1 @c Copyright (C) 1988,89,92-97,1998 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
6 @chapter Target Description Macros
7 @cindex machine description macros
8 @cindex target description macros
9 @cindex macros, target description
10 @cindex @file{tm.h} macros
12 In addition to the file @file{@var{machine}.md}, a machine description
13 includes a C header file conventionally given the name
14 @file{@var{machine}.h}. This header file defines numerous macros
15 that convey the information about the target machine that does not fit
16 into the scheme of the @file{.md} file. The file @file{tm.h} should be
17 a link to @file{@var{machine}.h}. The header file @file{config.h}
18 includes @file{tm.h} and most compiler source files include
22 * Driver:: Controlling how the driver runs the compilation passes.
23 * Run-time Target:: Defining @samp{-m} options like @samp{-m68000} and @samp{-m68020}.
24 * Storage Layout:: Defining sizes and alignments of data.
25 * Type Layout:: Defining sizes and properties of basic user data types.
26 * Registers:: Naming and describing the hardware registers.
27 * Register Classes:: Defining the classes of hardware registers.
28 * Stack and Calling:: Defining which way the stack grows and by how much.
29 * Varargs:: Defining the varargs macros.
30 * Trampolines:: Code set up at run time to enter a nested function.
31 * Library Calls:: Controlling how library routines are implicitly called.
32 * Addressing Modes:: Defining addressing modes valid for memory operands.
33 * Condition Code:: Defining how insns update the condition code.
34 * Costs:: Defining relative costs of different operations.
35 * Sections:: Dividing storage into text, data, and other sections.
36 * PIC:: Macros for position independent code.
37 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
38 * Debugging Info:: Defining the format of debugging output.
39 * Cross-compilation:: Handling floating point for cross-compilers.
40 * Misc:: Everything else.
44 @section Controlling the Compilation Driver, @file{gcc}
46 @cindex controlling the compilation driver
48 @c prevent bad page break with this line
49 You can control the compilation driver.
52 @findex SWITCH_TAKES_ARG
53 @item SWITCH_TAKES_ARG (@var{char})
54 A C expression which determines whether the option @samp{-@var{char}}
55 takes arguments. The value should be the number of arguments that
56 option takes--zero, for many options.
58 By default, this macro is defined as
59 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
60 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
61 wish to add additional options which take arguments. Any redefinition
62 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
65 @findex WORD_SWITCH_TAKES_ARG
66 @item WORD_SWITCH_TAKES_ARG (@var{name})
67 A C expression which determines whether the option @samp{-@var{name}}
68 takes arguments. The value should be the number of arguments that
69 option takes--zero, for many options. This macro rather than
70 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
72 By default, this macro is defined as
73 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
74 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
75 wish to add additional options which take arguments. Any redefinition
76 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
79 @findex SWITCHES_NEED_SPACES
80 @item SWITCHES_NEED_SPACES
81 A string-valued C expression which enumerates the options for which
82 the linker needs a space between the option and its argument.
84 If this macro is not defined, the default value is @code{""}.
88 A C string constant that tells the GNU CC driver program options to
89 pass to CPP. It can also specify how to translate options you
90 give to GNU CC into options for GNU CC to pass to the CPP.
92 Do not define this macro if it does not need to do anything.
94 @findex NO_BUILTIN_SIZE_TYPE
95 @item NO_BUILTIN_SIZE_TYPE
96 If this macro is defined, the preprocessor will not define the builtin macro
97 @code{__SIZE_TYPE__}. The macro @code{__SIZE_TYPE__} must then be defined
98 by @code{CPP_SPEC} instead.
100 This should be defined if @code{SIZE_TYPE} depends on target dependent flags
101 which are not accessible to the preprocessor. Otherwise, it should not
104 @findex NO_BUILTIN_PTRDIFF_TYPE
105 @item NO_BUILTIN_PTRDIFF_TYPE
106 If this macro is defined, the preprocessor will not define the builtin macro
107 @code{__PTRDIFF_TYPE__}. The macro @code{__PTRDIFF_TYPE__} must then be
108 defined by @code{CPP_SPEC} instead.
110 This should be defined if @code{PTRDIFF_TYPE} depends on target dependent flags
111 which are not accessible to the preprocessor. Otherwise, it should not
114 @findex SIGNED_CHAR_SPEC
115 @item SIGNED_CHAR_SPEC
116 A C string constant that tells the GNU CC driver program options to
117 pass to CPP. By default, this macro is defined to pass the option
118 @samp{-D__CHAR_UNSIGNED__} to CPP if @code{char} will be treated as
119 @code{unsigned char} by @code{cc1}.
121 Do not define this macro unless you need to override the default
126 A C string constant that tells the GNU CC driver program options to
127 pass to @code{cc1}. It can also specify how to translate options you
128 give to GNU CC into options for GNU CC to pass to the @code{cc1}.
130 Do not define this macro if it does not need to do anything.
134 A C string constant that tells the GNU CC driver program options to
135 pass to @code{cc1plus}. It can also specify how to translate options you
136 give to GNU CC into options for GNU CC to pass to the @code{cc1plus}.
138 Do not define this macro if it does not need to do anything.
142 A C string constant that tells the GNU CC driver program options to
143 pass to the assembler. It can also specify how to translate options
144 you give to GNU CC into options for GNU CC to pass to the assembler.
145 See the file @file{sun3.h} for an example of this.
147 Do not define this macro if it does not need to do anything.
149 @findex ASM_FINAL_SPEC
151 A C string constant that tells the GNU CC driver program how to
152 run any programs which cleanup after the normal assembler.
153 Normally, this is not needed. See the file @file{mips.h} for
156 Do not define this macro if it does not need to do anything.
160 A C string constant that tells the GNU CC driver program options to
161 pass to the linker. It can also specify how to translate options you
162 give to GNU CC into options for GNU CC to pass to the linker.
164 Do not define this macro if it does not need to do anything.
168 Another C string constant used much like @code{LINK_SPEC}. The difference
169 between the two is that @code{LIB_SPEC} is used at the end of the
170 command given to the linker.
172 If this macro is not defined, a default is provided that
173 loads the standard C library from the usual place. See @file{gcc.c}.
177 Another C string constant that tells the GNU CC driver program
178 how and when to place a reference to @file{libgcc.a} into the
179 linker command line. This constant is placed both before and after
180 the value of @code{LIB_SPEC}.
182 If this macro is not defined, the GNU CC driver provides a default that
183 passes the string @samp{-lgcc} to the linker unless the @samp{-shared}
186 @findex STARTFILE_SPEC
188 Another C string constant used much like @code{LINK_SPEC}. The
189 difference between the two is that @code{STARTFILE_SPEC} is used at
190 the very beginning of the command given to the linker.
192 If this macro is not defined, a default is provided that loads the
193 standard C startup file from the usual place. See @file{gcc.c}.
