGMA
8140 Define this macro (to a value of 1) if you want the System V style
8141 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8142 [=<value>]} to be supported by gcc.
8144 The pack pragma specifies the maximum alignment (in bytes) of fields
8145 within a structure, in much the same way as the @samp{__aligned__} and
8146 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8147 the behaviour to the default.
8149 The weak pragma only works if @code{SUPPORTS_WEAK} and
8150 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8151 of specifically named weak labels, optionally with a value.
8153 @findex HANDLE_PRAGMA_PACK_PUSH_POP
8156 @item HANDLE_PRAGMA_PACK_PUSH_POP
8157 Define this macro (to a value of 1) if you want to support the Win32
8158 style pragmas @samp{#pragma pack(push,<n>)} and @samp{#pragma
8159 pack(pop)}. The pack(push,<n>) pragma specifies the maximum alignment
8160 (in bytes) of fields within a structure, in much the same way as the
8161 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8162 pack value of zero resets the behaviour to the default. Successive
8163 invocations of this pragma cause the previous values to be stacked, so
8164 that invocations of @samp{#pragma pack(pop)} will return to the previous
8167 @findex VALID_MACHINE_DECL_ATTRIBUTE
8168 @item VALID_MACHINE_DECL_ATTRIBUTE (@var{decl}, @var{attributes}, @var{identifier}, @var{args})
8169 If defined, a C expression whose value is nonzero if @var{identifier} with
8170 arguments @var{args} is a valid machine specific attribute for @var{decl}.
8171 The attributes in @var{attributes} have previously been assigned to @var{decl}.
8173 @findex VALID_MACHINE_TYPE_ATTRIBUTE
8174 @item VALID_MACHINE_TYPE_ATTRIBUTE (@var{type}, @var{attributes}, @var{identifier}, @var{args})
8175 If defined, a C expression whose value is nonzero if @var{identifier} with
8176 arguments @var{args} is a valid machine specific attribute for @var{type}.
8177 The attributes in @var{attributes} have previously been assigned to @var{type}.
8179 @findex COMP_TYPE_ATTRIBUTES
8180 @item COMP_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
8181 If defined, a C expression whose value is zero if the attributes on
8182 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8183 and two if they are nearly compatible (which causes a warning to be
8186 @findex SET_DEFAULT_TYPE_ATTRIBUTES
8187 @item SET_DEFAULT_TYPE_ATTRIBUTES (@var{type})
8188 If defined, a C statement that assigns default attributes to
8189 newly defined @var{type}.
8191 @findex MERGE_MACHINE_TYPE_ATTRIBUTES
8192 @item MERGE_MACHINE_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
8193 Define this macro if the merging of type attributes needs special handling.
8194 If defined, the result is a list of the combined TYPE_ATTRIBUTES of
8195 @var{type1} and @var{type2}. It is assumed that comptypes has already been
8196 called and returned 1.
8198 @findex MERGE_MACHINE_DECL_ATTRIBUTES
8199 @item MERGE_MACHINE_DECL_ATTRIBUTES (@var{olddecl}, @var{newdecl})
8200 Define this macro if the merging of decl attributes needs special handling.
8201 If defined, the result is a list of the combined DECL_MACHINE_ATTRIBUTES of
8202 @var{olddecl} and @var{newdecl}. @var{newdecl} is a duplicate declaration
8203 of @var{olddecl}. Examples of when this is needed are when one attribute
8204 overrides another, or when an attribute is nullified by a subsequent
8207 @findex INSERT_ATTRIBUTES
8208 @item INSERT_ATTRIBUTES (@var{node}, @var{attr_ptr}, @var{prefix_ptr})
8209 Define this macro if you want to be able to add attributes to a decl
8210 when it is being created. This is normally useful for backends which
8211 wish to implement a pragma by using the attributes which correspond to
8212 the pragma's effect. The @var{node} argument is the decl which is being
8213 created. The @var{attr_ptr} argument is a pointer to the attribute list
8214 for this decl. The @var{prefix_ptr} is a pointer to the list of
8215 attributes that have appeared after the specifiers and modifiers of the
8216 declaration, but before the declaration proper.
