1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000
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 * Storage Layout:: Defining sizes and alignments of data.
26 * Type Layout:: Defining sizes and properties of basic user data types.
27 * Registers:: Naming and describing the hardware registers.
28 * Register Classes:: Defining the classes of hardware registers.
29 * Stack and Calling:: Defining which way the stack grows and by how much.
30 * Varargs:: Defining the varargs macros.
31 * Trampolines:: Code set up at run time to enter a nested function.
32 * Library Calls:: Controlling how library routines are implicitly called.
33 * Addressing Modes:: Defining addressing modes valid for memory operands.
34 * Condition Code:: Defining how insns update the condition code.
35 * Costs:: Defining relative costs of different operations.
36 * Sections:: Dividing storage into text, data, and other sections.
37 * PIC:: Macros for position independent code.
38 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
39 * Debugging Info:: Defining the format of debugging output.
40 * Cross-compilation:: Handling floating point for cross-compilers.
41 * Mode Switching:: Insertion of mode-switching instructions.
42 * Misc:: Everything else.
46 @section Controlling the Compilation Driver, @file{gcc}
48 @cindex controlling the compilation driver
50 @c prevent bad page break with this line
51 You can control the compilation driver.
54 @findex SWITCH_TAKES_ARG
55 @item SWITCH_TAKES_ARG (@var{char})
56 A C expression which determines whether the option @samp{-@var{char}}
57 takes arguments. The value should be the number of arguments that
58 option takes--zero, for many options.
60 By default, this macro is defined as
61 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
62 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
63 wish to add additional options which take arguments. Any redefinition
64 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
67 @findex WORD_SWITCH_TAKES_ARG
68 @item WORD_SWITCH_TAKES_ARG (@var{name})
69 A C expression which determines whether the option @samp{-@var{name}}
70 takes arguments. The value should be the number of arguments that
71 option takes--zero, for many options. This macro rather than
72 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
74 By default, this macro is defined as
75 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
76 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
77 wish to add additional options which take arguments. Any redefinition
78 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
81 @findex SWITCH_CURTAILS_COMPILATION
82 @item SWITCH_CURTAILS_COMPILATION (@var{char})
83 A C expression which determines whether the option @samp{-@var{char}}
84 stops compilation before the generation of an executable. The value is
85 boolean, non-zero if the option does stop an executable from being
86 generated, zero otherwise.
88 By default, this macro is defined as
89 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
90 options properly. You need not define
91 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
92 options which affect the generation of an executable. Any redefinition
93 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
94 for additional options.
96 @findex SWITCHES_NEED_SPACES
97 @item SWITCHES_NEED_SPACES
98 A string-valued C expression which enumerates the options for which
99 the linker needs a space between the option and its argument.
101 If this macro is not defined, the default value is @code{""}.
105 A C string constant that tells the GCC driver program options to
106 pass to CPP. It can also specify how to translate options you
107 give to GCC into options for GCC to pass to the CPP.
109 Do not define this macro if it does not need to do anything.
111 @findex NO_BUILTIN_SIZE_TYPE
112 @item NO_BUILTIN_SIZE_TYPE
113 If this macro is defined, the preprocessor will not define the builtin macro
114 @code{__SIZE_TYPE__}. The macro @code{__SIZE_TYPE__} must then be defined
115 by @code{CPP_SPEC} instead.
117 This should be defined if @code{SIZE_TYPE} depends on target dependent flags
118 which are not accessible to the preprocessor. Otherwise, it should not
121 @findex NO_BUILTIN_PTRDIFF_TYPE
122 @item NO_BUILTIN_PTRDIFF_TYPE
123 If this macro is defined, the preprocessor will not define the builtin macro
124 @code{__PTRDIFF_TYPE__}. The macro @code{__PTRDIFF_TYPE__} must then be
125 defined by @code{CPP_SPEC} instead.
127 This should be defined if @code{PTRDIFF_TYPE} depends on target dependent flags
128 which are not accessible to the preprocessor. Otherwise, it should not
131 @findex NO_BUILTIN_WCHAR_TYPE
132 @item NO_BUILTIN_WCHAR_TYPE
133 If this macro is defined, the preprocessor will not define the builtin macro
134 @code{__WCHAR_TYPE__}. The macro @code{__WCHAR_TYPE__} must then be
135 defined by @code{CPP_SPEC} instead.
137 This should be defined if @code{WCHAR_TYPE} depends on target dependent flags
138 which are not accessible to the preprocessor. Otherwise, it should not
141 @findex SIGNED_CHAR_SPEC
142 @item SIGNED_CHAR_SPEC
143 A C string constant that tells the GCC driver program options to
144 pass to CPP. By default, this macro is defined to pass the option
145 @samp{-D__CHAR_UNSIGNED__} to CPP if @code{char} will be treated as
146 @code{unsigned char} by @code{cc1}.
148 Do not define this macro unless you need to override the default
153 A C string constant that tells the GCC driver program options to
154 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
156 It can also specify how to translate options you give to GCC into options
157 for GCC to pass to front ends..
159 Do not define this macro if it does not need to do anything.
163 A C string constant that tells the GCC driver program options to
164 pass to @code{cc1plus}. It can also specify how to translate options you
165 give to GCC into options for GCC to pass to the @code{cc1plus}.
167 Do not define this macro if it does not need to do anything.
168 Note that everything defined in CC1_SPEC is already passed to
169 @code{cc1plus} so there is no need to duplicate the contents of
170 CC1_SPEC in CC1PLUS_SPEC.
174 A C string constant that tells the GCC driver program options to
175 pass to the assembler. It can also specify how to translate options
176 you give to GCC into options for GCC to pass to the assembler.
177 See the file @file{sun3.h} for an example of this.
179 Do not define this macro if it does not need to do anything.
181 @findex ASM_FINAL_SPEC
183 A C string constant that tells the GCC driver program how to
184 run any programs which cleanup after the normal assembler.
185 Normally, this is not needed. See the file @file{mips.h} for
188 Do not define this macro if it does not need to do anything.
192 A C string constant that tells the GCC driver program options to
193 pass to the linker. It can also specify how to translate options you
194 give to GCC into options for GCC to pass to the linker.
196 Do not define this macro if it does not need to do anything.
200 Another C string constant used much like @code{LINK_SPEC}. The difference
201 between the two is that @code{LIB_SPEC} is used at the end of the
202 command given to the linker.
204 If this macro is not defined, a default is provided that
205 loads the standard C library from the usual place. See @file{gcc.c}.
209 Another C string constant that tells the GCC driver program
210 how and when to place a reference to @file{libgcc.a} into the
211 linker command line. This constant is placed both before and after
212 the value of @code{LIB_SPEC}.
214 If this macro is not defined, the GCC driver provides a default that
215 passes the string @samp{-lgcc} to the linker unless the @samp{-shared}
218 @findex STARTFILE_SPEC
220 Another C string constant used much like @code{LINK_SPEC}. The
221 difference between the two is that @code{STARTFILE_SPEC} is used at
222 the very beginning of the command given to the linker.
224 If this macro is not defined, a default is provided that loads the
225 standard C startup file from the usual place. See @file{gcc.c}.
229 Another C string constant used much like @code{LINK_SPEC}. The
230 difference between the two is that @code{ENDFILE_SPEC} is used at
231 the very end of the command given to the linker.
233 Do not define this macro if it does not need to do anything.
237 Define this macro to provide additional specifications to put in the
238 @file{specs} file that can be used in various specifications like
241 The definition should be an initializer for an array of structures,
242 containing a string constant, that defines the specification name, and a
243 string constant that provides the specification.
245 Do not define this macro if it does not need to do anything.
247 @code{EXTRA_SPECS} is useful when an architecture contains several
248 related targets, which have various @code{..._SPECS} which are similar
249 to each other, and the maintainer would like one central place to keep
252 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
253 define either @code{_CALL_SYSV} when the System V calling sequence is
254 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
257 The @file{config/rs6000/rs6000.h} target file defines:
260 #define EXTRA_SPECS \
261 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
263 #define CPP_SYS_DEFAULT ""
266 The @file{config/rs6000/sysv.h} target file defines:
270 "%@{posix: -D_POSIX_SOURCE @} \
271 %@{mcall-sysv: -D_CALL_SYSV @} %@{mcall-aix: -D_CALL_AIX @} \
272 %@{!mcall-sysv: %@{!mcall-aix: %(cpp_sysv_default) @}@} \
273 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
275 #undef CPP_SYSV_DEFAULT
276 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
279 while the @file{config/rs6000/eabiaix.h} target file defines
280 @code{CPP_SYSV_DEFAULT} as:
283 #undef CPP_SYSV_DEFAULT
284 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
287 @findex LINK_LIBGCC_SPECIAL
288 @item LINK_LIBGCC_SPECIAL
289 Define this macro if the driver program should find the library
290 @file{libgcc.a} itself and should not pass @samp{-L} options to the
291 linker. If you do not define this macro, the driver program will pass
292 the argument @samp{-lgcc} to tell the linker to do the search and will
293 pass @samp{-L} options to it.
295 @findex LINK_LIBGCC_SPECIAL_1
296 @item LINK_LIBGCC_SPECIAL_1
297 Define this macro if the driver program should find the library
298 @file{libgcc.a}. If you do not define this macro, the driver program will pass
299 the argument @samp{-lgcc} to tell the linker to do the search.
300 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
301 not affect @samp{-L} options.
303 @findex LINK_COMMAND_SPEC
304 @item LINK_COMMAND_SPEC
305 A C string constant giving the complete command line need to execute the
306 linker. When you do this, you will need to update your port each time a
307 change is made to the link command line within @file{gcc.c}. Therefore,
308 define this macro only if you need to completely redefine the command
309 line for invoking the linker and there is no other way to accomplish
312 @findex MULTILIB_DEFAULTS
313 @item MULTILIB_DEFAULTS
314 Define this macro as a C expression for the initializer of an array of
315 string to tell the driver program which options are defaults for this
316 target and thus do not need to be handled specially when using
317 @code{MULTILIB_OPTIONS}.
319 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
320 the target makefile fragment or if none of the options listed in
321 @code{MULTILIB_OPTIONS} are set by default.
322 @xref{Target Fragment}.
324 @findex RELATIVE_PREFIX_NOT_LINKDIR
325 @item RELATIVE_PREFIX_NOT_LINKDIR
326 Define this macro to tell @code{gcc} that it should only translate
327 a @samp{-B} prefix into a @samp{-L} linker option if the prefix
328 indicates an absolute file name.
330 @findex STANDARD_EXEC_PREFIX
331 @item STANDARD_EXEC_PREFIX
332 Define this macro as a C string constant if you wish to override the
333 standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
334 try when searching for the executable files of the compiler.
336 @findex MD_EXEC_PREFIX
338 If defined, this macro is an additional prefix to try after
339 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
340 when the @samp{-b} option is used, or the compiler is built as a cross
341 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
342 to the list of directories used to find the assembler in @file{configure.in}.
344 @findex STANDARD_STARTFILE_PREFIX
345 @item STANDARD_STARTFILE_PREFIX
346 Define this macro as a C string constant if you wish to override the
347 standard choice of @file{/usr/local/lib/} as the default prefix to
348 try when searching for startup files such as @file{crt0.o}.
350 @findex MD_STARTFILE_PREFIX
351 @item MD_STARTFILE_PREFIX
352 If defined, this macro supplies an additional prefix to try after the
353 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
354 @samp{-b} option is used, or when the compiler is built as a cross
357 @findex MD_STARTFILE_PREFIX_1
358 @item MD_STARTFILE_PREFIX_1
359 If defined, this macro supplies yet another prefix to try after the
360 standard prefixes. It is not searched when the @samp{-b} option is
361 used, or when the compiler is built as a cross compiler.
363 @findex INIT_ENVIRONMENT
364 @item INIT_ENVIRONMENT
365 Define this macro as a C string constant if you wish to set environment
366 variables for programs called by the driver, such as the assembler and
367 loader. The driver passes the value of this macro to @code{putenv} to
368 initialize the necessary environment variables.
370 @findex LOCAL_INCLUDE_DIR
371 @item LOCAL_INCLUDE_DIR
372 Define this macro as a C string constant if you wish to override the
373 standard choice of @file{/usr/local/include} as the default prefix to
374 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
375 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
377 Cross compilers do not use this macro and do not search either
378 @file{/usr/local/include} or its replacement.
380 @findex SYSTEM_INCLUDE_DIR
381 @item SYSTEM_INCLUDE_DIR
382 Define this macro as a C string constant if you wish to specify a
383 system-specific directory to search for header files before the standard
384 directory. @code{SYSTEM_INCLUDE_DIR} comes before
385 @code{STANDARD_INCLUDE_DIR} in the search order.
387 Cross compilers do not use this macro and do not search the directory
390 @findex STANDARD_INCLUDE_DIR
391 @item STANDARD_INCLUDE_DIR
392 Define this macro as a C string constant if you wish to override the
393 standard choice of @file{/usr/include} as the default prefix to
394 try when searching for header files.
396 Cross compilers do not use this macro and do not search either
397 @file{/usr/include} or its replacement.
399 @findex STANDARD_INCLUDE_COMPONENT
400 @item STANDARD_INCLUDE_COMPONENT
401 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
402 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
403 If you do not define this macro, no component is used.
405 @findex INCLUDE_DEFAULTS
406 @item INCLUDE_DEFAULTS
407 Define this macro if you wish to override the entire default search path
408 for include files. For a native compiler, the default search path
409 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
410 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
411 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
412 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
413 and specify private search areas for GCC. The directory
414 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
416 The definition should be an initializer for an array of structures.
417 Each array element should have four elements: the directory name (a
418 string constant), the component name, and flag for C++-only directories,
419 and a flag showing that the includes in the directory don't need to be
420 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
421 the array with a null element.
423 The component name denotes what GNU package the include file is part of,
424 if any, in all upper-case letters. For example, it might be @samp{GCC}
425 or @samp{BINUTILS}. If the package is part of the a vendor-supplied
426 operating system, code the component name as @samp{0}.
428 @findex STRUCT_FORCE_BLK
429 @item STRUCT_FORCE_BLK (@var{field})
430 Return 1 if a structure containing @var{field} should be accessed using
433 Normally, this is not needed. See the file @file{c4x.h} for an example
434 of how to use this macro to prevent a structure having a floating point
435 field from being accessed in an integer mode.
438 For example, here is the definition used for VAX/VMS:
441 #define INCLUDE_DEFAULTS \
443 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
444 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
445 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
452 Here is the order of prefixes tried for exec files:
456 Any prefixes specified by the user with @samp{-B}.
459 The environment variable @code{GCC_EXEC_PREFIX}, if any.
462 The directories specified by the environment variable @code{COMPILER_PATH}.
465 The macro @code{STANDARD_EXEC_PREFIX}.
468 @file{/usr/lib/gcc/}.
471 The macro @code{MD_EXEC_PREFIX}, if any.
474 Here is the order of prefixes tried for startfiles:
478 Any prefixes specified by the user with @samp{-B}.
481 The environment variable @code{GCC_EXEC_PREFIX}, if any.
484 The directories specified by the environment variable @code{LIBRARY_PATH}
485 (or port-specific name; native only, cross compilers do not use this).
488 The macro @code{STANDARD_EXEC_PREFIX}.
491 @file{/usr/lib/gcc/}.
494 The macro @code{MD_EXEC_PREFIX}, if any.
497 The macro @code{MD_STARTFILE_PREFIX}, if any.
500 The macro @code{STANDARD_STARTFILE_PREFIX}.
509 @node Run-time Target
510 @section Run-time Target Specification
511 @cindex run-time target specification
512 @cindex predefined macros
513 @cindex target specifications
515 @c prevent bad page break with this line
516 Here are run-time target specifications.
519 @findex CPP_PREDEFINES
521 Define this to be a string constant containing @samp{-D} options to
522 define the predefined macros that identify this machine and system.
523 These macros will be predefined unless the @samp{-ansi} option is
526 In addition, a parallel set of macros are predefined, whose names are
527 made by appending @samp{__} at the beginning and at the end. These
528 @samp{__} macros are permitted by the ANSI standard, so they are
529 predefined regardless of whether @samp{-ansi} is specified.
531 For example, on the Sun, one can use the following value:
534 "-Dmc68000 -Dsun -Dunix"
537 The result is to define the macros @code{__mc68000__}, @code{__sun__}
538 and @code{__unix__} unconditionally, and the macros @code{mc68000},
539 @code{sun} and @code{unix} provided @samp{-ansi} is not specified.
541 @findex extern int target_flags
542 @item extern int target_flags;
543 This declaration should be present.
545 @cindex optional hardware or system features
546 @cindex features, optional, in system conventions
548 This series of macros is to allow compiler command arguments to
549 enable or disable the use of optional features of the target machine.
550 For example, one machine description serves both the 68000 and
551 the 68020; a command argument tells the compiler whether it should
552 use 68020-only instructions or not. This command argument works
553 by means of a macro @code{TARGET_68020} that tests a bit in
556 Define a macro @code{TARGET_@var{featurename}} for each such option.
557 Its definition should test a bit in @code{target_flags}; for example:
560 #define TARGET_68020 (target_flags & 1)
563 One place where these macros are used is in the condition-expressions
564 of instruction patterns. Note how @code{TARGET_68020} appears
565 frequently in the 68000 machine description file, @file{m68k.md}.
566 Another place they are used is in the definitions of the other
567 macros in the @file{@var{machine}.h} file.
569 @findex TARGET_SWITCHES
570 @item TARGET_SWITCHES
571 This macro defines names of command options to set and clear
572 bits in @code{target_flags}. Its definition is an initializer
573 with a subgrouping for each command option.
575 Each subgrouping contains a string constant, that defines the option
576 name, a number, which contains the bits to set in
577 @code{target_flags}, and a second string which is the description
578 displayed by --help. If the number is negative then the bits specified
579 by the number are cleared instead of being set. If the description
580 string is present but empty, then no help information will be displayed
581 for that option, but it will not count as an undocumented option. The
582 actual option name is made by appending @samp{-m} to the specified name.
584 One of the subgroupings should have a null string. The number in
585 this grouping is the default value for @code{target_flags}. Any
586 target options act starting with that value.
588 Here is an example which defines @samp{-m68000} and @samp{-m68020}
589 with opposite meanings, and picks the latter as the default:
592 #define TARGET_SWITCHES \
593 @{ @{ "68020", 1, "" @}, \
594 @{ "68000", -1, "Compile for the 68000" @}, \
598 @findex TARGET_OPTIONS
600 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
601 options that have values. Its definition is an initializer with a
602 subgrouping for each command option.
604 Each subgrouping contains a string constant, that defines the fixed part
605 of the option name, the address of a variable, and a description string.
606 The variable, type @code{char *}, is set to the variable part of the
607 given option if the fixed part matches. The actual option name is made
608 by appending @samp{-m} to the specified name.
610 Here is an example which defines @samp{-mshort-data-@var{number}}. If the
611 given option is @samp{-mshort-data-512}, the variable @code{m88k_short_data}
612 will be set to the string @code{"512"}.
615 extern char *m88k_short_data;
616 #define TARGET_OPTIONS \
617 @{ @{ "short-data-", &m88k_short_data, "Specify the size of the short data section" @} @}
620 @findex TARGET_VERSION
622 This macro is a C statement to print on @code{stderr} a string
623 describing the particular machine description choice. Every machine
624 description should define @code{TARGET_VERSION}. For example:
628 #define TARGET_VERSION \
629 fprintf (stderr, " (68k, Motorola syntax)");
631 #define TARGET_VERSION \
632 fprintf (stderr, " (68k, MIT syntax)");
636 @findex OVERRIDE_OPTIONS
637 @item OVERRIDE_OPTIONS
638 Sometimes certain combinations of command options do not make sense on
639 a particular target machine. You can define a macro
640 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
641 defined, is executed once just after all the command options have been
644 Don't use this macro to turn on various extra optimizations for
645 @samp{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
647 @findex OPTIMIZATION_OPTIONS
648 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
649 Some machines may desire to change what optimizations are performed for
650 various optimization levels. This macro, if defined, is executed once
651 just after the optimization level is determined and before the remainder
652 of the command options have been parsed. Values set in this macro are
653 used as the default values for the other command line options.
655 @var{level} is the optimization level specified; 2 if @samp{-O2} is
656 specified, 1 if @samp{-O} is specified, and 0 if neither is specified.
658 @var{size} is non-zero if @samp{-Os} is specified and zero otherwise.
660 You should not use this macro to change options that are not
661 machine-specific. These should uniformly selected by the same
662 optimization level on all supported machines. Use this macro to enable
663 machine-specific optimizations.
665 @strong{Do not examine @code{write_symbols} in
666 this macro!} The debugging options are not supposed to alter the
669 @findex CAN_DEBUG_WITHOUT_FP
670 @item CAN_DEBUG_WITHOUT_FP
671 Define this macro if debugging can be performed even without a frame
672 pointer. If this macro is defined, GCC will turn on the
673 @samp{-fomit-frame-pointer} option whenever @samp{-O} is specified.
677 @section Storage Layout
678 @cindex storage layout
680 Note that the definitions of the macros in this table which are sizes or
681 alignments measured in bits do not need to be constant. They can be C
682 expressions that refer to static variables, such as the @code{target_flags}.
683 @xref{Run-time Target}.
686 @findex BITS_BIG_ENDIAN
687 @item BITS_BIG_ENDIAN
688 Define this macro to have the value 1 if the most significant bit in a
689 byte has the lowest number; otherwise define it to have the value zero.
690 This means that bit-field instructions count from the most significant
691 bit. If the machine has no bit-field instructions, then this must still
692 be defined, but it doesn't matter which value it is defined to. This
693 macro need not be a constant.
695 This macro does not affect the way structure fields are packed into
696 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
698 @findex BYTES_BIG_ENDIAN
699 @item BYTES_BIG_ENDIAN
700 Define this macro to have the value 1 if the most significant byte in a
701 word has the lowest number. This macro need not be a constant.
703 @findex WORDS_BIG_ENDIAN
704 @item WORDS_BIG_ENDIAN
705 Define this macro to have the value 1 if, in a multiword object, the
706 most significant word has the lowest number. This applies to both
707 memory locations and registers; GCC fundamentally assumes that the
708 order of words in memory is the same as the order in registers. This
709 macro need not be a constant.
711 @findex LIBGCC2_WORDS_BIG_ENDIAN
712 @item LIBGCC2_WORDS_BIG_ENDIAN
713 Define this macro if WORDS_BIG_ENDIAN is not constant. This must be a
714 constant value with the same meaning as WORDS_BIG_ENDIAN, which will be
715 used only when compiling libgcc2.c. Typically the value will be set
716 based on preprocessor defines.
718 @findex FLOAT_WORDS_BIG_ENDIAN
719 @item FLOAT_WORDS_BIG_ENDIAN
720 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
721 @code{TFmode} floating point numbers are stored in memory with the word
722 containing the sign bit at the lowest address; otherwise define it to
723 have the value 0. This macro need not be a constant.
725 You need not define this macro if the ordering is the same as for
728 @findex BITS_PER_UNIT
730 Define this macro to be the number of bits in an addressable storage
731 unit (byte); normally 8.
733 @findex BITS_PER_WORD
735 Number of bits in a word; normally 32.
737 @findex MAX_BITS_PER_WORD
738 @item MAX_BITS_PER_WORD
739 Maximum number of bits in a word. If this is undefined, the default is
740 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
741 largest value that @code{BITS_PER_WORD} can have at run-time.
743 @findex UNITS_PER_WORD
745 Number of storage units in a word; normally 4.
747 @findex MIN_UNITS_PER_WORD
748 @item MIN_UNITS_PER_WORD
749 Minimum number of units in a word. If this is undefined, the default is
750 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
751 smallest value that @code{UNITS_PER_WORD} can have at run-time.
755 Width of a pointer, in bits. You must specify a value no wider than the
756 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
757 you must define @code{POINTERS_EXTEND_UNSIGNED}.
759 @findex POINTERS_EXTEND_UNSIGNED
760 @item POINTERS_EXTEND_UNSIGNED
761 A C expression whose value is nonzero if pointers that need to be
762 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
763 be zero-extended and zero if they are to be sign-extended.
765 You need not define this macro if the @code{POINTER_SIZE} is equal
766 to the width of @code{Pmode}.
769 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
770 A macro to update @var{m} and @var{unsignedp} when an object whose type
771 is @var{type} and which has the specified mode and signedness is to be
772 stored in a register. This macro is only called when @var{type} is a
775 On most RISC machines, which only have operations that operate on a full
776 register, define this macro to set @var{m} to @code{word_mode} if
777 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
778 cases, only integer modes should be widened because wider-precision
779 floating-point operations are usually more expensive than their narrower
782 For most machines, the macro definition does not change @var{unsignedp}.
783 However, some machines, have instructions that preferentially handle
784 either signed or unsigned quantities of certain modes. For example, on
785 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
786 sign-extend the result to 64 bits. On such machines, set
787 @var{unsignedp} according to which kind of extension is more efficient.
789 Do not define this macro if it would never modify @var{m}.
791 @findex PROMOTE_FUNCTION_ARGS
792 @item PROMOTE_FUNCTION_ARGS
793 Define this macro if the promotion described by @code{PROMOTE_MODE}
794 should also be done for outgoing function arguments.
796 @findex PROMOTE_FUNCTION_RETURN
797 @item PROMOTE_FUNCTION_RETURN
798 Define this macro if the promotion described by @code{PROMOTE_MODE}
799 should also be done for the return value of functions.
801 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
802 promotions done by @code{PROMOTE_MODE}.
804 @findex PROMOTE_FOR_CALL_ONLY
805 @item PROMOTE_FOR_CALL_ONLY
806 Define this macro if the promotion described by @code{PROMOTE_MODE}
807 should @emph{only} be performed for outgoing function arguments or
808 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
809 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
811 @findex PARM_BOUNDARY
813 Normal alignment required for function parameters on the stack, in
814 bits. All stack parameters receive at least this much alignment
815 regardless of data type. On most machines, this is the same as the
818 @findex STACK_BOUNDARY
820 Define this macro if there is a guaranteed alignment for the stack
821 pointer on this machine. The definition is a C expression
822 for the desired alignment (measured in bits). This value is used as a
823 default if PREFERRED_STACK_BOUNDARY is not defined.
825 @findex PREFERRED_STACK_BOUNDARY
826 @item PREFERRED_STACK_BOUNDARY
827 Define this macro if you wish to preserve a certain alignment for
828 the stack pointer. The definition is a C expression
829 for the desired alignment (measured in bits). If STACK_BOUNDARY is
830 also defined, this macro must evaluate to a value equal to or larger
833 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
834 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
835 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
836 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
837 be momentarily unaligned while pushing arguments.
839 @findex FUNCTION_BOUNDARY
840 @item FUNCTION_BOUNDARY
841 Alignment required for a function entry point, in bits.
843 @findex BIGGEST_ALIGNMENT
844 @item BIGGEST_ALIGNMENT
845 Biggest alignment that any data type can require on this machine, in bits.
847 @findex MINIMUM_ATOMIC_ALIGNMENT
848 @item MINIMUM_ATOMIC_ALIGNMENT
849 If defined, the smallest alignment, in bits, that can be given to an
850 object that can be referenced in one operation, without disturbing any
851 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
852 on machines that don't have byte or half-word store operations.
854 @findex BIGGEST_FIELD_ALIGNMENT
855 @item BIGGEST_FIELD_ALIGNMENT
856 Biggest alignment that any structure field can require on this machine,
857 in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
858 structure fields only.
860 @findex ADJUST_FIELD_ALIGN
861 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
862 An expression for the alignment of a structure field @var{field} if the
863 alignment computed in the usual way is @var{computed}. GCC uses
864 this value instead of the value in @code{BIGGEST_ALIGNMENT} or
865 @code{BIGGEST_FIELD_ALIGNMENT}, if defined, for structure fields only.
867 @findex MAX_OFILE_ALIGNMENT
868 @item MAX_OFILE_ALIGNMENT
869 Biggest alignment supported by the object file format of this machine.
870 Use this macro to limit the alignment which can be specified using the
871 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
872 the default value is @code{BIGGEST_ALIGNMENT}.
874 @findex DATA_ALIGNMENT
875 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
876 If defined, a C expression to compute the alignment for a variables in
877 the static store. @var{type} is the data type, and @var{basic-align} is
878 the alignment that the object would ordinarily have. The value of this
879 macro is used instead of that alignment to align the object.
881 If this macro is not defined, then @var{basic-align} is used.
884 One use of this macro is to increase alignment of medium-size data to
885 make it all fit in fewer cache lines. Another is to cause character
886 arrays to be word-aligned so that @code{strcpy} calls that copy
887 constants to character arrays can be done inline.
889 @findex CONSTANT_ALIGNMENT
890 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
891 If defined, a C expression to compute the alignment given to a constant
892 that is being placed in memory. @var{constant} is the constant and
893 @var{basic-align} is the alignment that the object would ordinarily
894 have. The value of this macro is used instead of that alignment to
897 If this macro is not defined, then @var{basic-align} is used.
899 The typical use of this macro is to increase alignment for string
900 constants to be word aligned so that @code{strcpy} calls that copy
901 constants can be done inline.
903 @findex LOCAL_ALIGNMENT
904 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
905 If defined, a C expression to compute the alignment for a variables in
906 the local store. @var{type} is the data type, and @var{basic-align} is
907 the alignment that the object would ordinarily have. The value of this
908 macro is used instead of that alignment to align the object.
910 If this macro is not defined, then @var{basic-align} is used.
912 One use of this macro is to increase alignment of medium-size data to
913 make it all fit in fewer cache lines.
915 @findex EMPTY_FIELD_BOUNDARY
916 @item EMPTY_FIELD_BOUNDARY
917 Alignment in bits to be given to a structure bit field that follows an
918 empty field such as @code{int : 0;}.
920 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
921 that results from an empty field.
923 @findex STRUCTURE_SIZE_BOUNDARY
924 @item STRUCTURE_SIZE_BOUNDARY
925 Number of bits which any structure or union's size must be a multiple of.
926 Each structure or union's size is rounded up to a multiple of this.
928 If you do not define this macro, the default is the same as
929 @code{BITS_PER_UNIT}.
931 @findex STRICT_ALIGNMENT
932 @item STRICT_ALIGNMENT
933 Define this macro to be the value 1 if instructions will fail to work
934 if given data not on the nominal alignment. If instructions will merely
935 go slower in that case, define this macro as 0.
937 @findex PCC_BITFIELD_TYPE_MATTERS
938 @item PCC_BITFIELD_TYPE_MATTERS
939 Define this if you wish to imitate the way many other C compilers handle
940 alignment of bitfields and the structures that contain them.
942 The behavior is that the type written for a bitfield (@code{int},
943 @code{short}, or other integer type) imposes an alignment for the
944 entire structure, as if the structure really did contain an ordinary
945 field of that type. In addition, the bitfield is placed within the
946 structure so that it would fit within such a field, not crossing a
949 Thus, on most machines, a bitfield whose type is written as @code{int}
950 would not cross a four-byte boundary, and would force four-byte
951 alignment for the whole structure. (The alignment used may not be four
952 bytes; it is controlled by the other alignment parameters.)
954 If the macro is defined, its definition should be a C expression;
955 a nonzero value for the expression enables this behavior.
957 Note that if this macro is not defined, or its value is zero, some
958 bitfields may cross more than one alignment boundary. The compiler can
959 support such references if there are @samp{insv}, @samp{extv}, and
960 @samp{extzv} insns that can directly reference memory.
962 The other known way of making bitfields work is to define
963 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
964 Then every structure can be accessed with fullwords.
966 Unless the machine has bitfield instructions or you define
967 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
968 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
970 If your aim is to make GCC use the same conventions for laying out
971 bitfields as are used by another compiler, here is how to investigate
972 what the other compiler does. Compile and run this program:
991 printf ("Size of foo1 is %d\n",
992 sizeof (struct foo1));
993 printf ("Size of foo2 is %d\n",
994 sizeof (struct foo2));
999 If this prints 2 and 5, then the compiler's behavior is what you would
1000 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1002 @findex BITFIELD_NBYTES_LIMITED
1003 @item BITFIELD_NBYTES_LIMITED
1004 Like PCC_BITFIELD_TYPE_MATTERS except that its effect is limited to
1005 aligning a bitfield within the structure.
1007 @findex ROUND_TYPE_SIZE
1008 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
1009 Define this macro as an expression for the overall size of a type
1010 (given by @var{type} as a tree node) when the size computed in the
1011 usual way is @var{computed} and the alignment is @var{specified}.
1013 The default is to round @var{computed} up to a multiple of @var{specified}.
1015 @findex ROUND_TYPE_SIZE_UNIT
1016 @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified})
1017 Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are
1018 specified in units (bytes). If you define @code{ROUND_TYPE_SIZE},
1019 you must also define this macro and they must be defined consistently
1022 @findex ROUND_TYPE_ALIGN
1023 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1024 Define this macro as an expression for the alignment of a type (given
1025 by @var{type} as a tree node) if the alignment computed in the usual
1026 way is @var{computed} and the alignment explicitly specified was
1029 The default is to use @var{specified} if it is larger; otherwise, use
1030 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1032 @findex MAX_FIXED_MODE_SIZE
1033 @item MAX_FIXED_MODE_SIZE
1034 An integer expression for the size in bits of the largest integer
1035 machine mode that should actually be used. All integer machine modes of
1036 this size or smaller can be used for structures and unions with the
1037 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1038 (DImode)} is assumed.
1040 @findex STACK_SAVEAREA_MODE
1041 @item STACK_SAVEAREA_MODE (@var{save_level})
1042 If defined, an expression of type @code{enum machine_mode} that
1043 specifies the mode of the save area operand of a
1044 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1045 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1046 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1047 having its mode specified.
1049 You need not define this macro if it always returns @code{Pmode}. You
1050 would most commonly define this macro if the
1051 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1054 @findex STACK_SIZE_MODE
1055 @item STACK_SIZE_MODE
1056 If defined, an expression of type @code{enum machine_mode} that
1057 specifies the mode of the size increment operand of an
1058 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1060 You need not define this macro if it always returns @code{word_mode}.
1061 You would most commonly define this macro if the @code{allocate_stack}
1062 pattern needs to support both a 32- and a 64-bit mode.
1064 @findex CHECK_FLOAT_VALUE
1065 @item CHECK_FLOAT_VALUE (@var{mode}, @var{value}, @var{overflow})
1066 A C statement to validate the value @var{value} (of type
1067 @code{double}) for mode @var{mode}. This means that you check whether
1068 @var{value} fits within the possible range of values for mode
1069 @var{mode} on this target machine. The mode @var{mode} is always
1070 a mode of class @code{MODE_FLOAT}. @var{overflow} is nonzero if
1071 the value is already known to be out of range.
1073 If @var{value} is not valid or if @var{overflow} is nonzero, you should
1074 set @var{overflow} to 1 and then assign some valid value to @var{value}.
1075 Allowing an invalid value to go through the compiler can produce
1076 incorrect assembler code which may even cause Unix assemblers to crash.
1078 This macro need not be defined if there is no work for it to do.
1080 @findex TARGET_FLOAT_FORMAT
1081 @item TARGET_FLOAT_FORMAT
1082 A code distinguishing the floating point format of the target machine.
1083 There are three defined values:
1086 @findex IEEE_FLOAT_FORMAT
1087 @item IEEE_FLOAT_FORMAT
1088 This code indicates IEEE floating point. It is the default; there is no
1089 need to define this macro when the format is IEEE.
1091 @findex VAX_FLOAT_FORMAT
1092 @item VAX_FLOAT_FORMAT
1093 This code indicates the peculiar format used on the Vax.
1095 @findex UNKNOWN_FLOAT_FORMAT
1096 @item UNKNOWN_FLOAT_FORMAT
1097 This code indicates any other format.
1100 The value of this macro is compared with @code{HOST_FLOAT_FORMAT}
1101 (@pxref{Config}) to determine whether the target machine has the same
1102 format as the host machine. If any other formats are actually in use on
1103 supported machines, new codes should be defined for them.
1105 The ordering of the component words of floating point values stored in
1106 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN} for the target
1107 machine and @code{HOST_FLOAT_WORDS_BIG_ENDIAN} for the host.
1109 @findex DEFAULT_VTABLE_THUNKS
1110 @item DEFAULT_VTABLE_THUNKS
1111 GCC supports two ways of implementing C++ vtables: traditional or with
1112 so-called ``thunks''. The flag @samp{-fvtable-thunk} chooses between them.