197 Another C string constant used much like @code{LINK_SPEC}. The
198 difference between the two is that @code{ENDFILE_SPEC} is used at
199 the very end of the command given to the linker.
201 Do not define this macro if it does not need to do anything.
205 Define this macro to provide additional specifications to put in the
206 @file{specs} file that can be used in various specifications like
209 The definition should be an initializer for an array of structures,
210 containing a string constant, that defines the specification name, and a
211 string constant that provides the specification.
213 Do not define this macro if it does not need to do anything.
215 @code{EXTRA_SPECS} is useful when an architecture contains several
216 related targets, which have various @code{..._SPECS} which are similar
217 to each other, and the maintainer would like one central place to keep
220 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
221 define either @code{_CALL_SYSV} when the System V calling sequence is
222 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
225 The @file{config/rs6000/rs6000.h} target file defines:
228 #define EXTRA_SPECS \
229 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
231 #define CPP_SYS_DEFAULT ""
234 The @file{config/rs6000/sysv.h} target file defines:
238 "%@{posix: -D_POSIX_SOURCE @} \
239 %@{mcall-sysv: -D_CALL_SYSV @} %@{mcall-aix: -D_CALL_AIX @} \
240 %@{!mcall-sysv: %@{!mcall-aix: %(cpp_sysv_default) @}@} \
241 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
243 #undef CPP_SYSV_DEFAULT
244 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
247 while the @file{config/rs6000/eabiaix.h} target file defines
248 @code{CPP_SYSV_DEFAULT} as:
251 #undef CPP_SYSV_DEFAULT
252 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
255 @findex LINK_LIBGCC_SPECIAL
256 @item LINK_LIBGCC_SPECIAL
257 Define this macro if the driver program should find the library
258 @file{libgcc.a} itself and should not pass @samp{-L} options to the
259 linker. If you do not define this macro, the driver program will pass
260 the argument @samp{-lgcc} to tell the linker to do the search and will
261 pass @samp{-L} options to it.
263 @findex LINK_LIBGCC_SPECIAL_1
264 @item LINK_LIBGCC_SPECIAL_1
265 Define this macro if the driver program should find the library
266 @file{libgcc.a}. If you do not define this macro, the driver program will pass
267 the argument @samp{-lgcc} to tell the linker to do the search.
268 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
269 not affect @samp{-L} options.
271 @findex MULTILIB_DEFAULTS
272 @item MULTILIB_DEFAULTS
273 Define this macro as a C expression for the initializer of an array of
274 string to tell the driver program which options are defaults for this
275 target and thus do not need to be handled specially when using
276 @code{MULTILIB_OPTIONS}.
278 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
279 the target makefile fragment or if none of the options listed in
280 @code{MULTILIB_OPTIONS} are set by default.
281 @xref{Target Fragment}.
283 @findex RELATIVE_PREFIX_NOT_LINKDIR
284 @item RELATIVE_PREFIX_NOT_LINKDIR
285 Define this macro to tell @code{gcc} that it should only translate
286 a @samp{-B} prefix into a @samp{-L} linker option if the prefix
287 indicates an absolute file name.
289 @findex STANDARD_EXEC_PREFIX
290 @item STANDARD_EXEC_PREFIX
291 Define this macro as a C string constant if you wish to override the
292 standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
293 try when searching for the executable files of the compiler.
295 @findex MD_EXEC_PREFIX
297 If defined, this macro is an additional prefix to try after
298 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
299 when the @samp{-b} option is used, or the compiler is built as a cross
302 @findex STANDARD_STARTFILE_PREFIX
303 @item STANDARD_STARTFILE_PREFIX
304 Define this macro as a C string constant if you wish to override the
305 standard choice of @file{/usr/local/lib/} as the default prefix to
306 try when searching for startup files such as @file{crt0.o}.
308 @findex MD_STARTFILE_PREFIX
309 @item MD_STARTFILE_PREFIX
310 If defined, this macro supplies an additional prefix to try after the
311 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
312 @samp{-b} option is used, or when the compiler is built as a cross
315 @findex MD_STARTFILE_PREFIX_1
316 @item MD_STARTFILE_PREFIX_1
317 If defined, this macro supplies yet another prefix to try after the
318 standard prefixes. It is not searched when the @samp{-b} option is
319 used, or when the compiler is built as a cross compiler.
321 @findex INIT_ENVIRONMENT
322 @item INIT_ENVIRONMENT
323 Define this macro as a C string constant if you wish to set environment
324 variables for programs called by the driver, such as the assembler and
325 loader. The driver passes the value of this macro to @code{putenv} to
326 initialize the necessary environment variables.
328 @findex LOCAL_INCLUDE_DIR
329 @item LOCAL_INCLUDE_DIR
330 Define this macro as a C string constant if you wish to override the
331 standard choice of @file{/usr/local/include} as the default prefix to
332 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
333 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
335 Cross compilers do not use this macro and do not search either
336 @file{/usr/local/include} or its replacement.
338 @findex SYSTEM_INCLUDE_DIR
339 @item SYSTEM_INCLUDE_DIR
340 Define this macro as a C string constant if you wish to specify a
341 system-specific directory to search for header files before the standard
342 directory. @code{SYSTEM_INCLUDE_DIR} comes before
343 @code{STANDARD_INCLUDE_DIR} in the search order.
345 Cross compilers do not use this macro and do not search the directory
348 @findex STANDARD_INCLUDE_DIR
349 @item STANDARD_INCLUDE_DIR
350 Define this macro as a C string constant if you wish to override the
351 standard choice of @file{/usr/include} as the default prefix to
352 try when searching for header files.
354 Cross compilers do not use this macro and do not search either
355 @file{/usr/include} or its replacement.
357 @findex STANDARD_INCLUDE_COMPONENT
358 @item STANDARD_INCLUDE_COMPONENT
359 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
360 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
361 If you do not define this macro, no component is used.
363 @findex INCLUDE_DEFAULTS
364 @item INCLUDE_DEFAULTS
365 Define this macro if you wish to override the entire default search path
366 for include files. For a native compiler, the default search path
367 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
368 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
369 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
370 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
371 and specify private search areas for GCC. The directory
372 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
374 The definition should be an initializer for an array of structures.
375 Each array element should have four elements: the directory name (a
376 string constant), the component name, and flag for C++-only directories,
377 and a flag showing that the includes in the directory don't need to be
378 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
379 the array with a null element.
381 The component name denotes what GNU package the include file is part of,
382 if any, in all upper-case letters. For example, it might be @samp{GCC}
383 or @samp{BINUTILS}. If the package is part of the a vendor-supplied
384 operating system, code the component name as @samp{0}.
387 For example, here is the definition used for VAX/VMS:
390 #define INCLUDE_DEFAULTS \
392 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
393 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
394 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
401 Here is the order of prefixes tried for exec files:
405 Any prefixes specified by the user with @samp{-B}.
408 The environment variable @code{GCC_EXEC_PREFIX}, if any.
411 The directories specified by the environment variable @code{COMPILER_PATH}.
414 The macro @code{STANDARD_EXEC_PREFIX}.
417 @file{/usr/lib/gcc/}.