8218 @findex SET_DEFAULT_DECL_ATTRIBUTES
8219 @item SET_DEFAULT_DECL_ATTRIBUTES (@var{decl}, @var{attributes})
8220 If defined, a C statement that assigns default attributes to
8221 newly defined @var{decl}.
8223 @findex DOLLARS_IN_IDENTIFIERS
8224 @item DOLLARS_IN_IDENTIFIERS
8225 Define this macro to control use of the character @samp{$} in identifier
8226 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
8227 1 is the default; there is no need to define this macro in that case.
8228 This macro controls the compiler proper; it does not affect the preprocessor.
8230 @findex NO_DOLLAR_IN_LABEL
8231 @item NO_DOLLAR_IN_LABEL
8232 Define this macro if the assembler does not accept the character
8233 @samp{$} in label names. By default constructors and destructors in
8234 G++ have @samp{$} in the identifiers. If this macro is defined,
8235 @samp{.} is used instead.
8237 @findex NO_DOT_IN_LABEL
8238 @item NO_DOT_IN_LABEL
8239 Define this macro if the assembler does not accept the character
8240 @samp{.} in label names. By default constructors and destructors in G++
8241 have names that use @samp{.}. If this macro is defined, these names
8242 are rewritten to avoid @samp{.}.
8244 @findex DEFAULT_MAIN_RETURN
8245 @item DEFAULT_MAIN_RETURN
8246 Define this macro if the target system expects every program's @code{main}
8247 function to return a standard ``success'' value by default (if no other
8248 value is explicitly returned).
8250 The definition should be a C statement (sans semicolon) to generate the
8251 appropriate rtl instructions. It is used only when compiling the end of
8256 Define this if the target system lacks the function @code{atexit}
8257 from the ISO C standard. If this macro is defined, a default definition
8258 will be provided to support C++. If @code{ON_EXIT} is not defined,
8259 a default @code{exit} function will also be provided.
8263 Define this macro if the target has another way to implement atexit
8264 functionality without replacing @code{exit}. For instance, SunOS 4 has
8265 a similar @code{on_exit} library function.
8267 The definition should be a functional macro which can be used just like
8268 the @code{atexit} function.
8272 Define this if your @code{exit} function needs to do something
8273 besides calling an external function @code{_cleanup} before
8274 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
8275 only needed if @code{NEED_ATEXIT} is defined and @code{ON_EXIT} is not
8278 @findex INSN_SETS_ARE_DELAYED
8279 @item INSN_SETS_ARE_DELAYED (@var{insn})
8280 Define this macro as a C expression that is nonzero if it is safe for the
8281 delay slot scheduler to place instructions in the delay slot of @var{insn},
8282 even if they appear to use a resource set or clobbered in @var{insn}.
8283 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8284 every @code{call_insn} has this behavior. On machines where some @code{insn}
8285 or @code{jump_insn} is really a function call and hence has this behavior,
8286 you should define this macro.
8288 You need not define this macro if it would always return zero.
8290 @findex INSN_REFERENCES_ARE_DELAYED
8291 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
8292 Define this macro as a C expression that is nonzero if it is safe for the
8293 delay slot scheduler to place instructions in the delay slot of @var{insn},
8294 even if they appear to set or clobber a resource referenced in @var{insn}.
8295 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8296 some @code{insn} or @code{jump_insn} is really a function call and its operands
8297 are registers whose use is actually in the subroutine it calls, you should
8298 define this macro. Doing so allows the delay slot scheduler to move
8299 instructions which copy arguments into the argument registers into the delay
8302 You need not define this macro if it would always return zero.
8304 @findex MACHINE_DEPENDENT_REORG
8305 @item MACHINE_DEPENDENT_REORG (@var{insn})
8306 In rare cases, correct code generation requires extra machine
8307 dependent processing between the second jump optimization pass and
8308 delayed branch scheduling. On those machines, define this macro as a C
8309 statement to act on the code starting at @var{insn}.
8311 @findex MULTIPLE_SYMBOL_SPACES
8312 @item MULTIPLE_SYMBOL_SPACES
8313 Define this macro if in some cases global symbols from one translation
8314 unit may not be bound to undefined symbols in another translation unit
8315 without user intervention. For instance, under Microsoft Windows
8316 symbols must be explicitly imported from shared libraries (DLLs).