1113 Define this macro to be a C expression for the default value of that flag.
1114 If @code{DEFAULT_VTABLE_THUNKS} is 0, GCC uses the traditional
1115 implementation by default. The ``thunk'' implementation is more efficient
1116 (especially if you have provided an implementation of
1117 @code{ASM_OUTPUT_MI_THUNK}, see @ref{Function Entry}), but is not binary
1118 compatible with code compiled using the traditional implementation.
1119 If you are writing a new port, define @code{DEFAULT_VTABLE_THUNKS} to 1.
1121 If you do not define this macro, the default for @samp{-fvtable-thunk} is 0.
1125 @section Layout of Source Language Data Types
1127 These macros define the sizes and other characteristics of the standard
1128 basic data types used in programs being compiled. Unlike the macros in
1129 the previous section, these apply to specific features of C and related
1130 languages, rather than to fundamental aspects of storage layout.
1133 @findex INT_TYPE_SIZE
1135 A C expression for the size in bits of the type @code{int} on the
1136 target machine. If you don't define this, the default is one word.
1138 @findex MAX_INT_TYPE_SIZE
1139 @item MAX_INT_TYPE_SIZE
1140 Maximum number for the size in bits of the type @code{int} on the target
1141 machine. If this is undefined, the default is @code{INT_TYPE_SIZE}.
1142 Otherwise, it is the constant value that is the largest value that
1143 @code{INT_TYPE_SIZE} can have at run-time. This is used in @code{cpp}.
1145 @findex SHORT_TYPE_SIZE
1146 @item SHORT_TYPE_SIZE
1147 A C expression for the size in bits of the type @code{short} on the
1148 target machine. If you don't define this, the default is half a word.
1149 (If this would be less than one storage unit, it is rounded up to one
1152 @findex LONG_TYPE_SIZE
1153 @item LONG_TYPE_SIZE
1154 A C expression for the size in bits of the type @code{long} on the
1155 target machine. If you don't define this, the default is one word.
1157 @findex MAX_LONG_TYPE_SIZE
1158 @item MAX_LONG_TYPE_SIZE
1159 Maximum number for the size in bits of the type @code{long} on the
1160 target machine. If this is undefined, the default is
1161 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1162 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1165 @findex LONG_LONG_TYPE_SIZE
1166 @item LONG_LONG_TYPE_SIZE
1167 A C expression for the size in bits of the type @code{long long} on the
1168 target machine. If you don't define this, the default is two
1169 words. If you want to support GNU Ada on your machine, the value of
1170 macro must be at least 64.
1172 @findex CHAR_TYPE_SIZE
1173 @item CHAR_TYPE_SIZE
1174 A C expression for the size in bits of the type @code{char} on the
1175 target machine. If you don't define this, the default is
1176 @code{BITS_PER_UNIT}.
1178 @findex MAX_CHAR_TYPE_SIZE
1179 @item MAX_CHAR_TYPE_SIZE
1180 Maximum number for the size in bits of the type @code{char} on the
1181 target machine. If this is undefined, the default is
1182 @code{CHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1183 largest value that @code{CHAR_TYPE_SIZE} can have at run-time. This is
1186 @findex FLOAT_TYPE_SIZE
1187 @item FLOAT_TYPE_SIZE
1188 A C expression for the size in bits of the type @code{float} on the
1189 target machine. If you don't define this, the default is one word.
1191 @findex DOUBLE_TYPE_SIZE
1192 @item DOUBLE_TYPE_SIZE
1193 A C expression for the size in bits of the type @code{double} on the
1194 target machine. If you don't define this, the default is two
1197 @findex LONG_DOUBLE_TYPE_SIZE
1198 @item LONG_DOUBLE_TYPE_SIZE
1199 A C expression for the size in bits of the type @code{long double} on
1200 the target machine. If you don't define this, the default is two
1203 @findex WIDEST_HARDWARE_FP_SIZE
1204 @item WIDEST_HARDWARE_FP_SIZE
1205 A C expression for the size in bits of the widest floating-point format
1206 supported by the hardware. If you define this macro, you must specify a
1207 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1208 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1211 @findex DEFAULT_SIGNED_CHAR
1212 @item DEFAULT_SIGNED_CHAR
1213 An expression whose value is 1 or 0, according to whether the type
1214 @code{char} should be signed or unsigned by default. The user can
1215 always override this default with the options @samp{-fsigned-char}
1216 and @samp{-funsigned-char}.
1218 @findex DEFAULT_SHORT_ENUMS
1219 @item DEFAULT_SHORT_ENUMS
1220 A C expression to determine whether to give an @code{enum} type
1221 only as many bytes as it takes to represent the range of possible values
1222 of that type. A nonzero value means to do that; a zero value means all
1223 @code{enum} types should be allocated like @code{int}.
1225 If you don't define the macro, the default is 0.
1229 A C expression for a string describing the name of the data type to use
1230 for size values. The typedef name @code{size_t} is defined using the
1231 contents of the string.
1233 The string can contain more than one keyword. If so, separate them with
1234 spaces, and write first any length keyword, then @code{unsigned} if
1235 appropriate, and finally @code{int}. The string must exactly match one
1236 of the data type names defined in the function
1237 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1238 omit @code{int} or change the order---that would cause the compiler to
1241 If you don't define this macro, the default is @code{"long unsigned
1244 @findex PTRDIFF_TYPE
1246 A C expression for a string describing the name of the data type to use
1247 for the result of subtracting two pointers. The typedef name
1248 @code{ptrdiff_t} is defined using the contents of the string. See
1249 @code{SIZE_TYPE} above for more information.
1251 If you don't define this macro, the default is @code{"long int"}.
1255 A C expression for a string describing the name of the data type to use
1256 for wide characters. The typedef name @code{wchar_t} is defined using
1257 the contents of the string. See @code{SIZE_TYPE} above for more
1260 If you don't define this macro, the default is @code{"int"}.
1262 @findex WCHAR_TYPE_SIZE
1263 @item WCHAR_TYPE_SIZE
1264 A C expression for the size in bits of the data type for wide
1265 characters. This is used in @code{cpp}, which cannot make use of
1268 @findex MAX_WCHAR_TYPE_SIZE
1269 @item MAX_WCHAR_TYPE_SIZE
1270 Maximum number for the size in bits of the data type for wide
1271 characters. If this is undefined, the default is
1272 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1273 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1276 @findex OBJC_INT_SELECTORS
1277 @item OBJC_INT_SELECTORS
1278 Define this macro if the type of Objective C selectors should be
1281 If this macro is not defined, then selectors should have the type
1282 @code{struct objc_selector *}.
1284 @findex OBJC_SELECTORS_WITHOUT_LABELS
1285 @item OBJC_SELECTORS_WITHOUT_LABELS
1286 Define this macro if the compiler can group all the selectors together
1287 into a vector and use just one label at the beginning of the vector.
1288 Otherwise, the compiler must give each selector its own assembler
1291 On certain machines, it is important to have a separate label for each
1292 selector because this enables the linker to eliminate duplicate selectors.
1296 A C constant expression for the integer value for escape sequence
1301 @findex TARGET_NEWLINE
1304 @itemx TARGET_NEWLINE
1305 C constant expressions for the integer values for escape sequences
1306 @samp{\b}, @samp{\t} and @samp{\n}.
1314 C constant expressions for the integer values for escape sequences
1315 @samp{\v}, @samp{\f} and @samp{\r}.
1319 @section Register Usage
1320 @cindex register usage
1322 This section explains how to describe what registers the target machine
1323 has, and how (in general) they can be used.
1325 The description of which registers a specific instruction can use is
1326 done with register classes; see @ref{Register Classes}. For information
1327 on using registers to access a stack frame, see @ref{Frame Registers}.
1328 For passing values in registers, see @ref{Register Arguments}.
1329 For returning values in registers, see @ref{Scalar Return}.
1332 * Register Basics:: Number and kinds of registers.
1333 * Allocation Order:: Order in which registers are allocated.
1334 * Values in Registers:: What kinds of values each reg can hold.
1335 * Leaf Functions:: Renumbering registers for leaf functions.
1336 * Stack Registers:: Handling a register stack such as 80387.
1339 @node Register Basics
1340 @subsection Basic Characteristics of Registers
1342 @c prevent bad page break with this line
1343 Registers have various characteristics.
1346 @findex FIRST_PSEUDO_REGISTER
1347 @item FIRST_PSEUDO_REGISTER
1348 Number of hardware registers known to the compiler. They receive
1349 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1350 pseudo register's number really is assigned the number
1351 @code{FIRST_PSEUDO_REGISTER}.
1353 @item FIXED_REGISTERS
1354 @findex FIXED_REGISTERS
1355 @cindex fixed register
1356 An initializer that says which registers are used for fixed purposes
1357 all throughout the compiled code and are therefore not available for
1358 general allocation. These would include the stack pointer, the frame
1359 pointer (except on machines where that can be used as a general
1360 register when no frame pointer is needed), the program counter on
1361 machines where that is considered one of the addressable registers,
1362 and any other numbered register with a standard use.
1364 This information is expressed as a sequence of numbers, separated by
1365 commas and surrounded by braces. The @var{n}th number is 1 if
1366 register @var{n} is fixed, 0 otherwise.
1368 The table initialized from this macro, and the table initialized by
1369 the following one, may be overridden at run time either automatically,
1370 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1371 the user with the command options @samp{-ffixed-@var{reg}},
1372 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}}.
1374 @findex CALL_USED_REGISTERS
1375 @item CALL_USED_REGISTERS
1376 @cindex call-used register
1377 @cindex call-clobbered register
1378 @cindex call-saved register
1379 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1380 clobbered (in general) by function calls as well as for fixed
1381 registers. This macro therefore identifies the registers that are not
1382 available for general allocation of values that must live across
1385 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1386 automatically saves it on function entry and restores it on function
1387 exit, if the register is used within the function.
1389 @findex HARD_REGNO_CALL_PART_CLOBBERED
1390 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1391 @cindex call-used register
1392 @cindex call-clobbered register
1393 @cindex call-saved register
1394 A C expression that is non-zero if it is not permissible to store a
1395 value of mode @var{mode} in hard register number @var{regno} across a
1396 call without some part of it being clobbered. For most machines this
1397 macro need not be defined. It is only required for machines that do not
1398 preserve the entire contents of a register across a call.
1400 @findex CONDITIONAL_REGISTER_USAGE
1402 @findex call_used_regs
1403 @item CONDITIONAL_REGISTER_USAGE
1404 Zero or more C statements that may conditionally modify four variables
1405 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs}
1406 (these three are of type @code{char []}) and @code{reg_class_contents}
1407 (of type @code{HARD_REG_SET}).
1408 Before the macro is called @code{fixed_regs}, @code{call_used_regs}
1409 and @code{reg_class_contents} have been initialized from
1410 @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS} and
1411 @code{REG_CLASS_CONTENTS}, respectively,
1412 @code{global_regs} has been cleared, and any @samp{-ffixed-@var{reg}},
1413 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}} command
1414 options have been applied.
1416 This is necessary in case the fixed or call-clobbered registers depend
1419 You need not define this macro if it has no work to do.
1421 @cindex disabling certain registers
1422 @cindex controlling register usage
1423 If the usage of an entire class of registers depends on the target
1424 flags, you may indicate this to GCC by using this macro to modify
1425 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1426 registers in the classes which should not be used by GCC. Also define
1427 the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1428 is called with a letter for a class that shouldn't be used.
1430 (However, if this class is not included in @code{GENERAL_REGS} and all
1431 of the insn patterns whose constraints permit this class are
1432 controlled by target switches, then GCC will automatically avoid using
1433 these registers when the target switches are opposed to them.)
1435 @findex NON_SAVING_SETJMP
1436 @item NON_SAVING_SETJMP
1437 If this macro is defined and has a nonzero value, it means that
1438 @code{setjmp} and related functions fail to save the registers, or that
1439 @code{longjmp} fails to restore them. To compensate, the compiler
1440 avoids putting variables in registers in functions that use
1443 @findex INCOMING_REGNO
1444 @item INCOMING_REGNO (@var{out})
1445 Define this macro if the target machine has register windows. This C
1446 expression returns the register number as seen by the called function
1447 corresponding to the register number @var{out} as seen by the calling
1448 function. Return @var{out} if register number @var{out} is not an
1451 @findex OUTGOING_REGNO
1452 @item OUTGOING_REGNO (@var{in})
1453 Define this macro if the target machine has register windows. This C
1454 expression returns the register number as seen by the calling function
1455 corresponding to the register number @var{in} as seen by the called
1456 function. Return @var{in} if register number @var{in} is not an inbound
1462 If the program counter has a register number, define this as that
1463 register number. Otherwise, do not define it.
1467 @node Allocation Order
1468 @subsection Order of Allocation of Registers
1469 @cindex order of register allocation
1470 @cindex register allocation order
1472 @c prevent bad page break with this line
1473 Registers are allocated in order.
1476 @findex REG_ALLOC_ORDER
1477 @item REG_ALLOC_ORDER
1478 If defined, an initializer for a vector of integers, containing the
1479 numbers of hard registers in the order in which GCC should prefer
1480 to use them (from most preferred to least).
1482 If this macro is not defined, registers are used lowest numbered first
1483 (all else being equal).
1485 One use of this macro is on machines where the highest numbered
1486 registers must always be saved and the save-multiple-registers
1487 instruction supports only sequences of consecutive registers. On such
1488 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1489 the highest numbered allocable register first.
1491 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1492 @item ORDER_REGS_FOR_LOCAL_ALLOC
1493 A C statement (sans semicolon) to choose the order in which to allocate
1494 hard registers for pseudo-registers local to a basic block.
1496 Store the desired register order in the array @code{reg_alloc_order}.
1497 Element 0 should be the register to allocate first; element 1, the next
1498 register; and so on.
1500 The macro body should not assume anything about the contents of
1501 @code{reg_alloc_order} before execution of the macro.
1503 On most machines, it is not necessary to define this macro.
1506 @node Values in Registers
1507 @subsection How Values Fit in Registers
1509 This section discusses the macros that describe which kinds of values
1510 (specifically, which machine modes) each register can hold, and how many
1511 consecutive registers are needed for a given mode.
1514 @findex HARD_REGNO_NREGS
1515 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1516 A C expression for the number of consecutive hard registers, starting
1517 at register number @var{regno}, required to hold a value of mode
1520 On a machine where all registers are exactly one word, a suitable
1521 definition of this macro is
1524 #define HARD_REGNO_NREGS(REGNO, MODE) \
1525 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1529 @findex ALTER_HARD_SUBREG
1530 @item ALTER_HARD_SUBREG (@var{tgt_mode}, @var{word}, @var{src_mode}, @var{regno})
1531 A C expression that returns an adjusted hard register number for
1534 (subreg:@var{tgt_mode} (reg:@var{src_mode} @var{regno}) @var{word})
1537 This may be needed if the target machine has mixed sized big-endian
1538 registers, like Sparc v9.
1540 @findex HARD_REGNO_MODE_OK
1541 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1542 A C expression that is nonzero if it is permissible to store a value
1543 of mode @var{mode} in hard register number @var{regno} (or in several
1544 registers starting with that one). For a machine where all registers
1545 are equivalent, a suitable definition is
1548 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1551 You need not include code to check for the numbers of fixed registers,
1552 because the allocation mechanism considers them to be always occupied.
1554 @cindex register pairs
1555 On some machines, double-precision values must be kept in even/odd
1556 register pairs. You can implement that by defining this macro to reject
1557 odd register numbers for such modes.
1559 The minimum requirement for a mode to be OK in a register is that the
1560 @samp{mov@var{mode}} instruction pattern support moves between the
1561 register and other hard register in the same class and that moving a
1562 value into the register and back out not alter it.
1564 Since the same instruction used to move @code{word_mode} will work for
1565 all narrower integer modes, it is not necessary on any machine for
1566 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1567 you define patterns @samp{movhi}, etc., to take advantage of this. This
1568 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1569 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1572 Many machines have special registers for floating point arithmetic.
1573 Often people assume that floating point machine modes are allowed only
1574 in floating point registers. This is not true. Any registers that
1575 can hold integers can safely @emph{hold} a floating point machine
1576 mode, whether or not floating arithmetic can be done on it in those
1577 registers. Integer move instructions can be used to move the values.
1579 On some machines, though, the converse is true: fixed-point machine
1580 modes may not go in floating registers. This is true if the floating
1581 registers normalize any value stored in them, because storing a
1582 non-floating value there would garble it. In this case,
1583 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1584 floating registers. But if the floating registers do not automatically
1585 normalize, if you can store any bit pattern in one and retrieve it
1586 unchanged without a trap, then any machine mode may go in a floating
1587 register, so you can define this macro to say so.
1589 The primary significance of special floating registers is rather that
1590 they are the registers acceptable in floating point arithmetic
1591 instructions. However, this is of no concern to
1592 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1593 constraints for those instructions.
1595 On some machines, the floating registers are especially slow to access,
1596 so that it is better to store a value in a stack frame than in such a
1597 register if floating point arithmetic is not being done. As long as the
1598 floating registers are not in class @code{GENERAL_REGS}, they will not
1599 be used unless some pattern's constraint asks for one.
1601 @findex MODES_TIEABLE_P
1602 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
1603 A C expression that is nonzero if a value of mode
1604 @var{mode1} is accessible in mode @var{mode2} without copying.
1606 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
1607 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
1608 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
1609 should be nonzero. If they differ for any @var{r}, you should define
1610 this macro to return zero unless some other mechanism ensures the
1611 accessibility of the value in a narrower mode.
1613 You should define this macro to return nonzero in as many cases as
1614 possible since doing so will allow GCC to perform better register
1617 @findex AVOID_CCMODE_COPIES
1618 @item AVOID_CCMODE_COPIES
1619 Define this macro if the compiler should avoid copies to/from @code{CCmode}
1620 registers. You should only define this macro if support for copying to/from
1621 @code{CCmode} is incomplete.
1624 @node Leaf Functions
1625 @subsection Handling Leaf Functions
1627 @cindex leaf functions
1628 @cindex functions, leaf
1629 On some machines, a leaf function (i.e., one which makes no calls) can run
1630 more efficiently if it does not make its own register window. Often this
1631 means it is required to receive its arguments in the registers where they
1632 are passed by the caller, instead of the registers where they would
1635 The special treatment for leaf functions generally applies only when
1636 other conditions are met; for example, often they may use only those
1637 registers for its own variables and temporaries. We use the term ``leaf
1638 function'' to mean a function that is suitable for this special
1639 handling, so that functions with no calls are not necessarily ``leaf
1642 GCC assigns register numbers before it knows whether the function is
1643 suitable for leaf function treatment. So it needs to renumber the
1644 registers in order to output a leaf function. The following macros
1648 @findex LEAF_REGISTERS
1649 @item LEAF_REGISTERS
1650 A C initializer for a vector, indexed by hard register number, which
1651 contains 1 for a register that is allowable in a candidate for leaf
1654 If leaf function treatment involves renumbering the registers, then the
1655 registers marked here should be the ones before renumbering---those that
1656 GCC would ordinarily allocate. The registers which will actually be
1657 used in the assembler code, after renumbering, should not be marked with 1
1660 Define this macro only if the target machine offers a way to optimize
1661 the treatment of leaf functions.
1663 @findex LEAF_REG_REMAP
1664 @item LEAF_REG_REMAP (@var{regno})
1665 A C expression whose value is the register number to which @var{regno}
1666 should be renumbered, when a function is treated as a leaf function.
1668 If @var{regno} is a register number which should not appear in a leaf
1669 function before renumbering, then the expression should yield -1, which
1670 will cause the compiler to abort.
1672 Define this macro only if the target machine offers a way to optimize the
1673 treatment of leaf functions, and registers need to be renumbered to do
1677 @findex current_function_is_leaf
1678 @findex current_function_uses_only_leaf_regs
1679 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
1680 treat leaf functions specially. They can test the C variable
1681 @code{current_function_is_leaf} which is nonzero for leaf functions.
1682 @code{current_function_is_leaf} is set prior to local register allocation
1683 and is valid for the remaining compiler passes. They can also test the C
1684 variable @code{current_function_uses_only_leaf_regs} which is nonzero for
1685 leaf functions which only use leaf registers.
1686 @code{current_function_uses_only_leaf_regs} is valid after reload and is
1687 only useful if @code{LEAF_REGISTERS} is defined.
1688 @c changed this to fix overfull. ALSO: why the "it" at the beginning
1689 @c of the next paragraph?! --mew 2feb93
1691 @node Stack Registers
1692 @subsection Registers That Form a Stack
1694 There are special features to handle computers where some of the
1695 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
1696 Stack registers are normally written by pushing onto the stack, and are
1697 numbered relative to the top of the stack.
1699 Currently, GCC can only handle one group of stack-like registers, and
1700 they must be consecutively numbered.
1705 Define this if the machine has any stack-like registers.
1707 @findex FIRST_STACK_REG
1708 @item FIRST_STACK_REG
1709 The number of the first stack-like register. This one is the top
1712 @findex LAST_STACK_REG
1713 @item LAST_STACK_REG
1714 The number of the last stack-like register. This one is the bottom of
1718 @node Register Classes
1719 @section Register Classes
1720 @cindex register class definitions
1721 @cindex class definitions, register
1723 On many machines, the numbered registers are not all equivalent.
1724 For example, certain registers may not be allowed for indexed addressing;
1725 certain registers may not be allowed in some instructions. These machine
1726 restrictions are described to the compiler using @dfn{register classes}.
1728 You define a number of register classes, giving each one a name and saying
1729 which of the registers belong to it. Then you can specify register classes
1730 that are allowed as operands to particular instruction patterns.
1734 In general, each register will belong to several classes. In fact, one
1735 class must be named @code{ALL_REGS} and contain all the registers. Another
1736 class must be named @code{NO_REGS} and contain no registers. Often the
1737 union of two classes will be another class; however, this is not required.
1739 @findex GENERAL_REGS
1740 One of the classes must be named @code{GENERAL_REGS}. There is nothing
1741 terribly special about the name, but the operand constraint letters
1742 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
1743 the same as @code{ALL_REGS}, just define it as a macro which expands
1746 Order the classes so that if class @var{x} is contained in class @var{y}
1747 then @var{x} has a lower class number than @var{y}.
1749 The way classes other than @code{GENERAL_REGS} are specified in operand
1750 constraints is through machine-dependent operand constraint letters.
1751 You can define such letters to correspond to various classes, then use
1752 them in operand constraints.
1754 You should define a class for the union of two classes whenever some
1755 instruction allows both classes. For example, if an instruction allows
1756 either a floating point (coprocessor) register or a general register for a
1757 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
1758 which includes both of them. Otherwise you will get suboptimal code.
1760 You must also specify certain redundant information about the register
1761 classes: for each class, which classes contain it and which ones are
1762 contained in it; for each pair of classes, the largest class contained
1765 When a value occupying several consecutive registers is expected in a
1766 certain class, all the registers used must belong to that class.
1767 Therefore, register classes cannot be used to enforce a requirement for
1768 a register pair to start with an even-numbered register. The way to
1769 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
1771 Register classes used for input-operands of bitwise-and or shift
1772 instructions have a special requirement: each such class must have, for
1773 each fixed-point machine mode, a subclass whose registers can transfer that
1774 mode to or from memory. For example, on some machines, the operations for
1775 single-byte values (@code{QImode}) are limited to certain registers. When
1776 this is so, each register class that is used in a bitwise-and or shift
1777 instruction must have a subclass consisting of registers from which
1778 single-byte values can be loaded or stored. This is so that
1779 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
1782 @findex enum reg_class
1783 @item enum reg_class
1784 An enumeral type that must be defined with all the register class names
1785 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
1786 must be the last register class, followed by one more enumeral value,
1787 @code{LIM_REG_CLASSES}, which is not a register class but rather
1788 tells how many classes there are.
1790 Each register class has a number, which is the value of casting
1791 the class name to type @code{int}. The number serves as an index
1792 in many of the tables described below.
1794 @findex N_REG_CLASSES
1796 The number of distinct register classes, defined as follows:
1799 #define N_REG_CLASSES (int) LIM_REG_CLASSES
1802 @findex REG_CLASS_NAMES
1803 @item REG_CLASS_NAMES
1804 An initializer containing the names of the register classes as C string
1805 constants. These names are used in writing some of the debugging dumps.
1807 @findex REG_CLASS_CONTENTS
1808 @item REG_CLASS_CONTENTS
1809 An initializer containing the contents of the register classes, as integers
1810 which are bit masks. The @var{n}th integer specifies the contents of class
1811 @var{n}. The way the integer @var{mask} is interpreted is that
1812 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
1814 When the machine has more than 32 registers, an integer does not suffice.
1815 Then the integers are replaced by sub-initializers, braced groupings containing
1816 several integers. Each sub-initializer must be suitable as an initializer
1817 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
1819 @findex REGNO_REG_CLASS
1820 @item REGNO_REG_CLASS (@var{regno})
1821 A C expression whose value is a register class containing hard register
1822 @var{regno}. In general there is more than one such class; choose a class
1823 which is @dfn{minimal}, meaning that no smaller class also contains the
1826 @findex BASE_REG_CLASS
1827 @item BASE_REG_CLASS
1828 A macro whose definition is the name of the class to which a valid
1829 base register must belong. A base register is one used in an address
1830 which is the register value plus a displacement.
1832 @findex INDEX_REG_CLASS
1833 @item INDEX_REG_CLASS
1834 A macro whose definition is the name of the class to which a valid
1835 index register must belong. An index register is one used in an
1836 address where its value is either multiplied by a scale factor or
1837 added to another register (as well as added to a displacement).
1839 @findex REG_CLASS_FROM_LETTER
1840 @item REG_CLASS_FROM_LETTER (@var{char})
1841 A C expression which defines the machine-dependent operand constraint
1842 letters for register classes. If @var{char} is such a letter, the
1843 value should be the register class corresponding to it. Otherwise,
1844 the value should be @code{NO_REGS}. The register letter @samp{r},
1845 corresponding to class @code{GENERAL_REGS}, will not be passed
1846 to this macro; you do not need to handle it.
1848 @findex REGNO_OK_FOR_BASE_P
1849 @item REGNO_OK_FOR_BASE_P (@var{num})
1850 A C expression which is nonzero if register number @var{num} is
1851 suitable for use as a base register in operand addresses. It may be
1852 either a suitable hard register or a pseudo register that has been
1853 allocated such a hard register.
1855 @findex REGNO_MODE_OK_FOR_BASE_P
1856 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
1857 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
1858 that expression may examine the mode of the memory reference in
1859 @var{mode}. You should define this macro if the mode of the memory
1860 reference affects whether a register may be used as a base register. If
1861 you define this macro, the compiler will use it instead of
1862 @code{REGNO_OK_FOR_BASE_P}.
1864 @findex REGNO_OK_FOR_INDEX_P
1865 @item REGNO_OK_FOR_INDEX_P (@var{num})
1866 A C expression which is nonzero if register number @var{num} is
1867 suitable for use as an index register in operand addresses. It may be
1868 either a suitable hard register or a pseudo register that has been
1869 allocated such a hard register.
1871 The difference between an index register and a base register is that
1872 the index register may be scaled. If an address involves the sum of
1873 two registers, neither one of them scaled, then either one may be
1874 labeled the ``base'' and the other the ``index''; but whichever
1875 labeling is used must fit the machine's constraints of which registers
1876 may serve in each capacity. The compiler will try both labelings,
1877 looking for one that is valid, and will reload one or both registers
1878 only if neither labeling works.
1880 @findex PREFERRED_RELOAD_CLASS
1881 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
1882 A C expression that places additional restrictions on the register class
1883 to use when it is necessary to copy value @var{x} into a register in class
1884 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
1885 another, smaller class. On many machines, the following definition is
1889 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
1892 Sometimes returning a more restrictive class makes better code. For
1893 example, on the 68000, when @var{x} is an integer constant that is in range
1894 for a @samp{moveq} instruction, the value of this macro is always
1895 @code{DATA_REGS} as long as @var{class} includes the data registers.
1896 Requiring a data register guarantees that a @samp{moveq} will be used.
1898 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
1899 you can force @var{x} into a memory constant. This is useful on
1900 certain machines where immediate floating values cannot be loaded into
1901 certain kinds of registers.
1903 @findex PREFERRED_OUTPUT_RELOAD_CLASS
1904 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
1905 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
1906 input reloads. If you don't define this macro, the default is to use
1907 @var{class}, unchanged.
1909 @findex LIMIT_RELOAD_CLASS
1910 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
1911 A C expression that places additional restrictions on the register class
1912 to use when it is necessary to be able to hold a value of mode
1913 @var{mode} in a reload register for which class @var{class} would
1916 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
1917 there are certain modes that simply can't go in certain reload classes.
1919 The value is a register class; perhaps @var{class}, or perhaps another,
1922 Don't define this macro unless the target machine has limitations which
1923 require the macro to do something nontrivial.
1925 @findex SECONDARY_RELOAD_CLASS
1926 @findex SECONDARY_INPUT_RELOAD_CLASS
1927 @findex SECONDARY_OUTPUT_RELOAD_CLASS
1928 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1929 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1930 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1931 Many machines have some registers that cannot be copied directly to or
1932 from memory or even from other types of registers. An example is the
1933 @samp{MQ} register, which on most machines, can only be copied to or
1934 from general registers, but not memory. Some machines allow copying all
1935 registers to and from memory, but require a scratch register for stores
1936 to some memory locations (e.g., those with symbolic address on the RT,
1937 and those with certain symbolic address on the Sparc when compiling
1938 PIC). In some cases, both an intermediate and a scratch register are
1941 You should define these macros to indicate to the reload phase that it may
1942 need to allocate at least one register for a reload in addition to the
1943 register to contain the data. Specifically, if copying @var{x} to a
1944 register @var{class} in @var{mode} requires an intermediate register,
1945 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
1946 largest register class all of whose registers can be used as
1947 intermediate registers or scratch registers.
1949 If copying a register @var{class} in @var{mode} to @var{x} requires an
1950 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
1951 should be defined to return the largest register class required. If the
1952 requirements for input and output reloads are the same, the macro
1953 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
1956 The values returned by these macros are often @code{GENERAL_REGS}.
1957 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
1958 can be directly copied to or from a register of @var{class} in
1959 @var{mode} without requiring a scratch register. Do not define this
1960 macro if it would always return @code{NO_REGS}.
1962 If a scratch register is required (either with or without an
1963 intermediate register), you should define patterns for
1964 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
1965 (@pxref{Standard Names}. These patterns, which will normally be
1966 implemented with a @code{define_expand}, should be similar to the
1967 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
1970 Define constraints for the reload register and scratch register that
1971 contain a single register class. If the original reload register (whose
1972 class is @var{class}) can meet the constraint given in the pattern, the
1973 value returned by these macros is used for the class of the scratch
1974 register. Otherwise, two additional reload registers are required.
1975 Their classes are obtained from the constraints in the insn pattern.
1977 @var{x} might be a pseudo-register or a @code{subreg} of a
1978 pseudo-register, which could either be in a hard register or in memory.
1979 Use @code{true_regnum} to find out; it will return -1 if the pseudo is
1980 in memory and the hard register number if it is in a register.
1982 These macros should not be used in the case where a particular class of
1983 registers can only be copied to memory and not to another class of
1984 registers. In that case, secondary reload registers are not needed and
1985 would not be helpful. Instead, a stack location must be used to perform
1986 the copy and the @code{mov@var{m}} pattern should use memory as a
1987 intermediate storage. This case often occurs between floating-point and
1990 @findex SECONDARY_MEMORY_NEEDED
1991 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
1992 Certain machines have the property that some registers cannot be copied
1993 to some other registers without using memory. Define this macro on
1994 those machines to be a C expression that is non-zero if objects of mode
1995 @var{m} in registers of @var{class1} can only be copied to registers of
1996 class @var{class2} by storing a register of @var{class1} into memory
1997 and loading that memory location into a register of @var{class2}.
1999 Do not define this macro if its value would always be zero.
2001 @findex SECONDARY_MEMORY_NEEDED_RTX
2002 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2003 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2004 allocates a stack slot for a memory location needed for register copies.
2005 If this macro is defined, the compiler instead uses the memory location
2006 defined by this macro.
2008 Do not define this macro if you do not define
2009 @code{SECONDARY_MEMORY_NEEDED}.
2011 @findex SECONDARY_MEMORY_NEEDED_MODE
2012 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2013 When the compiler needs a secondary memory location to copy between two
2014 registers of mode @var{mode}, it normally allocates sufficient memory to
2015 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2016 load operations in a mode that many bits wide and whose class is the
2017 same as that of @var{mode}.
2019 This is right thing to do on most machines because it ensures that all
2020 bits of the register are copied and prevents accesses to the registers
2021 in a narrower mode, which some machines prohibit for floating-point
2024 However, this default behavior is not correct on some machines, such as
2025 the DEC Alpha, that store short integers in floating-point registers
2026 differently than in integer registers. On those machines, the default
2027 widening will not work correctly and you must define this macro to
2028 suppress that widening in some cases. See the file @file{alpha.h} for
2031 Do not define this macro if you do not define
2032 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2033 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2035 @findex SMALL_REGISTER_CLASSES
2036 @item SMALL_REGISTER_CLASSES
2037 On some machines, it is risky to let hard registers live across arbitrary
2038 insns. Typically, these machines have instructions that require values
2039 to be in specific registers (like an accumulator), and reload will fail
2040 if the required hard register is used for another purpose across such an
2043 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a non-zero
2044 value on these machines. When this macro has a non-zero value, the
2045 compiler will try to minimize the lifetime of hard registers.
2047 It is always safe to define this macro with a non-zero value, but if you
2048 unnecessarily define it, you will reduce the amount of optimizations
2049 that can be performed in some cases. If you do not define this macro
2050 with a non-zero value when it is required, the compiler will run out of
2051 spill registers and print a fatal error message. For most machines, you
2052 should not define this macro at all.
2054 @findex CLASS_LIKELY_SPILLED_P
2055 @item CLASS_LIKELY_SPILLED_P (@var{class})
2056 A C expression whose value is nonzero if pseudos that have been assigned
2057 to registers of class @var{class} would likely be spilled because
2058 registers of @var{class} are needed for spill registers.
2060 The default value of this macro returns 1 if @var{class} has exactly one
2061 register and zero otherwise. On most machines, this default should be
2062 used. Only define this macro to some other expression if pseudos
2063 allocated by @file{local-alloc.c} end up in memory because their hard
2064 registers were needed for spill registers. If this macro returns nonzero
2065 for those classes, those pseudos will only be allocated by
2066 @file{global.c}, which knows how to reallocate the pseudo to another
2067 register. If there would not be another register available for
2068 reallocation, you should not change the definition of this macro since
2069 the only effect of such a definition would be to slow down register
2072 @findex CLASS_MAX_NREGS
2073 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2074 A C expression for the maximum number of consecutive registers
2075 of class @var{class} needed to hold a value of mode @var{mode}.
2077 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2078 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2079 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2080 @var{mode})} for all @var{regno} values in the class @var{class}.
2082 This macro helps control the handling of multiple-word values
2085 @item CLASS_CANNOT_CHANGE_SIZE
2086 If defined, a C expression for a class that contains registers which the
2087 compiler must always access in a mode that is the same size as the mode
2088 in which it loaded the register.
2090 For the example, loading 32-bit integer or floating-point objects into
2091 floating-point registers on the Alpha extends them to 64-bits.
2092 Therefore loading a 64-bit object and then storing it as a 32-bit object
2093 does not store the low-order 32-bits, as would be the case for a normal
2094 register. Therefore, @file{alpha.h} defines this macro as
2098 Three other special macros describe which operands fit which constraint
2102 @findex CONST_OK_FOR_LETTER_P
2103 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2104 A C expression that defines the machine-dependent operand constraint
2105 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2106 particular ranges of integer values. If @var{c} is one of those
2107 letters, the expression should check that @var{value}, an integer, is in
2108 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2109 not one of those letters, the value should be 0 regardless of
2112 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2113 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2114 A C expression that defines the machine-dependent operand constraint
2115 letters that specify particular ranges of @code{const_double} values
2116 (@samp{G} or @samp{H}).
2118 If @var{c} is one of those letters, the expression should check that
2119 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2120 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2121 letters, the value should be 0 regardless of @var{value}.
2123 @code{const_double} is used for all floating-point constants and for
2124 @code{DImode} fixed-point constants. A given letter can accept either
2125 or both kinds of values. It can use @code{GET_MODE} to distinguish
2126 between these kinds.