420 The macro @code{MD_EXEC_PREFIX}, if any.
423 Here is the order of prefixes tried for startfiles:
427 Any prefixes specified by the user with @samp{-B}.
430 The environment variable @code{GCC_EXEC_PREFIX}, if any.
433 The directories specified by the environment variable @code{LIBRARY_PATH}
434 (native only, cross compilers do not use this).
437 The macro @code{STANDARD_EXEC_PREFIX}.
440 @file{/usr/lib/gcc/}.
443 The macro @code{MD_EXEC_PREFIX}, if any.
446 The macro @code{MD_STARTFILE_PREFIX}, if any.
449 The macro @code{STANDARD_STARTFILE_PREFIX}.
458 @node Run-time Target
459 @section Run-time Target Specification
460 @cindex run-time target specification
461 @cindex predefined macros
462 @cindex target specifications
464 @c prevent bad page break with this line
465 Here are run-time target specifications.
468 @findex CPP_PREDEFINES
470 Define this to be a string constant containing @samp{-D} options to
471 define the predefined macros that identify this machine and system.
472 These macros will be predefined unless the @samp{-ansi} option is
475 In addition, a parallel set of macros are predefined, whose names are
476 made by appending @samp{__} at the beginning and at the end. These
477 @samp{__} macros are permitted by the ANSI standard, so they are
478 predefined regardless of whether @samp{-ansi} is specified.
480 For example, on the Sun, one can use the following value:
483 "-Dmc68000 -Dsun -Dunix"
486 The result is to define the macros @code{__mc68000__}, @code{__sun__}
487 and @code{__unix__} unconditionally, and the macros @code{mc68000},
488 @code{sun} and @code{unix} provided @samp{-ansi} is not specified.
490 @findex extern int target_flags
491 @item extern int target_flags;
492 This declaration should be present.
494 @cindex optional hardware or system features
495 @cindex features, optional, in system conventions
497 This series of macros is to allow compiler command arguments to
498 enable or disable the use of optional features of the target machine.
499 For example, one machine description serves both the 68000 and
500 the 68020; a command argument tells the compiler whether it should
501 use 68020-only instructions or not. This command argument works
502 by means of a macro @code{TARGET_68020} that tests a bit in
505 Define a macro @code{TARGET_@var{featurename}} for each such option.
506 Its definition should test a bit in @code{target_flags}; for example:
509 #define TARGET_68020 (target_flags & 1)
512 One place where these macros are used is in the condition-expressions
513 of instruction patterns. Note how @code{TARGET_68020} appears
514 frequently in the 68000 machine description file, @file{m68k.md}.
515 Another place they are used is in the definitions of the other
516 macros in the @file{@var{machine}.h} file.
518 @findex TARGET_SWITCHES
519 @item TARGET_SWITCHES
520 This macro defines names of command options to set and clear
521 bits in @code{target_flags}. Its definition is an initializer
522 with a subgrouping for each command option.
524 Each subgrouping contains a string constant, that defines the option
525 name, and a number, which contains the bits to set in
526 @code{target_flags}. A negative number says to clear bits instead;
527 the negative of the number is which bits to clear. The actual option
528 name is made by appending @samp{-m} to the specified name.
530 One of the subgroupings should have a null string. The number in
531 this grouping is the default value for @code{target_flags}. Any
532 target options act starting with that value.
534 Here is an example which defines @samp{-m68000} and @samp{-m68020}
535 with opposite meanings, and picks the latter as the default:
538 #define TARGET_SWITCHES \
539 @{ @{ "68020", 1@}, \
544 @findex TARGET_OPTIONS
546 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
547 options that have values. Its definition is an initializer with a
548 subgrouping for each command option.
550 Each subgrouping contains a string constant, that defines the fixed part
551 of the option name, and the address of a variable. The variable, type
552 @code{char *}, is set to the variable part of the given option if the fixed
553 part matches. The actual option name is made by appending @samp{-m} to the
556 Here is an example which defines @samp{-mshort-data-@var{number}}. If the
557 given option is @samp{-mshort-data-512}, the variable @code{m88k_short_data}
558 will be set to the string @code{"512"}.
561 extern char *m88k_short_data;
562 #define TARGET_OPTIONS \
563 @{ @{ "short-data-", &m88k_short_data @} @}
566 @findex TARGET_VERSION
568 This macro is a C statement to print on @code{stderr} a string
569 describing the particular machine description choice. Every machine
570 description should define @code{TARGET_VERSION}. For example:
574 #define TARGET_VERSION \
575 fprintf (stderr, " (68k, Motorola syntax)");
577 #define TARGET_VERSION \
578 fprintf (stderr, " (68k, MIT syntax)");
582 @findex OVERRIDE_OPTIONS
583 @item OVERRIDE_OPTIONS
584 Sometimes certain combinations of command options do not make sense on
585 a particular target machine. You can define a macro
586 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
587 defined, is executed once just after all the command options have been
590 Don't use this macro to turn on various extra optimizations for
591 @samp{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
593 @findex OPTIMIZATION_OPTIONS
594 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
595 Some machines may desire to change what optimizations are performed for
596 various optimization levels. This macro, if defined, is executed once
597 just after the optimization level is determined and before the remainder
598 of the command options have been parsed. Values set in this macro are
599 used as the default values for the other command line options.
601 @var{level} is the optimization level specified; 2 if @samp{-O2} is
602 specified, 1 if @samp{-O} is specified, and 0 if neither is specified.
604 @var{size} is non-zero if @samp{-Os} is specified and zero otherwise.
606 You should not use this macro to change options that are not
607 machine-specific. These should uniformly selected by the same
608 optimization level on all supported machines. Use this macro to enable
609 machine-specific optimizations.
611 @strong{Do not examine @code{write_symbols} in
612 this macro!} The debugging options are not supposed to alter the
615 @findex CAN_DEBUG_WITHOUT_FP
616 @item CAN_DEBUG_WITHOUT_FP
617 Define this macro if debugging can be performed even without a frame
618 pointer. If this macro is defined, GNU CC will turn on the
619 @samp{-fomit-frame-pointer} option whenever @samp{-O} is specified.
623 @section Storage Layout
624 @cindex storage layout
626 Note that the definitions of the macros in this table which are sizes or
627 alignments measured in bits do not need to be constant. They can be C
628 expressions that refer to static variables, such as the @code{target_flags}.
629 @xref{Run-time Target}.
632 @findex BITS_BIG_ENDIAN
633 @item BITS_BIG_ENDIAN
634 Define this macro to have the value 1 if the most significant bit in a
635 byte has the lowest number; otherwise define it to have the value zero.