8318 @findex MD_ASM_CLOBBERS
8319 @item MD_ASM_CLOBBERS
8320 A C statement that adds to @var{CLOBBERS} @code{STRING_CST} trees for
8321 any hard regs the port wishes to automatically clobber for all asms.
8325 A C expression that returns how many instructions can be issued at the
8326 same time if the machine is a superscalar machine.
8328 @findex MD_SCHED_INIT
8329 @item MD_SCHED_INIT (@var{file}, @var{verbose}, @var{max_ready})
8330 A C statement which is executed by the scheduler at the
8331 beginning of each block of instructions that are to be scheduled.
8332 @var{file} is either a null pointer, or a stdio stream to write any
8333 debug output to. @var{verbose} is the verbose level provided by
8334 @samp{-fsched-verbose-}@var{n}. @var{max_ready} is the maximum number
8335 of insns in the current scheduling region that can be live at the same
8336 time. This can be used to allocate scratch space if it is needed.
8338 @findex MD_SCHED_FINISH
8339 @item MD_SCHED_FINISH (@var{file}, @var{verbose})
8340 A C statement which is executed by the scheduler at the end of each block
8341 of instructions that are to be scheduled. It can be used to perform
8342 cleanup of any actions done by the other scheduling macros.
8343 @var{file} is either a null pointer, or a stdio stream to write any
8344 debug output to. @var{verbose} is the verbose level provided by
8345 @samp{-fsched-verbose-}@var{n}.
8347 @findex MD_SCHED_REORDER
8348 @item MD_SCHED_REORDER (@var{file}, @var{verbose}, @var{ready}, @var{n_ready}, @var{clock}, @var{can_issue_more})
8349 A C statement which is executed by the scheduler after it
8350 has scheduled the ready list to allow the machine description to reorder
8351 it (for example to combine two small instructions together on
8352 @samp{VLIW} machines). @var{file} is either a null pointer, or a stdio
8353 stream to write any debug output to. @var{verbose} is the verbose level
8354 provided by @samp{-fsched-verbose-}@var{n}. @var{ready} is a pointer to
8355 the ready list of instructions that are ready to be scheduled.
8356 @var{n_ready} is the number of elements in the ready list. The
8357 scheduler reads the ready list in reverse order, starting with
8358 @var{ready}[@var{n_ready}-1] and going to @var{ready}[0]. @var{clock}
8359 is the timer tick of the scheduler. @var{can_issue_more} is an output
8360 parameter that is set to the number of insns that can issue this clock;
8361 normally this is just @code{issue_rate}. See also @samp{MD_SCHED_REORDER2}.
8363 @findex MD_SCHED_REORDER2
8364 @item MD_SCHED_REORDER2 (@var{file}, @var{verbose}, @var{ready}, @var{n_ready}, @var{clock}, @var{can_issue_more})
8365 Like @samp{MD_SCHED_REORDER}, but called at a different time. While the
8366 @samp{MD_SCHED_REORDER} macro is called whenever the scheduler starts a
8367 new cycle, this macro is used immediately after @samp{MD_SCHED_VARIABLE_ISSUE}
8368 is called; it can reorder the ready list and set @var{can_issue_more} to
8369 determine whether there are more insns to be scheduled in the same cycle.
8370 Defining this macro can be useful if there are frequent situations where
8371 scheduling one insn causes other insns to become ready in the same cycle,
8372 these other insns can then be taken into account properly.
8374 @findex MD_SCHED_VARIABLE_ISSUE
8375 @item MD_SCHED_VARIABLE_ISSUE (@var{file}, @var{verbose}, @var{insn}, @var{more})
8376 A C statement which is executed by the scheduler after it
8377 has scheduled an insn from the ready list. @var{file} is either a null
8378 pointer, or a stdio stream to write any debug output to. @var{verbose}
8379 is the verbose level provided by @samp{-fsched-verbose-}@var{n}.