2128 @findex EXTRA_CONSTRAINT
2129 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2130 A C expression that defines the optional machine-dependent constraint
2131 letters (@samp{Q}, @samp{R}, @samp{S}, @samp{T}, @samp{U}) that can
2132 be used to segregate specific types of operands, usually memory
2133 references, for the target machine. Normally this macro will not be
2134 defined. If it is required for a particular target machine, it should
2135 return 1 if @var{value} corresponds to the operand type represented by
2136 the constraint letter @var{c}. If @var{c} is not defined as an extra
2137 constraint, the value returned should be 0 regardless of @var{value}.
2139 For example, on the ROMP, load instructions cannot have their output in r0 if
2140 the memory reference contains a symbolic address. Constraint letter
2141 @samp{Q} is defined as representing a memory address that does
2142 @emph{not} contain a symbolic address. An alternative is specified with
2143 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2144 alternative specifies @samp{m} on the input and a register class that
2145 does not include r0 on the output.
2148 @node Stack and Calling
2149 @section Stack Layout and Calling Conventions
2150 @cindex calling conventions
2152 @c prevent bad page break with this line
2153 This describes the stack layout and calling conventions.
2161 * Register Arguments::
2163 * Aggregate Return::
2170 @subsection Basic Stack Layout
2171 @cindex stack frame layout
2172 @cindex frame layout
2174 @c prevent bad page break with this line
2175 Here is the basic stack layout.
2178 @findex STACK_GROWS_DOWNWARD
2179 @item STACK_GROWS_DOWNWARD
2180 Define this macro if pushing a word onto the stack moves the stack
2181 pointer to a smaller address.
2183 When we say, ``define this macro if @dots{},'' it means that the
2184 compiler checks this macro only with @code{#ifdef} so the precise
2185 definition used does not matter.
2187 @findex FRAME_GROWS_DOWNWARD
2188 @item FRAME_GROWS_DOWNWARD
2189 Define this macro if the addresses of local variable slots are at negative
2190 offsets from the frame pointer.
2192 @findex ARGS_GROW_DOWNWARD
2193 @item ARGS_GROW_DOWNWARD
2194 Define this macro if successive arguments to a function occupy decreasing
2195 addresses on the stack.
2197 @findex STARTING_FRAME_OFFSET
2198 @item STARTING_FRAME_OFFSET
2199 Offset from the frame pointer to the first local variable slot to be allocated.
2201 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2202 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2203 Otherwise, it is found by adding the length of the first slot to the
2204 value @code{STARTING_FRAME_OFFSET}.
2205 @c i'm not sure if the above is still correct.. had to change it to get
2206 @c rid of an overfull. --mew 2feb93
2208 @findex STACK_POINTER_OFFSET
2209 @item STACK_POINTER_OFFSET
2210 Offset from the stack pointer register to the first location at which
2211 outgoing arguments are placed. If not specified, the default value of
2212 zero is used. This is the proper value for most machines.
2214 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2215 the first location at which outgoing arguments are placed.
2217 @findex FIRST_PARM_OFFSET
2218 @item FIRST_PARM_OFFSET (@var{fundecl})
2219 Offset from the argument pointer register to the first argument's
2220 address. On some machines it may depend on the data type of the
2223 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2224 the first argument's address.
2226 @findex STACK_DYNAMIC_OFFSET
2227 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2228 Offset from the stack pointer register to an item dynamically allocated
2229 on the stack, e.g., by @code{alloca}.
2231 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2232 length of the outgoing arguments. The default is correct for most
2233 machines. See @file{function.c} for details.
2235 @findex DYNAMIC_CHAIN_ADDRESS
2236 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2237 A C expression whose value is RTL representing the address in a stack
2238 frame where the pointer to the caller's frame is stored. Assume that
2239 @var{frameaddr} is an RTL expression for the address of the stack frame
2242 If you don't define this macro, the default is to return the value
2243 of @var{frameaddr}---that is, the stack frame address is also the
2244 address of the stack word that points to the previous frame.
2246 @findex SETUP_FRAME_ADDRESSES
2247 @item SETUP_FRAME_ADDRESSES
2248 If defined, a C expression that produces the machine-specific code to
2249 setup the stack so that arbitrary frames can be accessed. For example,
2250 on the Sparc, we must flush all of the register windows to the stack
2251 before we can access arbitrary stack frames. You will seldom need to
2254 @findex BUILTIN_SETJMP_FRAME_VALUE
2255 @item BUILTIN_SETJMP_FRAME_VALUE
2256 If defined, a C expression that contains an rtx that is used to store
2257 the address of the current frame into the built in @code{setjmp} buffer.
2258 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2259 machines. One reason you may need to define this macro is if
2260 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2262 @findex RETURN_ADDR_RTX
2263 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2264 A C expression whose value is RTL representing the value of the return
2265 address for the frame @var{count} steps up from the current frame, after
2266 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2267 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2268 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2270 The value of the expression must always be the correct address when
2271 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2272 determine the return address of other frames.
2274 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2275 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2276 Define this if the return address of a particular stack frame is accessed
2277 from the frame pointer of the previous stack frame.
2279 @findex INCOMING_RETURN_ADDR_RTX
2280 @item INCOMING_RETURN_ADDR_RTX
2281 A C expression whose value is RTL representing the location of the
2282 incoming return address at the beginning of any function, before the
2283 prologue. This RTL is either a @code{REG}, indicating that the return
2284 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2287 You only need to define this macro if you want to support call frame
2288 debugging information like that provided by DWARF 2.
2290 @findex INCOMING_FRAME_SP_OFFSET
2291 @item INCOMING_FRAME_SP_OFFSET
2292 A C expression whose value is an integer giving the offset, in bytes,
2293 from the value of the stack pointer register to the top of the stack
2294 frame at the beginning of any function, before the prologue. The top of
2295 the frame is defined to be the value of the stack pointer in the
2296 previous frame, just before the call instruction.
2298 You only need to define this macro if you want to support call frame
2299 debugging information like that provided by DWARF 2.
2301 @findex ARG_POINTER_CFA_OFFSET
2302 @item ARG_POINTER_CFA_OFFSET
2303 A C expression whose value is an integer giving the offset, in bytes,
2304 from the argument pointer to the canonical frame address (cfa). The
2305 final value should coincide with that calculated by
2306 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2307 during virtual register instantiation.
2309 You only need to define this macro if you want to support call frame
2310 debugging information like that provided by DWARF 2.
2314 Define this macro if the stack size for the target is very small. This
2315 has the effect of disabling gcc's builtin @samp{alloca}, though
2316 @samp{__builtin_alloca} is not affected.
2319 @node Stack Checking
2320 @subsection Specifying How Stack Checking is Done
2322 GCC will check that stack references are within the boundaries of
2323 the stack, if the @samp{-fstack-check} is specified, in one of three ways:
2327 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2328 will assume that you have arranged for stack checking to be done at
2329 appropriate places in the configuration files, e.g., in
2330 @code{FUNCTION_PROLOGUE}. GCC will do not other special processing.
2333 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
2334 called @code{check_stack} in your @file{md} file, GCC will call that
2335 pattern with one argument which is the address to compare the stack
2336 value against. You must arrange for this pattern to report an error if
2337 the stack pointer is out of range.
2340 If neither of the above are true, GCC will generate code to periodically
2341 ``probe'' the stack pointer using the values of the macros defined below.
2344 Normally, you will use the default values of these macros, so GCC
2345 will use the third approach.
2348 @findex STACK_CHECK_BUILTIN
2349 @item STACK_CHECK_BUILTIN
2350 A nonzero value if stack checking is done by the configuration files in a
2351 machine-dependent manner. You should define this macro if stack checking
2352 is require by the ABI of your machine or if you would like to have to stack
2353 checking in some more efficient way than GCC's portable approach.
2354 The default value of this macro is zero.
2356 @findex STACK_CHECK_PROBE_INTERVAL
2357 @item STACK_CHECK_PROBE_INTERVAL
2358 An integer representing the interval at which GCC must generate stack
2359 probe instructions. You will normally define this macro to be no larger
2360 than the size of the ``guard pages'' at the end of a stack area. The
2361 default value of 4096 is suitable for most systems.
2363 @findex STACK_CHECK_PROBE_LOAD
2364 @item STACK_CHECK_PROBE_LOAD
2365 A integer which is nonzero if GCC should perform the stack probe
2366 as a load instruction and zero if GCC should use a store instruction.
2367 The default is zero, which is the most efficient choice on most systems.
2369 @findex STACK_CHECK_PROTECT
2370 @item STACK_CHECK_PROTECT
2371 The number of bytes of stack needed to recover from a stack overflow,
2372 for languages where such a recovery is supported. The default value of
2373 75 words should be adequate for most machines.
2375 @findex STACK_CHECK_MAX_FRAME_SIZE
2376 @item STACK_CHECK_MAX_FRAME_SIZE
2377 The maximum size of a stack frame, in bytes. GCC will generate probe
2378 instructions in non-leaf functions to ensure at least this many bytes of
2379 stack are available. If a stack frame is larger than this size, stack
2380 checking will not be reliable and GCC will issue a warning. The
2381 default is chosen so that GCC only generates one instruction on most
2382 systems. You should normally not change the default value of this macro.
2384 @findex STACK_CHECK_FIXED_FRAME_SIZE
2385 @item STACK_CHECK_FIXED_FRAME_SIZE
2386 GCC uses this value to generate the above warning message. It
2387 represents the amount of fixed frame used by a function, not including
2388 space for any callee-saved registers, temporaries and user variables.
2389 You need only specify an upper bound for this amount and will normally
2390 use the default of four words.
2392 @findex STACK_CHECK_MAX_VAR_SIZE
2393 @item STACK_CHECK_MAX_VAR_SIZE
2394 The maximum size, in bytes, of an object that GCC will place in the
2395 fixed area of the stack frame when the user specifies
2396 @samp{-fstack-check}.
2397 GCC computed the default from the values of the above macros and you will
2398 normally not need to override that default.
2402 @node Frame Registers
2403 @subsection Registers That Address the Stack Frame
2405 @c prevent bad page break with this line
2406 This discusses registers that address the stack frame.
2409 @findex STACK_POINTER_REGNUM
2410 @item STACK_POINTER_REGNUM
2411 The register number of the stack pointer register, which must also be a
2412 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2413 the hardware determines which register this is.
2415 @findex FRAME_POINTER_REGNUM
2416 @item FRAME_POINTER_REGNUM
2417 The register number of the frame pointer register, which is used to
2418 access automatic variables in the stack frame. On some machines, the
2419 hardware determines which register this is. On other machines, you can
2420 choose any register you wish for this purpose.
2422 @findex HARD_FRAME_POINTER_REGNUM
2423 @item HARD_FRAME_POINTER_REGNUM
2424 On some machines the offset between the frame pointer and starting
2425 offset of the automatic variables is not known until after register
2426 allocation has been done (for example, because the saved registers are
2427 between these two locations). On those machines, define
2428 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
2429 be used internally until the offset is known, and define
2430 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
2431 used for the frame pointer.
2433 You should define this macro only in the very rare circumstances when it
2434 is not possible to calculate the offset between the frame pointer and
2435 the automatic variables until after register allocation has been
2436 completed. When this macro is defined, you must also indicate in your
2437 definition of @code{ELIMINABLE_REGS} how to eliminate
2438 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
2439 or @code{STACK_POINTER_REGNUM}.
2441 Do not define this macro if it would be the same as
2442 @code{FRAME_POINTER_REGNUM}.
2444 @findex ARG_POINTER_REGNUM
2445 @item ARG_POINTER_REGNUM
2446 The register number of the arg pointer register, which is used to access
2447 the function's argument list. On some machines, this is the same as the
2448 frame pointer register. On some machines, the hardware determines which
2449 register this is. On other machines, you can choose any register you
2450 wish for this purpose. If this is not the same register as the frame
2451 pointer register, then you must mark it as a fixed register according to
2452 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2453 (@pxref{Elimination}).
2455 @findex RETURN_ADDRESS_POINTER_REGNUM
2456 @item RETURN_ADDRESS_POINTER_REGNUM
2457 The register number of the return address pointer register, which is used to
2458 access the current function's return address from the stack. On some
2459 machines, the return address is not at a fixed offset from the frame
2460 pointer or stack pointer or argument pointer. This register can be defined
2461 to point to the return address on the stack, and then be converted by
2462 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
2464 Do not define this macro unless there is no other way to get the return
2465 address from the stack.
2467 @findex STATIC_CHAIN_REGNUM
2468 @findex STATIC_CHAIN_INCOMING_REGNUM
2469 @item STATIC_CHAIN_REGNUM
2470 @itemx STATIC_CHAIN_INCOMING_REGNUM
2471 Register numbers used for passing a function's static chain pointer. If
2472 register windows are used, the register number as seen by the called
2473 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
2474 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
2475 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
2476 not be defined.@refill
2478 The static chain register need not be a fixed register.
2480 If the static chain is passed in memory, these macros should not be
2481 defined; instead, the next two macros should be defined.
2483 @findex STATIC_CHAIN
2484 @findex STATIC_CHAIN_INCOMING
2486 @itemx STATIC_CHAIN_INCOMING
2487 If the static chain is passed in memory, these macros provide rtx giving
2488 @code{mem} expressions that denote where they are stored.
2489 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
2490 as seen by the calling and called functions, respectively. Often the former
2491 will be at an offset from the stack pointer and the latter at an offset from
2492 the frame pointer.@refill
2494 @findex stack_pointer_rtx
2495 @findex frame_pointer_rtx
2496 @findex arg_pointer_rtx
2497 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
2498 @code{arg_pointer_rtx} will have been initialized prior to the use of these
2499 macros and should be used to refer to those items.
2501 If the static chain is passed in a register, the two previous macros should
2506 @subsection Eliminating Frame Pointer and Arg Pointer
2508 @c prevent bad page break with this line
2509 This is about eliminating the frame pointer and arg pointer.
2512 @findex FRAME_POINTER_REQUIRED
2513 @item FRAME_POINTER_REQUIRED
2514 A C expression which is nonzero if a function must have and use a frame
2515 pointer. This expression is evaluated in the reload pass. If its value is
2516 nonzero the function will have a frame pointer.
2518 The expression can in principle examine the current function and decide
2519 according to the facts, but on most machines the constant 0 or the
2520 constant 1 suffices. Use 0 when the machine allows code to be generated
2521 with no frame pointer, and doing so saves some time or space. Use 1
2522 when there is no possible advantage to avoiding a frame pointer.
2524 In certain cases, the compiler does not know how to produce valid code
2525 without a frame pointer. The compiler recognizes those cases and
2526 automatically gives the function a frame pointer regardless of what
2527 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
2530 In a function that does not require a frame pointer, the frame pointer
2531 register can be allocated for ordinary usage, unless you mark it as a
2532 fixed register. See @code{FIXED_REGISTERS} for more information.
2534 @findex INITIAL_FRAME_POINTER_OFFSET
2535 @findex get_frame_size
2536 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
2537 A C statement to store in the variable @var{depth-var} the difference
2538 between the frame pointer and the stack pointer values immediately after
2539 the function prologue. The value would be computed from information
2540 such as the result of @code{get_frame_size ()} and the tables of
2541 registers @code{regs_ever_live} and @code{call_used_regs}.
2543 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
2544 need not be defined. Otherwise, it must be defined even if
2545 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
2546 case, you may set @var{depth-var} to anything.
2548 @findex ELIMINABLE_REGS
2549 @item ELIMINABLE_REGS
2550 If defined, this macro specifies a table of register pairs used to
2551 eliminate unneeded registers that point into the stack frame. If it is not
2552 defined, the only elimination attempted by the compiler is to replace
2553 references to the frame pointer with references to the stack pointer.
2555 The definition of this macro is a list of structure initializations, each
2556 of which specifies an original and replacement register.
2558 On some machines, the position of the argument pointer is not known until
2559 the compilation is completed. In such a case, a separate hard register
2560 must be used for the argument pointer. This register can be eliminated by
2561 replacing it with either the frame pointer or the argument pointer,
2562 depending on whether or not the frame pointer has been eliminated.
2564 In this case, you might specify:
2566 #define ELIMINABLE_REGS \
2567 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
2568 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
2569 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
2572 Note that the elimination of the argument pointer with the stack pointer is
2573 specified first since that is the preferred elimination.
2575 @findex CAN_ELIMINATE
2576 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
2577 A C expression that returns non-zero if the compiler is allowed to try
2578 to replace register number @var{from-reg} with register number
2579 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
2580 is defined, and will usually be the constant 1, since most of the cases
2581 preventing register elimination are things that the compiler already
2584 @findex INITIAL_ELIMINATION_OFFSET
2585 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
2586 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
2587 specifies the initial difference between the specified pair of
2588 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
2591 @findex LONGJMP_RESTORE_FROM_STACK
2592 @item LONGJMP_RESTORE_FROM_STACK
2593 Define this macro if the @code{longjmp} function restores registers from
2594 the stack frames, rather than from those saved specifically by
2595 @code{setjmp}. Certain quantities must not be kept in registers across
2596 a call to @code{setjmp} on such machines.
2599 @node Stack Arguments
2600 @subsection Passing Function Arguments on the Stack
2601 @cindex arguments on stack
2602 @cindex stack arguments
2604 The macros in this section control how arguments are passed
2605 on the stack. See the following section for other macros that
2606 control passing certain arguments in registers.
2609 @findex PROMOTE_PROTOTYPES
2610 @item PROMOTE_PROTOTYPES
2611 A C expression whose value is nonzero if an argument declared in
2612 a prototype as an integral type smaller than @code{int} should
2613 actually be passed as an @code{int}. In addition to avoiding
2614 errors in certain cases of mismatch, it also makes for better
2615 code on certain machines. If the macro is not defined in target
2616 header files, it defaults to 0.
2618 @findex PUSH_ROUNDING
2619 @item PUSH_ROUNDING (@var{npushed})
2620 A C expression that is the number of bytes actually pushed onto the
2621 stack when an instruction attempts to push @var{npushed} bytes.
2623 If the target machine does not have a push instruction, do not define
2624 this macro. That directs GCC to use an alternate strategy: to
2625 allocate the entire argument block and then store the arguments into
2628 On some machines, the definition
2631 #define PUSH_ROUNDING(BYTES) (BYTES)
2635 will suffice. But on other machines, instructions that appear
2636 to push one byte actually push two bytes in an attempt to maintain
2637 alignment. Then the definition should be
2640 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
2643 @findex ACCUMULATE_OUTGOING_ARGS
2644 @findex current_function_outgoing_args_size
2645 @item ACCUMULATE_OUTGOING_ARGS
2646 If defined, the maximum amount of space required for outgoing arguments
2647 will be computed and placed into the variable
2648 @code{current_function_outgoing_args_size}. No space will be pushed
2649 onto the stack for each call; instead, the function prologue should
2650 increase the stack frame size by this amount.
2652 Defining both @code{PUSH_ROUNDING} and @code{ACCUMULATE_OUTGOING_ARGS}
2655 @findex REG_PARM_STACK_SPACE
2656 @item REG_PARM_STACK_SPACE (@var{fndecl})
2657 Define this macro if functions should assume that stack space has been
2658 allocated for arguments even when their values are passed in
2661 The value of this macro is the size, in bytes, of the area reserved for
2662 arguments passed in registers for the function represented by @var{fndecl},
2663 which can be zero if GCC is calling a library function.
2665 This space can be allocated by the caller, or be a part of the
2666 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
2668 @c above is overfull. not sure what to do. --mew 5feb93 did
2669 @c something, not sure if it looks good. --mew 10feb93
2671 @findex MAYBE_REG_PARM_STACK_SPACE
2672 @findex FINAL_REG_PARM_STACK_SPACE
2673 @item MAYBE_REG_PARM_STACK_SPACE
2674 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
2675 Define these macros in addition to the one above if functions might
2676 allocate stack space for arguments even when their values are passed
2677 in registers. These should be used when the stack space allocated
2678 for arguments in registers is not a simple constant independent of the
2679 function declaration.
2681 The value of the first macro is the size, in bytes, of the area that
2682 we should initially assume would be reserved for arguments passed in registers.
2684 The value of the second macro is the actual size, in bytes, of the area
2685 that will be reserved for arguments passed in registers. This takes two
2686 arguments: an integer representing the number of bytes of fixed sized
2687 arguments on the stack, and a tree representing the number of bytes of
2688 variable sized arguments on the stack.
2690 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
2691 called for libcall functions, the current function, or for a function
2692 being called when it is known that such stack space must be allocated.
2693 In each case this value can be easily computed.
2695 When deciding whether a called function needs such stack space, and how
2696 much space to reserve, GCC uses these two macros instead of
2697 @code{REG_PARM_STACK_SPACE}.
2699 @findex OUTGOING_REG_PARM_STACK_SPACE
2700 @item OUTGOING_REG_PARM_STACK_SPACE
2701 Define this if it is the responsibility of the caller to allocate the area
2702 reserved for arguments passed in registers.
2704 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
2705 whether the space for these arguments counts in the value of
2706 @code{current_function_outgoing_args_size}.
2708 @findex STACK_PARMS_IN_REG_PARM_AREA
2709 @item STACK_PARMS_IN_REG_PARM_AREA
2710 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
2711 stack parameters don't skip the area specified by it.
2712 @c i changed this, makes more sens and it should have taken care of the
2713 @c overfull.. not as specific, tho. --mew 5feb93
2715 Normally, when a parameter is not passed in registers, it is placed on the
2716 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
2717 suppresses this behavior and causes the parameter to be passed on the
2718 stack in its natural location.
2720 @findex RETURN_POPS_ARGS
2721 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
2722 A C expression that should indicate the number of bytes of its own
2723 arguments that a function pops on returning, or 0 if the
2724 function pops no arguments and the caller must therefore pop them all
2725 after the function returns.
2727 @var{fundecl} is a C variable whose value is a tree node that describes
2728 the function in question. Normally it is a node of type
2729 @code{FUNCTION_DECL} that describes the declaration of the function.
2730 From this you can obtain the DECL_MACHINE_ATTRIBUTES of the function.
2732 @var{funtype} is a C variable whose value is a tree node that
2733 describes the function in question. Normally it is a node of type
2734 @code{FUNCTION_TYPE} that describes the data type of the function.
2735 From this it is possible to obtain the data types of the value and
2736 arguments (if known).
2738 When a call to a library function is being considered, @var{fundecl}
2739 will contain an identifier node for the library function. Thus, if
2740 you need to distinguish among various library functions, you can do so
2741 by their names. Note that ``library function'' in this context means
2742 a function used to perform arithmetic, whose name is known specially
2743 in the compiler and was not mentioned in the C code being compiled.
2745 @var{stack-size} is the number of bytes of arguments passed on the
2746 stack. If a variable number of bytes is passed, it is zero, and
2747 argument popping will always be the responsibility of the calling function.
2749 On the Vax, all functions always pop their arguments, so the definition
2750 of this macro is @var{stack-size}. On the 68000, using the standard
2751 calling convention, no functions pop their arguments, so the value of
2752 the macro is always 0 in this case. But an alternative calling
2753 convention is available in which functions that take a fixed number of
2754 arguments pop them but other functions (such as @code{printf}) pop
2755 nothing (the caller pops all). When this convention is in use,
2756 @var{funtype} is examined to determine whether a function takes a fixed
2757 number of arguments.
2760 @node Register Arguments
2761 @subsection Passing Arguments in Registers
2762 @cindex arguments in registers
2763 @cindex registers arguments
2765 This section describes the macros which let you control how various
2766 types of arguments are passed in registers or how they are arranged in
2770 @findex FUNCTION_ARG
2771 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2772 A C expression that controls whether a function argument is passed
2773 in a register, and which register.
2775 The arguments are @var{cum}, which summarizes all the previous
2776 arguments; @var{mode}, the machine mode of the argument; @var{type},
2777 the data type of the argument as a tree node or 0 if that is not known
2778 (which happens for C support library functions); and @var{named},
2779 which is 1 for an ordinary argument and 0 for nameless arguments that
2780 correspond to @samp{@dots{}} in the called function's prototype.
2782 The value of the expression is usually either a @code{reg} RTX for the
2783 hard register in which to pass the argument, or zero to pass the
2784 argument on the stack.
2786 For machines like the Vax and 68000, where normally all arguments are
2787 pushed, zero suffices as a definition.
2789 The value of the expression can also be a @code{parallel} RTX. This is
2790 used when an argument is passed in multiple locations. The mode of the
2791 of the @code{parallel} should be the mode of the entire argument. The
2792 @code{parallel} holds any number of @code{expr_list} pairs; each one
2793 describes where part of the argument is passed. In each
2794 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
2795 register in which to pass this part of the argument, and the mode of the
2796 register RTX indicates how large this part of the argument is. The
2797 second operand of the @code{expr_list} is a @code{const_int} which gives
2798 the offset in bytes into the entire argument of where this part starts.
2799 As a special exception the first @code{expr_list} in the @code{parallel}
2800 RTX may have a first operand of zero. This indicates that the entire
2801 argument is also stored on the stack.
2803 @cindex @file{stdarg.h} and register arguments
2804 The usual way to make the ANSI library @file{stdarg.h} work on a machine
2805 where some arguments are usually passed in registers, is to cause
2806 nameless arguments to be passed on the stack instead. This is done
2807 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
2809 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
2810 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
2811 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
2812 in the definition of this macro to determine if this argument is of a
2813 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
2814 is not defined and @code{FUNCTION_ARG} returns non-zero for such an
2815 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
2816 defined, the argument will be computed in the stack and then loaded into
2819 @findex MUST_PASS_IN_STACK
2820 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
2821 Define as a C expression that evaluates to nonzero if we do not know how
2822 to pass TYPE solely in registers. The file @file{expr.h} defines a
2823 definition that is usually appropriate, refer to @file{expr.h} for additional
2826 @findex FUNCTION_INCOMING_ARG
2827 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2828 Define this macro if the target machine has ``register windows'', so
2829 that the register in which a function sees an arguments is not
2830 necessarily the same as the one in which the caller passed the
2833 For such machines, @code{FUNCTION_ARG} computes the register in which
2834 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
2835 be defined in a similar fashion to tell the function being called
2836 where the arguments will arrive.
2838 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
2839 serves both purposes.@refill
2841 @findex FUNCTION_ARG_PARTIAL_NREGS
2842 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
2843 A C expression for the number of words, at the beginning of an
2844 argument, must be put in registers. The value must be zero for
2845 arguments that are passed entirely in registers or that are entirely
2846 pushed on the stack.
2848 On some machines, certain arguments must be passed partially in
2849 registers and partially in memory. On these machines, typically the
2850 first @var{n} words of arguments are passed in registers, and the rest
2851 on the stack. If a multi-word argument (a @code{double} or a
2852 structure) crosses that boundary, its first few words must be passed
2853 in registers and the rest must be pushed. This macro tells the
2854 compiler when this occurs, and how many of the words should go in
2857 @code{FUNCTION_ARG} for these arguments should return the first
2858 register to be used by the caller for this argument; likewise
2859 @code{FUNCTION_INCOMING_ARG}, for the called function.
2861 @findex FUNCTION_ARG_PASS_BY_REFERENCE
2862 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2863 A C expression that indicates when an argument must be passed by reference.
2864 If nonzero for an argument, a copy of that argument is made in memory and a
2865 pointer to the argument is passed instead of the argument itself.
2866 The pointer is passed in whatever way is appropriate for passing a pointer
2869 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
2870 definition of this macro might be
2872 #define FUNCTION_ARG_PASS_BY_REFERENCE\
2873 (CUM, MODE, TYPE, NAMED) \
2874 MUST_PASS_IN_STACK (MODE, TYPE)
2876 @c this is *still* too long. --mew 5feb93
2878 @findex FUNCTION_ARG_CALLEE_COPIES
2879 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
2880 If defined, a C expression that indicates when it is the called function's
2881 responsibility to make a copy of arguments passed by invisible reference.
2882 Normally, the caller makes a copy and passes the address of the copy to the
2883 routine being called. When FUNCTION_ARG_CALLEE_COPIES is defined and is
2884 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
2885 ``live'' value. The called function must not modify this value. If it can be
2886 determined that the value won't be modified, it need not make a copy;
2887 otherwise a copy must be made.
2889 @findex CUMULATIVE_ARGS
2890 @item CUMULATIVE_ARGS
2891 A C type for declaring a variable that is used as the first argument of
2892 @code{FUNCTION_ARG} and other related values. For some target machines,
2893 the type @code{int} suffices and can hold the number of bytes of
2896 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
2897 arguments that have been passed on the stack. The compiler has other
2898 variables to keep track of that. For target machines on which all
2899 arguments are passed on the stack, there is no need to store anything in
2900 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
2901 should not be empty, so use @code{int}.
2903 @findex INIT_CUMULATIVE_ARGS
2904 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
2905 A C statement (sans semicolon) for initializing the variable @var{cum}
2906 for the state at the beginning of the argument list. The variable has
2907 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
2908 for the data type of the function which will receive the args, or 0
2909 if the args are to a compiler support library function. The value of
2910 @var{indirect} is nonzero when processing an indirect call, for example
2911 a call through a function pointer. The value of @var{indirect} is zero
2912 for a call to an explicitly named function, a library function call, or when
2913 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
2916 When processing a call to a compiler support library function,
2917 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
2918 contains the name of the function, as a string. @var{libname} is 0 when
2919 an ordinary C function call is being processed. Thus, each time this
2920 macro is called, either @var{libname} or @var{fntype} is nonzero, but
2921 never both of them at once.
2923 @findex INIT_CUMULATIVE_INCOMING_ARGS
2924 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
2925 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
2926 finding the arguments for the function being compiled. If this macro is
2927 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
2929 The value passed for @var{libname} is always 0, since library routines
2930 with special calling conventions are never compiled with GCC. The
2931 argument @var{libname} exists for symmetry with
2932 @code{INIT_CUMULATIVE_ARGS}.
2933 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
2934 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
2936 @findex FUNCTION_ARG_ADVANCE
2937 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2938 A C statement (sans semicolon) to update the summarizer variable
2939 @var{cum} to advance past an argument in the argument list. The
2940 values @var{mode}, @var{type} and @var{named} describe that argument.
2941 Once this is done, the variable @var{cum} is suitable for analyzing
2942 the @emph{following} argument with @code{FUNCTION_ARG}, etc.@refill
2944 This macro need not do anything if the argument in question was passed
2945 on the stack. The compiler knows how to track the amount of stack space
2946 used for arguments without any special help.
2948 @findex FUNCTION_ARG_PADDING
2949 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
2950 If defined, a C expression which determines whether, and in which direction,
2951 to pad out an argument with extra space. The value should be of type
2952 @code{enum direction}: either @code{upward} to pad above the argument,
2953 @code{downward} to pad below, or @code{none} to inhibit padding.
2955 The @emph{amount} of padding is always just enough to reach the next
2956 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
2959 This macro has a default definition which is right for most systems.
2960 For little-endian machines, the default is to pad upward. For
2961 big-endian machines, the default is to pad downward for an argument of
2962 constant size shorter than an @code{int}, and upward otherwise.
2964 @findex PAD_VARARGS_DOWN
2965 @item PAD_VARARGS_DOWN
2966 If defined, a C expression which determines whether the default
2967 implementation of va_arg will attempt to pad down before reading the
2968 next argument, if that argument is smaller than its aligned space as
2969 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
2970 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
2972 @findex FUNCTION_ARG_BOUNDARY
2973 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
2974 If defined, a C expression that gives the alignment boundary, in bits,
2975 of an argument with the specified mode and type. If it is not defined,
2976 @code{PARM_BOUNDARY} is used for all arguments.
2978 @findex FUNCTION_ARG_REGNO_P
2979 @item FUNCTION_ARG_REGNO_P (@var{regno})
2980 A C expression that is nonzero if @var{regno} is the number of a hard
2981 register in which function arguments are sometimes passed. This does
2982 @emph{not} include implicit arguments such as the static chain and
2983 the structure-value address. On many machines, no registers can be
2984 used for this purpose since all function arguments are pushed on the
2987 @findex LOAD_ARGS_REVERSED
2988 @item LOAD_ARGS_REVERSED
2989 If defined, the order in which arguments are loaded into their
2990 respective argument registers is reversed so that the last
2991 argument is loaded first. This macro only affects arguments
2992 passed in registers.
2997 @subsection How Scalar Function Values Are Returned
2998 @cindex return values in registers
2999 @cindex values, returned by functions
3000 @cindex scalars, returned as values
3002 This section discusses the macros that control returning scalars as
3003 values---values that can fit in registers.
3006 @findex TRADITIONAL_RETURN_FLOAT
3007 @item TRADITIONAL_RETURN_FLOAT
3008 Define this macro if @samp{-traditional} should not cause functions
3009 declared to return @code{float} to convert the value to @code{double}.
3011 @findex FUNCTION_VALUE
3012 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3013 A C expression to create an RTX representing the place where a
3014 function returns a value of data type @var{valtype}. @var{valtype} is
3015 a tree node representing a data type. Write @code{TYPE_MODE
3016 (@var{valtype})} to get the machine mode used to represent that type.
3017 On many machines, only the mode is relevant. (Actually, on most
3018 machines, scalar values are returned in the same place regardless of
3021 The value of the expression is usually a @code{reg} RTX for the hard
3022 register where the return value is stored. The value can also be a
3023 @code{parallel} RTX, if the return value is in multiple places. See
3024 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3026 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3027 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3030 If the precise function being called is known, @var{func} is a tree
3031 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3032 pointer. This makes it possible to use a different value-returning
3033 convention for specific functions when all their calls are
3036 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3037 types, because these are returned in another way. See
3038 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3040 @findex FUNCTION_OUTGOING_VALUE
3041 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3042 Define this macro if the target machine has ``register windows''
3043 so that the register in which a function returns its value is not
3044 the same as the one in which the caller sees the value.
3046 For such machines, @code{FUNCTION_VALUE} computes the register in which
3047 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3048 defined in a similar fashion to tell the function where to put the
3051 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3052 @code{FUNCTION_VALUE} serves both purposes.@refill
3054 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3055 aggregate data types, because these are returned in another way. See
3056 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3058 @findex LIBCALL_VALUE
3059 @item LIBCALL_VALUE (@var{mode})
3060 A C expression to create an RTX representing the place where a library
3061 function returns a value of mode @var{mode}. If the precise function
3062 being called is known, @var{func} is a tree node
3063 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3064 pointer. This makes it possible to use a different value-returning
3065 convention for specific functions when all their calls are
3068 Note that ``library function'' in this context means a compiler
3069 support routine, used to perform arithmetic, whose name is known
3070 specially by the compiler and was not mentioned in the C code being
3073 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3074 data types, because none of the library functions returns such types.
3076 @findex FUNCTION_VALUE_REGNO_P
3077 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3078 A C expression that is nonzero if @var{regno} is the number of a hard
3079 register in which the values of called function may come back.
3081 A register whose use for returning values is limited to serving as the
3082 second of a pair (for a value of type @code{double}, say) need not be
3083 recognized by this macro. So for most machines, this definition
3087 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3090 If the machine has register windows, so that the caller and the called
3091 function use different registers for the return value, this macro
3092 should recognize only the caller's register numbers.
3094 @findex APPLY_RESULT_SIZE
3095 @item APPLY_RESULT_SIZE
3096 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3097 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3098 saving and restoring an arbitrary return value.
3101 @node Aggregate Return
3102 @subsection How Large Values Are Returned
3103 @cindex aggregates as return values
3104 @cindex large return values
3105 @cindex returning aggregate values
3106 @cindex structure value address
3108 When a function value's mode is @code{BLKmode} (and in some other
3109 cases), the value is not returned according to @code{FUNCTION_VALUE}
3110 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3111 block of memory in which the value should be stored. This address
3112 is called the @dfn{structure value address}.
3114 This section describes how to control returning structure values in
3118 @findex RETURN_IN_MEMORY
3119 @item RETURN_IN_MEMORY (@var{type})
3120 A C expression which can inhibit the returning of certain function
3121 values in registers, based on the type of value. A nonzero value says
3122 to return the function value in memory, just as large structures are
3123 always returned. Here @var{type} will be a C expression of type
3124 @code{tree}, representing the data type of the value.
3126 Note that values of mode @code{BLKmode} must be explicitly handled
3127 by this macro. Also, the option @samp{-fpcc-struct-return}
3128 takes effect regardless of this macro. On most systems, it is
3129 possible to leave the macro undefined; this causes a default
3130 definition to be used, whose value is the constant 1 for @code{BLKmode}
3131 values, and 0 otherwise.