636 This means that bit-field instructions count from the most significant
637 bit. If the machine has no bit-field instructions, then this must still
638 be defined, but it doesn't matter which value it is defined to. This
639 macro need not be a constant.
641 This macro does not affect the way structure fields are packed into
642 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
644 @findex BYTES_BIG_ENDIAN
645 @item BYTES_BIG_ENDIAN
646 Define this macro to have the value 1 if the most significant byte in a
647 word has the lowest number. This macro need not be a constant.
649 @findex WORDS_BIG_ENDIAN
650 @item WORDS_BIG_ENDIAN
651 Define this macro to have the value 1 if, in a multiword object, the
652 most significant word has the lowest number. This applies to both
653 memory locations and registers; GNU CC fundamentally assumes that the
654 order of words in memory is the same as the order in registers. This
655 macro need not be a constant.
657 @findex LIBGCC2_WORDS_BIG_ENDIAN
658 @item LIBGCC2_WORDS_BIG_ENDIAN
659 Define this macro if WORDS_BIG_ENDIAN is not constant. This must be a
660 constant value with the same meaning as WORDS_BIG_ENDIAN, which will be
661 used only when compiling libgcc2.c. Typically the value will be set
662 based on preprocessor defines.
664 @findex FLOAT_WORDS_BIG_ENDIAN
665 @item FLOAT_WORDS_BIG_ENDIAN
666 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
667 @code{TFmode} floating point numbers are stored in memory with the word
668 containing the sign bit at the lowest address; otherwise define it to
669 have the value 0. This macro need not be a constant.
671 You need not define this macro if the ordering is the same as for
674 @findex BITS_PER_UNIT
676 Define this macro to be the number of bits in an addressable storage
677 unit (byte); normally 8.
679 @findex BITS_PER_WORD
681 Number of bits in a word; normally 32.
683 @findex MAX_BITS_PER_WORD
684 @item MAX_BITS_PER_WORD
685 Maximum number of bits in a word. If this is undefined, the default is
686 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
687 largest value that @code{BITS_PER_WORD} can have at run-time.
689 @findex UNITS_PER_WORD
691 Number of storage units in a word; normally 4.
693 @findex MIN_UNITS_PER_WORD
694 @item MIN_UNITS_PER_WORD
695 Minimum number of units in a word. If this is undefined, the default is
696 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
697 smallest value that @code{UNITS_PER_WORD} can have at run-time.
701 Width of a pointer, in bits. You must specify a value no wider than the
702 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
703 you must define @code{POINTERS_EXTEND_UNSIGNED}.
705 @findex POINTERS_EXTEND_UNSIGNED
706 @item POINTERS_EXTEND_UNSIGNED
707 A C expression whose value is nonzero if pointers that need to be
708 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode}
709 are sign-extended and zero if they are zero-extended.
711 You need not define this macro if the @code{POINTER_SIZE} is equal
712 to the width of @code{Pmode}.
715 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
716 A macro to update @var{m} and @var{unsignedp} when an object whose type
717 is @var{type} and which has the specified mode and signedness is to be
718 stored in a register. This macro is only called when @var{type} is a
721 On most RISC machines, which only have operations that operate on a full
722 register, define this macro to set @var{m} to @code{word_mode} if
723 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
724 cases, only integer modes should be widened because wider-precision
725 floating-point operations are usually more expensive than their narrower
728 For most machines, the macro definition does not change @var{unsignedp}.
729 However, some machines, have instructions that preferentially handle
730 either signed or unsigned quantities of certain modes. For example, on
731 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
732 sign-extend the result to 64 bits. On such machines, set
733 @var{unsignedp} according to which kind of extension is more efficient.
735 Do not define this macro if it would never modify @var{m}.
737 @findex PROMOTE_FUNCTION_ARGS
738 @item PROMOTE_FUNCTION_ARGS
739 Define this macro if the promotion described by @code{PROMOTE_MODE}
740 should also be done for outgoing function arguments.
742 @findex PROMOTE_FUNCTION_RETURN
743 @item PROMOTE_FUNCTION_RETURN
744 Define this macro if the promotion described by @code{PROMOTE_MODE}
745 should also be done for the return value of functions.
747 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
748 promotions done by @code{PROMOTE_MODE}.
750 @findex PROMOTE_FOR_CALL_ONLY
751 @item PROMOTE_FOR_CALL_ONLY
752 Define this macro if the promotion described by @code{PROMOTE_MODE}
753 should @emph{only} be performed for outgoing function arguments or
754 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
755 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
757 @findex PARM_BOUNDARY
759 Normal alignment required for function parameters on the stack, in
760 bits. All stack parameters receive at least this much alignment
761 regardless of data type. On most machines, this is the same as the
764 @findex STACK_BOUNDARY
766 Define this macro if you wish to preserve a certain alignment for
767 the stack pointer. The definition is a C expression
768 for the desired alignment (measured in bits).
770 @cindex @code{PUSH_ROUNDING}, interaction with @code{STACK_BOUNDARY}
771 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
772 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies a
773 less strict alignment than @code{STACK_BOUNDARY}, the stack may be
774 momentarily unaligned while pushing arguments.
776 @findex FUNCTION_BOUNDARY
777 @item FUNCTION_BOUNDARY
778 Alignment required for a function entry point, in bits.
780 @findex BIGGEST_ALIGNMENT
781 @item BIGGEST_ALIGNMENT
782 Biggest alignment that any data type can require on this machine, in bits.
784 @findex MINIMUM_ATOMIC_ALIGNMENT
785 @item MINIMUM_ATOMIC_ALIGNMENT
786 If defined, the smallest alignment, in bits, that can be given to an
787 object that can be referenced in one operation, without disturbing any
788 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
789 on machines that don't have byte or half-word store operations.
791 @findex BIGGEST_FIELD_ALIGNMENT
792 @item BIGGEST_FIELD_ALIGNMENT
793 Biggest alignment that any structure field can require on this machine,
794 in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
795 structure fields only.
797 @findex ADJUST_FIELD_ALIGN
798 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
799 An expression for the alignment of a structure field @var{field} if the
800 alignment computed in the usual way is @var{computed}. GNU CC uses
801 this value instead of the value in @code{BIGGEST_ALIGNMENT} or
802 @code{BIGGEST_FIELD_ALIGNMENT}, if defined, for structure fields only.
804 @findex MAX_OFILE_ALIGNMENT
805 @item MAX_OFILE_ALIGNMENT
806 Biggest alignment supported by the object file format of this machine.
807 Use this macro to limit the alignment which can be specified using the
808 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
809 the default value is @code{BIGGEST_ALIGNMENT}.
811 @findex DATA_ALIGNMENT
812 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
813 If defined, a C expression to compute the alignment for a static
814 variable. @var{type} is the data type, and @var{basic-align} is the
815 alignment that the object would ordinarily have. The value of this
816 macro is used instead of that alignment to align the object.