8380 @var{insn} is the instruction that was scheduled. @var{more} is the
8381 number of instructions that can be issued in the current cycle. The
8382 @samp{MD_SCHED_VARIABLE_ISSUE} macro is responsible for updating the
8383 value of @var{more} (typically by @var{more}--).
8385 @findex MAX_INTEGER_COMPUTATION_MODE
8386 @item MAX_INTEGER_COMPUTATION_MODE
8387 Define this to the largest integer machine mode which can be used for
8388 operations other than load, store and copy operations.
8390 You need only define this macro if the target holds values larger than
8391 @code{word_mode} in general purpose registers. Most targets should not define
8394 @findex MATH_LIBRARY
8396 Define this macro as a C string constant for the linker argument to link
8397 in the system math library, or @samp{""} if the target does not have a
8398 separate math library.
8400 You need only define this macro if the default of @samp{"-lm"} is wrong.
8402 @findex LIBRARY_PATH_ENV
8403 @item LIBRARY_PATH_ENV
8404 Define this macro as a C string constant for the environment variable that
8405 specifies where the linker should look for libraries.
8407 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
8410 @findex TARGET_HAS_F_SETLKW
8411 @item TARGET_HAS_F_SETLKW
8412 Define this macro if the target supports file locking with fcntl / F_SETLKW.
8413 Note that this functionality is part of POSIX.
8414 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
8415 to use file locking when exiting a program, which avoids race conditions
8416 if the program has forked.
8418 @findex MAX_CONDITIONAL_EXECUTE
8419 @item MAX_CONDITIONAL_EXECUTE
8421 A C expression for the maximum number of instructions to execute via
8422 conditional execution instructions instead of a branch. A value of
8423 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8424 1 if it does use cc0.
8426 @findex IFCVT_MODIFY_TESTS
8427 @item IFCVT_MODIFY_TESTS
8428 A C expression to modify the tests in @code{TRUE_EXPR}, and
8429 @code{FALSE_EXPPR} for use in converting insns in @code{TEST_BB},
8430 @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB} basic blocks to
8431 conditional execution. Set either @code{TRUE_EXPR} or @code{FALSE_EXPR}
8432 to a null pointer if the tests cannot be converted.
8434 @findex IFCVT_MODIFY_INSN
8435 @item IFCVT_MODIFY_INSN
8436 A C expression to modify the @code{PATTERN} of an @code{INSN} that is to
8437 be converted to conditional execution format.
8439 @findex IFCVT_MODIFY_FINAL
8440 @item IFCVT_MODIFY_FINAL
8441 A C expression to perform any final machine dependent modifications in
8442 converting code to conditional execution in the basic blocks
8443 @code{TEST_BB}, @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB}.
8445 @findex IFCVT_MODIFY_CANCEL
8446 @item IFCVT_MODIFY_CANCEL
8447 A C expression to cancel any machine dependent modifications in
8448 converting code to conditional execution in the basic blocks
8449 @code{TEST_BB}, @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB}.
8451 @findex MD_INIT_BUILTINS
8452 @item MD_INIT_BUILTINS
8453 Define this macro if you have any machine-specific builtin functions that
8454 need to be defined. It should be a C expression that performs the
8457 Machine specific builtins can be useful to expand special machine
8458 instructions that would otherwise not normally be generated because
8459 they have no equivalent in the source language (for example, SIMD vector
8460 instructions or prefetch instructions).
8462 To create a builtin function, call the function @code{builtin_function}
8463 which is defined by the language frontend. You can use any type nodes set
8464 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
8465 only language frontends that use these two functions will use
8466 @samp{MD_INIT_BUILTINS}.
8468 @findex MD_EXPAND_BUILTIN
8469 @item MD_EXPAND_BUILTIN(@var{exp}, @var{target}, @var{subtarget}, @var{mode}, @var{ignore})
8471 Expand a call to a machine specific builtin that was set up by
8472 @samp{MD_INIT_BUILTINS}. @var{exp} is the expression for the function call;
8473 the result should go to @var{target} if that is convenient, and have mode
8474 @var{mode} if that is convenient. @var{subtarget} may be used as the target
8475 for computing one of @var{exp}'s operands. @var{ignore} is nonzero if the value
8477 This macro should return the result of the call to the builtin.