3133 Do not use this macro to indicate that structures and unions should always
3134 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3137 @findex DEFAULT_PCC_STRUCT_RETURN
3138 @item DEFAULT_PCC_STRUCT_RETURN
3139 Define this macro to be 1 if all structure and union return values must be
3140 in memory. Since this results in slower code, this should be defined
3141 only if needed for compatibility with other compilers or with an ABI.
3142 If you define this macro to be 0, then the conventions used for structure
3143 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3145 If not defined, this defaults to the value 1.
3147 @findex STRUCT_VALUE_REGNUM
3148 @item STRUCT_VALUE_REGNUM
3149 If the structure value address is passed in a register, then
3150 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3152 @findex STRUCT_VALUE
3154 If the structure value address is not passed in a register, define
3155 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3156 where the address is passed. If it returns 0, the address is passed as
3157 an ``invisible'' first argument.
3159 @findex STRUCT_VALUE_INCOMING_REGNUM
3160 @item STRUCT_VALUE_INCOMING_REGNUM
3161 On some architectures the place where the structure value address
3162 is found by the called function is not the same place that the
3163 caller put it. This can be due to register windows, or it could
3164 be because the function prologue moves it to a different place.
3166 If the incoming location of the structure value address is in a
3167 register, define this macro as the register number.
3169 @findex STRUCT_VALUE_INCOMING
3170 @item STRUCT_VALUE_INCOMING
3171 If the incoming location is not a register, then you should define
3172 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3173 called function should find the value. If it should find the value on
3174 the stack, define this to create a @code{mem} which refers to the frame
3175 pointer. A definition of 0 means that the address is passed as an
3176 ``invisible'' first argument.
3178 @findex PCC_STATIC_STRUCT_RETURN
3179 @item PCC_STATIC_STRUCT_RETURN
3180 Define this macro if the usual system convention on the target machine
3181 for returning structures and unions is for the called function to return
3182 the address of a static variable containing the value.
3184 Do not define this if the usual system convention is for the caller to
3185 pass an address to the subroutine.
3187 This macro has effect in @samp{-fpcc-struct-return} mode, but it does
3188 nothing when you use @samp{-freg-struct-return} mode.
3192 @subsection Caller-Saves Register Allocation
3194 If you enable it, GCC can save registers around function calls. This
3195 makes it possible to use call-clobbered registers to hold variables that
3196 must live across calls.
3199 @findex DEFAULT_CALLER_SAVES
3200 @item DEFAULT_CALLER_SAVES
3201 Define this macro if function calls on the target machine do not preserve
3202 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3203 for all registers. When defined, this macro enables @samp{-fcaller-saves}
3204 by default for all optimization levels. It has no effect for optimization
3205 levels 2 and higher, where @samp{-fcaller-saves} is the default.
3207 @findex CALLER_SAVE_PROFITABLE
3208 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3209 A C expression to determine whether it is worthwhile to consider placing
3210 a pseudo-register in a call-clobbered hard register and saving and
3211 restoring it around each function call. The expression should be 1 when
3212 this is worth doing, and 0 otherwise.
3214 If you don't define this macro, a default is used which is good on most
3215 machines: @code{4 * @var{calls} < @var{refs}}.
3217 @findex HARD_REGNO_CALLER_SAVE_MODE
3218 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3219 A C expression specifying which mode is required for saving @var{nregs}
3220 of a pseudo-register in call-clobbered hard register @var{regno}. If
3221 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3222 returned. For most machines this macro need not be defined since GCC
3223 will select the smallest suitable mode.
3226 @node Function Entry
3227 @subsection Function Entry and Exit
3228 @cindex function entry and exit
3232 This section describes the macros that output function entry
3233 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3236 @findex FUNCTION_PROLOGUE
3237 @item FUNCTION_PROLOGUE (@var{file}, @var{size})
3238 A C compound statement that outputs the assembler code for entry to a
3239 function. The prologue is responsible for setting up the stack frame,
3240 initializing the frame pointer register, saving registers that must be
3241 saved, and allocating @var{size} additional bytes of storage for the
3242 local variables. @var{size} is an integer. @var{file} is a stdio
3243 stream to which the assembler code should be output.
3245 The label for the beginning of the function need not be output by this
3246 macro. That has already been done when the macro is run.
3248 @findex regs_ever_live
3249 To determine which registers to save, the macro can refer to the array
3250 @code{regs_ever_live}: element @var{r} is nonzero if hard register
3251 @var{r} is used anywhere within the function. This implies the function
3252 prologue should save register @var{r}, provided it is not one of the
3253 call-used registers. (@code{FUNCTION_EPILOGUE} must likewise use
3254 @code{regs_ever_live}.)
3256 On machines that have ``register windows'', the function entry code does
3257 not save on the stack the registers that are in the windows, even if
3258 they are supposed to be preserved by function calls; instead it takes
3259 appropriate steps to ``push'' the register stack, if any non-call-used
3260 registers are used in the function.
3262 @findex frame_pointer_needed
3263 On machines where functions may or may not have frame-pointers, the
3264 function entry code must vary accordingly; it must set up the frame
3265 pointer if one is wanted, and not otherwise. To determine whether a
3266 frame pointer is in wanted, the macro can refer to the variable
3267 @code{frame_pointer_needed}. The variable's value will be 1 at run
3268 time in a function that needs a frame pointer. @xref{Elimination}.
3270 The function entry code is responsible for allocating any stack space
3271 required for the function. This stack space consists of the regions
3272 listed below. In most cases, these regions are allocated in the
3273 order listed, with the last listed region closest to the top of the
3274 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
3275 the highest address if it is not defined). You can use a different order
3276 for a machine if doing so is more convenient or required for
3277 compatibility reasons. Except in cases where required by standard
3278 or by a debugger, there is no reason why the stack layout used by GCC
3279 need agree with that used by other compilers for a machine.
3283 @findex current_function_pretend_args_size
3284 A region of @code{current_function_pretend_args_size} bytes of
3285 uninitialized space just underneath the first argument arriving on the
3286 stack. (This may not be at the very start of the allocated stack region
3287 if the calling sequence has pushed anything else since pushing the stack
3288 arguments. But usually, on such machines, nothing else has been pushed
3289 yet, because the function prologue itself does all the pushing.) This
3290 region is used on machines where an argument may be passed partly in
3291 registers and partly in memory, and, in some cases to support the
3292 features in @file{varargs.h} and @file{stdargs.h}.
3295 An area of memory used to save certain registers used by the function.
3296 The size of this area, which may also include space for such things as
3297 the return address and pointers to previous stack frames, is
3298 machine-specific and usually depends on which registers have been used
3299 in the function. Machines with register windows often do not require
3303 A region of at least @var{size} bytes, possibly rounded up to an allocation
3304 boundary, to contain the local variables of the function. On some machines,
3305 this region and the save area may occur in the opposite order, with the
3306 save area closer to the top of the stack.
3309 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3310 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3311 @code{current_function_outgoing_args_size} bytes to be used for outgoing
3312 argument lists of the function. @xref{Stack Arguments}.
3315 Normally, it is necessary for the macros @code{FUNCTION_PROLOGUE} and
3316 @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C
3317 variable @code{current_function_is_leaf} is nonzero for such a function.
3319 @findex EXIT_IGNORE_STACK
3320 @item EXIT_IGNORE_STACK
3321 Define this macro as a C expression that is nonzero if the return
3322 instruction or the function epilogue ignores the value of the stack
3323 pointer; in other words, if it is safe to delete an instruction to
3324 adjust the stack pointer before a return from the function.
3326 Note that this macro's value is relevant only for functions for which
3327 frame pointers are maintained. It is never safe to delete a final
3328 stack adjustment in a function that has no frame pointer, and the
3329 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3331 @findex EPILOGUE_USES
3332 @item EPILOGUE_USES (@var{regno})
3333 Define this macro as a C expression that is nonzero for registers that are
3334 used by the epilogue or the @samp{return} pattern. The stack and frame
3335 pointer registers are already be assumed to be used as needed.
3337 @findex FUNCTION_EPILOGUE
3338 @item FUNCTION_EPILOGUE (@var{file}, @var{size})
3339 A C compound statement that outputs the assembler code for exit from a
3340 function. The epilogue is responsible for restoring the saved
3341 registers and stack pointer to their values when the function was
3342 called, and returning control to the caller. This macro takes the
3343 same arguments as the macro @code{FUNCTION_PROLOGUE}, and the
3344 registers to restore are determined from @code{regs_ever_live} and
3345 @code{CALL_USED_REGISTERS} in the same way.
3347 On some machines, there is a single instruction that does all the work
3348 of returning from the function. On these machines, give that
3349 instruction the name @samp{return} and do not define the macro
3350 @code{FUNCTION_EPILOGUE} at all.
3352 Do not define a pattern named @samp{return} if you want the
3353 @code{FUNCTION_EPILOGUE} to be used. If you want the target switches
3354 to control whether return instructions or epilogues are used, define a
3355 @samp{return} pattern with a validity condition that tests the target
3356 switches appropriately. If the @samp{return} pattern's validity
3357 condition is false, epilogues will be used.
3359 On machines where functions may or may not have frame-pointers, the
3360 function exit code must vary accordingly. Sometimes the code for these
3361 two cases is completely different. To determine whether a frame pointer
3362 is wanted, the macro can refer to the variable
3363 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
3364 a function that needs a frame pointer.
3366 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
3367 treat leaf functions specially. The C variable @code{current_function_is_leaf}
3368 is nonzero for such a function. @xref{Leaf Functions}.
3370 On some machines, some functions pop their arguments on exit while
3371 others leave that for the caller to do. For example, the 68020 when
3372 given @samp{-mrtd} pops arguments in functions that take a fixed
3373 number of arguments.
3375 @findex current_function_pops_args
3376 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
3377 functions pop their own arguments. @code{FUNCTION_EPILOGUE} needs to
3378 know what was decided. The variable that is called
3379 @code{current_function_pops_args} is the number of bytes of its
3380 arguments that a function should pop. @xref{Scalar Return}.
3381 @c what is the "its arguments" in the above sentence referring to, pray
3382 @c tell? --mew 5feb93
3384 @findex DELAY_SLOTS_FOR_EPILOGUE
3385 @item DELAY_SLOTS_FOR_EPILOGUE
3386 Define this macro if the function epilogue contains delay slots to which
3387 instructions from the rest of the function can be ``moved''. The
3388 definition should be a C expression whose value is an integer
3389 representing the number of delay slots there.
3391 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
3392 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
3393 A C expression that returns 1 if @var{insn} can be placed in delay
3394 slot number @var{n} of the epilogue.
3396 The argument @var{n} is an integer which identifies the delay slot now
3397 being considered (since different slots may have different rules of
3398 eligibility). It is never negative and is always less than the number
3399 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
3400 If you reject a particular insn for a given delay slot, in principle, it
3401 may be reconsidered for a subsequent delay slot. Also, other insns may
3402 (at least in principle) be considered for the so far unfilled delay
3405 @findex current_function_epilogue_delay_list
3406 @findex final_scan_insn
3407 The insns accepted to fill the epilogue delay slots are put in an RTL
3408 list made with @code{insn_list} objects, stored in the variable
3409 @code{current_function_epilogue_delay_list}. The insn for the first
3410 delay slot comes first in the list. Your definition of the macro
3411 @code{FUNCTION_EPILOGUE} should fill the delay slots by outputting the
3412 insns in this list, usually by calling @code{final_scan_insn}.
3414 You need not define this macro if you did not define
3415 @code{DELAY_SLOTS_FOR_EPILOGUE}.
3417 @findex ASM_OUTPUT_MI_THUNK
3418 @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function})
3419 A C compound statement that outputs the assembler code for a thunk
3420 function, used to implement C++ virtual function calls with multiple
3421 inheritance. The thunk acts as a wrapper around a virtual function,
3422 adjusting the implicit object parameter before handing control off to
3425 First, emit code to add the integer @var{delta} to the location that
3426 contains the incoming first argument. Assume that this argument
3427 contains a pointer, and is the one used to pass the @code{this} pointer
3428 in C++. This is the incoming argument @emph{before} the function prologue,
3429 e.g. @samp{%o0} on a sparc. The addition must preserve the values of
3430 all other incoming arguments.
3432 After the addition, emit code to jump to @var{function}, which is a
3433 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
3434 not touch the return address. Hence returning from @var{FUNCTION} will
3435 return to whoever called the current @samp{thunk}.
3437 The effect must be as if @var{function} had been called directly with
3438 the adjusted first argument. This macro is responsible for emitting all
3439 of the code for a thunk function; @code{FUNCTION_PROLOGUE} and
3440 @code{FUNCTION_EPILOGUE} are not invoked.
3442 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
3443 have already been extracted from it.) It might possibly be useful on
3444 some targets, but probably not.
3446 If you do not define this macro, the target-independent code in the C++
3447 frontend will generate a less efficient heavyweight thunk that calls
3448 @var{function} instead of jumping to it. The generic approach does
3449 not support varargs.
3453 @subsection Generating Code for Profiling
3454 @cindex profiling, code generation
3456 These macros will help you generate code for profiling.
3459 @findex FUNCTION_PROFILER
3460 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
3461 A C statement or compound statement to output to @var{file} some
3462 assembler code to call the profiling subroutine @code{mcount}.
3463 Before calling, the assembler code must load the address of a
3464 counter variable into a register where @code{mcount} expects to
3465 find the address. The name of this variable is @samp{LP} followed
3466 by the number @var{labelno}, so you would generate the name using
3467 @samp{LP%d} in a @code{fprintf}.
3470 The details of how the address should be passed to @code{mcount} are
3471 determined by your operating system environment, not by GCC. To
3472 figure them out, compile a small program for profiling using the
3473 system's installed C compiler and look at the assembler code that
3476 @findex PROFILE_BEFORE_PROLOGUE
3477 @item PROFILE_BEFORE_PROLOGUE
3478 Define this macro if the code for function profiling should come before
3479 the function prologue. Normally, the profiling code comes after.
3481 @findex FUNCTION_BLOCK_PROFILER
3482 @vindex profile_block_flag
3483 @item FUNCTION_BLOCK_PROFILER (@var{file}, @var{labelno})
3484 A C statement or compound statement to output to @var{file} some
3485 assembler code to initialize basic-block profiling for the current
3486 object module. The global compile flag @code{profile_block_flag}
3487 distinguishes two profile modes.
3490 @findex __bb_init_func
3491 @item profile_block_flag != 2
3492 Output code to call the subroutine @code{__bb_init_func} once per
3493 object module, passing it as its sole argument the address of a block
3494 allocated in the object module.
3496 The name of the block is a local symbol made with this statement:
3499 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3502 Of course, since you are writing the definition of
3503 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3504 can take a short cut in the definition of this macro and use the name
3505 that you know will result.
3507 The first word of this block is a flag which will be nonzero if the
3508 object module has already been initialized. So test this word first,
3509 and do not call @code{__bb_init_func} if the flag is
3510 nonzero. BLOCK_OR_LABEL contains a unique number which may be used to
3511 generate a label as a branch destination when @code{__bb_init_func}
3514 Described in assembler language, the code to be output looks like:
3524 @findex __bb_init_trace_func
3525 @item profile_block_flag == 2
3526 Output code to call the subroutine @code{__bb_init_trace_func}
3527 and pass two parameters to it. The first parameter is the same as
3528 for @code{__bb_init_func}. The second parameter is the number of the
3529 first basic block of the function as given by BLOCK_OR_LABEL. Note
3530 that @code{__bb_init_trace_func} has to be called, even if the object
3531 module has been initialized already.
3533 Described in assembler language, the code to be output looks like:
3536 parameter2 <- BLOCK_OR_LABEL
3537 call __bb_init_trace_func
3541 @findex BLOCK_PROFILER
3542 @vindex profile_block_flag
3543 @item BLOCK_PROFILER (@var{file}, @var{blockno})
3544 A C statement or compound statement to output to @var{file} some
3545 assembler code to increment the count associated with the basic
3546 block number @var{blockno}. The global compile flag
3547 @code{profile_block_flag} distinguishes two profile modes.
3550 @item profile_block_flag != 2
3551 Output code to increment the counter directly. Basic blocks are
3552 numbered separately from zero within each compilation. The count
3553 associated with block number @var{blockno} is at index
3554 @var{blockno} in a vector of words; the name of this array is a local
3555 symbol made with this statement:
3558 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 2);
3561 @c This paragraph is the same as one a few paragraphs up.
3562 @c That is not an error.
3563 Of course, since you are writing the definition of
3564 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3565 can take a short cut in the definition of this macro and use the name
3566 that you know will result.
3568 Described in assembler language, the code to be output looks like:
3571 inc (LPBX2+4*BLOCKNO)
3575 @findex __bb_trace_func
3576 @item profile_block_flag == 2
3577 Output code to initialize the global structure @code{__bb} and
3578 call the function @code{__bb_trace_func}, which will increment the
3581 @code{__bb} consists of two words. In the first word, the current
3582 basic block number, as given by BLOCKNO, has to be stored. In
3583 the second word, the address of a block allocated in the object
3584 module has to be stored. The address is given by the label created
3585 with this statement:
3588 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3591 Described in assembler language, the code to be output looks like:
3593 move BLOCKNO -> (__bb)
3594 move LPBX0 -> (__bb+4)
3595 call __bb_trace_func
3599 @findex FUNCTION_BLOCK_PROFILER_EXIT
3600 @findex __bb_trace_ret
3601 @vindex profile_block_flag
3602 @item FUNCTION_BLOCK_PROFILER_EXIT (@var{file})
3603 A C statement or compound statement to output to @var{file}
3604 assembler code to call function @code{__bb_trace_ret}. The
3605 assembler code should only be output
3606 if the global compile flag @code{profile_block_flag} == 2. This
3607 macro has to be used at every place where code for returning from
3608 a function is generated (e.g. @code{FUNCTION_EPILOGUE}). Although
3609 you have to write the definition of @code{FUNCTION_EPILOGUE}
3610 as well, you have to define this macro to tell the compiler, that
3611 the proper call to @code{__bb_trace_ret} is produced.
3613 @findex MACHINE_STATE_SAVE
3614 @findex __bb_init_trace_func
3615 @findex __bb_trace_func
3616 @findex __bb_trace_ret
3617 @item MACHINE_STATE_SAVE (@var{id})
3618 A C statement or compound statement to save all registers, which may
3619 be clobbered by a function call, including condition codes. The
3620 @code{asm} statement will be mostly likely needed to handle this
3621 task. Local labels in the assembler code can be concatenated with the
3622 string @var{id}, to obtain a unique label name.
3624 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3625 @code{FUNCTION_EPILOGUE} must be saved in the macros
3626 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3627 @code{BLOCK_PROFILER} prior calling @code{__bb_init_trace_func},
3628 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3630 @findex MACHINE_STATE_RESTORE
3631 @findex __bb_init_trace_func
3632 @findex __bb_trace_func
3633 @findex __bb_trace_ret
3634 @item MACHINE_STATE_RESTORE (@var{id})
3635 A C statement or compound statement to restore all registers, including
3636 condition codes, saved by @code{MACHINE_STATE_SAVE}.
3638 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3639 @code{FUNCTION_EPILOGUE} must be restored in the macros
3640 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3641 @code{BLOCK_PROFILER} after calling @code{__bb_init_trace_func},
3642 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3644 @findex BLOCK_PROFILER_CODE
3645 @item BLOCK_PROFILER_CODE
3646 A C function or functions which are needed in the library to
3647 support block profiling.
3651 @section Implementing the Varargs Macros
3652 @cindex varargs implementation
3654 GCC comes with an implementation of @file{varargs.h} and
3655 @file{stdarg.h} that work without change on machines that pass arguments
3656 on the stack. Other machines require their own implementations of
3657 varargs, and the two machine independent header files must have
3658 conditionals to include it.
3660 ANSI @file{stdarg.h} differs from traditional @file{varargs.h} mainly in
3661 the calling convention for @code{va_start}. The traditional
3662 implementation takes just one argument, which is the variable in which
3663 to store the argument pointer. The ANSI implementation of
3664 @code{va_start} takes an additional second argument. The user is
3665 supposed to write the last named argument of the function here.
3667 However, @code{va_start} should not use this argument. The way to find
3668 the end of the named arguments is with the built-in functions described
3672 @findex __builtin_saveregs
3673 @item __builtin_saveregs ()
3674 Use this built-in function to save the argument registers in memory so
3675 that the varargs mechanism can access them. Both ANSI and traditional
3676 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3677 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
3679 On some machines, @code{__builtin_saveregs} is open-coded under the
3680 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
3681 it calls a routine written in assembler language, found in
3684 Code generated for the call to @code{__builtin_saveregs} appears at the
3685 beginning of the function, as opposed to where the call to
3686 @code{__builtin_saveregs} is written, regardless of what the code is.
3687 This is because the registers must be saved before the function starts
3688 to use them for its own purposes.
3689 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3692 @findex __builtin_args_info
3693 @item __builtin_args_info (@var{category})
3694 Use this built-in function to find the first anonymous arguments in
3697 In general, a machine may have several categories of registers used for
3698 arguments, each for a particular category of data types. (For example,
3699 on some machines, floating-point registers are used for floating-point
3700 arguments while other arguments are passed in the general registers.)
3701 To make non-varargs functions use the proper calling convention, you
3702 have defined the @code{CUMULATIVE_ARGS} data type to record how many
3703 registers in each category have been used so far
3705 @code{__builtin_args_info} accesses the same data structure of type
3706 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
3707 with it, with @var{category} specifying which word to access. Thus, the
3708 value indicates the first unused register in a given category.
3710 Normally, you would use @code{__builtin_args_info} in the implementation
3711 of @code{va_start}, accessing each category just once and storing the
3712 value in the @code{va_list} object. This is because @code{va_list} will
3713 have to update the values, and there is no way to alter the
3714 values accessed by @code{__builtin_args_info}.
3716 @findex __builtin_next_arg
3717 @item __builtin_next_arg (@var{lastarg})
3718 This is the equivalent of @code{__builtin_args_info}, for stack
3719 arguments. It returns the address of the first anonymous stack
3720 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
3721 returns the address of the location above the first anonymous stack
3722 argument. Use it in @code{va_start} to initialize the pointer for
3723 fetching arguments from the stack. Also use it in @code{va_start} to
3724 verify that the second parameter @var{lastarg} is the last named argument
3725 of the current function.
3727 @findex __builtin_classify_type
3728 @item __builtin_classify_type (@var{object})
3729 Since each machine has its own conventions for which data types are
3730 passed in which kind of register, your implementation of @code{va_arg}
3731 has to embody these conventions. The easiest way to categorize the
3732 specified data type is to use @code{__builtin_classify_type} together
3733 with @code{sizeof} and @code{__alignof__}.
3735 @code{__builtin_classify_type} ignores the value of @var{object},
3736 considering only its data type. It returns an integer describing what
3737 kind of type that is---integer, floating, pointer, structure, and so on.
3739 The file @file{typeclass.h} defines an enumeration that you can use to
3740 interpret the values of @code{__builtin_classify_type}.
3743 These machine description macros help implement varargs:
3746 @findex EXPAND_BUILTIN_SAVEREGS
3747 @item EXPAND_BUILTIN_SAVEREGS ()
3748 If defined, is a C expression that produces the machine-specific code
3749 for a call to @code{__builtin_saveregs}. This code will be moved to the
3750 very beginning of the function, before any parameter access are made.
3751 The return value of this function should be an RTX that contains the
3752 value to use as the return of @code{__builtin_saveregs}.
3754 @findex SETUP_INCOMING_VARARGS
3755 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
3756 This macro offers an alternative to using @code{__builtin_saveregs} and
3757 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
3758 anonymous register arguments into the stack so that all the arguments
3759 appear to have been passed consecutively on the stack. Once this is
3760 done, you can use the standard implementation of varargs that works for
3761 machines that pass all their arguments on the stack.
3763 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
3764 structure, containing the values that are obtained after processing the
3765 named arguments. The arguments @var{mode} and @var{type} describe the
3766 last named argument---its machine mode and its data type as a tree node.
3768 The macro implementation should do two things: first, push onto the
3769 stack all the argument registers @emph{not} used for the named
3770 arguments, and second, store the size of the data thus pushed into the
3771 @code{int}-valued variable whose name is supplied as the argument
3772 @var{pretend_args_size}. The value that you store here will serve as
3773 additional offset for setting up the stack frame.
3775 Because you must generate code to push the anonymous arguments at
3776 compile time without knowing their data types,
3777 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
3778 a single category of argument register and use it uniformly for all data
3781 If the argument @var{second_time} is nonzero, it means that the
3782 arguments of the function are being analyzed for the second time. This
3783 happens for an inline function, which is not actually compiled until the
3784 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
3785 not generate any instructions in this case.
3787 @findex STRICT_ARGUMENT_NAMING
3788 @item STRICT_ARGUMENT_NAMING
3789 Define this macro to be a nonzero value if the location where a function
3790 argument is passed depends on whether or not it is a named argument.
3792 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
3793 is set for varargs and stdarg functions. If this macro returns a
3794 nonzero value, the @var{named} argument is always true for named
3795 arguments, and false for unnamed arguments. If it returns a value of
3796 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
3797 are treated as named. Otherwise, all named arguments except the last
3798 are treated as named.
3800 You need not define this macro if it always returns zero.
3802 @findex PRETEND_OUTGOING_VARARGS_NAMED
3803 @item PRETEND_OUTGOING_VARARGS_NAMED
3804 If you need to conditionally change ABIs so that one works with
3805 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
3806 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
3807 defined, then define this macro to return nonzero if
3808 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
3809 Otherwise, you should not define this macro.
3813 @section Trampolines for Nested Functions
3814 @cindex trampolines for nested functions
3815 @cindex nested functions, trampolines for
3817 A @dfn{trampoline} is a small piece of code that is created at run time
3818 when the address of a nested function is taken. It normally resides on
3819 the stack, in the stack frame of the containing function. These macros
3820 tell GCC how to generate code to allocate and initialize a
3823 The instructions in the trampoline must do two things: load a constant
3824 address into the static chain register, and jump to the real address of
3825 the nested function. On CISC machines such as the m68k, this requires
3826 two instructions, a move immediate and a jump. Then the two addresses
3827 exist in the trampoline as word-long immediate operands. On RISC
3828 machines, it is often necessary to load each address into a register in
3829 two parts. Then pieces of each address form separate immediate
3832 The code generated to initialize the trampoline must store the variable
3833 parts---the static chain value and the function address---into the
3834 immediate operands of the instructions. On a CISC machine, this is
3835 simply a matter of copying each address to a memory reference at the
3836 proper offset from the start of the trampoline. On a RISC machine, it
3837 may be necessary to take out pieces of the address and store them
3841 @findex TRAMPOLINE_TEMPLATE
3842 @item TRAMPOLINE_TEMPLATE (@var{file})
3843 A C statement to output, on the stream @var{file}, assembler code for a
3844 block of data that contains the constant parts of a trampoline. This
3845 code should not include a label---the label is taken care of
3848 If you do not define this macro, it means no template is needed
3849 for the target. Do not define this macro on systems where the block move
3850 code to copy the trampoline into place would be larger than the code
3851 to generate it on the spot.
3853 @findex TRAMPOLINE_SECTION
3854 @item TRAMPOLINE_SECTION
3855 The name of a subroutine to switch to the section in which the
3856 trampoline template is to be placed (@pxref{Sections}). The default is
3857 a value of @samp{readonly_data_section}, which places the trampoline in
3858 the section containing read-only data.
3860 @findex TRAMPOLINE_SIZE
3861 @item TRAMPOLINE_SIZE
3862 A C expression for the size in bytes of the trampoline, as an integer.
3864 @findex TRAMPOLINE_ALIGNMENT
3865 @item TRAMPOLINE_ALIGNMENT
3866 Alignment required for trampolines, in bits.
3868 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
3869 is used for aligning trampolines.
3871 @findex INITIALIZE_TRAMPOLINE
3872 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
3873 A C statement to initialize the variable parts of a trampoline.
3874 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
3875 an RTX for the address of the nested function; @var{static_chain} is an
3876 RTX for the static chain value that should be passed to the function
3879 @findex ALLOCATE_TRAMPOLINE
3880 @item ALLOCATE_TRAMPOLINE (@var{fp})
3881 A C expression to allocate run-time space for a trampoline. The
3882 expression value should be an RTX representing a memory reference to the
3883 space for the trampoline.
3885 @cindex @code{FUNCTION_EPILOGUE} and trampolines
3886 @cindex @code{FUNCTION_PROLOGUE} and trampolines
3887 If this macro is not defined, by default the trampoline is allocated as
3888 a stack slot. This default is right for most machines. The exceptions
3889 are machines where it is impossible to execute instructions in the stack
3890 area. On such machines, you may have to implement a separate stack,
3891 using this macro in conjunction with @code{FUNCTION_PROLOGUE} and
3892 @code{FUNCTION_EPILOGUE}.
3894 @var{fp} points to a data structure, a @code{struct function}, which
3895 describes the compilation status of the immediate containing function of
3896 the function which the trampoline is for. Normally (when
3897 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
3898 trampoline is in the stack frame of this containing function. Other
3899 allocation strategies probably must do something analogous with this
3903 Implementing trampolines is difficult on many machines because they have
3904 separate instruction and data caches. Writing into a stack location
3905 fails to clear the memory in the instruction cache, so when the program
3906 jumps to that location, it executes the old contents.
3908 Here are two possible solutions. One is to clear the relevant parts of
3909 the instruction cache whenever a trampoline is set up. The other is to
3910 make all trampolines identical, by having them jump to a standard
3911 subroutine. The former technique makes trampoline execution faster; the
3912 latter makes initialization faster.
3914 To clear the instruction cache when a trampoline is initialized, define
3915 the following macros which describe the shape of the cache.
3918 @findex INSN_CACHE_SIZE
3919 @item INSN_CACHE_SIZE
3920 The total size in bytes of the cache.
3922 @findex INSN_CACHE_LINE_WIDTH
3923 @item INSN_CACHE_LINE_WIDTH
3924 The length in bytes of each cache line. The cache is divided into cache
3925 lines which are disjoint slots, each holding a contiguous chunk of data
3926 fetched from memory. Each time data is brought into the cache, an
3927 entire line is read at once. The data loaded into a cache line is
3928 always aligned on a boundary equal to the line size.
3930 @findex INSN_CACHE_DEPTH
3931 @item INSN_CACHE_DEPTH
3932 The number of alternative cache lines that can hold any particular memory
3936 Alternatively, if the machine has system calls or instructions to clear
3937 the instruction cache directly, you can define the following macro.
3940 @findex CLEAR_INSN_CACHE
3941 @item CLEAR_INSN_CACHE (@var{BEG}, @var{END})
3942 If defined, expands to a C expression clearing the @emph{instruction
3943 cache} in the specified interval. If it is not defined, and the macro
3944 INSN_CACHE_SIZE is defined, some generic code is generated to clear the
3945 cache. The definition of this macro would typically be a series of
3946 @code{asm} statements. Both @var{BEG} and @var{END} are both pointer
3950 To use a standard subroutine, define the following macro. In addition,
3951 you must make sure that the instructions in a trampoline fill an entire
3952 cache line with identical instructions, or else ensure that the
3953 beginning of the trampoline code is always aligned at the same point in
3954 its cache line. Look in @file{m68k.h} as a guide.
3957 @findex TRANSFER_FROM_TRAMPOLINE
3958 @item TRANSFER_FROM_TRAMPOLINE
3959 Define this macro if trampolines need a special subroutine to do their
3960 work. The macro should expand to a series of @code{asm} statements
3961 which will be compiled with GCC. They go in a library function named
3962 @code{__transfer_from_trampoline}.
3964 If you need to avoid executing the ordinary prologue code of a compiled
3965 C function when you jump to the subroutine, you can do so by placing a
3966 special label of your own in the assembler code. Use one @code{asm}
3967 statement to generate an assembler label, and another to make the label
3968 global. Then trampolines can use that label to jump directly to your
3969 special assembler code.
3973 @section Implicit Calls to Library Routines
3974 @cindex library subroutine names
3975 @cindex @file{libgcc.a}
3977 @c prevent bad page break with this line
3978 Here is an explanation of implicit calls to library routines.
3981 @findex MULSI3_LIBCALL
3982 @item MULSI3_LIBCALL
3983 A C string constant giving the name of the function to call for
3984 multiplication of one signed full-word by another. If you do not
3985 define this macro, the default name is used, which is @code{__mulsi3},
3986 a function defined in @file{libgcc.a}.
3988 @findex DIVSI3_LIBCALL
3989 @item DIVSI3_LIBCALL
3990 A C string constant giving the name of the function to call for
3991 division of one signed full-word by another. If you do not define
3992 this macro, the default name is used, which is @code{__divsi3}, a
3993 function defined in @file{libgcc.a}.
3995 @findex UDIVSI3_LIBCALL
3996 @item UDIVSI3_LIBCALL
3997 A C string constant giving the name of the function to call for
3998 division of one unsigned full-word by another. If you do not define
3999 this macro, the default name is used, which is @code{__udivsi3}, a
4000 function defined in @file{libgcc.a}.
4002 @findex MODSI3_LIBCALL
4003 @item MODSI3_LIBCALL
4004 A C string constant giving the name of the function to call for the
4005 remainder in division of one signed full-word by another. If you do
4006 not define this macro, the default name is used, which is
4007 @code{__modsi3}, a function defined in @file{libgcc.a}.
4009 @findex UMODSI3_LIBCALL
4010 @item UMODSI3_LIBCALL
4011 A C string constant giving the name of the function to call for the
4012 remainder in division of one unsigned full-word by another. If you do
4013 not define this macro, the default name is used, which is
4014 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4016 @findex MULDI3_LIBCALL
4017 @item MULDI3_LIBCALL
4018 A C string constant giving the name of the function to call for
4019 multiplication of one signed double-word by another. If you do not
4020 define this macro, the default name is used, which is @code{__muldi3},
4021 a function defined in @file{libgcc.a}.
4023 @findex DIVDI3_LIBCALL
4024 @item DIVDI3_LIBCALL
4025 A C string constant giving the name of the function to call for
4026 division of one signed double-word by another. If you do not define
4027 this macro, the default name is used, which is @code{__divdi3}, a
4028 function defined in @file{libgcc.a}.
4030 @findex UDIVDI3_LIBCALL
4031 @item UDIVDI3_LIBCALL
4032 A C string constant giving the name of the function to call for
4033 division of one unsigned full-word by another. If you do not define
4034 this macro, the default name is used, which is @code{__udivdi3}, a
4035 function defined in @file{libgcc.a}.
4037 @findex MODDI3_LIBCALL
4038 @item MODDI3_LIBCALL
4039 A C string constant giving the name of the function to call for the
4040 remainder in division of one signed double-word by another. If you do
4041 not define this macro, the default name is used, which is
4042 @code{__moddi3}, a function defined in @file{libgcc.a}.
4044 @findex UMODDI3_LIBCALL
4045 @item UMODDI3_LIBCALL
4046 A C string constant giving the name of the function to call for the
4047 remainder in division of one unsigned full-word by another. If you do
4048 not define this macro, the default name is used, which is
4049 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4051 @findex INIT_TARGET_OPTABS
4052 @item INIT_TARGET_OPTABS
4053 Define this macro as a C statement that declares additional library
4054 routines renames existing ones. @code{init_optabs} calls this macro after
4055 initializing all the normal library routines.
4057 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4058 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4059 Define this macro as a C statement that returns nonzero if a call to
4060 the floating point comparison library function will return a boolean
4061 value that indicates the result of the comparison. It should return
4062 zero if one of gcc's own libgcc functions is called.
4064 Most ports don't need to define this macro.
4067 @cindex @code{EDOM}, implicit usage
4069 The value of @code{EDOM} on the target machine, as a C integer constant
4070 expression. If you don't define this macro, GCC does not attempt to
4071 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4072 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4075 If you do not define @code{TARGET_EDOM}, then compiled code reports
4076 domain errors by calling the library function and letting it report the
4077 error. If mathematical functions on your system use @code{matherr} when
4078 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4079 that @code{matherr} is used normally.
4081 @findex GEN_ERRNO_RTX
4082 @cindex @code{errno}, implicit usage
4084 Define this macro as a C expression to create an rtl expression that
4085 refers to the global ``variable'' @code{errno}. (On certain systems,
4086 @code{errno} may not actually be a variable.) If you don't define this
4087 macro, a reasonable default is used.