818 If this macro is not defined, then @var{basic-align} is used.
821 One use of this macro is to increase alignment of medium-size data to
822 make it all fit in fewer cache lines. Another is to cause character
823 arrays to be word-aligned so that @code{strcpy} calls that copy
824 constants to character arrays can be done inline.
826 @findex CONSTANT_ALIGNMENT
827 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
828 If defined, a C expression to compute the alignment given to a constant
829 that is being placed in memory. @var{constant} is the constant and
830 @var{basic-align} is the alignment that the object would ordinarily
831 have. The value of this macro is used instead of that alignment to
834 If this macro is not defined, then @var{basic-align} is used.
836 The typical use of this macro is to increase alignment for string
837 constants to be word aligned so that @code{strcpy} calls that copy
838 constants can be done inline.
840 @findex EMPTY_FIELD_BOUNDARY
841 @item EMPTY_FIELD_BOUNDARY
842 Alignment in bits to be given to a structure bit field that follows an
843 empty field such as @code{int : 0;}.
845 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
846 that results from an empty field.
848 @findex STRUCTURE_SIZE_BOUNDARY
849 @item STRUCTURE_SIZE_BOUNDARY
850 Number of bits which any structure or union's size must be a multiple of.
851 Each structure or union's size is rounded up to a multiple of this.
853 If you do not define this macro, the default is the same as
854 @code{BITS_PER_UNIT}.
856 @findex STRICT_ALIGNMENT
857 @item STRICT_ALIGNMENT
858 Define this macro to be the value 1 if instructions will fail to work
859 if given data not on the nominal alignment. If instructions will merely
860 go slower in that case, define this macro as 0.
862 @findex PCC_BITFIELD_TYPE_MATTERS
863 @item PCC_BITFIELD_TYPE_MATTERS
864 Define this if you wish to imitate the way many other C compilers handle
865 alignment of bitfields and the structures that contain them.
867 The behavior is that the type written for a bitfield (@code{int},
868 @code{short}, or other integer type) imposes an alignment for the
869 entire structure, as if the structure really did contain an ordinary
870 field of that type. In addition, the bitfield is placed within the
871 structure so that it would fit within such a field, not crossing a
874 Thus, on most machines, a bitfield whose type is written as @code{int}
875 would not cross a four-byte boundary, and would force four-byte
876 alignment for the whole structure. (The alignment used may not be four
877 bytes; it is controlled by the other alignment parameters.)
879 If the macro is defined, its definition should be a C expression;
880 a nonzero value for the expression enables this behavior.
882 Note that if this macro is not defined, or its value is zero, some
883 bitfields may cross more than one alignment boundary. The compiler can
884 support such references if there are @samp{insv}, @samp{extv}, and
885 @samp{extzv} insns that can directly reference memory.
887 The other known way of making bitfields work is to define
888 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
889 Then every structure can be accessed with fullwords.
891 Unless the machine has bitfield instructions or you define
892 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
893 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
895 If your aim is to make GNU CC use the same conventions for laying out
896 bitfields as are used by another compiler, here is how to investigate
897 what the other compiler does. Compile and run this program:
916 printf ("Size of foo1 is %d\n",
917 sizeof (struct foo1));
918 printf ("Size of foo2 is %d\n",
919 sizeof (struct foo2));
924 If this prints 2 and 5, then the compiler's behavior is what you would
925 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
927 @findex BITFIELD_NBYTES_LIMITED
928 @item BITFIELD_NBYTES_LIMITED
929 Like PCC_BITFIELD_TYPE_MATTERS except that its effect is limited to
930 aligning a bitfield within the structure.
932 @findex ROUND_TYPE_SIZE
933 @item ROUND_TYPE_SIZE (@var{struct}, @var{size}, @var{align})
934 Define this macro as an expression for the overall size of a structure
935 (given by @var{struct} as a tree node) when the size computed from the
936 fields is @var{size} and the alignment is @var{align}.
938 The default is to round @var{size} up to a multiple of @var{align}.
940 @findex ROUND_TYPE_ALIGN
941 @item ROUND_TYPE_ALIGN (@var{struct}, @var{computed}, @var{specified})
942 Define this macro as an expression for the alignment of a structure
943 (given by @var{struct} as a tree node) if the alignment computed in the
944 usual way is @var{computed} and the alignment explicitly specified was
947 The default is to use @var{specified} if it is larger; otherwise, use
948 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
950 @findex MAX_FIXED_MODE_SIZE
951 @item MAX_FIXED_MODE_SIZE
952 An integer expression for the size in bits of the largest integer
953 machine mode that should actually be used. All integer machine modes of
954 this size or smaller can be used for structures and unions with the
955 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
956 (DImode)} is assumed.
958 @findex CHECK_FLOAT_VALUE
959 @item CHECK_FLOAT_VALUE (@var{mode}, @var{value}, @var{overflow})
960 A C statement to validate the value @var{value} (of type
961 @code{double}) for mode @var{mode}. This means that you check whether
962 @var{value} fits within the possible range of values for mode
963 @var{mode} on this target machine. The mode @var{mode} is always
964 a mode of class @code{MODE_FLOAT}. @var{overflow} is nonzero if
965 the value is already known to be out of range.
967 If @var{value} is not valid or if @var{overflow} is nonzero, you should
968 set @var{overflow} to 1 and then assign some valid value to @var{value}.
969 Allowing an invalid value to go through the compiler can produce
970 incorrect assembler code which may even cause Unix assemblers to crash.
972 This macro need not be defined if there is no work for it to do.
974 @findex TARGET_FLOAT_FORMAT
975 @item TARGET_FLOAT_FORMAT
976 A code distinguishing the floating point format of the target machine.
977 There are three defined values:
980 @findex IEEE_FLOAT_FORMAT
981 @item IEEE_FLOAT_FORMAT
982 This code indicates IEEE floating point. It is the default; there is no
983 need to define this macro when the format is IEEE.
985 @findex VAX_FLOAT_FORMAT
986 @item VAX_FLOAT_FORMAT
987 This code indicates the peculiar format used on the Vax.
989 @findex UNKNOWN_FLOAT_FORMAT
990 @item UNKNOWN_FLOAT_FORMAT
991 This code indicates any other format.
994 The value of this macro is compared with @code{HOST_FLOAT_FORMAT}
995 (@pxref{Config}) to determine whether the target machine has the same
996 format as the host machine. If any other formats are actually in use on
997 supported machines, new codes should be defined for them.
999 The ordering of the component words of floating point values stored in
1000 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN} for the target
1001 machine and @code{HOST_FLOAT_WORDS_BIG_ENDIAN} for the host.
1003 @findex DEFAULT_VTABLE_THUNKS
1004 @item DEFAULT_VTABLE_THUNKS
1005 GNU CC supports two ways of implementing C++ vtables: traditional or with
1006 so-called ``thunks''. The flag @samp{-fvtable-thunk} chooses between them.
1007 Define this macro to be a C expression for the default value of that flag.
1008 If @code{DEFAULT_VTABLE_THUNKS} is 0, GNU CC uses the traditional
1009 implementation by default. The ``thunk'' implementation is more efficient
1010 (especially if you have provided an implementation of
1011 @code{ASM_OUTPUT_MI_THUNK}, see @ref{Function Entry}), but is not binary
1012 compatible with code compiled using the traditional implementation.
1013 If you are writing a new ports, define @code{DEFAULT_VTABLE_THUNKS} to 1.
1015 If you do not define this macro, the default for @samp{-fvtable-thunk} is 0.
1019 @section Layout of Source Language Data Types
1021 These macros define the sizes and other characteristics of the standard
1022 basic data types used in programs being compiled. Unlike the macros in
1023 the previous section, these apply to specific features of C and related
1024 languages, rather than to fundamental aspects of storage layout.
1027 @findex INT_TYPE_SIZE
1029 A C expression for the size in bits of the type @code{int} on the
1030 target machine. If you don't define this, the default is one word.
1032 @findex MAX_INT_TYPE_SIZE
1033 @item MAX_INT_TYPE_SIZE
1034 Maximum number for the size in bits of the type @code{int} on the target
1035 machine. If this is undefined, the default is @code{INT_TYPE_SIZE}.
1036 Otherwise, it is the constant value that is the largest value that
1037 @code{INT_TYPE_SIZE} can have at run-time. This is used in @code{cpp}.
1039 @findex SHORT_TYPE_SIZE
1040 @item SHORT_TYPE_SIZE
1041 A C expression for the size in bits of the type @code{short} on the
1042 target machine. If you don't define this, the default is half a word.
1043 (If this would be less than one storage unit, it is rounded up to one
1046 @findex LONG_TYPE_SIZE
1047 @item LONG_TYPE_SIZE
1048 A C expression for the size in bits of the type @code{long} on the
1049 target machine. If you don't define this, the default is one word.
1051 @findex MAX_LONG_TYPE_SIZE
1052 @item MAX_LONG_TYPE_SIZE
1053 Maximum number for the size in bits of the type @code{long} on the
1054 target machine. If this is undefined, the default is
1055 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1056 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1059 @findex LONG_LONG_TYPE_SIZE
1060 @item LONG_LONG_TYPE_SIZE
1061 A C expression for the size in bits of the type @code{long long} on the
1062 target machine. If you don't define this, the default is two
1063 words. If you want to support GNU Ada on your machine, the value of
1064 macro must be at least 64.
1066 @findex CHAR_TYPE_SIZE
1067 @item CHAR_TYPE_SIZE
1068 A C expression for the size in bits of the type @code{char} on the
1069 target machine. If you don't define this, the default is one quarter
1070 of a word. (If this would be less than one storage unit, it is rounded up
1073 @findex MAX_CHAR_TYPE_SIZE
1074 @item MAX_CHAR_TYPE_SIZE
1075 Maximum number for the size in bits of the type @code{char} on the
1076 target machine. If this is undefined, the default is
1077 @code{CHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1078 largest value that @code{CHAR_TYPE_SIZE} can have at run-time. This is
1081 @findex FLOAT_TYPE_SIZE
1082 @item FLOAT_TYPE_SIZE
1083 A C expression for the size in bits of the type @code{float} on the
1084 target machine. If you don't define this, the default is one word.
1086 @findex DOUBLE_TYPE_SIZE
1087 @item DOUBLE_TYPE_SIZE
1088 A C expression for the size in bits of the type @code{double} on the
1089 target machine. If you don't define this, the default is two
1092 @findex LONG_DOUBLE_TYPE_SIZE
1093 @item LONG_DOUBLE_TYPE_SIZE
1094 A C expression for the size in bits of the type @code{long double} on
1095 the target machine. If you don't define this, the default is two
1098 @findex WIDEST_HARDWARE_FP_SIZE
1099 @item WIDEST_HARDWARE_FP_SIZE
1100 A C expression for the size in bits of the widest floating-point format
1101 supported by the hardware. If you define this macro, you must specify a
1102 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1103 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1106 @findex DEFAULT_SIGNED_CHAR
1107 @item DEFAULT_SIGNED_CHAR
1108 An expression whose value is 1 or 0, according to whether the type
1109 @code{char} should be signed or unsigned by default. The user can
1110 always override this default with the options @samp{-fsigned-char}
1111 and @samp{-funsigned-char}.
1113 @findex DEFAULT_SHORT_ENUMS
1114 @item DEFAULT_SHORT_ENUMS
1115 A C expression to determine whether to give an @code{enum} type
1116 only as many bytes as it takes to represent the range of possible values
1117 of that type. A nonzero value means to do that; a zero value means all
1118 @code{enum} types should be allocated like @code{int}.
1120 If you don't define the macro, the default is 0.
1124 A C expression for a string describing the name of the data type to use
1125 for size values. The typedef name @code{size_t} is defined using the
1126 contents of the string.
1128 The string can contain more than one keyword. If so, separate them with
1129 spaces, and write first any length keyword, then @code{unsigned} if
1130 appropriate, and finally @code{int}. The string must exactly match one
1131 of the data type names defined in the function
1132 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1133 omit @code{int} or change the order---that would cause the compiler to
1136 If you don't define this macro, the default is @code{"long unsigned
1139 @findex PTRDIFF_TYPE
1141 A C expression for a string describing the name of the data type to use
1142 for the result of subtracting two pointers. The typedef name
1143 @code{ptrdiff_t} is defined using the contents of the string. See
1144 @code{SIZE_TYPE} above for more information.
1146 If you don't define this macro, the default is @code{"long int"}.
1150 A C expression for a string describing the name of the data type to use
1151 for wide characters. The typedef name @code{wchar_t} is defined using
1152 the contents of the string. See @code{SIZE_TYPE} above for more
1155 If you don't define this macro, the default is @code{"int"}.
1157 @findex WCHAR_TYPE_SIZE
1158 @item WCHAR_TYPE_SIZE
1159 A C expression for the size in bits of the data type for wide
1160 characters. This is used in @code{cpp}, which cannot make use of
1163 @findex MAX_WCHAR_TYPE_SIZE
1164 @item MAX_WCHAR_TYPE_SIZE
1165 Maximum number for the size in bits of the data type for wide
1166 characters. If this is undefined, the default is
1167 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1168 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1171 @findex OBJC_INT_SELECTORS
1172 @item OBJC_INT_SELECTORS
1173 Define this macro if the type of Objective C selectors should be
1176 If this macro is not defined, then selectors should have the type
1177 @code{struct objc_selector *}.
1179 @findex OBJC_SELECTORS_WITHOUT_LABELS
1180 @item OBJC_SELECTORS_WITHOUT_LABELS
1181 Define this macro if the compiler can group all the selectors together
1182 into a vector and use just one label at the beginning of the vector.