4089 @findex TARGET_MEM_FUNCTIONS
4090 @cindex @code{bcopy}, implicit usage
4091 @cindex @code{memcpy}, implicit usage
4092 @cindex @code{bzero}, implicit usage
4093 @cindex @code{memset}, implicit usage
4094 @item TARGET_MEM_FUNCTIONS
4095 Define this macro if GCC should generate calls to the System V
4096 (and ANSI C) library functions @code{memcpy} and @code{memset}
4097 rather than the BSD functions @code{bcopy} and @code{bzero}.
4099 @findex LIBGCC_NEEDS_DOUBLE
4100 @item LIBGCC_NEEDS_DOUBLE
4101 Define this macro if only @code{float} arguments cannot be passed to
4102 library routines (so they must be converted to @code{double}). This
4103 macro affects both how library calls are generated and how the library
4104 routines in @file{libgcc1.c} accept their arguments. It is useful on
4105 machines where floating and fixed point arguments are passed
4106 differently, such as the i860.
4108 @findex FLOAT_ARG_TYPE
4109 @item FLOAT_ARG_TYPE
4110 Define this macro to override the type used by the library routines to
4111 pick up arguments of type @code{float}. (By default, they use a union
4112 of @code{float} and @code{int}.)
4114 The obvious choice would be @code{float}---but that won't work with
4115 traditional C compilers that expect all arguments declared as @code{float}
4116 to arrive as @code{double}. To avoid this conversion, the library routines
4117 ask for the value as some other type and then treat it as a @code{float}.
4119 On some systems, no other type will work for this. For these systems,
4120 you must use @code{LIBGCC_NEEDS_DOUBLE} instead, to force conversion of
4121 the values @code{double} before they are passed.
4124 @item FLOATIFY (@var{passed-value})
4125 Define this macro to override the way library routines redesignate a
4126 @code{float} argument as a @code{float} instead of the type it was
4127 passed as. The default is an expression which takes the @code{float}
4130 @findex FLOAT_VALUE_TYPE
4131 @item FLOAT_VALUE_TYPE
4132 Define this macro to override the type used by the library routines to
4133 return values that ought to have type @code{float}. (By default, they
4136 The obvious choice would be @code{float}---but that won't work with
4137 traditional C compilers gratuitously convert values declared as
4138 @code{float} into @code{double}.
4141 @item INTIFY (@var{float-value})
4142 Define this macro to override the way the value of a
4143 @code{float}-returning library routine should be packaged in order to
4144 return it. These functions are actually declared to return type
4145 @code{FLOAT_VALUE_TYPE} (normally @code{int}).
4147 These values can't be returned as type @code{float} because traditional
4148 C compilers would gratuitously convert the value to a @code{double}.
4150 A local variable named @code{intify} is always available when the macro
4151 @code{INTIFY} is used. It is a union of a @code{float} field named
4152 @code{f} and a field named @code{i} whose type is
4153 @code{FLOAT_VALUE_TYPE} or @code{int}.
4155 If you don't define this macro, the default definition works by copying
4156 the value through that union.
4158 @findex nongcc_SI_type
4159 @item nongcc_SI_type
4160 Define this macro as the name of the data type corresponding to
4161 @code{SImode} in the system's own C compiler.
4163 You need not define this macro if that type is @code{long int}, as it usually
4166 @findex nongcc_word_type
4167 @item nongcc_word_type
4168 Define this macro as the name of the data type corresponding to the
4169 word_mode in the system's own C compiler.
4171 You need not define this macro if that type is @code{long int}, as it usually
4174 @findex perform_@dots{}
4175 @item perform_@dots{}
4176 Define these macros to supply explicit C statements to carry out various
4177 arithmetic operations on types @code{float} and @code{double} in the
4178 library routines in @file{libgcc1.c}. See that file for a full list
4179 of these macros and their arguments.
4181 On most machines, you don't need to define any of these macros, because
4182 the C compiler that comes with the system takes care of doing them.
4184 @findex NEXT_OBJC_RUNTIME
4185 @item NEXT_OBJC_RUNTIME
4186 Define this macro to generate code for Objective C message sending using
4187 the calling convention of the NeXT system. This calling convention
4188 involves passing the object, the selector and the method arguments all
4189 at once to the method-lookup library function.
4191 The default calling convention passes just the object and the selector
4192 to the lookup function, which returns a pointer to the method.
4195 @node Addressing Modes
4196 @section Addressing Modes
4197 @cindex addressing modes
4199 @c prevent bad page break with this line
4200 This is about addressing modes.
4203 @findex HAVE_POST_INCREMENT
4204 @item HAVE_POST_INCREMENT
4205 A C expression that is nonzero the machine supports post-increment addressing.
4207 @findex HAVE_PRE_INCREMENT
4208 @findex HAVE_POST_DECREMENT
4209 @findex HAVE_PRE_DECREMENT
4210 @item HAVE_PRE_INCREMENT
4211 @itemx HAVE_POST_DECREMENT
4212 @itemx HAVE_PRE_DECREMENT
4213 Similar for other kinds of addressing.
4215 @findex CONSTANT_ADDRESS_P
4216 @item CONSTANT_ADDRESS_P (@var{x})
4217 A C expression that is 1 if the RTX @var{x} is a constant which
4218 is a valid address. On most machines, this can be defined as
4219 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4220 in which constant addresses are supported.
4223 @code{CONSTANT_P} accepts integer-values expressions whose values are
4224 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4225 @code{high} expressions and @code{const} arithmetic expressions, in
4226 addition to @code{const_int} and @code{const_double} expressions.
4228 @findex MAX_REGS_PER_ADDRESS
4229 @item MAX_REGS_PER_ADDRESS
4230 A number, the maximum number of registers that can appear in a valid
4231 memory address. Note that it is up to you to specify a value equal to
4232 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4235 @findex GO_IF_LEGITIMATE_ADDRESS
4236 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4237 A C compound statement with a conditional @code{goto @var{label};}
4238 executed if @var{x} (an RTX) is a legitimate memory address on the
4239 target machine for a memory operand of mode @var{mode}.
4241 It usually pays to define several simpler macros to serve as
4242 subroutines for this one. Otherwise it may be too complicated to
4245 This macro must exist in two variants: a strict variant and a
4246 non-strict one. The strict variant is used in the reload pass. It
4247 must be defined so that any pseudo-register that has not been
4248 allocated a hard register is considered a memory reference. In
4249 contexts where some kind of register is required, a pseudo-register
4250 with no hard register must be rejected.
4252 The non-strict variant is used in other passes. It must be defined to
4253 accept all pseudo-registers in every context where some kind of
4254 register is required.
4256 @findex REG_OK_STRICT
4257 Compiler source files that want to use the strict variant of this
4258 macro define the macro @code{REG_OK_STRICT}. You should use an
4259 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4260 in that case and the non-strict variant otherwise.
4262 Subroutines to check for acceptable registers for various purposes (one
4263 for base registers, one for index registers, and so on) are typically
4264 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4265 Then only these subroutine macros need have two variants; the higher
4266 levels of macros may be the same whether strict or not.@refill
4268 Normally, constant addresses which are the sum of a @code{symbol_ref}
4269 and an integer are stored inside a @code{const} RTX to mark them as
4270 constant. Therefore, there is no need to recognize such sums
4271 specifically as legitimate addresses. Normally you would simply
4272 recognize any @code{const} as legitimate.
4274 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4275 sums that are not marked with @code{const}. It assumes that a naked
4276 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4277 naked constant sums as illegitimate addresses, so that none of them will
4278 be given to @code{PRINT_OPERAND_ADDRESS}.
4280 @cindex @code{ENCODE_SECTION_INFO} and address validation
4281 On some machines, whether a symbolic address is legitimate depends on
4282 the section that the address refers to. On these machines, define the
4283 macro @code{ENCODE_SECTION_INFO} to store the information into the
4284 @code{symbol_ref}, and then check for it here. When you see a
4285 @code{const}, you will have to look inside it to find the
4286 @code{symbol_ref} in order to determine the section. @xref{Assembler
4289 @findex saveable_obstack
4290 The best way to modify the name string is by adding text to the
4291 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4292 the new name in @code{saveable_obstack}. You will have to modify
4293 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4294 output the name accordingly, and define @code{STRIP_NAME_ENCODING} to
4295 access the original name string.
4297 You can check the information stored here into the @code{symbol_ref} in
4298 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4299 @code{PRINT_OPERAND_ADDRESS}.
4301 @findex REG_OK_FOR_BASE_P
4302 @item REG_OK_FOR_BASE_P (@var{x})
4303 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4304 RTX) is valid for use as a base register. For hard registers, it
4305 should always accept those which the hardware permits and reject the
4306 others. Whether the macro accepts or rejects pseudo registers must be
4307 controlled by @code{REG_OK_STRICT} as described above. This usually
4308 requires two variant definitions, of which @code{REG_OK_STRICT}
4309 controls the one actually used.
4311 @findex REG_MODE_OK_FOR_BASE_P
4312 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4313 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4314 that expression may examine the mode of the memory reference in
4315 @var{mode}. You should define this macro if the mode of the memory
4316 reference affects whether a register may be used as a base register. If
4317 you define this macro, the compiler will use it instead of
4318 @code{REG_OK_FOR_BASE_P}.
4320 @findex REG_OK_FOR_INDEX_P
4321 @item REG_OK_FOR_INDEX_P (@var{x})
4322 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4323 RTX) is valid for use as an index register.
4325 The difference between an index register and a base register is that
4326 the index register may be scaled. If an address involves the sum of
4327 two registers, neither one of them scaled, then either one may be
4328 labeled the ``base'' and the other the ``index''; but whichever
4329 labeling is used must fit the machine's constraints of which registers
4330 may serve in each capacity. The compiler will try both labelings,
4331 looking for one that is valid, and will reload one or both registers
4332 only if neither labeling works.
4334 @findex LEGITIMIZE_ADDRESS
4335 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4336 A C compound statement that attempts to replace @var{x} with a valid
4337 memory address for an operand of mode @var{mode}. @var{win} will be a
4338 C statement label elsewhere in the code; the macro definition may use
4341 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4345 to avoid further processing if the address has become legitimate.
4347 @findex break_out_memory_refs
4348 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4349 and @var{oldx} will be the operand that was given to that function to produce
4352 The code generated by this macro should not alter the substructure of
4353 @var{x}. If it transforms @var{x} into a more legitimate form, it
4354 should assign @var{x} (which will always be a C variable) a new value.
4356 It is not necessary for this macro to come up with a legitimate
4357 address. The compiler has standard ways of doing so in all cases. In
4358 fact, it is safe for this macro to do nothing. But often a
4359 machine-dependent strategy can generate better code.
4361 @findex LEGITIMIZE_RELOAD_ADDRESS
4362 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4363 A C compound statement that attempts to replace @var{x}, which is an address
4364 that needs reloading, with a valid memory address for an operand of mode
4365 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4366 It is not necessary to define this macro, but it might be useful for
4367 performance reasons.
4369 For example, on the i386, it is sometimes possible to use a single
4370 reload register instead of two by reloading a sum of two pseudo
4371 registers into a register. On the other hand, for number of RISC
4372 processors offsets are limited so that often an intermediate address
4373 needs to be generated in order to address a stack slot. By defining
4374 LEGITIMIZE_RELOAD_ADDRESS appropriately, the intermediate addresses
4375 generated for adjacent some stack slots can be made identical, and thus
4378 @emph{Note}: This macro should be used with caution. It is necessary
4379 to know something of how reload works in order to effectively use this,
4380 and it is quite easy to produce macros that build in too much knowledge
4381 of reload internals.
4383 @emph{Note}: This macro must be able to reload an address created by a
4384 previous invocation of this macro. If it fails to handle such addresses
4385 then the compiler may generate incorrect code or abort.
4388 The macro definition should use @code{push_reload} to indicate parts that
4389 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4390 suitable to be passed unaltered to @code{push_reload}.
4392 The code generated by this macro must not alter the substructure of
4393 @var{x}. If it transforms @var{x} into a more legitimate form, it
4394 should assign @var{x} (which will always be a C variable) a new value.
4395 This also applies to parts that you change indirectly by calling
4398 @findex strict_memory_address_p
4399 The macro definition may use @code{strict_memory_address_p} to test if
4400 the address has become legitimate.
4403 If you want to change only a part of @var{x}, one standard way of doing
4404 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4405 single level of rtl. Thus, if the part to be changed is not at the
4406 top level, you'll need to replace first the top leve
4407 It is not necessary for this macro to come up with a legitimate
4408 address; but often a machine-dependent strategy can generate better code.
4410 @findex GO_IF_MODE_DEPENDENT_ADDRESS
4411 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4412 A C statement or compound statement with a conditional @code{goto
4413 @var{label};} executed if memory address @var{x} (an RTX) can have
4414 different meanings depending on the machine mode of the memory
4415 reference it is used for or if the address is valid for some modes
4418 Autoincrement and autodecrement addresses typically have mode-dependent
4419 effects because the amount of the increment or decrement is the size
4420 of the operand being addressed. Some machines have other mode-dependent
4421 addresses. Many RISC machines have no mode-dependent addresses.
4423 You may assume that @var{addr} is a valid address for the machine.
4425 @findex LEGITIMATE_CONSTANT_P
4426 @item LEGITIMATE_CONSTANT_P (@var{x})
4427 A C expression that is nonzero if @var{x} is a legitimate constant for
4428 an immediate operand on the target machine. You can assume that
4429 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4430 @samp{1} is a suitable definition for this macro on machines where
4431 anything @code{CONSTANT_P} is valid.@refill
4434 @node Condition Code
4435 @section Condition Code Status
4436 @cindex condition code status
4438 @c prevent bad page break with this line
4439 This describes the condition code status.
4442 The file @file{conditions.h} defines a variable @code{cc_status} to
4443 describe how the condition code was computed (in case the interpretation of
4444 the condition code depends on the instruction that it was set by). This
4445 variable contains the RTL expressions on which the condition code is
4446 currently based, and several standard flags.
4448 Sometimes additional machine-specific flags must be defined in the machine
4449 description header file. It can also add additional machine-specific
4450 information by defining @code{CC_STATUS_MDEP}.
4453 @findex CC_STATUS_MDEP
4454 @item CC_STATUS_MDEP
4455 C code for a data type which is used for declaring the @code{mdep}
4456 component of @code{cc_status}. It defaults to @code{int}.
4458 This macro is not used on machines that do not use @code{cc0}.
4460 @findex CC_STATUS_MDEP_INIT
4461 @item CC_STATUS_MDEP_INIT
4462 A C expression to initialize the @code{mdep} field to ``empty''.
4463 The default definition does nothing, since most machines don't use
4464 the field anyway. If you want to use the field, you should probably
4465 define this macro to initialize it.
4467 This macro is not used on machines that do not use @code{cc0}.
4469 @findex NOTICE_UPDATE_CC
4470 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4471 A C compound statement to set the components of @code{cc_status}
4472 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4473 this macro's responsibility to recognize insns that set the condition
4474 code as a byproduct of other activity as well as those that explicitly
4477 This macro is not used on machines that do not use @code{cc0}.
4479 If there are insns that do not set the condition code but do alter
4480 other machine registers, this macro must check to see whether they
4481 invalidate the expressions that the condition code is recorded as
4482 reflecting. For example, on the 68000, insns that store in address
4483 registers do not set the condition code, which means that usually
4484 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4485 insns. But suppose that the previous insn set the condition code
4486 based on location @samp{a4@@(102)} and the current insn stores a new
4487 value in @samp{a4}. Although the condition code is not changed by
4488 this, it will no longer be true that it reflects the contents of
4489 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4490 @code{cc_status} in this case to say that nothing is known about the
4491 condition code value.
4493 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4494 with the results of peephole optimization: insns whose patterns are
4495 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4496 constants which are just the operands. The RTL structure of these
4497 insns is not sufficient to indicate what the insns actually do. What
4498 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4499 @code{CC_STATUS_INIT}.
4501 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4502 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4503 @samp{cc}. This avoids having detailed information about patterns in
4504 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4506 @findex EXTRA_CC_MODES
4507 @item EXTRA_CC_MODES
4508 A list of additional modes for condition code values in registers
4509 (@pxref{Jump Patterns}). This macro should expand to a sequence of
4510 calls of the macro @code{CC} separated by white space. @code{CC} takes
4511 two arguments. The first is the enumeration name of the mode, which
4512 should begin with @samp{CC} and end with @samp{mode}. The second is a C
4513 string giving the printable name of the mode; it should be the same as
4514 the first argument, but with the trailing @samp{mode} removed.
4516 You should only define this macro if additional modes are required.
4518 A sample definition of @code{EXTRA_CC_MODES} is:
4520 #define EXTRA_CC_MODES \
4521 CC(CC_NOOVmode, "CC_NOOV") \
4522 CC(CCFPmode, "CCFP") \
4523 CC(CCFPEmode, "CCFPE")
4526 @findex SELECT_CC_MODE
4527 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4528 Returns a mode from class @code{MODE_CC} to be used when comparison
4529 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
4530 example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
4531 @pxref{Jump Patterns} for a description of the reason for this
4535 #define SELECT_CC_MODE(OP,X,Y) \
4536 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4537 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
4538 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4539 || GET_CODE (X) == NEG) \
4540 ? CC_NOOVmode : CCmode))
4543 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
4545 @findex CANONICALIZE_COMPARISON
4546 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
4547 On some machines not all possible comparisons are defined, but you can
4548 convert an invalid comparison into a valid one. For example, the Alpha
4549 does not have a @code{GT} comparison, but you can use an @code{LT}
4550 comparison instead and swap the order of the operands.
4552 On such machines, define this macro to be a C statement to do any
4553 required conversions. @var{code} is the initial comparison code
4554 and @var{op0} and @var{op1} are the left and right operands of the
4555 comparison, respectively. You should modify @var{code}, @var{op0}, and
4556 @var{op1} as required.
4558 GCC will not assume that the comparison resulting from this macro is
4559 valid but will see if the resulting insn matches a pattern in the
4562 You need not define this macro if it would never change the comparison
4565 @findex REVERSIBLE_CC_MODE
4566 @item REVERSIBLE_CC_MODE (@var{mode})
4567 A C expression whose value is one if it is always safe to reverse a
4568 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4569 can ever return @var{mode} for a floating-point inequality comparison,
4570 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4572 You need not define this macro if it would always returns zero or if the
4573 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4574 For example, here is the definition used on the Sparc, where floating-point
4575 inequality comparisons are always given @code{CCFPEmode}:
4578 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
4584 @section Describing Relative Costs of Operations
4585 @cindex costs of instructions
4586 @cindex relative costs
4587 @cindex speed of instructions
4589 These macros let you describe the relative speed of various operations
4590 on the target machine.
4594 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
4595 A part of a C @code{switch} statement that describes the relative costs
4596 of constant RTL expressions. It must contain @code{case} labels for
4597 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
4598 @code{label_ref} and @code{const_double}. Each case must ultimately
4599 reach a @code{return} statement to return the relative cost of the use
4600 of that kind of constant value in an expression. The cost may depend on
4601 the precise value of the constant, which is available for examination in
4602 @var{x}, and the rtx code of the expression in which it is contained,
4603 found in @var{outer_code}.
4605 @var{code} is the expression code---redundant, since it can be
4606 obtained with @code{GET_CODE (@var{x})}.
4609 @findex COSTS_N_INSNS
4610 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4611 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
4612 This can be used, for example, to indicate how costly a multiply
4613 instruction is. In writing this macro, you can use the construct
4614 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
4615 instructions. @var{outer_code} is the code of the expression in which
4616 @var{x} is contained.
4618 This macro is optional; do not define it if the default cost assumptions
4619 are adequate for the target machine.
4621 @findex DEFAULT_RTX_COSTS
4622 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4623 This macro, if defined, is called for any case not handled by the
4624 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
4625 to put case labels into the macro, but the code, or any functions it
4626 calls, must assume that the RTL in @var{x} could be of any type that has
4627 not already been handled. The arguments are the same as for
4628 @code{RTX_COSTS}, and the macro should execute a return statement giving
4629 the cost of any RTL expressions that it can handle. The default cost
4630 calculation is used for any RTL for which this macro does not return a
4633 This macro is optional; do not define it if the default cost assumptions
4634 are adequate for the target machine.
4636 @findex ADDRESS_COST
4637 @item ADDRESS_COST (@var{address})
4638 An expression giving the cost of an addressing mode that contains
4639 @var{address}. If not defined, the cost is computed from
4640 the @var{address} expression and the @code{CONST_COSTS} values.
4642 For most CISC machines, the default cost is a good approximation of the
4643 true cost of the addressing mode. However, on RISC machines, all
4644 instructions normally have the same length and execution time. Hence
4645 all addresses will have equal costs.
4647 In cases where more than one form of an address is known, the form with
4648 the lowest cost will be used. If multiple forms have the same, lowest,
4649 cost, the one that is the most complex will be used.
4651 For example, suppose an address that is equal to the sum of a register
4652 and a constant is used twice in the same basic block. When this macro
4653 is not defined, the address will be computed in a register and memory
4654 references will be indirect through that register. On machines where
4655 the cost of the addressing mode containing the sum is no higher than
4656 that of a simple indirect reference, this will produce an additional
4657 instruction and possibly require an additional register. Proper
4658 specification of this macro eliminates this overhead for such machines.
4660 Similar use of this macro is made in strength reduction of loops.
4662 @var{address} need not be valid as an address. In such a case, the cost
4663 is not relevant and can be any value; invalid addresses need not be
4664 assigned a different cost.
4666 On machines where an address involving more than one register is as
4667 cheap as an address computation involving only one register, defining
4668 @code{ADDRESS_COST} to reflect this can cause two registers to be live
4669 over a region of code where only one would have been if
4670 @code{ADDRESS_COST} were not defined in that manner. This effect should
4671 be considered in the definition of this macro. Equivalent costs should
4672 probably only be given to addresses with different numbers of registers
4673 on machines with lots of registers.
4675 This macro will normally either not be defined or be defined as a
4678 @findex REGISTER_MOVE_COST
4679 @item REGISTER_MOVE_COST (@var{from}, @var{to})
4680 A C expression for the cost of moving data from a register in class
4681 @var{from} to one in class @var{to}. The classes are expressed using
4682 the enumeration values such as @code{GENERAL_REGS}. A value of 2 is the
4683 default; other values are interpreted relative to that.
4685 It is not required that the cost always equal 2 when @var{from} is the
4686 same as @var{to}; on some machines it is expensive to move between
4687 registers if they are not general registers.
4689 If reload sees an insn consisting of a single @code{set} between two
4690 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4691 classes returns a value of 2, reload does not check to ensure that the
4692 constraints of the insn are met. Setting a cost of other than 2 will
4693 allow reload to verify that the constraints are met. You should do this
4694 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4696 @findex MEMORY_MOVE_COST
4697 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4698 A C expression for the cost of moving data of mode @var{mode} between a
4699 register of class @var{class} and memory; @var{in} is zero if the value
4700 is to be written to memory, non-zero if it is to be read in. This cost
4701 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
4702 registers and memory is more expensive than between two registers, you
4703 should define this macro to express the relative cost.
4705 If you do not define this macro, GCC uses a default cost of 4 plus
4706 the cost of copying via a secondary reload register, if one is
4707 needed. If your machine requires a secondary reload register to copy
4708 between memory and a register of @var{class} but the reload mechanism is
4709 more complex than copying via an intermediate, define this macro to
4710 reflect the actual cost of the move.
4712 GCC defines the function @code{memory_move_secondary_cost} if
4713 secondary reloads are needed. It computes the costs due to copying via
4714 a secondary register. If your machine copies from memory using a
4715 secondary register in the conventional way but the default base value of
4716 4 is not correct for your machine, define this macro to add some other
4717 value to the result of that function. The arguments to that function
4718 are the same as to this macro.
4722 A C expression for the cost of a branch instruction. A value of 1 is
4723 the default; other values are interpreted relative to that.
4726 Here are additional macros which do not specify precise relative costs,
4727 but only that certain actions are more expensive than GCC would
4731 @findex SLOW_BYTE_ACCESS
4732 @item SLOW_BYTE_ACCESS
4733 Define this macro as a C expression which is nonzero if accessing less
4734 than a word of memory (i.e. a @code{char} or a @code{short}) is no
4735 faster than accessing a word of memory, i.e., if such access
4736 require more than one instruction or if there is no difference in cost
4737 between byte and (aligned) word loads.
4739 When this macro is not defined, the compiler will access a field by
4740 finding the smallest containing object; when it is defined, a fullword
4741 load will be used if alignment permits. Unless bytes accesses are
4742 faster than word accesses, using word accesses is preferable since it
4743 may eliminate subsequent memory access if subsequent accesses occur to
4744 other fields in the same word of the structure, but to different bytes.
4746 @findex SLOW_ZERO_EXTEND
4747 @item SLOW_ZERO_EXTEND
4748 Define this macro if zero-extension (of a @code{char} or @code{short}
4749 to an @code{int}) can be done faster if the destination is a register
4750 that is known to be zero.
4752 If you define this macro, you must have instruction patterns that
4753 recognize RTL structures like this:
4756 (set (strict_low_part (subreg:QI (reg:SI @dots{}) 0)) @dots{})
4760 and likewise for @code{HImode}.
4762 @findex SLOW_UNALIGNED_ACCESS
4763 @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
4764 Define this macro to be the value 1 if memory accesses described by the
4765 @var{mode} and @var{alignment} parameters have a cost many times greater
4766 than aligned accesses, for example if they are emulated in a trap
4769 When this macro is non-zero, the compiler will act as if
4770 @code{STRICT_ALIGNMENT} were non-zero when generating code for block
4771 moves. This can cause significantly more instructions to be produced.
4772 Therefore, do not set this macro non-zero if unaligned accesses only add a
4773 cycle or two to the time for a memory access.
4775 If the value of this macro is always zero, it need not be defined. If
4776 this macro is defined, it should produce a non-zero value when
4777 @code{STRICT_ALIGNMENT} is non-zero.
4779 @findex DONT_REDUCE_ADDR
4780 @item DONT_REDUCE_ADDR
4781 Define this macro to inhibit strength reduction of memory addresses.
4782 (On some machines, such strength reduction seems to do harm rather
4787 The threshold of number of scalar memory-to-memory move insns, @emph{below}
4788 which a sequence of insns should be generated instead of a
4789 string move insn or a library call. Increasing the value will always
4790 make code faster, but eventually incurs high cost in increased code size.
4792 Note that on machines where the corresponding move insn is a
4793 @code{define_expand} that emits a sequence of insns, this macro counts
4794 the number of such sequences.
4796 If you don't define this, a reasonable default is used.
4798 @findex MOVE_BY_PIECES_P
4799 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
4800 A C expression used to determine whether @code{move_by_pieces} will be used to
4801 copy a chunk of memory, or whether some other block move mechanism
4802 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
4803 than @code{MOVE_RATIO}.
4805 @findex MOVE_MAX_PIECES
4806 @item MOVE_MAX_PIECES
4807 A C expression used by @code{move_by_pieces} to determine the largest unit
4808 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
4810 @findex USE_LOAD_POST_INCREMENT
4811 @item USE_LOAD_POST_INCREMENT (@var{mode})
4812 A C expression used to determine whether a load postincrement is a good
4813 thing to use for a given mode. Defaults to the value of
4814 @code{HAVE_POST_INCREMENT}.
4816 @findex USE_LOAD_POST_DECREMENT
4817 @item USE_LOAD_POST_DECREMENT (@var{mode})
4818 A C expression used to determine whether a load postdecrement is a good
4819 thing to use for a given mode. Defaults to the value of
4820 @code{HAVE_POST_DECREMENT}.
4822 @findex USE_LOAD_PRE_INCREMENT
4823 @item USE_LOAD_PRE_INCREMENT (@var{mode})
4824 A C expression used to determine whether a load preincrement is a good
4825 thing to use for a given mode. Defaults to the value of
4826 @code{HAVE_PRE_INCREMENT}.
4828 @findex USE_LOAD_PRE_DECREMENT
4829 @item USE_LOAD_PRE_DECREMENT (@var{mode})
4830 A C expression used to determine whether a load predecrement is a good
4831 thing to use for a given mode. Defaults to the value of
4832 @code{HAVE_PRE_DECREMENT}.
4834 @findex USE_STORE_POST_INCREMENT
4835 @item USE_STORE_POST_INCREMENT (@var{mode})
4836 A C expression used to determine whether a store postincrement is a good
4837 thing to use for a given mode. Defaults to the value of
4838 @code{HAVE_POST_INCREMENT}.
4840 @findex USE_STORE_POST_DECREMENT
4841 @item USE_STORE_POST_DECREMENT (@var{mode})
4842 A C expression used to determine whether a store postdeccrement is a good
4843 thing to use for a given mode. Defaults to the value of
4844 @code{HAVE_POST_DECREMENT}.
4846 @findex USE_STORE_PRE_INCREMENT
4847 @item USE_STORE_PRE_INCREMENT (@var{mode})
4848 This macro is used to determine whether a store preincrement is a good
4849 thing to use for a given mode. Defaults to the value of
4850 @code{HAVE_PRE_INCREMENT}.
4852 @findex USE_STORE_PRE_DECREMENT
4853 @item USE_STORE_PRE_DECREMENT (@var{mode})
4854 This macro is used to determine whether a store predecrement is a good
4855 thing to use for a given mode. Defaults to the value of
4856 @code{HAVE_PRE_DECREMENT}.
4858 @findex NO_FUNCTION_CSE
4859 @item NO_FUNCTION_CSE
4860 Define this macro if it is as good or better to call a constant
4861 function address than to call an address kept in a register.
4863 @findex NO_RECURSIVE_FUNCTION_CSE
4864 @item NO_RECURSIVE_FUNCTION_CSE
4865 Define this macro if it is as good or better for a function to call
4866 itself with an explicit address than to call an address kept in a
4870 @item ADJUST_COST (@var{insn}, @var{link}, @var{dep_insn}, @var{cost})
4871 A C statement (sans semicolon) to update the integer variable @var{cost}
4872 based on the relationship between @var{insn} that is dependent on
4873 @var{dep_insn} through the dependence @var{link}. The default is to
4874 make no adjustment to @var{cost}. This can be used for example to
4875 specify to the scheduler that an output- or anti-dependence does not
4876 incur the same cost as a data-dependence.
4878 @findex ADJUST_PRIORITY
4879 @item ADJUST_PRIORITY (@var{insn})
4880 A C statement (sans semicolon) to update the integer scheduling
4881 priority @code{INSN_PRIORITY(@var{insn})}. Reduce the priority
4882 to execute the @var{insn} earlier, increase the priority to execute
4883 @var{insn} later. Do not define this macro if you do not need to
4884 adjust the scheduling priorities of insns.
4888 @section Dividing the Output into Sections (Texts, Data, @dots{})
4889 @c the above section title is WAY too long. maybe cut the part between
4890 @c the (...)? --mew 10feb93
4892 An object file is divided into sections containing different types of
4893 data. In the most common case, there are three sections: the @dfn{text
4894 section}, which holds instructions and read-only data; the @dfn{data
4895 section}, which holds initialized writable data; and the @dfn{bss
4896 section}, which holds uninitialized data. Some systems have other kinds
4899 The compiler must tell the assembler when to switch sections. These
4900 macros control what commands to output to tell the assembler this. You
4901 can also define additional sections.
4904 @findex TEXT_SECTION_ASM_OP
4905 @item TEXT_SECTION_ASM_OP
4906 A C expression whose value is a string containing the assembler
4907 operation that should precede instructions and read-only data. Normally
4908 @code{".text"} is right.
4910 @findex DATA_SECTION_ASM_OP
4911 @item DATA_SECTION_ASM_OP
4912 A C expression whose value is a string containing the assembler
4913 operation to identify the following data as writable initialized data.
4914 Normally @code{".data"} is right.
4916 @findex SHARED_SECTION_ASM_OP
4917 @item SHARED_SECTION_ASM_OP
4918 If defined, a C expression whose value is a string containing the
4919 assembler operation to identify the following data as shared data. If
4920 not defined, @code{DATA_SECTION_ASM_OP} will be used.
4922 @findex BSS_SECTION_ASM_OP
4923 @item BSS_SECTION_ASM_OP
4924 If defined, a C expression whose value is a string containing the
4925 assembler operation to identify the following data as uninitialized global
4926 data. If not defined, and neither @code{ASM_OUTPUT_BSS} nor
4927 @code{ASM_OUTPUT_ALIGNED_BSS} are defined, uninitialized global data will be
4928 output in the data section if @samp{-fno-common} is passed, otherwise
4929 @code{ASM_OUTPUT_COMMON} will be used.
4931 @findex SHARED_BSS_SECTION_ASM_OP
4932 @item SHARED_BSS_SECTION_ASM_OP
4933 If defined, a C expression whose value is a string containing the
4934 assembler operation to identify the following data as uninitialized global
4935 shared data. If not defined, and @code{BSS_SECTION_ASM_OP} is, the latter
4938 @findex INIT_SECTION_ASM_OP
4939 @item INIT_SECTION_ASM_OP
4940 If defined, a C expression whose value is a string containing the
4941 assembler operation to identify the following data as initialization
4942 code. If not defined, GCC will assume such a section does not
4945 @findex EXTRA_SECTIONS
4948 @item EXTRA_SECTIONS
4949 A list of names for sections other than the standard two, which are
4950 @code{in_text} and @code{in_data}. You need not define this macro
4951 on a system with no other sections (that GCC needs to use).
4953 @findex EXTRA_SECTION_FUNCTIONS
4954 @findex text_section
4955 @findex data_section
4956 @item EXTRA_SECTION_FUNCTIONS
4957 One or more functions to be defined in @file{varasm.c}. These
4958 functions should do jobs analogous to those of @code{text_section} and
4959 @code{data_section}, for your additional sections. Do not define this
4960 macro if you do not define @code{EXTRA_SECTIONS}.
4962 @findex READONLY_DATA_SECTION
4963 @item READONLY_DATA_SECTION
4964 On most machines, read-only variables, constants, and jump tables are
4965 placed in the text section. If this is not the case on your machine,
4966 this macro should be defined to be the name of a function (either
4967 @code{data_section} or a function defined in @code{EXTRA_SECTIONS}) that
4968 switches to the section to be used for read-only items.
4970 If these items should be placed in the text section, this macro should
4973 @findex SELECT_SECTION
4974 @item SELECT_SECTION (@var{exp}, @var{reloc})
4975 A C statement or statements to switch to the appropriate section for
4976 output of @var{exp}. You can assume that @var{exp} is either a
4977 @code{VAR_DECL} node or a constant of some sort. @var{reloc}
4978 indicates whether the initial value of @var{exp} requires link-time
4979 relocations. Select the section by calling @code{text_section} or one
4980 of the alternatives for other sections.
4982 Do not define this macro if you put all read-only variables and
4983 constants in the read-only data section (usually the text section).
4985 @findex SELECT_RTX_SECTION
4986 @item SELECT_RTX_SECTION (@var{mode}, @var{rtx})
4987 A C statement or statements to switch to the appropriate section for
4988 output of @var{rtx} in mode @var{mode}. You can assume that @var{rtx}
4989 is some kind of constant in RTL. The argument @var{mode} is redundant
4990 except in the case of a @code{const_int} rtx. Select the section by
4991 calling @code{text_section} or one of the alternatives for other
4994 Do not define this macro if you put all constants in the read-only
4997 @findex JUMP_TABLES_IN_TEXT_SECTION
4998 @item JUMP_TABLES_IN_TEXT_SECTION
4999 Define this macro to be an expression with a non-zero value if jump
5000 tables (for @code{tablejump} insns) should be output in the text
5001 section, along with the assembler instructions. Otherwise, the
5002 readonly data section is used.
5004 This macro is irrelevant if there is no separate readonly data section.
5006 @findex ENCODE_SECTION_INFO
5007 @item ENCODE_SECTION_INFO (@var{decl})
5008 Define this macro if references to a symbol must be treated differently
5009 depending on something about the variable or function named by the
5010 symbol (such as what section it is in).
5012 The macro definition, if any, is executed immediately after the rtl for
5013 @var{decl} has been created and stored in @code{DECL_RTL (@var{decl})}.
5014 The value of the rtl will be a @code{mem} whose address is a
5017 @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO}
5018 The usual thing for this macro to do is to record a flag in the
5019 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
5020 modified name string in the @code{symbol_ref} (if one bit is not enough
5023 @findex STRIP_NAME_ENCODING
5024 @item STRIP_NAME_ENCODING (@var{var}, @var{sym_name})
5025 Decode @var{sym_name} and store the real name part in @var{var}, sans
5026 the characters that encode section info. Define this macro if
5027 @code{ENCODE_SECTION_INFO} alters the symbol's name string.