1183 Otherwise, the compiler must give each selector its own assembler
1186 On certain machines, it is important to have a separate label for each
1187 selector because this enables the linker to eliminate duplicate selectors.
1191 A C constant expression for the integer value for escape sequence
1196 @findex TARGET_NEWLINE
1199 @itemx TARGET_NEWLINE
1200 C constant expressions for the integer values for escape sequences
1201 @samp{\b}, @samp{\t} and @samp{\n}.
1209 C constant expressions for the integer values for escape sequences
1210 @samp{\v}, @samp{\f} and @samp{\r}.
1214 @section Register Usage
1215 @cindex register usage
1217 This section explains how to describe what registers the target machine
1218 has, and how (in general) they can be used.
1220 The description of which registers a specific instruction can use is
1221 done with register classes; see @ref{Register Classes}. For information
1222 on using registers to access a stack frame, see @ref{Frame Registers}.
1223 For passing values in registers, see @ref{Register Arguments}.
1224 For returning values in registers, see @ref{Scalar Return}.
1227 * Register Basics:: Number and kinds of registers.
1228 * Allocation Order:: Order in which registers are allocated.
1229 * Values in Registers:: What kinds of values each reg can hold.
1230 * Leaf Functions:: Renumbering registers for leaf functions.
1231 * Stack Registers:: Handling a register stack such as 80387.
1232 * Obsolete Register Macros:: Macros formerly used for the 80387.
1235 @node Register Basics
1236 @subsection Basic Characteristics of Registers
1238 @c prevent bad page break with this line
1239 Registers have various characteristics.
1242 @findex FIRST_PSEUDO_REGISTER
1243 @item FIRST_PSEUDO_REGISTER
1244 Number of hardware registers known to the compiler. They receive
1245 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1246 pseudo register's number really is assigned the number
1247 @code{FIRST_PSEUDO_REGISTER}.
1249 @item FIXED_REGISTERS
1250 @findex FIXED_REGISTERS
1251 @cindex fixed register
1252 An initializer that says which registers are used for fixed purposes
1253 all throughout the compiled code and are therefore not available for
1254 general allocation. These would include the stack pointer, the frame
1255 pointer (except on machines where that can be used as a general
1256 register when no frame pointer is needed), the program counter on
1257 machines where that is considered one of the addressable registers,
1258 and any other numbered register with a standard use.
1260 This information is expressed as a sequence of numbers, separated by
1261 commas and surrounded by braces. The @var{n}th number is 1 if
1262 register @var{n} is fixed, 0 otherwise.
1264 The table initialized from this macro, and the table initialized by
1265 the following one, may be overridden at run time either automatically,
1266 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1267 the user with the command options @samp{-ffixed-@var{reg}},
1268 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}}.
1270 @findex CALL_USED_REGISTERS
1271 @item CALL_USED_REGISTERS
1272 @cindex call-used register
1273 @cindex call-clobbered register
1274 @cindex call-saved register
1275 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1276 clobbered (in general) by function calls as well as for fixed
1277 registers. This macro therefore identifies the registers that are not
1278 available for general allocation of values that must live across
1281 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1282 automatically saves it on function entry and restores it on function
1283 exit, if the register is used within the function.
1285 @findex CONDITIONAL_REGISTER_USAGE
1287 @findex call_used_regs
1288 @item CONDITIONAL_REGISTER_USAGE
1289 Zero or more C statements that may conditionally modify two variables
1290 @code{fixed_regs} and @code{call_used_regs} (both of type @code{char
1291 []}) after they have been initialized from the two preceding macros.
1293 This is necessary in case the fixed or call-clobbered registers depend
1296 You need not define this macro if it has no work to do.
1298 @cindex disabling certain registers
1299 @cindex controlling register usage
1300 If the usage of an entire class of registers depends on the target
1301 flags, you may indicate this to GCC by using this macro to modify
1302 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1303 registers in the classes which should not be used by GCC. Also define
1304 the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1305 is called with a letter for a class that shouldn't be used.
1307 (However, if this class is not included in @code{GENERAL_REGS} and all
1308 of the insn patterns whose constraints permit this class are
1309 controlled by target switches, then GCC will automatically avoid using
1310 these registers when the target switches are opposed to them.)
1312 @findex NON_SAVING_SETJMP
1313 @item NON_SAVING_SETJMP
1314 If this macro is defined and has a nonzero value, it means that
1315 @code{setjmp} and related functions fail to save the registers, or that
1316 @code{longjmp} fails to restore them. To compensate, the compiler
1317 avoids putting variables in registers in functions that use
1320 @findex INCOMING_REGNO
1321 @item INCOMING_REGNO (@var{out})
1322 Define this macro if the target machine has register windows. This C
1323 expression returns the register number as seen by the called function
1324 corresponding to the register number @var{out} as seen by the calling
1325 function. Return @var{out} if register number @var{out} is not an
1328 @findex OUTGOING_REGNO
1329 @item OUTGOING_REGNO (@var{in})
1330 Define this macro if the target machine has register windows. This C
1331 expression returns the register number as seen by the calling function
1332 corresponding to the register number @var{in} as seen by the called
1333 function. Return @var{in} if register number @var{in} is not an inbound
1339 If the program counter has a register number, define this as that
1340 register number. Otherwise, do not define it.
1344 @node Allocation Order
1345 @subsection Order of Allocation of Registers
1346 @cindex order of register allocation
1347 @cindex register allocation order
1349 @c prevent bad page break with this line
1350 Registers are allocated in order.
1353 @findex REG_ALLOC_ORDER
1354 @item REG_ALLOC_ORDER
1355 If defined, an initializer for a vector of integers, containing the
1356 numbers of hard registers in the order in which GNU CC should prefer
1357 to use them (from most preferred to least).
1359 If this macro is not defined, registers are used lowest numbered first
1360 (all else being equal).
1362 One use of this macro is on machines where the highest numbered
1363 registers must always be saved and the save-multiple-registers
1364 instruction supports only sequences of consecutive registers. On such
1365 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1366 the highest numbered allocable register first.
1368 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1369 @item ORDER_REGS_FOR_LOCAL_ALLOC
1370 A C statement (sans semicolon) to choose the order in which to allocate
1371 hard registers for pseudo-registers local to a basic block.
1373 Store the desired register order in the array @code{reg_alloc_order}.
1374 Element 0 should be the register to allocate first; element 1, the next
1375 register; and so on.
1377 The macro body should not assume anything about the contents of
1378 @code{reg_alloc_order} before execution of the macro.
1380 On most machines, it is not necessary to define this macro.
1383 @node Values in Registers
1384 @subsection How Values Fit in Registers
1386 This section discusses the macros that describe which kinds of values
1387 (specifically, which machine modes) each register can hold, and how many
1388 consecutive registers are needed for a given mode.