5029 @findex UNIQUE_SECTION_P
5030 @item UNIQUE_SECTION_P (@var{decl})
5031 A C expression which evaluates to true if @var{decl} should be placed
5032 into a unique section for some target-specific reason. If you do not
5033 define this macro, the default is @samp{0}. Note that the flag
5034 @samp{-ffunction-sections} will also cause functions to be placed into
5037 @findex UNIQUE_SECTION
5038 @item UNIQUE_SECTION (@var{decl}, @var{reloc})
5039 A C statement to build up a unique section name, expressed as a
5040 STRING_CST node, and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5041 @var{reloc} indicates whether the initial value of @var{exp} requires
5042 link-time relocations. If you do not define this macro, GCC will use
5043 the symbol name prefixed by @samp{.} as the section name. Note - this
5044 macro can now be called for unitialised data items as well as
5045 initialised data and functions.
5049 @section Position Independent Code
5050 @cindex position independent code
5053 This section describes macros that help implement generation of position
5054 independent code. Simply defining these macros is not enough to
5055 generate valid PIC; you must also add support to the macros
5056 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5057 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5058 @samp{movsi} to do something appropriate when the source operand
5059 contains a symbolic address. You may also need to alter the handling of
5060 switch statements so that they use relative addresses.
5061 @c i rearranged the order of the macros above to try to force one of
5062 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5065 @findex PIC_OFFSET_TABLE_REGNUM
5066 @item PIC_OFFSET_TABLE_REGNUM
5067 The register number of the register used to address a table of static
5068 data addresses in memory. In some cases this register is defined by a
5069 processor's ``application binary interface'' (ABI). When this macro
5070 is defined, RTL is generated for this register once, as with the stack
5071 pointer and frame pointer registers. If this macro is not defined, it
5072 is up to the machine-dependent files to allocate such a register (if
5075 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5076 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5077 Define this macro if the register defined by
5078 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
5079 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5081 @findex FINALIZE_PIC
5083 By generating position-independent code, when two different programs (A
5084 and B) share a common library (libC.a), the text of the library can be
5085 shared whether or not the library is linked at the same address for both
5086 programs. In some of these environments, position-independent code
5087 requires not only the use of different addressing modes, but also
5088 special code to enable the use of these addressing modes.
5090 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
5091 codes once the function is being compiled into assembly code, but not
5092 before. (It is not done before, because in the case of compiling an
5093 inline function, it would lead to multiple PIC prologues being
5094 included in functions which used inline functions and were compiled to
5097 @findex LEGITIMATE_PIC_OPERAND_P
5098 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
5099 A C expression that is nonzero if @var{x} is a legitimate immediate
5100 operand on the target machine when generating position independent code.
5101 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5102 check this. You can also assume @var{flag_pic} is true, so you need not
5103 check it either. You need not define this macro if all constants
5104 (including @code{SYMBOL_REF}) can be immediate operands when generating
5105 position independent code.
5108 @node Assembler Format
5109 @section Defining the Output Assembler Language
5111 This section describes macros whose principal purpose is to describe how
5112 to write instructions in assembler language--rather than what the
5116 * File Framework:: Structural information for the assembler file.
5117 * Data Output:: Output of constants (numbers, strings, addresses).
5118 * Uninitialized Data:: Output of uninitialized variables.
5119 * Label Output:: Output and generation of labels.
5120 * Initialization:: General principles of initialization
5121 and termination routines.
5122 * Macros for Initialization::
5123 Specific macros that control the handling of
5124 initialization and termination routines.
5125 * Instruction Output:: Output of actual instructions.
5126 * Dispatch Tables:: Output of jump tables.
5127 * Exception Region Output:: Output of exception region code.
5128 * Alignment Output:: Pseudo ops for alignment and skipping data.
5131 @node File Framework
5132 @subsection The Overall Framework of an Assembler File
5133 @cindex assembler format
5134 @cindex output of assembler code
5136 @c prevent bad page break with this line
5137 This describes the overall framework of an assembler file.
5140 @findex ASM_FILE_START
5141 @item ASM_FILE_START (@var{stream})
5142 A C expression which outputs to the stdio stream @var{stream}
5143 some appropriate text to go at the start of an assembler file.
5145 Normally this macro is defined to output a line containing
5146 @samp{#NO_APP}, which is a comment that has no effect on most
5147 assemblers but tells the GNU assembler that it can save time by not
5148 checking for certain assembler constructs.
5150 On systems that use SDB, it is necessary to output certain commands;
5151 see @file{attasm.h}.
5153 @findex ASM_FILE_END
5154 @item ASM_FILE_END (@var{stream})
5155 A C expression which outputs to the stdio stream @var{stream}
5156 some appropriate text to go at the end of an assembler file.
5158 If this macro is not defined, the default is to output nothing
5159 special at the end of the file. Most systems don't require any
5162 On systems that use SDB, it is necessary to output certain commands;
5163 see @file{attasm.h}.
5165 @findex ASM_IDENTIFY_GCC
5166 @item ASM_IDENTIFY_GCC (@var{file})
5167 A C statement to output assembler commands which will identify
5168 the object file as having been compiled with GCC (or another
5171 If you don't define this macro, the string @samp{gcc_compiled.:}
5172 is output. This string is calculated to define a symbol which,
5173 on BSD systems, will never be defined for any other reason.
5174 GDB checks for the presence of this symbol when reading the
5175 symbol table of an executable.
5177 On non-BSD systems, you must arrange communication with GDB in
5178 some other fashion. If GDB is not used on your system, you can
5179 define this macro with an empty body.
5181 @findex ASM_COMMENT_START
5182 @item ASM_COMMENT_START
5183 A C string constant describing how to begin a comment in the target
5184 assembler language. The compiler assumes that the comment will end at
5185 the end of the line.
5189 A C string constant for text to be output before each @code{asm}
5190 statement or group of consecutive ones. Normally this is
5191 @code{"#APP"}, which is a comment that has no effect on most
5192 assemblers but tells the GNU assembler that it must check the lines
5193 that follow for all valid assembler constructs.
5197 A C string constant for text to be output after each @code{asm}
5198 statement or group of consecutive ones. Normally this is
5199 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5200 time-saving assumptions that are valid for ordinary compiler output.
5202 @findex ASM_OUTPUT_SOURCE_FILENAME
5203 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5204 A C statement to output COFF information or DWARF debugging information
5205 which indicates that filename @var{name} is the current source file to
5206 the stdio stream @var{stream}.
5208 This macro need not be defined if the standard form of output
5209 for the file format in use is appropriate.
5211 @findex OUTPUT_QUOTED_STRING
5212 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5213 A C statement to output the string @var{string} to the stdio stream
5214 @var{stream}. If you do not call the function @code{output_quoted_string}
5215 in your config files, GCC will only call it to output filenames to
5216 the assembler source. So you can use it to canonicalize the format
5217 of the filename using this macro.
5219 @findex ASM_OUTPUT_SOURCE_LINE
5220 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
5221 A C statement to output DBX or SDB debugging information before code
5222 for line number @var{line} of the current source file to the
5223 stdio stream @var{stream}.
5225 This macro need not be defined if the standard form of debugging
5226 information for the debugger in use is appropriate.
5228 @findex ASM_OUTPUT_IDENT
5229 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
5230 A C statement to output something to the assembler file to handle a
5231 @samp{#ident} directive containing the text @var{string}. If this
5232 macro is not defined, nothing is output for a @samp{#ident} directive.
5234 @findex ASM_OUTPUT_SECTION_NAME
5235 @item ASM_OUTPUT_SECTION_NAME (@var{stream}, @var{decl}, @var{name}, @var{reloc})
5236 A C statement to output something to the assembler file to switch to section
5237 @var{name} for object @var{decl} which is either a @code{FUNCTION_DECL}, a
5238 @code{VAR_DECL} or @code{NULL_TREE}. @var{reloc}
5239 indicates whether the initial value of @var{exp} requires link-time
5240 relocations. Some target formats do not support
5241 arbitrary sections. Do not define this macro in such cases.
5243 At present this macro is only used to support section attributes.
5244 When this macro is undefined, section attributes are disabled.
5246 @findex OBJC_PROLOGUE
5248 A C statement to output any assembler statements which are required to
5249 precede any Objective C object definitions or message sending. The
5250 statement is executed only when compiling an Objective C program.
5255 @subsection Output of Data
5257 @c prevent bad page break with this line
5258 This describes data output.
5261 @findex ASM_OUTPUT_LONG_DOUBLE
5262 @findex ASM_OUTPUT_DOUBLE
5263 @findex ASM_OUTPUT_FLOAT
5264 @item ASM_OUTPUT_LONG_DOUBLE (@var{stream}, @var{value})
5265 @itemx ASM_OUTPUT_DOUBLE (@var{stream}, @var{value})
5266 @itemx ASM_OUTPUT_FLOAT (@var{stream}, @var{value})
5267 @itemx ASM_OUTPUT_THREE_QUARTER_FLOAT (@var{stream}, @var{value})
5268 @itemx ASM_OUTPUT_SHORT_FLOAT (@var{stream}, @var{value})
5269 @itemx ASM_OUTPUT_BYTE_FLOAT (@var{stream}, @var{value})
5270 A C statement to output to the stdio stream @var{stream} an assembler
5271 instruction to assemble a floating-point constant of @code{TFmode},
5272 @code{DFmode}, @code{SFmode}, @code{TQFmode}, @code{HFmode}, or
5273 @code{QFmode}, respectively, whose value is @var{value}. @var{value}
5274 will be a C expression of type @code{REAL_VALUE_TYPE}. Macros such as
5275 @code{REAL_VALUE_TO_TARGET_DOUBLE} are useful for writing these
5278 @findex ASM_OUTPUT_QUADRUPLE_INT
5279 @findex ASM_OUTPUT_DOUBLE_INT
5280 @findex ASM_OUTPUT_INT
5281 @findex ASM_OUTPUT_SHORT
5282 @findex ASM_OUTPUT_CHAR
5283 @findex output_addr_const
5284 @item ASM_OUTPUT_QUADRUPLE_INT (@var{stream}, @var{exp})
5285 @itemx ASM_OUTPUT_DOUBLE_INT (@var{stream}, @var{exp})
5286 @itemx ASM_OUTPUT_INT (@var{stream}, @var{exp})
5287 @itemx ASM_OUTPUT_SHORT (@var{stream}, @var{exp})
5288 @itemx ASM_OUTPUT_CHAR (@var{stream}, @var{exp})
5289 A C statement to output to the stdio stream @var{stream} an assembler
5290 instruction to assemble an integer of 16, 8, 4, 2 or 1 bytes,
5291 respectively, whose value is @var{value}. The argument @var{exp} will
5292 be an RTL expression which represents a constant value. Use
5293 @samp{output_addr_const (@var{stream}, @var{exp})} to output this value
5294 as an assembler expression.@refill
5296 For sizes larger than @code{UNITS_PER_WORD}, if the action of a macro
5297 would be identical to repeatedly calling the macro corresponding to
5298 a size of @code{UNITS_PER_WORD}, once for each word, you need not define
5301 @findex ASM_OUTPUT_BYTE
5302 @item ASM_OUTPUT_BYTE (@var{stream}, @var{value})
5303 A C statement to output to the stdio stream @var{stream} an assembler
5304 instruction to assemble a single byte containing the number @var{value}.
5308 A C string constant giving the pseudo-op to use for a sequence of
5309 single-byte constants. If this macro is not defined, the default is
5312 @findex ASM_OUTPUT_ASCII
5313 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5314 A C statement to output to the stdio stream @var{stream} an assembler
5315 instruction to assemble a string constant containing the @var{len}
5316 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5317 @code{char *} and @var{len} a C expression of type @code{int}.
5319 If the assembler has a @code{.ascii} pseudo-op as found in the
5320 Berkeley Unix assembler, do not define the macro
5321 @code{ASM_OUTPUT_ASCII}.
5323 @findex CONSTANT_POOL_BEFORE_FUNCTION
5324 @item CONSTANT_POOL_BEFORE_FUNCTION
5325 You may define this macro as a C expression. You should define the
5326 expression to have a non-zero value if GCC should output the constant
5327 pool for a function before the code for the function, or a zero value if
5328 GCC should output the constant pool after the function. If you do
5329 not define this macro, the usual case, GCC will output the constant
5330 pool before the function.
5332 @findex ASM_OUTPUT_POOL_PROLOGUE
5333 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
5334 A C statement to output assembler commands to define the start of the
5335 constant pool for a function. @var{funname} is a string giving
5336 the name of the function. Should the return type of the function
5337 be required, it can be obtained via @var{fundecl}. @var{size}
5338 is the size, in bytes, of the constant pool that will be written
5339 immediately after this call.
5341 If no constant-pool prefix is required, the usual case, this macro need
5344 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
5345 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5346 A C statement (with or without semicolon) to output a constant in the
5347 constant pool, if it needs special treatment. (This macro need not do
5348 anything for RTL expressions that can be output normally.)
5350 The argument @var{file} is the standard I/O stream to output the
5351 assembler code on. @var{x} is the RTL expression for the constant to
5352 output, and @var{mode} is the machine mode (in case @var{x} is a
5353 @samp{const_int}). @var{align} is the required alignment for the value
5354 @var{x}; you should output an assembler directive to force this much
5357 The argument @var{labelno} is a number to use in an internal label for
5358 the address of this pool entry. The definition of this macro is
5359 responsible for outputting the label definition at the proper place.
5360 Here is how to do this:
5363 ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
5366 When you output a pool entry specially, you should end with a
5367 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5368 entry from being output a second time in the usual manner.
5370 You need not define this macro if it would do nothing.
5372 @findex CONSTANT_AFTER_FUNCTION_P
5373 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
5374 Define this macro as a C expression which is nonzero if the constant
5375 @var{exp}, of type @code{tree}, should be output after the code for a
5376 function. The compiler will normally output all constants before the
5377 function; you need not define this macro if this is OK.
5379 @findex ASM_OUTPUT_POOL_EPILOGUE
5380 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5381 A C statement to output assembler commands to at the end of the constant
5382 pool for a function. @var{funname} is a string giving the name of the
5383 function. Should the return type of the function be required, you can
5384 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5385 constant pool that GCC wrote immediately before this call.
5387 If no constant-pool epilogue is required, the usual case, you need not
5390 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
5391 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
5392 Define this macro as a C expression which is nonzero if @var{C} is
5393 used as a logical line separator by the assembler.
5395 If you do not define this macro, the default is that only
5396 the character @samp{;} is treated as a logical line separator.
5399 @findex ASM_OPEN_PAREN
5400 @findex ASM_CLOSE_PAREN
5401 @item ASM_OPEN_PAREN
5402 @itemx ASM_CLOSE_PAREN
5403 These macros are defined as C string constant, describing the syntax
5404 in the assembler for grouping arithmetic expressions. The following
5405 definitions are correct for most assemblers:
5408 #define ASM_OPEN_PAREN "("
5409 #define ASM_CLOSE_PAREN ")"
5413 These macros are provided by @file{real.h} for writing the definitions
5414 of @code{ASM_OUTPUT_DOUBLE} and the like:
5417 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5418 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5419 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5420 @findex REAL_VALUE_TO_TARGET_SINGLE
5421 @findex REAL_VALUE_TO_TARGET_DOUBLE
5422 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
5423 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
5424 floating point representation, and store its bit pattern in the array of
5425 @code{long int} whose address is @var{l}. The number of elements in the
5426 output array is determined by the size of the desired target floating
5427 point data type: 32 bits of it go in each @code{long int} array
5428 element. Each array element holds 32 bits of the result, even if
5429 @code{long int} is wider than 32 bits on the host machine.
5431 The array element values are designed so that you can print them out
5432 using @code{fprintf} in the order they should appear in the target
5435 @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
5436 @findex REAL_VALUE_TO_DECIMAL
5437 This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
5438 decimal number and stores it as a string into @var{string}.
5439 You must pass, as @var{string}, the address of a long enough block
5440 of space to hold the result.
5442 The argument @var{format} is a @code{printf}-specification that serves
5443 as a suggestion for how to format the output string.
5446 @node Uninitialized Data
5447 @subsection Output of Uninitialized Variables
5449 Each of the macros in this section is used to do the whole job of
5450 outputting a single uninitialized variable.
5453 @findex ASM_OUTPUT_COMMON
5454 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5455 A C statement (sans semicolon) to output to the stdio stream
5456 @var{stream} the assembler definition of a common-label named
5457 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5458 is the size rounded up to whatever alignment the caller wants.
5460 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5461 output the name itself; before and after that, output the additional
5462 assembler syntax for defining the name, and a newline.
5464 This macro controls how the assembler definitions of uninitialized
5465 common global variables are output.
5467 @findex ASM_OUTPUT_ALIGNED_COMMON
5468 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5469 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5470 separate, explicit argument. If you define this macro, it is used in
5471 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5472 handling the required alignment of the variable. The alignment is specified
5473 as the number of bits.
5475 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
5476 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5477 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5478 variable to be output, if there is one, or @code{NULL_TREE} if there
5479 is no corresponding variable. If you define this macro, GCC will use it
5480 in place of both @code{ASM_OUTPUT_COMMON} and
5481 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5482 the variable's decl in order to chose what to output.
5484 @findex ASM_OUTPUT_SHARED_COMMON
5485 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5486 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
5487 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
5490 @findex ASM_OUTPUT_BSS
5491 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5492 A C statement (sans semicolon) to output to the stdio stream
5493 @var{stream} the assembler definition of uninitialized global @var{decl} named
5494 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5495 is the size rounded up to whatever alignment the caller wants.
5497 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
5498 defining this macro. If unable, use the expression
5499 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5500 before and after that, output the additional assembler syntax for defining
5501 the name, and a newline.
5503 This macro controls how the assembler definitions of uninitialized global
5504 variables are output. This macro exists to properly support languages like
5505 @code{c++} which do not have @code{common} data. However, this macro currently
5506 is not defined for all targets. If this macro and
5507 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
5508 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
5509 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
5511 @findex ASM_OUTPUT_ALIGNED_BSS
5512 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5513 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
5514 separate, explicit argument. If you define this macro, it is used in
5515 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
5516 handling the required alignment of the variable. The alignment is specified
5517 as the number of bits.
5519 Try to use function @code{asm_output_aligned_bss} defined in file
5520 @file{varasm.c} when defining this macro.
5522 @findex ASM_OUTPUT_SHARED_BSS
5523 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5524 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
5525 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
5528 @findex ASM_OUTPUT_LOCAL
5529 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5530 A C statement (sans semicolon) to output to the stdio stream
5531 @var{stream} the assembler definition of a local-common-label named
5532 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5533 is the size rounded up to whatever alignment the caller wants.
5535 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5536 output the name itself; before and after that, output the additional
5537 assembler syntax for defining the name, and a newline.
5539 This macro controls how the assembler definitions of uninitialized
5540 static variables are output.
5542 @findex ASM_OUTPUT_ALIGNED_LOCAL
5543 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5544 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5545 separate, explicit argument. If you define this macro, it is used in
5546 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5547 handling the required alignment of the variable. The alignment is specified
5548 as the number of bits.
5550 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
5551 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5552 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5553 variable to be output, if there is one, or @code{NULL_TREE} if there
5554 is no corresponding variable. If you define this macro, GCC will use it
5555 in place of both @code{ASM_OUTPUT_DECL} and
5556 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5557 the variable's decl in order to chose what to output.
5559 @findex ASM_OUTPUT_SHARED_LOCAL
5560 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5561 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
5562 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
5567 @subsection Output and Generation of Labels
5569 @c prevent bad page break with this line
5570 This is about outputting labels.
5573 @findex ASM_OUTPUT_LABEL
5574 @findex assemble_name
5575 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5576 A C statement (sans semicolon) to output to the stdio stream
5577 @var{stream} the assembler definition of a label named @var{name}.
5578 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5579 output the name itself; before and after that, output the additional
5580 assembler syntax for defining the name, and a newline.
5582 @findex ASM_DECLARE_FUNCTION_NAME
5583 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5584 A C statement (sans semicolon) to output to the stdio stream
5585 @var{stream} any text necessary for declaring the name @var{name} of a
5586 function which is being defined. This macro is responsible for
5587 outputting the label definition (perhaps using
5588 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
5589 @code{FUNCTION_DECL} tree node representing the function.
5591 If this macro is not defined, then the function name is defined in the
5592 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5594 @findex ASM_DECLARE_FUNCTION_SIZE
5595 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5596 A C statement (sans semicolon) to output to the stdio stream
5597 @var{stream} any text necessary for declaring the size of a function
5598 which is being defined. The argument @var{name} is the name of the
5599 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5600 representing the function.
5602 If this macro is not defined, then the function size is not defined.
5604 @findex ASM_DECLARE_OBJECT_NAME
5605 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5606 A C statement (sans semicolon) to output to the stdio stream
5607 @var{stream} any text necessary for declaring the name @var{name} of an
5608 initialized variable which is being defined. This macro must output the
5609 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
5610 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
5612 If this macro is not defined, then the variable name is defined in the
5613 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5615 @findex ASM_DECLARE_REGISTER_GLOBAL
5616 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
5617 A C statement (sans semicolon) to output to the stdio stream
5618 @var{stream} any text necessary for claiming a register @var{regno}
5619 for a global variable @var{decl} with name @var{name}.
5621 If you don't define this macro, that is equivalent to defining it to do
5624 @findex ASM_FINISH_DECLARE_OBJECT
5625 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5626 A C statement (sans semicolon) to finish up declaring a variable name
5627 once the compiler has processed its initializer fully and thus has had a
5628 chance to determine the size of an array when controlled by an
5629 initializer. This is used on systems where it's necessary to declare
5630 something about the size of the object.
5632 If you don't define this macro, that is equivalent to defining it to do
5635 @findex ASM_GLOBALIZE_LABEL
5636 @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
5637 A C statement (sans semicolon) to output to the stdio stream
5638 @var{stream} some commands that will make the label @var{name} global;
5639 that is, available for reference from other files. Use the expression
5640 @code{assemble_name (@var{stream}, @var{name})} to output the name
5641 itself; before and after that, output the additional assembler syntax
5642 for making that name global, and a newline.
5644 @findex ASM_WEAKEN_LABEL
5645 @item ASM_WEAKEN_LABEL
5646 A C statement (sans semicolon) to output to the stdio stream
5647 @var{stream} some commands that will make the label @var{name} weak;
5648 that is, available for reference from other files but only used if
5649 no other definition is available. Use the expression
5650 @code{assemble_name (@var{stream}, @var{name})} to output the name
5651 itself; before and after that, output the additional assembler syntax
5652 for making that name weak, and a newline.
5654 If you don't define this macro, GCC will not support weak
5655 symbols and you should not define the @code{SUPPORTS_WEAK} macro.
5657 @findex SUPPORTS_WEAK
5659 A C expression which evaluates to true if the target supports weak symbols.
5661 If you don't define this macro, @file{defaults.h} provides a default
5662 definition. If @code{ASM_WEAKEN_LABEL} is defined, the default
5663 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5664 you want to control weak symbol support with a compiler flag such as
5667 @findex MAKE_DECL_ONE_ONLY (@var{decl})
5668 @item MAKE_DECL_ONE_ONLY
5669 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5670 public symbol such that extra copies in multiple translation units will
5671 be discarded by the linker. Define this macro if your object file
5672 format provides support for this concept, such as the @samp{COMDAT}
5673 section flags in the Microsoft Windows PE/COFF format, and this support
5674 requires changes to @var{decl}, such as putting it in a separate section.
5676 @findex SUPPORTS_ONE_ONLY
5677 @item SUPPORTS_ONE_ONLY
5678 A C expression which evaluates to true if the target supports one-only
5681 If you don't define this macro, @file{varasm.c} provides a default
5682 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
5683 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5684 you want to control one-only symbol support with a compiler flag, or if
5685 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
5686 be emitted as one-only.
5688 @findex ASM_OUTPUT_EXTERNAL
5689 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
5690 A C statement (sans semicolon) to output to the stdio stream
5691 @var{stream} any text necessary for declaring the name of an external
5692 symbol named @var{name} which is referenced in this compilation but
5693 not defined. The value of @var{decl} is the tree node for the
5696 This macro need not be defined if it does not need to output anything.
5697 The GNU assembler and most Unix assemblers don't require anything.
5699 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
5700 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
5701 A C statement (sans semicolon) to output on @var{stream} an assembler
5702 pseudo-op to declare a library function name external. The name of the
5703 library function is given by @var{symref}, which has type @code{rtx} and
5704 is a @code{symbol_ref}.
5706 This macro need not be defined if it does not need to output anything.
5707 The GNU assembler and most Unix assemblers don't require anything.
5709 @findex ASM_OUTPUT_LABELREF
5710 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
5711 A C statement (sans semicolon) to output to the stdio stream
5712 @var{stream} a reference in assembler syntax to a label named
5713 @var{name}. This should add @samp{_} to the front of the name, if that
5714 is customary on your operating system, as it is in most Berkeley Unix
5715 systems. This macro is used in @code{assemble_name}.
5717 @ignore @c Seems not to exist anymore.
5718 @findex ASM_OUTPUT_LABELREF_AS_INT
5719 @item ASM_OUTPUT_LABELREF_AS_INT (@var{file}, @var{label})
5720 Define this macro for systems that use the program @code{collect2}.
5721 The definition should be a C statement to output a word containing
5722 a reference to the label @var{label}.
5725 @findex ASM_OUTPUT_INTERNAL_LABEL
5726 @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
5727 A C statement to output to the stdio stream @var{stream} a label whose
5728 name is made from the string @var{prefix} and the number @var{num}.
5730 It is absolutely essential that these labels be distinct from the labels
5731 used for user-level functions and variables. Otherwise, certain programs
5732 will have name conflicts with internal labels.
5734 It is desirable to exclude internal labels from the symbol table of the
5735 object file. Most assemblers have a naming convention for labels that
5736 should be excluded; on many systems, the letter @samp{L} at the
5737 beginning of a label has this effect. You should find out what
5738 convention your system uses, and follow it.
5740 The usual definition of this macro is as follows:
5743 fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
5746 @findex ASM_OUTPUT_ALTERNATE_LABEL_NAME
5747 @item ASM_OUTPUT_ALTERNATE_LABEL_NAME (@var{stream}, @var{string})
5748 A C statement to output to the stdio stream @var{stream} the string
5751 The default definition of this macro is as follows:
5754 fprintf (@var{stream}, "%s:\n", LABEL_ALTERNATE_NAME (INSN))
5757 @findex ASM_GENERATE_INTERNAL_LABEL
5758 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
5759 A C statement to store into the string @var{string} a label whose name
5760 is made from the string @var{prefix} and the number @var{num}.
5762 This string, when output subsequently by @code{assemble_name}, should
5763 produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
5764 with the same @var{prefix} and @var{num}.
5766 If the string begins with @samp{*}, then @code{assemble_name} will
5767 output the rest of the string unchanged. It is often convenient for
5768 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
5769 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
5770 to output the string, and may change it. (Of course,
5771 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
5772 you should know what it does on your machine.)
5774 @findex ASM_FORMAT_PRIVATE_NAME
5775 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
5776 A C expression to assign to @var{outvar} (which is a variable of type
5777 @code{char *}) a newly allocated string made from the string
5778 @var{name} and the number @var{number}, with some suitable punctuation
5779 added. Use @code{alloca} to get space for the string.
5781 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
5782 produce an assembler label for an internal static variable whose name is
5783 @var{name}. Therefore, the string must be such as to result in valid
5784 assembler code. The argument @var{number} is different each time this
5785 macro is executed; it prevents conflicts between similarly-named
5786 internal static variables in different scopes.
5788 Ideally this string should not be a valid C identifier, to prevent any
5789 conflict with the user's own symbols. Most assemblers allow periods
5790 or percent signs in assembler symbols; putting at least one of these
5791 between the name and the number will suffice.
5793 @findex ASM_OUTPUT_DEF
5794 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
5795 A C statement to output to the stdio stream @var{stream} assembler code
5796 which defines (equates) the symbol @var{name} to have the value @var{value}.
5799 If SET_ASM_OP is defined, a default definition is provided which is
5800 correct for most systems.
5802 @findex ASM_OUTPUT_DEF_FROM_DECLS
5803 @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
5804 A C statement to output to the stdio stream @var{stream} assembler code
5805 which defines (equates) the symbol whoes tree node is @var{decl_of_name}
5806 to have the value of the tree node @var{decl_of_value}. This macro will
5807 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
5808 the tree nodes are available.
5810 @findex ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL
5811 @item ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL (@var{stream}, @var{symbol}, @var{high}, @var{low})
5812 A C statement to output to the stdio stream @var{stream} assembler code
5813 which defines (equates) the symbol @var{symbol} to have a value equal to
5814 the difference of the two symbols @var{high} and @var{low}, i.e.
5815 @var{high} minus @var{low}. GCC guarantees that the symbols @var{high}
5816 and @var{low} are already known by the assembler so that the difference
5817 resolves into a constant.
5820 If SET_ASM_OP is defined, a default definition is provided which is
5821 correct for most systems.
5823 @findex ASM_OUTPUT_WEAK_ALIAS
5824 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
5825 A C statement to output to the stdio stream @var{stream} assembler code
5826 which defines (equates) the weak symbol @var{name} to have the value
5829 Define this macro if the target only supports weak aliases; define
5830 ASM_OUTPUT_DEF instead if possible.
5832 @findex OBJC_GEN_METHOD_LABEL
5833 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
5834 Define this macro to override the default assembler names used for
5835 Objective C methods.
5837 The default name is a unique method number followed by the name of the
5838 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
5839 the category is also included in the assembler name (e.g.@:
5842 These names are safe on most systems, but make debugging difficult since
5843 the method's selector is not present in the name. Therefore, particular
5844 systems define other ways of computing names.
5846 @var{buf} is an expression of type @code{char *} which gives you a
5847 buffer in which to store the name; its length is as long as
5848 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
5849 50 characters extra.
5851 The argument @var{is_inst} specifies whether the method is an instance
5852 method or a class method; @var{class_name} is the name of the class;
5853 @var{cat_name} is the name of the category (or NULL if the method is not
5854 in a category); and @var{sel_name} is the name of the selector.
5856 On systems where the assembler can handle quoted names, you can use this
5857 macro to provide more human-readable names.
5860 @node Initialization
5861 @subsection How Initialization Functions Are Handled
5862 @cindex initialization routines
5863 @cindex termination routines
5864 @cindex constructors, output of
5865 @cindex destructors, output of
5867 The compiled code for certain languages includes @dfn{constructors}
5868 (also called @dfn{initialization routines})---functions to initialize
5869 data in the program when the program is started. These functions need
5870 to be called before the program is ``started''---that is to say, before
5871 @code{main} is called.
5873 Compiling some languages generates @dfn{destructors} (also called
5874 @dfn{termination routines}) that should be called when the program
5877 To make the initialization and termination functions work, the compiler
5878 must output something in the assembler code to cause those functions to
5879 be called at the appropriate time. When you port the compiler to a new
5880 system, you need to specify how to do this.
5882 There are two major ways that GCC currently supports the execution of
5883 initialization and termination functions. Each way has two variants.
5884 Much of the structure is common to all four variations.
5886 @findex __CTOR_LIST__
5887 @findex __DTOR_LIST__
5888 The linker must build two lists of these functions---a list of
5889 initialization functions, called @code{__CTOR_LIST__}, and a list of
5890 termination functions, called @code{__DTOR_LIST__}.
5892 Each list always begins with an ignored function pointer (which may hold
5893 0, @minus{}1, or a count of the function pointers after it, depending on
5894 the environment). This is followed by a series of zero or more function
5895 pointers to constructors (or destructors), followed by a function
5896 pointer containing zero.
5898 Depending on the operating system and its executable file format, either
5899 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
5900 time and exit time. Constructors are called in reverse order of the
5901 list; destructors in forward order.
5903 The best way to handle static constructors works only for object file
5904 formats which provide arbitrarily-named sections. A section is set
5905 aside for a list of constructors, and another for a list of destructors.
5906 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
5907 object file that defines an initialization function also puts a word in
5908 the constructor section to point to that function. The linker
5909 accumulates all these words into one contiguous @samp{.ctors} section.
5910 Termination functions are handled similarly.
5912 To use this method, you need appropriate definitions of the macros
5913 @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}. Usually
5914 you can get them by including @file{svr4.h}.
5916 When arbitrary sections are available, there are two variants, depending
5917 upon how the code in @file{crtstuff.c} is called. On systems that
5918 support an @dfn{init} section which is executed at program startup,
5919 parts of @file{crtstuff.c} are compiled into that section. The
5920 program is linked by the @code{gcc} driver like this:
5923 ld -o @var{output_file} crtbegin.o @dots{} crtend.o -lgcc
5926 The head of a function (@code{__do_global_ctors}) appears in the init
5927 section of @file{crtbegin.o}; the remainder of the function appears in
5928 the init section of @file{crtend.o}. The linker will pull these two
5929 parts of the section together, making a whole function. If any of the
5930 user's object files linked into the middle of it contribute code, then that
5931 code will be executed as part of the body of @code{__do_global_ctors}.
5933 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
5936 If no init section is available, do not define
5937 @code{INIT_SECTION_ASM_OP}. Then @code{__do_global_ctors} is built into
5938 the text section like all other functions, and resides in
5939 @file{libgcc.a}. When GCC compiles any function called @code{main}, it
5940 inserts a procedure call to @code{__main} as the first executable code
5941 after the function prologue. The @code{__main} function, also defined
5942 in @file{libgcc2.c}, simply calls @file{__do_global_ctors}.
5944 In file formats that don't support arbitrary sections, there are again
5945 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
5946 and an `a.out' format must be used. In this case,
5947 @code{ASM_OUTPUT_CONSTRUCTOR} is defined to produce a @code{.stabs}
5948 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
5949 and with the address of the void function containing the initialization
5950 code as its value. The GNU linker recognizes this as a request to add
5951 the value to a ``set''; the values are accumulated, and are eventually
5952 placed in the executable as a vector in the format described above, with
5953 a leading (ignored) count and a trailing zero element.
5954 @code{ASM_OUTPUT_DESTRUCTOR} is handled similarly. Since no init
5955 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
5956 the compilation of @code{main} to call @code{__main} as above, starting
5957 the initialization process.
5959 The last variant uses neither arbitrary sections nor the GNU linker.
5960 This is preferable when you want to do dynamic linking and when using
5961 file formats which the GNU linker does not support, such as `ECOFF'. In
5962 this case, @code{ASM_OUTPUT_CONSTRUCTOR} does not produce an
5963 @code{N_SETT} symbol; initialization and termination functions are
5964 recognized simply by their names. This requires an extra program in the
5965 linkage step, called @code{collect2}. This program pretends to be the
5966 linker, for use with GCC; it does its job by running the ordinary
5967 linker, but also arranges to include the vectors of initialization and
5968 termination functions. These functions are called via @code{__main} as
5971 Choosing among these configuration options has been simplified by a set
5972 of operating-system-dependent files in the @file{config} subdirectory.
5973 These files define all of the relevant parameters. Usually it is
5974 sufficient to include one into your specific machine-dependent
5975 configuration file. These files are:
5979 For operating systems using the `a.out' format.
5982 For operating systems using the `MachO' format.
5985 For System V Release 3 and similar systems using `COFF' format.
5988 For System V Release 4 and similar systems using `ELF' format.
5991 For the VMS operating system.
5995 The following section describes the specific macros that control and
5996 customize the handling of initialization and termination functions.
5999 @node Macros for Initialization
6000 @subsection Macros Controlling Initialization Routines
6002 Here are the macros that control how the compiler handles initialization
6003 and termination functions:
6006 @findex INIT_SECTION_ASM_OP
6007 @item INIT_SECTION_ASM_OP
6008 If defined, a C string constant for the assembler operation to identify
6009 the following data as initialization code. If not defined, GCC will
6010 assume such a section does not exist. When you are using special
6011 sections for initialization and termination functions, this macro also
6012 controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to run the
6013 initialization functions.
6015 @item HAS_INIT_SECTION
6016 @findex HAS_INIT_SECTION
6017 If defined, @code{main} will not call @code{__main} as described above.
6018 This macro should be defined for systems that control the contents of the
6019 init section on a symbol-by-symbol basis, such as OSF/1, and should not
6020 be defined explicitly for systems that support
6021 @code{INIT_SECTION_ASM_OP}.
6023 @item LD_INIT_SWITCH
6024 @findex LD_INIT_SWITCH
6025 If defined, a C string constant for a switch that tells the linker that
6026 the following symbol is an initialization routine.
6028 @item LD_FINI_SWITCH
6029 @findex LD_FINI_SWITCH
6030 If defined, a C string constant for a switch that tells the linker that
6031 the following symbol is a finalization routine.
6034 @findex INVOKE__main
6035 If defined, @code{main} will call @code{__main} despite the presence of
6036 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6037 where the init section is not actually run automatically, but is still
6038 useful for collecting the lists of constructors and destructors.
6040 @item ASM_OUTPUT_CONSTRUCTOR (@var{stream}, @var{name})
6041 @findex ASM_OUTPUT_CONSTRUCTOR
6042 Define this macro as a C statement to output on the stream @var{stream}
6043 the assembler code to arrange to call the function named @var{name} at
6044 initialization time.
6046 Assume that @var{name} is the name of a C function generated
6047 automatically by the compiler. This function takes no arguments. Use
6048 the function @code{assemble_name} to output the name @var{name}; this
6049 performs any system-specific syntactic transformations such as adding an
6052 If you don't define this macro, nothing special is output to arrange to
6053 call the function. This is correct when the function will be called in
6054 some other manner---for example, by means of the @code{collect2} program,
6055 which looks through the symbol table to find these functions by their
6058 @item ASM_OUTPUT_DESTRUCTOR (@var{stream}, @var{name})
6059 @findex ASM_OUTPUT_DESTRUCTOR
6060 This is like @code{ASM_OUTPUT_CONSTRUCTOR} but used for termination
6061 functions rather than initialization functions.
6063 When @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR} are
6064 defined, the initializaiton routine generated for the generated object
6065 file will have static linkage.
6068 If your system uses @code{collect2} as the means of processing
6069 constructors, then that program normally uses @code{nm} to scan an
6070 object file for constructor functions to be called. On such systems you
6071 must not define @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}
6072 as the object file's initialization routine must have global scope.
6074 On certain kinds of systems, you can define these macros to make
6075 @code{collect2} work faster (and, in some cases, make it work at all):
6078 @findex OBJECT_FORMAT_COFF
6079 @item OBJECT_FORMAT_COFF
6080 Define this macro if the system uses COFF (Common Object File Format)
6081 object files, so that @code{collect2} can assume this format and scan
6082 object files directly for dynamic constructor/destructor functions.
6084 @findex OBJECT_FORMAT_ROSE
6085 @item OBJECT_FORMAT_ROSE
6086 Define this macro if the system uses ROSE format object files, so that
6087 @code{collect2} can assume this format and scan object files directly
6088 for dynamic constructor/destructor functions.
6090 These macros are effective only in a native compiler; @code{collect2} as
6091 part of a cross compiler always uses @code{nm} for the target machine.
6093 @findex REAL_NM_FILE_NAME
6094 @item REAL_NM_FILE_NAME
6095 Define this macro as a C string constant containing the file name to use
6096 to execute @code{nm}. The default is to search the path normally for
6099 If your system supports shared libraries and has a program to list the
6100 dynamic dependencies of a given library or executable, you can define
6101 these macros to enable support for running initialization and
6102 termination functions in shared libraries:
6106 Define this macro to a C string constant containing the name of the
6107 program which lists dynamic dependencies, like @code{"ldd"} under SunOS 4.
6109 @findex PARSE_LDD_OUTPUT
6110 @item PARSE_LDD_OUTPUT (@var{PTR})
6111 Define this macro to be C code that extracts filenames from the output
6112 of the program denoted by @code{LDD_SUFFIX}. @var{PTR} is a variable
6113 of type @code{char *} that points to the beginning of a line of output
6114 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6115 code must advance @var{PTR} to the beginning of the filename on that
6116 line. Otherwise, it must set @var{PTR} to @code{NULL}.
6120 @node Instruction Output
6121 @subsection Output of Assembler Instructions
6123 @c prevent bad page break with this line
6124 This describes assembler instruction output.
6127 @findex REGISTER_NAMES
6128 @item REGISTER_NAMES
6129 A C initializer containing the assembler's names for the machine
6130 registers, each one as a C string constant. This is what translates
6131 register numbers in the compiler into assembler language.
6133 @findex ADDITIONAL_REGISTER_NAMES
6134 @item ADDITIONAL_REGISTER_NAMES
6135 If defined, a C initializer for an array of structures containing a name
6136 and a register number. This macro defines additional names for hard
6137 registers, thus allowing the @code{asm} option in declarations to refer
6138 to registers using alternate names.
6140 @findex ASM_OUTPUT_OPCODE
6141 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6142 Define this macro if you are using an unusual assembler that
6143 requires different names for the machine instructions.
6145 The definition is a C statement or statements which output an
6146 assembler instruction opcode to the stdio stream @var{stream}. The
6147 macro-operand @var{ptr} is a variable of type @code{char *} which
6148 points to the opcode name in its ``internal'' form---the form that is
6149 written in the machine description. The definition should output the
6150 opcode name to @var{stream}, performing any translation you desire, and
6151 increment the variable @var{ptr} to point at the end of the opcode
6152 so that it will not be output twice.
6154 In fact, your macro definition may process less than the entire opcode
6155 name, or more than the opcode name; but if you want to process text
6156 that includes @samp{%}-sequences to substitute operands, you must take
6157 care of the substitution yourself. Just be sure to increment
6158 @var{ptr} over whatever text should not be output normally.
6160 @findex recog_operand
6161 If you need to look at the operand values, they can be found as the
6162 elements of @code{recog_operand}.
6164 If the macro definition does nothing, the instruction is output
6167 @findex FINAL_PRESCAN_INSN
6168 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6169 If defined, a C statement to be executed just prior to the output of
6170 assembler code for @var{insn}, to modify the extracted operands so
6171 they will be output differently.
6173 Here the argument @var{opvec} is the vector containing the operands
6174 extracted from @var{insn}, and @var{noperands} is the number of
6175 elements of the vector which contain meaningful data for this insn.
6176 The contents of this vector are what will be used to convert the insn
6177 template into assembler code, so you can change the assembler output
6178 by changing the contents of the vector.
6180 This macro is useful when various assembler syntaxes share a single
6181 file of instruction patterns; by defining this macro differently, you
6182 can cause a large class of instructions to be output differently (such
6183 as with rearranged operands). Naturally, variations in assembler
6184 syntax affecting individual insn patterns ought to be handled by
6185 writing conditional output routines in those patterns.
6187 If this macro is not defined, it is equivalent to a null statement.
6189 @findex FINAL_PRESCAN_LABEL
6190 @item FINAL_PRESCAN_LABEL
6191 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
6192 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
6193 @var{noperands} will be zero.
6195 @findex PRINT_OPERAND
6196 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6197 A C compound statement to output to stdio stream @var{stream} the
6198 assembler syntax for an instruction operand @var{x}. @var{x} is an
6201 @var{code} is a value that can be used to specify one of several ways
6202 of printing the operand. It is used when identical operands must be
6203 printed differently depending on the context. @var{code} comes from
6204 the @samp{%} specification that was used to request printing of the
6205 operand. If the specification was just @samp{%@var{digit}} then
6206 @var{code} is 0; if the specification was @samp{%@var{ltr}
6207 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6210 If @var{x} is a register, this macro should print the register's name.
6211 The names can be found in an array @code{reg_names} whose type is
6212 @code{char *[]}. @code{reg_names} is initialized from
6213 @code{REGISTER_NAMES}.
6215 When the machine description has a specification @samp{%@var{punct}}
6216 (a @samp{%} followed by a punctuation character), this macro is called
6217 with a null pointer for @var{x} and the punctuation character for
6220 @findex PRINT_OPERAND_PUNCT_VALID_P
6221 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6222 A C expression which evaluates to true if @var{code} is a valid
6223 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6224 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6225 punctuation characters (except for the standard one, @samp{%}) are used
6228 @findex PRINT_OPERAND_ADDRESS
6229 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6230 A C compound statement to output to stdio stream @var{stream} the
6231 assembler syntax for an instruction operand that is a memory reference
6232 whose address is @var{x}. @var{x} is an RTL expression.
6234 @cindex @code{ENCODE_SECTION_INFO} usage
6235 On some machines, the syntax for a symbolic address depends on the
6236 section that the address refers to. On these machines, define the macro
6237 @code{ENCODE_SECTION_INFO} to store the information into the
6238 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
6240 @findex DBR_OUTPUT_SEQEND
6241 @findex dbr_sequence_length
6242 @item DBR_OUTPUT_SEQEND(@var{file})
6243 A C statement, to be executed after all slot-filler instructions have
6244 been output. If necessary, call @code{dbr_sequence_length} to
6245 determine the number of slots filled in a sequence (zero if not
6246 currently outputting a sequence), to decide how many no-ops to output,
6249 Don't define this macro if it has nothing to do, but it is helpful in
6250 reading assembly output if the extent of the delay sequence is made
6251 explicit (e.g. with white space).
6253 @findex final_sequence
6254 Note that output routines for instructions with delay slots must be
6255 prepared to deal with not being output as part of a sequence (i.e.
6256 when the scheduling pass is not run, or when no slot fillers could be
6257 found.) The variable @code{final_sequence} is null when not
6258 processing a sequence, otherwise it contains the @code{sequence} rtx
6261 @findex REGISTER_PREFIX
6262 @findex LOCAL_LABEL_PREFIX
6263 @findex USER_LABEL_PREFIX
6264 @findex IMMEDIATE_PREFIX
6266 @item REGISTER_PREFIX
6267 @itemx LOCAL_LABEL_PREFIX
6268 @itemx USER_LABEL_PREFIX
6269 @itemx IMMEDIATE_PREFIX
6270 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6271 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6272 @file{final.c}). These are useful when a single @file{md} file must
6273 support multiple assembler formats. In that case, the various @file{tm.h}
6274 files can define these macros differently.
6276 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
6277 @findex ASM_FPRINTF_EXTENSIONS
6278 If defiend this macro should expand to a series of @code{case}
6279 statements which will be parsed inside the @code{switch} statement of
6280 the @code{asm_fprintf} function. This allows targets to define extra
6281 printf formats which may useful when generating their assembler
6282 statements. Noet that upper case letters are reserved for future
6283 generic extensions to asm_fprintf, and so are not available to target
6284 specific code. The output file is given by the parameter @var{file}.
6285 The varargs input pointer is @var{argptr} and the rest of the format
6286 string, starting the character after the one that is being switched
6287 upon, is pointed to by @var{format}.
6289 @findex ASSEMBLER_DIALECT
6290 @item ASSEMBLER_DIALECT
6291 If your target supports multiple dialects of assembler language (such as
6292 different opcodes), define this macro as a C expression that gives the
6293 numeric index of the assembler language dialect to use, with zero as the
6296 If this macro is defined, you may use constructs of the form
6297 @samp{@{option0|option1|option2@dots{}@}} in the output
6298 templates of patterns (@pxref{Output Template}) or in the first argument
6299 of @code{asm_fprintf}. This construct outputs @samp{option0},
6300 @samp{option1} or @samp{option2}, etc., if the value of
6301 @code{ASSEMBLER_DIALECT} is zero, one or two, etc. Any special
6302 characters within these strings retain their usual meaning.
6304 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6305 @samp{@}} do not have any special meaning when used in templates or
6306 operands to @code{asm_fprintf}.
6308 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6309 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6310 the variations in assembler language syntax with that mechanism. Define
6311 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6312 if the syntax variant are larger and involve such things as different
6313 opcodes or operand order.
6315 @findex ASM_OUTPUT_REG_PUSH
6316 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6317 A C expression to output to @var{stream} some assembler code
6318 which will push hard register number @var{regno} onto the stack.
6319 The code need not be optimal, since this macro is used only when
6322 @findex ASM_OUTPUT_REG_POP
6323 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6324 A C expression to output to @var{stream} some assembler code
6325 which will pop hard register number @var{regno} off of the stack.
6326 The code need not be optimal, since this macro is used only when
6330 @node Dispatch Tables
6331 @subsection Output of Dispatch Tables
6333 @c prevent bad page break with this line
6334 This concerns dispatch tables.
6337 @cindex dispatch table
6338 @findex ASM_OUTPUT_ADDR_DIFF_ELT
6339 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6340 A C statement to output to the stdio stream @var{stream} an assembler
6341 pseudo-instruction to generate a difference between two labels.
6342 @var{value} and @var{rel} are the numbers of two internal labels. The
6343 definitions of these labels are output using
6344 @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
6345 way here. For example,
6348 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6349 @var{value}, @var{rel})
6352 You must provide this macro on machines where the addresses in a
6353 dispatch table are relative to the table's own address. If defined, GNU
6354 CC will also use this macro on all machines when producing PIC.
6355 @var{body} is the body of the ADDR_DIFF_VEC; it is provided so that the
6356 mode and flags can be read.
6358 @findex ASM_OUTPUT_ADDR_VEC_ELT
6359 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6360 This macro should be provided on machines where the addresses
6361 in a dispatch table are absolute.
6363 The definition should be a C statement to output to the stdio stream
6364 @var{stream} an assembler pseudo-instruction to generate a reference to
6365 a label. @var{value} is the number of an internal label whose
6366 definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
6370 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6373 @findex ASM_OUTPUT_CASE_LABEL
6374 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6375 Define this if the label before a jump-table needs to be output
6376 specially. The first three arguments are the same as for
6377 @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
6378 jump-table which follows (a @code{jump_insn} containing an
6379 @code{addr_vec} or @code{addr_diff_vec}).
6381 This feature is used on system V to output a @code{swbeg} statement
6384 If this macro is not defined, these labels are output with
6385 @code{ASM_OUTPUT_INTERNAL_LABEL}.
6387 @findex ASM_OUTPUT_CASE_END
6388 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6389 Define this if something special must be output at the end of a
6390 jump-table. The definition should be a C statement to be executed
6391 after the assembler code for the table is written. It should write
6392 the appropriate code to stdio stream @var{stream}. The argument
6393 @var{table} is the jump-table insn, and @var{num} is the label-number
6394 of the preceding label.
6396 If this macro is not defined, nothing special is output at the end of
6400 @node Exception Region Output
6401 @subsection Assembler Commands for Exception Regions
6403 @c prevent bad page break with this line
6405 This describes commands marking the start and the end of an exception
6409 @findex ASM_OUTPUT_EH_REGION_BEG
6410 @item ASM_OUTPUT_EH_REGION_BEG ()
6411 A C expression to output text to mark the start of an exception region.
6413 This macro need not be defined on most platforms.
6415 @findex ASM_OUTPUT_EH_REGION_END
6416 @item ASM_OUTPUT_EH_REGION_END ()
6417 A C expression to output text to mark the end of an exception region.
6419 This macro need not be defined on most platforms.
6421 @findex EXCEPTION_SECTION
6422 @item EXCEPTION_SECTION ()
6423 A C expression to switch to the section in which the main
6424 exception table is to be placed (@pxref{Sections}). The default is a
6425 section named @code{.gcc_except_table} on machines that support named
6426 sections via @code{ASM_OUTPUT_SECTION_NAME}, otherwise if @samp{-fpic}
6427 or @samp{-fPIC} is in effect, the @code{data_section}, otherwise the
6428 @code{readonly_data_section}.
6430 @findex EH_FRAME_SECTION_ASM_OP
6431 @item EH_FRAME_SECTION_ASM_OP
6432 If defined, a C string constant for the assembler operation to switch to
6433 the section for exception handling frame unwind information. If not
6434 defined, GCC will provide a default definition if the target supports
6435 named sections. @file{crtstuff.c} uses this macro to switch to the
6436 appropriate section.
6438 You should define this symbol if your target supports DWARF 2 frame
6439 unwind information and the default definition does not work.
6441 @findex OMIT_EH_TABLE
6442 @item OMIT_EH_TABLE ()
6443 A C expression that is nonzero if the normal exception table output
6446 This macro need not be defined on most platforms.
6448 @findex EH_TABLE_LOOKUP
6449 @item EH_TABLE_LOOKUP ()
6450 Alternate runtime support for looking up an exception at runtime and
6451 finding the associated handler, if the default method won't work.
6453 This macro need not be defined on most platforms.
6455 @findex DOESNT_NEED_UNWINDER
6456 @item DOESNT_NEED_UNWINDER
6457 A C expression that decides whether or not the current function needs to
6458 have a function unwinder generated for it. See the file @code{except.c}
6459 for details on when to define this, and how.
6461 @findex MASK_RETURN_ADDR
6462 @item MASK_RETURN_ADDR
6463 An rtx used to mask the return address found via RETURN_ADDR_RTX, so
6464 that it does not contain any extraneous set bits in it.
6466 @findex DWARF2_UNWIND_INFO
6467 @item DWARF2_UNWIND_INFO
6468 Define this macro to 0 if your target supports DWARF 2 frame unwind
6469 information, but it does not yet work with exception handling.
6470 Otherwise, if your target supports this information (if it defines
6471 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
6472 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
6475 If this macro is defined to 1, the DWARF 2 unwinder will be the default
6476 exception handling mechanism; otherwise, setjmp/longjmp will be used by
6479 If this macro is defined to anything, the DWARF 2 unwinder will be used
6480 instead of inline unwinders and __unwind_function in the non-setjmp case.
6484 @node Alignment Output
6485 @subsection Assembler Commands for Alignment
6487 @c prevent bad page break with this line
6488 This describes commands for alignment.
6491 @findex LABEL_ALIGN_AFTER_BARRIER
6492 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
6493 The alignment (log base 2) to put in front of @var{label}, which follows
6496 This macro need not be defined if you don't want any special alignment
6497 to be done at such a time. Most machine descriptions do not currently
6500 Unless it's necessary to inspect the @var{label} parameter, it is better
6501 to set the variable @var{align_jumps} in the target's
6502 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6503 selection in @var{align_jumps} in a @code{LABEL_ALIGN_AFTER_BARRIER}
6506 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6507 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6508 The maximum number of bytes to skip when applying
6509 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
6510 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6513 @item LOOP_ALIGN (@var{label})
6514 The alignment (log base 2) to put in front of @var{label}, which follows
6515 a NOTE_INSN_LOOP_BEG note.
6517 This macro need not be defined if you don't want any special alignment
6518 to be done at such a time. Most machine descriptions do not currently
6521 Unless it's necessary to inspect the @var{label} parameter, it is better
6522 to set the variable @var{align_loops} in the target's
6523 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6524 selection in @var{align_loops} in a @code{LOOP_ALIGN} implementation.
6526 @findex LOOP_ALIGN_MAX_SKIP
6527 @item LOOP_ALIGN_MAX_SKIP
6528 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
6529 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6532 @item LABEL_ALIGN (@var{label})
6533 The alignment (log base 2) to put in front of @var{label}.
6534 If LABEL_ALIGN_AFTER_BARRIER / LOOP_ALIGN specify a different alignment,
6535 the maximum of the specified values is used.
6537 Unless it's necessary to inspect the @var{label} parameter, it is better
6538 to set the variable @var{align_labels} in the target's
6539 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6540 selection in @var{align_labels} in a @code{LABEL_ALIGN} implementation.
6542 @findex LABEL_ALIGN_MAX_SKIP
6543 @item LABEL_ALIGN_MAX_SKIP
6544 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
6545 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6547 @findex ASM_OUTPUT_SKIP
6548 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6549 A C statement to output to the stdio stream @var{stream} an assembler
6550 instruction to advance the location counter by @var{nbytes} bytes.
6551 Those bytes should be zero when loaded. @var{nbytes} will be a C
6552 expression of type @code{int}.
6554 @findex ASM_NO_SKIP_IN_TEXT
6555 @item ASM_NO_SKIP_IN_TEXT
6556 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6557 text section because it fails to put zeros in the bytes that are skipped.
6558 This is true on many Unix systems, where the pseudo--op to skip bytes
6559 produces no-op instructions rather than zeros when used in the text
6562 @findex ASM_OUTPUT_ALIGN
6563 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6564 A C statement to output to the stdio stream @var{stream} an assembler
6565 command to advance the location counter to a multiple of 2 to the
6566 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
6568 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
6569 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
6570 A C statement to output to the stdio stream @var{stream} an assembler
6571 command to advance the location counter to a multiple of 2 to the
6572 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
6573 satisfy the alignment request. @var{power} and @var{max_skip} will be
6574 a C expression of type @code{int}.
6578 @node Debugging Info
6579 @section Controlling Debugging Information Format
6581 @c prevent bad page break with this line
6582 This describes how to specify debugging information.
6585 * All Debuggers:: Macros that affect all debugging formats uniformly.
6586 * DBX Options:: Macros enabling specific options in DBX format.
6587 * DBX Hooks:: Hook macros for varying DBX format.
6588 * File Names and DBX:: Macros controlling output of file names in DBX format.
6589 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
6593 @subsection Macros Affecting All Debugging Formats
6595 @c prevent bad page break with this line
6596 These macros affect all debugging formats.
6599 @findex DBX_REGISTER_NUMBER
6600 @item DBX_REGISTER_NUMBER (@var{regno})
6601 A C expression that returns the DBX register number for the compiler
6602 register number @var{regno}. In simple cases, the value of this
6603 expression may be @var{regno} itself. But sometimes there are some
6604 registers that the compiler knows about and DBX does not, or vice
6605 versa. In such cases, some register may need to have one number in
6606 the compiler and another for DBX.
6608 If two registers have consecutive numbers inside GCC, and they can be
6609 used as a pair to hold a multiword value, then they @emph{must} have
6610 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
6611 Otherwise, debuggers will be unable to access such a pair, because they
6612 expect register pairs to be consecutive in their own numbering scheme.
6614 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
6615 does not preserve register pairs, then what you must do instead is
6616 redefine the actual register numbering scheme.
6618 @findex DEBUGGER_AUTO_OFFSET
6619 @item DEBUGGER_AUTO_OFFSET (@var{x})
6620 A C expression that returns the integer offset value for an automatic
6621 variable having address @var{x} (an RTL expression). The default
6622 computation assumes that @var{x} is based on the frame-pointer and
6623 gives the offset from the frame-pointer. This is required for targets
6624 that produce debugging output for DBX or COFF-style debugging output
6625 for SDB and allow the frame-pointer to be eliminated when the
6626 @samp{-g} options is used.
6628 @findex DEBUGGER_ARG_OFFSET
6629 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6630 A C expression that returns the integer offset value for an argument
6631 having address @var{x} (an RTL expression). The nominal offset is
6634 @findex PREFERRED_DEBUGGING_TYPE
6635 @item PREFERRED_DEBUGGING_TYPE
6636 A C expression that returns the type of debugging output GCC should
6637 produce when the user specifies just @samp{-g}. Define
6638 this if you have arranged for GCC to support more than one format of
6639 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
6640 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG}, and
6643 When the user specifies @samp{-ggdb}, GCC normally also uses the
6644 value of this macro to select the debugging output format, but with two
6645 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
6646 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
6647 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
6648 defined, GCC uses @code{DBX_DEBUG}.
6650 The value of this macro only affects the default debugging output; the
6651 user can always get a specific type of output by using @samp{-gstabs},
6652 @samp{-gcoff}, @samp{-gdwarf-1}, @samp{-gdwarf-2}, or @samp{-gxcoff}.
6656 @subsection Specific Options for DBX Output
6658 @c prevent bad page break with this line
6659 These are specific options for DBX output.
6662 @findex DBX_DEBUGGING_INFO
6663 @item DBX_DEBUGGING_INFO
6664 Define this macro if GCC should produce debugging output for DBX
6665 in response to the @samp{-g} option.
6667 @findex XCOFF_DEBUGGING_INFO
6668 @item XCOFF_DEBUGGING_INFO
6669 Define this macro if GCC should produce XCOFF format debugging output
6670 in response to the @samp{-g} option. This is a variant of DBX format.
6672 @findex DEFAULT_GDB_EXTENSIONS
6673 @item DEFAULT_GDB_EXTENSIONS
6674 Define this macro to control whether GCC should by default generate
6675 GDB's extended version of DBX debugging information (assuming DBX-format
6676 debugging information is enabled at all). If you don't define the
6677 macro, the default is 1: always generate the extended information
6678 if there is any occasion to.
6680 @findex DEBUG_SYMS_TEXT
6681 @item DEBUG_SYMS_TEXT
6682 Define this macro if all @code{.stabs} commands should be output while
6683 in the text section.
6685 @findex ASM_STABS_OP
6687 A C string constant naming the assembler pseudo op to use instead of
6688 @code{.stabs} to define an ordinary debugging symbol. If you don't
6689 define this macro, @code{.stabs} is used. This macro applies only to
6690 DBX debugging information format.
6692 @findex ASM_STABD_OP
6694 A C string constant naming the assembler pseudo op to use instead of
6695 @code{.stabd} to define a debugging symbol whose value is the current
6696 location. If you don't define this macro, @code{.stabd} is used.
6697 This macro applies only to DBX debugging information format.
6699 @findex ASM_STABN_OP
6701 A C string constant naming the assembler pseudo op to use instead of
6702 @code{.stabn} to define a debugging symbol with no name. If you don't
6703 define this macro, @code{.stabn} is used. This macro applies only to
6704 DBX debugging information format.
6706 @findex DBX_NO_XREFS
6708 Define this macro if DBX on your system does not support the construct
6709 @samp{xs@var{tagname}}. On some systems, this construct is used to
6710 describe a forward reference to a structure named @var{tagname}.
6711 On other systems, this construct is not supported at all.
6713 @findex DBX_CONTIN_LENGTH
6714 @item DBX_CONTIN_LENGTH
6715 A symbol name in DBX-format debugging information is normally
6716 continued (split into two separate @code{.stabs} directives) when it
6717 exceeds a certain length (by default, 80 characters). On some
6718 operating systems, DBX requires this splitting; on others, splitting
6719 must not be done. You can inhibit splitting by defining this macro
6720 with the value zero. You can override the default splitting-length by
6721 defining this macro as an expression for the length you desire.
6723 @findex DBX_CONTIN_CHAR
6724 @item DBX_CONTIN_CHAR
6725 Normally continuation is indicated by adding a @samp{\} character to
6726 the end of a @code{.stabs} string when a continuation follows. To use
6727 a different character instead, define this macro as a character
6728 constant for the character you want to use. Do not define this macro
6729 if backslash is correct for your system.
6731 @findex DBX_STATIC_STAB_DATA_SECTION
6732 @item DBX_STATIC_STAB_DATA_SECTION
6733 Define this macro if it is necessary to go to the data section before
6734 outputting the @samp{.stabs} pseudo-op for a non-global static
6737 @findex DBX_TYPE_DECL_STABS_CODE
6738 @item DBX_TYPE_DECL_STABS_CODE
6739 The value to use in the ``code'' field of the @code{.stabs} directive
6740 for a typedef. The default is @code{N_LSYM}.
6742 @findex DBX_STATIC_CONST_VAR_CODE
6743 @item DBX_STATIC_CONST_VAR_CODE
6744 The value to use in the ``code'' field of the @code{.stabs} directive
6745 for a static variable located in the text section. DBX format does not
6746 provide any ``right'' way to do this. The default is @code{N_FUN}.
6748 @findex DBX_REGPARM_STABS_CODE
6749 @item DBX_REGPARM_STABS_CODE
6750 The value to use in the ``code'' field of the @code{.stabs} directive
6751 for a parameter passed in registers. DBX format does not provide any
6752 ``right'' way to do this. The default is @code{N_RSYM}.
6754 @findex DBX_REGPARM_STABS_LETTER
6755 @item DBX_REGPARM_STABS_LETTER
6756 The letter to use in DBX symbol data to identify a symbol as a parameter
6757 passed in registers. DBX format does not customarily provide any way to
6758 do this. The default is @code{'P'}.
6760 @findex DBX_MEMPARM_STABS_LETTER
6761 @item DBX_MEMPARM_STABS_LETTER
6762 The letter to use in DBX symbol data to identify a symbol as a stack
6763 parameter. The default is @code{'p'}.
6765 @findex DBX_FUNCTION_FIRST
6766 @item DBX_FUNCTION_FIRST
6767 Define this macro if the DBX information for a function and its
6768 arguments should precede the assembler code for the function. Normally,
6769 in DBX format, the debugging information entirely follows the assembler
6772 @findex DBX_LBRAC_FIRST
6773 @item DBX_LBRAC_FIRST
6774 Define this macro if the @code{N_LBRAC} symbol for a block should
6775 precede the debugging information for variables and functions defined in
6776 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
6779 @findex DBX_BLOCKS_FUNCTION_RELATIVE
6780 @item DBX_BLOCKS_FUNCTION_RELATIVE
6781 Define this macro if the value of a symbol describing the scope of a
6782 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
6783 of the enclosing function. Normally, GNU C uses an absolute address.
6785 @findex DBX_USE_BINCL
6787 Define this macro if GNU C should generate @code{N_BINCL} and
6788 @code{N_EINCL} stabs for included header files, as on Sun systems. This
6789 macro also directs GNU C to output a type number as a pair of a file
6790 number and a type number within the file. Normally, GNU C does not
6791 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
6792 number for a type number.
6796 @subsection Open-Ended Hooks for DBX Format
6798 @c prevent bad page break with this line
6799 These are hooks for DBX format.
6802 @findex DBX_OUTPUT_LBRAC
6803 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
6804 Define this macro to say how to output to @var{stream} the debugging
6805 information for the start of a scope level for variable names. The
6806 argument @var{name} is the name of an assembler symbol (for use with
6807 @code{assemble_name}) whose value is the address where the scope begins.
6809 @findex DBX_OUTPUT_RBRAC
6810 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
6811 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
6813 @findex DBX_OUTPUT_ENUM
6814 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
6815 Define this macro if the target machine requires special handling to
6816 output an enumeration type. The definition should be a C statement
6817 (sans semicolon) to output the appropriate information to @var{stream}
6818 for the type @var{type}.
6820 @findex DBX_OUTPUT_FUNCTION_END
6821 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
6822 Define this macro if the target machine requires special output at the
6823 end of the debugging information for a function. The definition should
6824 be a C statement (sans semicolon) to output the appropriate information
6825 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
6828 @findex DBX_OUTPUT_STANDARD_TYPES
6829 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
6830 Define this macro if you need to control the order of output of the
6831 standard data types at the beginning of compilation. The argument
6832 @var{syms} is a @code{tree} which is a chain of all the predefined
6833 global symbols, including names of data types.
6835 Normally, DBX output starts with definitions of the types for integers
6836 and characters, followed by all the other predefined types of the
6837 particular language in no particular order.
6839 On some machines, it is necessary to output different particular types
6840 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
6841 those symbols in the necessary order. Any predefined types that you
6842 don't explicitly output will be output afterward in no particular order.
6844 Be careful not to define this macro so that it works only for C. There
6845 are no global variables to access most of the built-in types, because
6846 another language may have another set of types. The way to output a
6847 particular type is to look through @var{syms} to see if you can find it.
6853 for (decl = syms; decl; decl = TREE_CHAIN (decl))
6854 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
6856 dbxout_symbol (decl);
6862 This does nothing if the expected type does not exist.
6864 See the function @code{init_decl_processing} in @file{c-decl.c} to find
6865 the names to use for all the built-in C types.
6867 Here is another way of finding a particular type:
6869 @c this is still overfull. --mew 10feb93
6873 for (decl = syms; decl; decl = TREE_CHAIN (decl))
6874 if (TREE_CODE (decl) == TYPE_DECL
6875 && (TREE_CODE (TREE_TYPE (decl))
6877 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
6878 && TYPE_UNSIGNED (TREE_TYPE (decl)))
6880 /* @r{This must be @code{unsigned short}.} */
6881 dbxout_symbol (decl);
6887 @findex NO_DBX_FUNCTION_END
6888 @item NO_DBX_FUNCTION_END
6889 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
6890 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extention construct.
6891 On those machines, define this macro to turn this feature off without
6892 disturbing the rest of the gdb extensions.
6896 @node File Names and DBX
6897 @subsection File Names in DBX Format
6899 @c prevent bad page break with this line
6900 This describes file names in DBX format.
6903 @findex DBX_WORKING_DIRECTORY
6904 @item DBX_WORKING_DIRECTORY
6905 Define this if DBX wants to have the current directory recorded in each
6908 Note that the working directory is always recorded if GDB extensions are
6911 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
6912 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
6913 A C statement to output DBX debugging information to the stdio stream
6914 @var{stream} which indicates that file @var{name} is the main source
6915 file---the file specified as the input file for compilation.
6916 This macro is called only once, at the beginning of compilation.
6918 This macro need not be defined if the standard form of output
6919 for DBX debugging information is appropriate.
6921 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
6922 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
6923 A C statement to output DBX debugging information to the stdio stream
6924 @var{stream} which indicates that the current directory during
6925 compilation is named @var{name}.
6927 This macro need not be defined if the standard form of output
6928 for DBX debugging information is appropriate.
6930 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
6931 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
6932 A C statement to output DBX debugging information at the end of
6933 compilation of the main source file @var{name}.
6935 If you don't define this macro, nothing special is output at the end
6936 of compilation, which is correct for most machines.
6938 @findex DBX_OUTPUT_SOURCE_FILENAME
6939 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6940 A C statement to output DBX debugging information to the stdio stream
6941 @var{stream} which indicates that file @var{name} is the current source
6942 file. This output is generated each time input shifts to a different
6943 source file as a result of @samp{#include}, the end of an included file,
6944 or a @samp{#line} command.
6946 This macro need not be defined if the standard form of output
6947 for DBX debugging information is appropriate.
6952 @subsection Macros for SDB and DWARF Output
6954 @c prevent bad page break with this line
6955 Here are macros for SDB and DWARF output.
6958 @findex SDB_DEBUGGING_INFO
6959 @item SDB_DEBUGGING_INFO
6960 Define this macro if GCC should produce COFF-style debugging output
6961 for SDB in response to the @samp{-g} option.
6963 @findex DWARF_DEBUGGING_INFO
6964 @item DWARF_DEBUGGING_INFO
6965 Define this macro if GCC should produce dwarf format debugging output
6966 in response to the @samp{-g} option.
6968 @findex DWARF2_DEBUGGING_INFO
6969 @item DWARF2_DEBUGGING_INFO
6970 Define this macro if GCC should produce dwarf version 2 format
6971 debugging output in response to the @samp{-g} option.
6973 To support optional call frame debugging information, you must also
6974 define @code{INCOMING_RETURN_ADDR_RTX} and either set
6975 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
6976 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
6977 as appropriate from @code{FUNCTION_PROLOGUE} if you don't.
6979 @findex DWARF2_FRAME_INFO
6980 @item DWARF2_FRAME_INFO
6981 Define this macro to a nonzero value if GCC should always output
6982 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
6983 (@pxref{Exception Region Output} is nonzero, GCC will output this
6984 information not matter how you define @code{DWARF2_FRAME_INFO}.
6986 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
6987 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
6988 Define this macro if the linker does not work with Dwarf version 2.
6989 Normally, if the user specifies only @samp{-ggdb} GCC will use Dwarf
6990 version 2 if available; this macro disables this. See the description
6991 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
6993 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
6994 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
6995 By default, the Dwarf 2 debugging information generator will generate a
6996 label to mark the beginning of the text section. If it is better simply
6997 to use the name of the text section itself, rather than an explicit label,
6998 to indicate the beginning of the text section, define this macro to zero.
7000 @findex DWARF2_ASM_LINE_DEBUG_INFO
7001 @item DWARF2_ASM_LINE_DEBUG_INFO
7002 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
7003 line debug info sections. This will result in much more compact line number
7004 tables, and hence is desirable if it works.
7006 @findex PUT_SDB_@dots{}
7007 @item PUT_SDB_@dots{}
7008 Define these macros to override the assembler syntax for the special
7009 SDB assembler directives. See @file{sdbout.c} for a list of these
7010 macros and their arguments. If the standard syntax is used, you need
7011 not define them yourself.
7015 Some assemblers do not support a semicolon as a delimiter, even between
7016 SDB assembler directives. In that case, define this macro to be the
7017 delimiter to use (usually @samp{\n}). It is not necessary to define
7018 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
7021 @findex SDB_GENERATE_FAKE
7022 @item SDB_GENERATE_FAKE
7023 Define this macro to override the usual method of constructing a dummy
7024 name for anonymous structure and union types. See @file{sdbout.c} for
7027 @findex SDB_ALLOW_UNKNOWN_REFERENCES
7028 @item SDB_ALLOW_UNKNOWN_REFERENCES
7029 Define this macro to allow references to unknown structure,
7030 union, or enumeration tags to be emitted. Standard COFF does not
7031 allow handling of unknown references, MIPS ECOFF has support for
7034 @findex SDB_ALLOW_FORWARD_REFERENCES
7035 @item SDB_ALLOW_FORWARD_REFERENCES
7036 Define this macro to allow references to structure, union, or
7037 enumeration tags that have not yet been seen to be handled. Some
7038 assemblers choke if forward tags are used, while some require it.
7041 @node Cross-compilation
7042 @section Cross Compilation and Floating Point
7043 @cindex cross compilation and floating point
7044 @cindex floating point and cross compilation
7046 While all modern machines use 2's complement representation for integers,
7047 there are a variety of representations for floating point numbers. This
7048 means that in a cross-compiler the representation of floating point numbers
7049 in the compiled program may be different from that used in the machine
7050 doing the compilation.
7053 Because different representation systems may offer different amounts of
7054 range and precision, the cross compiler cannot safely use the host
7055 machine's floating point arithmetic. Therefore, floating point constants
7056 must be represented in the target machine's format. This means that the
7057 cross compiler cannot use @code{atof} to parse a floating point constant;
7058 it must have its own special routine to use instead. Also, constant
7059 folding must emulate the target machine's arithmetic (or must not be done
7062 The macros in the following table should be defined only if you are cross
7063 compiling between different floating point formats.
7065 Otherwise, don't define them. Then default definitions will be set up which
7066 use @code{double} as the data type, @code{==} to test for equality, etc.
7068 You don't need to worry about how many times you use an operand of any
7069 of these macros. The compiler never uses operands which have side effects.
7072 @findex REAL_VALUE_TYPE
7073 @item REAL_VALUE_TYPE
7074 A macro for the C data type to be used to hold a floating point value
7075 in the target machine's format. Typically this would be a
7076 @code{struct} containing an array of @code{int}.
7078 @findex REAL_VALUES_EQUAL
7079 @item REAL_VALUES_EQUAL (@var{x}, @var{y})
7080 A macro for a C expression which compares for equality the two values,
7081 @var{x} and @var{y}, both of type @code{REAL_VALUE_TYPE}.
7083 @findex REAL_VALUES_LESS
7084 @item REAL_VALUES_LESS (@var{x}, @var{y})
7085 A macro for a C expression which tests whether @var{x} is less than
7086 @var{y}, both values being of type @code{REAL_VALUE_TYPE} and
7087 interpreted as floating point numbers in the target machine's
7090 @findex REAL_VALUE_LDEXP
7092 @item REAL_VALUE_LDEXP (@var{x}, @var{scale})
7093 A macro for a C expression which performs the standard library
7094 function @code{ldexp}, but using the target machine's floating point
7095 representation. Both @var{x} and the value of the expression have
7096 type @code{REAL_VALUE_TYPE}. The second argument, @var{scale}, is an
7099 @findex REAL_VALUE_FIX
7100 @item REAL_VALUE_FIX (@var{x})
7101 A macro whose definition is a C expression to convert the target-machine
7102 floating point value @var{x} to a signed integer. @var{x} has type
7103 @code{REAL_VALUE_TYPE}.
7105 @findex REAL_VALUE_UNSIGNED_FIX
7106 @item REAL_VALUE_UNSIGNED_FIX (@var{x})
7107 A macro whose definition is a C expression to convert the target-machine
7108 floating point value @var{x} to an unsigned integer. @var{x} has type
7109 @code{REAL_VALUE_TYPE}.
7111 @findex REAL_VALUE_RNDZINT
7112 @item REAL_VALUE_RNDZINT (@var{x})
7113 A macro whose definition is a C expression to round the target-machine
7114 floating point value @var{x} towards zero to an integer value (but still
7115 as a floating point number). @var{x} has type @code{REAL_VALUE_TYPE},
7116 and so does the value.
7118 @findex REAL_VALUE_UNSIGNED_RNDZINT
7119 @item REAL_VALUE_UNSIGNED_RNDZINT (@var{x})
7120 A macro whose definition is a C expression to round the target-machine
7121 floating point value @var{x} towards zero to an unsigned integer value
7122 (but still represented as a floating point number). @var{x} has type
7123 @code{REAL_VALUE_TYPE}, and so does the value.
7125 @findex REAL_VALUE_ATOF
7126 @item REAL_VALUE_ATOF (@var{string}, @var{mode})
7127 A macro for a C expression which converts @var{string}, an expression of
7128 type @code{char *}, into a floating point number in the target machine's
7129 representation for mode @var{mode}. The value has type
7130 @code{REAL_VALUE_TYPE}.
7132 @findex REAL_INFINITY
7134 Define this macro if infinity is a possible floating point value, and
7135 therefore division by 0 is legitimate.
7137 @findex REAL_VALUE_ISINF
7139 @item REAL_VALUE_ISINF (@var{x})
7140 A macro for a C expression which determines whether @var{x}, a floating
7141 point value, is infinity. The value has type @code{int}.
7142 By default, this is defined to call @code{isinf}.
7144 @findex REAL_VALUE_ISNAN
7146 @item REAL_VALUE_ISNAN (@var{x})
7147 A macro for a C expression which determines whether @var{x}, a floating
7148 point value, is a ``nan'' (not-a-number). The value has type
7149 @code{int}. By default, this is defined to call @code{isnan}.
7152 @cindex constant folding and floating point
7153 Define the following additional macros if you want to make floating
7154 point constant folding work while cross compiling. If you don't
7155 define them, cross compilation is still possible, but constant folding
7156 will not happen for floating point values.
7159 @findex REAL_ARITHMETIC
7160 @item REAL_ARITHMETIC (@var{output}, @var{code}, @var{x}, @var{y})
7161 A macro for a C statement which calculates an arithmetic operation of
7162 the two floating point values @var{x} and @var{y}, both of type
7163 @code{REAL_VALUE_TYPE} in the target machine's representation, to
7164 produce a result of the same type and representation which is stored
7165 in @var{output} (which will be a variable).
7167 The operation to be performed is specified by @var{code}, a tree code
7168 which will always be one of the following: @code{PLUS_EXPR},
7169 @code{MINUS_EXPR}, @code{MULT_EXPR}, @code{RDIV_EXPR},
7170 @code{MAX_EXPR}, @code{MIN_EXPR}.@refill
7172 @cindex overflow while constant folding
7173 The expansion of this macro is responsible for checking for overflow.
7174 If overflow happens, the macro expansion should execute the statement
7175 @code{return 0;}, which indicates the inability to perform the
7176 arithmetic operation requested.
7178 @findex REAL_VALUE_NEGATE
7179 @item REAL_VALUE_NEGATE (@var{x})
7180 A macro for a C expression which returns the negative of the floating
7181 point value @var{x}. Both @var{x} and the value of the expression
7182 have type @code{REAL_VALUE_TYPE} and are in the target machine's
7183 floating point representation.
7185 There is no way for this macro to report overflow, since overflow
7186 can't happen in the negation operation.
7188 @findex REAL_VALUE_TRUNCATE
7189 @item REAL_VALUE_TRUNCATE (@var{mode}, @var{x})
7190 A macro for a C expression which converts the floating point value
7191 @var{x} to mode @var{mode}.
7193 Both @var{x} and the value of the expression are in the target machine's
7194 floating point representation and have type @code{REAL_VALUE_TYPE}.
7195 However, the value should have an appropriate bit pattern to be output
7196 properly as a floating constant whose precision accords with mode
7199 There is no way for this macro to report overflow.
7201 @findex REAL_VALUE_TO_INT
7202 @item REAL_VALUE_TO_INT (@var{low}, @var{high}, @var{x})
7203 A macro for a C expression which converts a floating point value
7204 @var{x} into a double-precision integer which is then stored into
7205 @var{low} and @var{high}, two variables of type @var{int}.
7207 @item REAL_VALUE_FROM_INT (@var{x}, @var{low}, @var{high}, @var{mode})
7208 @findex REAL_VALUE_FROM_INT
7209 A macro for a C expression which converts a double-precision integer
7210 found in @var{low} and @var{high}, two variables of type @var{int},
7211 into a floating point value which is then stored into @var{x}.
7212 The value is in the target machine's representation for mode @var{mode}
7213 and has the type @code{REAL_VALUE_TYPE}.
7216 @node Mode Switching
7217 @section Mode Switching Instructions
7218 @cindex mode switching
7219 The following macros control mode switching optimizations:
7222 @findex OPTIMIZE_MODE_SWITCHING
7223 @item OPTIMIZE_MODE_SWITCHING (@var{entity})
7224 Define this macro if the port needs extra instructions inserted for mode
7225 switching in an optimizing compilation.
7227 For an example, the SH4 can perform both single and double precision
7228 floating point operations, but to perform a single precision operation,
7229 the FPSCR PR bit has to be cleared, while for a double precision
7230 operation, this bit has to be set. Changing the PR bit requires a general
7231 purpose register as a scratch register, hence these FPSCR sets have to
7232 be inserted before reload, i.e. you can't put this into instruction emitting
7233 or MACHINE_DEPENDENT_REORG.
7235 You can have multiple entities that are mode-switched, and select at run time
7236 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
7237 return non-zero for any @var{entity} that that needs mode-switching.
7238 If you define this macro, you also have to define
7239 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
7240 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
7241 @code{MODE_AT_ENTRY} and @code{MODE_USES_IN_EXIT_BLOCK} are optional.
7243 @findex NUM_MODES_FOR_MODE_SWITCHING
7244 @item NUM_MODES_FOR_MODE_SWITCHING
7245 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
7246 initializer for an array of integers. Each initializer element
7247 N refers to an entity that needs mode switching, and specifies the number
7248 of different modes that might need to be set for this entity.
7249 The position of the initializer in the initializer - starting counting at
7250 zero - determines the integer that is used to refer to the mode-switched
7252 In macros that take mode arguments / yield a mode result, modes are
7253 represented as numbers 0 .. N - 1. N is used to specify that no mode
7254 switch is needed / supplied.
7256 @findex MODE_USES_IN_EXIT_BLOCK
7257 @item MODE_USES_IN_EXIT_BLOCK
7258 If this macro is defined, it is called for each exit block when mode switching
7259 optimization is performed. Its return value should be the pattern of an insn,
7260 or a sequence of insns. It is emitted before the return insn / use insns at
7261 the end of the exit block.
7263 This is done before insns are examined for their need of any mode switching.
7266 @item MODE_NEEDED (@var{entity}, @var{insn})
7267 @var{entity} is an integer specifying a mode-switched entity. If
7268 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
7269 return an integer value not larger than the corresponding element in
7270 NUM_MODES_FOR_MODE_SWITCHING, to denote the mode that @var{entity} must
7271 be switched into prior to the execution of INSN.
7273 @findex MODE_AT_ENTRY
7274 @item MODE_AT_ENTRY (@var{entity})
7275 If this macro is defined, it is evaluated for every @var{entity} that needs
7276 mode switching. It should evaluate to an integer, which is a mode that
7277 @var{entity} is assumed to be switched to at function entry.
7279 @findex MODE_PRIORITY_TO_MODE
7280 @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
7281 This macro specifies the order in which modes for ENTITY are processed.
7282 0 is the highest priority, NUM_MODES_FOR_MODE_SWITCHING[ENTITY] - 1 the
7283 lowest. The value of the macro should be an integer designating a mode
7284 for ENTITY. For any fixed @var{entity}, @code{mode_priority_to_mode}
7285 (@var{entity}, @var{n}) shall be a bijection in 0 ..
7286 @code{num_modes_for_mode_switching}[@var{entity}] - 1 .
7288 @findex EMIT_MODE_SET
7289 @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
7290 Generate one or more insns to set @var{entity} to @var{mode}.
7291 @var{hard_reg_live} is the set of hard registers live at the point where
7292 the insn(s) are to be inserted.
7296 @section Miscellaneous Parameters
7297 @cindex parameters, miscellaneous
7299 @c prevent bad page break with this line
7300 Here are several miscellaneous parameters.
7303 @item PREDICATE_CODES
7304 @findex PREDICATE_CODES
7305 Define this if you have defined special-purpose predicates in the file
7306 @file{@var{machine}.c}. This macro is called within an initializer of an
7307 array of structures. The first field in the structure is the name of a
7308 predicate and the second field is an array of rtl codes. For each
7309 predicate, list all rtl codes that can be in expressions matched by the
7310 predicate. The list should have a trailing comma. Here is an example
7311 of two entries in the list for a typical RISC machine:
7314 #define PREDICATE_CODES \
7315 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
7316 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
7319 Defining this macro does not affect the generated code (however,
7320 incorrect definitions that omit an rtl code that may be matched by the
7321 predicate can cause the compiler to malfunction). Instead, it allows
7322 the table built by @file{genrecog} to be more compact and efficient,
7323 thus speeding up the compiler. The most important predicates to include
7324 in the list specified by this macro are those used in the most insn
7327 @item SPECIAL_MODE_PREDICATES
7328 @findex SPECIAL_MODE_PREDICATES
7329 Define this if you have special predicates that know special things
7330 about modes. Genrecog will warn about certain forms of
7331 @code{match_operand} without a mode; if the operand predicate is
7332 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
7335 Here is an example from the IA-32 port (@code{ext_register_operand}
7336 specially checks for @code{HImode} or @code{SImode} in preparation
7337 for a byte extraction from @code{%ah} etc.).
7340 #define SPECIAL_MODE_PREDICATES \
7341 "ext_register_operand",
7344 @findex CASE_VECTOR_MODE
7345 @item CASE_VECTOR_MODE
7346 An alias for a machine mode name. This is the machine mode that
7347 elements of a jump-table should have.
7349 @findex CASE_VECTOR_SHORTEN_MODE
7350 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7351 Optional: return the preferred mode for an @code{addr_diff_vec}
7352 when the minimum and maximum offset are known. If you define this,
7353 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7354 To make this work, you also have to define INSN_ALIGN and
7355 make the alignment for @code{addr_diff_vec} explicit.
7356 The @var{body} argument is provided so that the offset_unsigned and scale
7357 flags can be updated.
7359 @findex CASE_VECTOR_PC_RELATIVE
7360 @item CASE_VECTOR_PC_RELATIVE
7361 Define this macro to be a C expression to indicate when jump-tables
7362 should contain relative addresses. If jump-tables never contain
7363 relative addresses, then you need not define this macro.
7365 @findex CASE_DROPS_THROUGH
7366 @item CASE_DROPS_THROUGH
7367 Define this if control falls through a @code{case} insn when the index
7368 value is out of range. This means the specified default-label is
7369 actually ignored by the @code{case} insn proper.
7371 @findex CASE_VALUES_THRESHOLD
7372 @item CASE_VALUES_THRESHOLD
7373 Define this to be the smallest number of different values for which it
7374 is best to use a jump-table instead of a tree of conditional branches.
7375 The default is four for machines with a @code{casesi} instruction and
7376 five otherwise. This is best for most machines.
7378 @findex WORD_REGISTER_OPERATIONS
7379 @item WORD_REGISTER_OPERATIONS
7380 Define this macro if operations between registers with integral mode
7381 smaller than a word are always performed on the entire register.
7382 Most RISC machines have this property and most CISC machines do not.
7384 @findex LOAD_EXTEND_OP
7385 @item LOAD_EXTEND_OP (@var{mode})
7386 Define this macro to be a C expression indicating when insns that read
7387 memory in @var{mode}, an integral mode narrower than a word, set the
7388 bits outside of @var{mode} to be either the sign-extension or the
7389 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7390 of @var{mode} for which the
7391 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7392 @code{NIL} for other modes.
7394 This macro is not called with @var{mode} non-integral or with a width
7395 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7396 value in this case. Do not define this macro if it would always return
7397 @code{NIL}. On machines where this macro is defined, you will normally
7398 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7400 @findex SHORT_IMMEDIATES_SIGN_EXTEND
7401 @item SHORT_IMMEDIATES_SIGN_EXTEND
7402 Define this macro if loading short immediate values into registers sign
7405 @findex IMPLICIT_FIX_EXPR
7406 @item IMPLICIT_FIX_EXPR
7407 An alias for a tree code that should be used by default for conversion
7408 of floating point values to fixed point. Normally,
7409 @code{FIX_ROUND_EXPR} is used.@refill
7411 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
7412 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
7413 Define this macro if the same instructions that convert a floating
7414 point number to a signed fixed point number also convert validly to an
7417 @findex EASY_DIV_EXPR
7419 An alias for a tree code that is the easiest kind of division to
7420 compile code for in the general case. It may be
7421 @code{TRUNC_DIV_EXPR}, @code{FLOOR_DIV_EXPR}, @code{CEIL_DIV_EXPR} or
7422 @code{ROUND_DIV_EXPR}. These four division operators differ in how
7423 they round the result to an integer. @code{EASY_DIV_EXPR} is used
7424 when it is permissible to use any of those kinds of division and the
7425 choice should be made on the basis of efficiency.@refill
7429 The maximum number of bytes that a single instruction can move quickly
7430 between memory and registers or between two memory locations.
7432 @findex MAX_MOVE_MAX
7434 The maximum number of bytes that a single instruction can move quickly
7435 between memory and registers or between two memory locations. If this
7436 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7437 constant value that is the largest value that @code{MOVE_MAX} can have
7440 @findex SHIFT_COUNT_TRUNCATED
7441 @item SHIFT_COUNT_TRUNCATED
7442 A C expression that is nonzero if on this machine the number of bits
7443 actually used for the count of a shift operation is equal to the number
7444 of bits needed to represent the size of the object being shifted. When
7445 this macro is non-zero, the compiler will assume that it is safe to omit
7446 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7447 truncates the count of a shift operation. On machines that have
7448 instructions that act on bitfields at variable positions, which may
7449 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7450 also enables deletion of truncations of the values that serve as
7451 arguments to bitfield instructions.
7453 If both types of instructions truncate the count (for shifts) and
7454 position (for bitfield operations), or if no variable-position bitfield
7455 instructions exist, you should define this macro.
7457 However, on some machines, such as the 80386 and the 680x0, truncation
7458 only applies to shift operations and not the (real or pretended)
7459 bitfield operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7460 such machines. Instead, add patterns to the @file{md} file that include
7461 the implied truncation of the shift instructions.
7463 You need not define this macro if it would always have the value of zero.
7465 @findex TRULY_NOOP_TRUNCATION
7466 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
7467 A C expression which is nonzero if on this machine it is safe to
7468 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
7469 bits (where @var{outprec} is smaller than @var{inprec}) by merely
7470 operating on it as if it had only @var{outprec} bits.
7472 On many machines, this expression can be 1.
7474 @c rearranged this, removed the phrase "it is reported that". this was
7475 @c to fix an overfull hbox. --mew 10feb93
7476 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
7477 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
7478 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
7479 such cases may improve things.
7481 @findex STORE_FLAG_VALUE
7482 @item STORE_FLAG_VALUE
7483 A C expression describing the value returned by a comparison operator
7484 with an integral mode and stored by a store-flag instruction
7485 (@samp{s@var{cond}}) when the condition is true. This description must
7486 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
7487 comparison operators whose results have a @code{MODE_INT} mode.
7489 A value of 1 or -1 means that the instruction implementing the
7490 comparison operator returns exactly 1 or -1 when the comparison is true
7491 and 0 when the comparison is false. Otherwise, the value indicates
7492 which bits of the result are guaranteed to be 1 when the comparison is
7493 true. This value is interpreted in the mode of the comparison
7494 operation, which is given by the mode of the first operand in the
7495 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
7496 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7499 If @code{STORE_FLAG_VALUE} is neither 1 or -1, the compiler will
7500 generate code that depends only on the specified bits. It can also
7501 replace comparison operators with equivalent operations if they cause
7502 the required bits to be set, even if the remaining bits are undefined.
7503 For example, on a machine whose comparison operators return an
7504 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7505 @samp{0x80000000}, saying that just the sign bit is relevant, the
7509 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7516 (ashift:SI @var{x} (const_int @var{n}))
7520 where @var{n} is the appropriate shift count to move the bit being
7521 tested into the sign bit.
7523 There is no way to describe a machine that always sets the low-order bit
7524 for a true value, but does not guarantee the value of any other bits,
7525 but we do not know of any machine that has such an instruction. If you
7526 are trying to port GCC to such a machine, include an instruction to
7527 perform a logical-and of the result with 1 in the pattern for the
7528 comparison operators and let us know
7530 (@pxref{Bug Reporting,,How to Report Bugs}).
7533 (@pxref{Bug Reporting,,How to Report Bugs,gcc.info,Using GCC}).
7536 Often, a machine will have multiple instructions that obtain a value
7537 from a comparison (or the condition codes). Here are rules to guide the
7538 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7543 Use the shortest sequence that yields a valid definition for
7544 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7545 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7546 comparison operators to do so because there may be opportunities to
7547 combine the normalization with other operations.
7550 For equal-length sequences, use a value of 1 or -1, with -1 being
7551 slightly preferred on machines with expensive jumps and 1 preferred on
7555 As a second choice, choose a value of @samp{0x80000001} if instructions
7556 exist that set both the sign and low-order bits but do not define the
7560 Otherwise, use a value of @samp{0x80000000}.
7563 Many machines can produce both the value chosen for
7564 @code{STORE_FLAG_VALUE} and its negation in the same number of
7565 instructions. On those machines, you should also define a pattern for
7566 those cases, e.g., one matching
7569 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7572 Some machines can also perform @code{and} or @code{plus} operations on
7573 condition code values with less instructions than the corresponding
7574 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
7575 machines, define the appropriate patterns. Use the names @code{incscc}
7576 and @code{decscc}, respectively, for the patterns which perform
7577 @code{plus} or @code{minus} operations on condition code values. See
7578 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
7579 find such instruction sequences on other machines.
7581 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7584 @findex FLOAT_STORE_FLAG_VALUE
7585 @item FLOAT_STORE_FLAG_VALUE (@var{mode})
7586 A C expression that gives a non-zero @code{REAL_VALUE_TYPE} value that is
7587 returned when comparison operators with floating-point results are true.
7588 Define this macro on machine that have comparison operations that return
7589 floating-point values. If there are no such operations, do not define
7594 An alias for the machine mode for pointers. On most machines, define
7595 this to be the integer mode corresponding to the width of a hardware
7596 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7597 On some machines you must define this to be one of the partial integer
7598 modes, such as @code{PSImode}.
7600 The width of @code{Pmode} must be at least as large as the value of
7601 @code{POINTER_SIZE}. If it is not equal, you must define the macro
7602 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7605 @findex FUNCTION_MODE
7607 An alias for the machine mode used for memory references to functions
7608 being called, in @code{call} RTL expressions. On most machines this
7609 should be @code{QImode}.
7611 @findex INTEGRATE_THRESHOLD
7612 @item INTEGRATE_THRESHOLD (@var{decl})
7613 A C expression for the maximum number of instructions above which the
7614 function @var{decl} should not be inlined. @var{decl} is a
7615 @code{FUNCTION_DECL} node.
7617 The default definition of this macro is 64 plus 8 times the number of
7618 arguments that the function accepts. Some people think a larger
7619 threshold should be used on RISC machines.
7621 @findex SCCS_DIRECTIVE
7622 @item SCCS_DIRECTIVE
7623 Define this if the preprocessor should ignore @code{#sccs} directives
7624 and print no error message.
7626 @findex NO_IMPLICIT_EXTERN_C
7627 @item NO_IMPLICIT_EXTERN_C
7628 Define this macro if the system header files support C++ as well as C.
7629 This macro inhibits the usual method of using system header files in
7630 C++, which is to pretend that the file's contents are enclosed in
7631 @samp{extern "C" @{@dots{}@}}.
7633 @findex HANDLE_PRAGMA
7636 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
7637 Define this macro if you want to implement any pragmas. If defined, it
7638 is a C expression whose value is 1 if the pragma was handled by the
7639 macro, zero otherwise. The argument @var{getc} is a function of type
7640 @samp{int (*)(void)} which will return the next character in the input
7641 stream, or EOF if no characters are left. The argument @var{ungetc} is
7642 a function of type @samp{void (*)(int)} which will push a character back
7643 into the input stream. The argument @var{name} is the word following
7644 #pragma in the input stream. The input stream pointer will be pointing
7645 just beyond the end of this word. The input stream should be left
7646 undistrubed if the expression returns zero, otherwise it should be
7647 pointing at the next character after the end of the pragma. Any
7648 characters remaining on the line will be ignored.
7650 It is generally a bad idea to implement new uses of @code{#pragma}. The
7651 only reason to define this macro is for compatibility with other
7652 compilers that do support @code{#pragma} for the sake of any user
7653 programs which already use it.
7655 If the pragma can be implemented by atttributes then the macro
7656 @samp{INSERT_ATTRIBUTES} might be a useful one to define as well.
7658 Note: older versions of this macro only had two arguments: @var{stream}
7659 and @var{token}. The macro was changed in order to allow it to work
7660 when gcc is built both with and without a cpp library.
7662 @findex HANDLE_SYSV_PRAGMA
7665 @item HANDLE_SYSV_PRAGMA
7666 Define this macro (to a value of 1) if you want the System V style
7667 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
7668 [=<value>]} to be supported by gcc.
7670 The pack pragma specifies the maximum alignment (in bytes) of fields
7671 within a structure, in much the same way as the @samp{__aligned__} and
7672 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
7673 the behaviour to the default.
7675 The weak pragma only works if @code{SUPPORTS_WEAK} and
7676 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
7677 of specifically named weak labels, optionally with a value.
7679 @findex HANDLE_PRAGMA_PACK_PUSH_POP
7682 @item HANDLE_PRAGMA_PACK_PUSH_POP
7683 Define this macro (to a value of 1) if you want to support the Win32
7684 style pragmas @samp{#pragma pack(push,<n>)} and @samp{#pragma
7685 pack(pop)}. The pack(push,<n>) pragma specifies the maximum alignment
7686 (in bytes) of fields within a structure, in much the same way as the
7687 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
7688 pack value of zero resets the behaviour to the default. Successive
7689 invocations of this pragma cause the previous values to be stacked, so
7690 that invocations of @samp{#pragma pack(pop)} will return to the previous
7693 @findex VALID_MACHINE_DECL_ATTRIBUTE
7694 @item VALID_MACHINE_DECL_ATTRIBUTE (@var{decl}, @var{attributes}, @var{identifier}, @var{args})
7695 If defined, a C expression whose value is nonzero if @var{identifier} with
7696 arguments @var{args} is a valid machine specific attribute for @var{decl}.
7697 The attributes in @var{attributes} have previously been assigned to @var{decl}.
7699 @findex VALID_MACHINE_TYPE_ATTRIBUTE
7700 @item VALID_MACHINE_TYPE_ATTRIBUTE (@var{type}, @var{attributes}, @var{identifier}, @var{args})
7701 If defined, a C expression whose value is nonzero if @var{identifier} with
7702 arguments @var{args} is a valid machine specific attribute for @var{type}.
7703 The attributes in @var{attributes} have previously been assigned to @var{type}.
7705 @findex COMP_TYPE_ATTRIBUTES
7706 @item COMP_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
7707 If defined, a C expression whose value is zero if the attributes on
7708 @var{type1} and @var{type2} are incompatible, one if they are compatible,
7709 and two if they are nearly compatible (which causes a warning to be
7712 @findex SET_DEFAULT_TYPE_ATTRIBUTES
7713 @item SET_DEFAULT_TYPE_ATTRIBUTES (@var{type})
7714 If defined, a C statement that assigns default attributes to
7715 newly defined @var{type}.
7717 @findex MERGE_MACHINE_TYPE_ATTRIBUTES
7718 @item MERGE_MACHINE_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
7719 Define this macro if the merging of type attributes needs special handling.
7720 If defined, the result is a list of the combined TYPE_ATTRIBUTES of
7721 @var{type1} and @var{type2}. It is assumed that comptypes has already been
7722 called and returned 1.
7724 @findex MERGE_MACHINE_DECL_ATTRIBUTES
7725 @item MERGE_MACHINE_DECL_ATTRIBUTES (@var{olddecl}, @var{newdecl})
7726 Define this macro if the merging of decl attributes needs special handling.
7727 If defined, the result is a list of the combined DECL_MACHINE_ATTRIBUTES of
7728 @var{olddecl} and @var{newdecl}. @var{newdecl} is a duplicate declaration
7729 of @var{olddecl}. Examples of when this is needed are when one attribute
7730 overrides another, or when an attribute is nullified by a subsequent
7733 @findex INSERT_ATTRIBUTES
7734 @item INSERT_ATTRIBUTES (@var{node}, @var{attr_ptr}, @var{prefix_ptr})
7735 Define this macro if you want to be able to add attributes to a decl
7736 when it is being created. This is normally useful for backends which
7737 wish to implement a pragma by using the attributes which correspond to
7738 the pragma's effect. The @var{node} argument is the decl which is being
7739 created. The @var{attr_ptr} argument is a pointer to the attribute list
7740 for this decl. The @var{prefix_ptr} is a pointer to the list of
7741 attributes that have appeared after the specifiers and modifiers of the
7742 declaration, but before the declaration proper.
7744 @findex SET_DEFAULT_DECL_ATTRIBUTES
7745 @item SET_DEFAULT_DECL_ATTRIBUTES (@var{decl}, @var{attributes})
7746 If defined, a C statement that assigns default attributes to
7747 newly defined @var{decl}.
7749 @findex DOLLARS_IN_IDENTIFIERS
7750 @item DOLLARS_IN_IDENTIFIERS
7751 Define this macro to control use of the character @samp{$} in identifier
7752 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
7753 1 is the default; there is no need to define this macro in that case.
7754 This macro controls the compiler proper; it does not affect the preprocessor.
7756 @findex NO_DOLLAR_IN_LABEL
7757 @item NO_DOLLAR_IN_LABEL
7758 Define this macro if the assembler does not accept the character
7759 @samp{$} in label names. By default constructors and destructors in
7760 G++ have @samp{$} in the identifiers. If this macro is defined,
7761 @samp{.} is used instead.
7763 @findex NO_DOT_IN_LABEL
7764 @item NO_DOT_IN_LABEL
7765 Define this macro if the assembler does not accept the character
7766 @samp{.} in label names. By default constructors and destructors in G++
7767 have names that use @samp{.}. If this macro is defined, these names
7768 are rewritten to avoid @samp{.}.
7770 @findex DEFAULT_MAIN_RETURN
7771 @item DEFAULT_MAIN_RETURN
7772 Define this macro if the target system expects every program's @code{main}
7773 function to return a standard ``success'' value by default (if no other
7774 value is explicitly returned).
7776 The definition should be a C statement (sans semicolon) to generate the
7777 appropriate rtl instructions. It is used only when compiling the end of
7782 Define this if the target system lacks the function @code{atexit}
7783 from the ANSI C standard. If this macro is defined, a default definition
7784 will be provided to support C++. If @code{ON_EXIT} is not defined,
7785 a default @code{exit} function will also be provided.
7789 Define this macro if the target has another way to implement atexit
7790 functionality without replacing @code{exit}. For instance, SunOS 4 has
7791 a similar @code{on_exit} library function.
7793 The definition should be a functional macro which can be used just like
7794 the @code{atexit} function.
7798 Define this if your @code{exit} function needs to do something
7799 besides calling an external function @code{_cleanup} before
7800 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
7801 only needed if neither @code{HAVE_ATEXIT} nor
7802 @code{INIT_SECTION_ASM_OP} are defined.
7804 @findex INSN_SETS_ARE_DELAYED
7805 @item INSN_SETS_ARE_DELAYED (@var{insn})
7806 Define this macro as a C expression that is nonzero if it is safe for the
7807 delay slot scheduler to place instructions in the delay slot of @var{insn},
7808 even if they appear to use a resource set or clobbered in @var{insn}.
7809 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
7810 every @code{call_insn} has this behavior. On machines where some @code{insn}
7811 or @code{jump_insn} is really a function call and hence has this behavior,
7812 you should define this macro.
7814 You need not define this macro if it would always return zero.
7816 @findex INSN_REFERENCES_ARE_DELAYED
7817 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
7818 Define this macro as a C expression that is nonzero if it is safe for the
7819 delay slot scheduler to place instructions in the delay slot of @var{insn},
7820 even if they appear to set or clobber a resource referenced in @var{insn}.
7821 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
7822 some @code{insn} or @code{jump_insn} is really a function call and its operands
7823 are registers whose use is actually in the subroutine it calls, you should
7824 define this macro. Doing so allows the delay slot scheduler to move
7825 instructions which copy arguments into the argument registers into the delay
7828 You need not define this macro if it would always return zero.
7830 @findex MACHINE_DEPENDENT_REORG
7831 @item MACHINE_DEPENDENT_REORG (@var{insn})
7832 In rare cases, correct code generation requires extra machine
7833 dependent processing between the second jump optimization pass and
7834 delayed branch scheduling. On those machines, define this macro as a C
7835 statement to act on the code starting at @var{insn}.
7837 @findex MULTIPLE_SYMBOL_SPACES
7838 @item MULTIPLE_SYMBOL_SPACES
7839 Define this macro if in some cases global symbols from one translation
7840 unit may not be bound to undefined symbols in another translation unit
7841 without user intervention. For instance, under Microsoft Windows
7842 symbols must be explicitly imported from shared libraries (DLLs).
7844 @findex MD_ASM_CLOBBERS
7845 @item MD_ASM_CLOBBERS
7846 A C statement that adds to @var{CLOBBERS} @code{STRING_CST} trees for
7847 any hard regs the port wishes to automatically clobber for all asms.
7851 A C expression that returns how many instructions can be issued at the
7852 same time if the machine is a superscalar machine.
7854 @findex MD_SCHED_INIT
7855 @item MD_SCHED_INIT (@var{file}, @var{verbose})
7856 A C statement which is executed by the scheduler at the
7857 beginning of each block of instructions that are to be scheduled.
7858 @var{file} is either a null pointer, or a stdio stream to write any
7859 debug output to. @var{verbose} is the verbose level provided by
7860 @samp{-fsched-verbose-}@var{n}.
7862 @findex MD_SCHED_REORDER
7863 @item MD_SCHED_REORDER (@var{file}, @var{verbose}, @var{ready}, @var{n_ready}, @var{clock}, @var{can_issue_more})
7864 A C statement which is executed by the scheduler after it
7865 has scheduled the ready list to allow the machine description to reorder
7866 it (for example to combine two small instructions together on
7867 @samp{VLIW} machines). @var{file} is either a null pointer, or a stdio
7868 stream to write any debug output to. @var{verbose} is the verbose level
7869 provided by @samp{-fsched-verbose-}@var{n}. @var{ready} is a pointer to
7870 the ready list of instructions that are ready to be scheduled.
7871 @var{n_ready} is the number of elements in the ready list. The
7872 scheduler reads the ready list in reverse order, starting with
7873 @var{ready}[@var{n_ready}-1] and going to @var{ready}[0]. @var{clock}
7874 is the timer tick of the scheduler. @var{can_issue_more} is an output
7875 parameter that is set to the number of insns that can issue this clock;
7876 normally this is just @code{issue_rate}.
7878 @findex MD_SCHED_VARIABLE_ISSUE
7879 @item MD_SCHED_VARIABLE_ISSUE (@var{file}, @var{verbose}, @var{insn}, @var{more})
7880 A C statement which is executed by the scheduler after it
7881 has scheduled an insn from the ready list. @var{file} is either a null
7882 pointer, or a stdio stream to write any debug output to. @var{verbose}
7883 is the verbose level provided by @samp{-fsched-verbose-}@var{n}.
7884 @var{insn} is the instruction that was scheduled. @var{more} is the
7885 number of instructions that can be issued in the current cycle. The
7886 @samp{MD_SCHED_VARIABLE_ISSUE} macro is responsible for updating the
7887 value of @var{more} (typically by @var{more}--).
7889 @findex MAX_INTEGER_COMPUTATION_MODE
7890 @item MAX_INTEGER_COMPUTATION_MODE
7891 Define this to the largest integer machine mode which can be used for
7892 operations other than load, store and copy operations.
7894 You need only define this macro if the target holds values larger than
7895 @code{word_mode} in general purpose registers. Most targets should not define
7898 @findex MATH_LIBRARY
7900 Define this macro as a C string constant for the linker argument to link
7901 in the system math library, or @samp{""} if the target does not have a
7902 separate math library.
7904 You need only define this macro if the default of @samp{"-lm"} is wrong.
7906 @findex LIBRARY_PATH_ENV
7907 @item LIBRARY_PATH_ENV
7908 Define this macro as a C string constant for the environment variable that
7909 specifies where the linker should look for libraries.
7911 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
7914 @findex TARGET_HAS_F_SETLKW
7915 @item TARGET_HAS_F_SETLKW
7916 Define this macro iff the target supports file locking with fcntl / F_SETLKW.
7917 Note that this functionality is part of POSIX.
7918 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
7919 to use file locking when exiting a program, which avoids race conditions
7920 if the program has forked.