1391 @findex HARD_REGNO_NREGS
1392 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1393 A C expression for the number of consecutive hard registers, starting
1394 at register number @var{regno}, required to hold a value of mode
1397 On a machine where all registers are exactly one word, a suitable
1398 definition of this macro is
1401 #define HARD_REGNO_NREGS(REGNO, MODE) \
1402 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1406 @findex HARD_REGNO_MODE_OK
1407 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1408 A C expression that is nonzero if it is permissible to store a value
1409 of mode @var{mode} in hard register number @var{regno} (or in several
1410 registers starting with that one). For a machine where all registers
1411 are equivalent, a suitable definition is
1414 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1417 You need not include code to check for the numbers of fixed registers,
1418 because the allocation mechanism considers them to be always occupied.
1420 @cindex register pairs
1421 On some machines, double-precision values must be kept in even/odd
1422 register pairs. You can implement that by defining this macro to reject
1423 odd register numbers for such modes.
1425 The minimum requirement for a mode to be OK in a register is that the
1426 @samp{mov@var{mode}} instruction pattern support moves between the
1427 register and other hard register in the same class and that moving a
1428 value into the register and back out not alter it.
1430 Since the same instruction used to move @code{word_mode} will work for
1431 all narrower integer modes, it is not necessary on any machine for
1432 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1433 you define patterns @samp{movhi}, etc., to take advantage of this. This
1434 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1435 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1438 Many machines have special registers for floating point arithmetic.
1439 Often people assume that floating point machine modes are allowed only
1440 in floating point registers. This is not true. Any registers that
1441 can hold integers can safely @emph{hold} a floating point machine
1442 mode, whether or not floating arithmetic can be done on it in those
1443 registers. Integer move instructions can be used to move the values.
1445 On some machines, though, the converse is true: fixed-point machine
1446 modes may not go in floating registers. This is true if the floating
1447 registers normalize any value stored in them, because storing a
1448 non-floating value there would garble it. In this case,
1449 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1450 floating registers. But if the floating registers do not automatically
1451 normalize, if you can store any bit pattern in one and retrieve it
1452 unchanged without a trap, then any machine mode may go in a floating
1453 register, so you can define this macro to say so.
1455 The primary significance of special floating registers is rather that
1456 they are the registers acceptable in floating point arithmetic
1457 instructions. However, this is of no concern to
1458 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1459 constraints for those instructions.
1461 On some machines, the floating registers are especially slow to access,
1462 so that it is better to store a value in a stack frame than in such a
1463 register if floating point arithmetic is not being done. As long as the
1464 floating registers are not in class @code{GENERAL_REGS}, they will not
1465 be used unless some pattern's constraint asks for one.
1467 @findex MODES_TIEABLE_P
1468 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
1469 A C expression that is nonzero if a value of mode
1470 @var{mode1} is accessible in mode @var{mode2} without copying.
1472 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
1473 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
1474 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
1475 should be nonzero. If they differ for any @var{r}, you should define
1476 this macro to return zero unless some other mechanism ensures the
1477 accessibility of the value in a narrower mode.
1479 You should define this macro to return nonzero in as many cases as
1480 possible since doing so will allow GNU CC to perform better register
1484 @node Leaf Functions
1485 @subsection Handling Leaf Functions
1487 @cindex leaf functions
1488 @cindex functions, leaf
1489 On some machines, a leaf function (i.e., one which makes no calls) can run
1490 more efficiently if it does not make its own register window. Often this
1491 means it is required to receive its arguments in the registers where they
1492 are passed by the caller, instead of the registers where they would
1495 The special treatment for leaf functions generally applies only when
1496 other conditions are met; for example, often they may use only those
1497 registers for its own variables and temporaries. We use the term ``leaf
1498 function'' to mean a function that is suitable for this special
1499 handling, so that functions with no calls are not necessarily ``leaf
1502 GNU CC assigns register numbers before it knows whether the function is
1503 suitable for leaf function treatment. So it needs to renumber the
1504 registers in order to output a leaf function. The following macros
1508 @findex LEAF_REGISTERS
1509 @item LEAF_REGISTERS
1510 A C initializer for a vector, indexed by hard register number, which
1511 contains 1 for a register that is allowable in a candidate for leaf
1514 If leaf function treatment involves renumbering the registers, then the
1515 registers marked here should be the ones before renumbering---those that
1516 GNU CC would ordinarily allocate. The registers which will actually be
1517 used in the assembler code, after renumbering, should not be marked with 1
1520 Define this macro only if the target machine offers a way to optimize
1521 the treatment of leaf functions.
1523 @findex LEAF_REG_REMAP
1524 @item LEAF_REG_REMAP (@var{regno})
1525 A C expression whose value is the register number to which @var{regno}
1526 should be renumbered, when a function is treated as a leaf function.
1528 If @var{regno} is a register number which should not appear in a leaf
1529 function before renumbering, then the expression should yield -1, which
1530 will cause the compiler to abort.
1532 Define this macro only if the target machine offers a way to optimize the
1533 treatment of leaf functions, and registers need to be renumbered to do
1537 @findex leaf_function
1538 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
1539 treat leaf functions specially. It can test the C variable
1540 @code{leaf_function} which is nonzero for leaf functions. (The variable
1541 @code{leaf_function} is defined only if @code{LEAF_REGISTERS} is
1543 @c changed this to fix overfull. ALSO: why the "it" at the beginning
1544 @c of the next paragraph?! --mew 2feb93
1546 @node Stack Registers
1547 @subsection Registers That Form a Stack
1549 There are special features to handle computers where some of the
1550 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
1551 Stack registers are normally written by pushing onto the stack, and are
1552 numbered relative to the top of the stack.
1554 Currently, GNU CC can only handle one group of stack-like registers, and
1555 they must be consecutively numbered.
1560 Define this if the machine has any stack-like registers.
1562 @findex FIRST_STACK_REG
1563 @item FIRST_STACK_REG
1564 The number of the first stack-like register. This one is the top
1567 @findex LAST_STACK_REG
1568 @item LAST_STACK_REG
1569 The number of the last stack-like register. This one is the bottom of
1573 @node Obsolete Register Macros
1574 @subsection Obsolete Macros for Controlling Register Usage
1576 These features do not work very well. They exist because they used to
1577 be required to generate correct code for the 80387 coprocessor of the
1578 80386. They are no longer used by that machine description and may be
1579 removed in a later version of the compiler. Don't use them!
1582 @findex OVERLAPPING_REGNO_P
1583 @item OVERLAPPING_REGNO_P (@var{regno})
1584 If defined, this is a C expression whose value is nonzero if hard
1585 register number @var{regno} is an overlapping register. This means a
1586 hard register which overlaps a hard register with a different number.
1587 (Such overlap is undesirable, but occasionally it allows a machine to
1588 be supported which otherwise could not be.) This macro must return
1589 nonzero for @emph{all} the registers which overlap each other. GNU CC
1590 can use an overlapping register only in certain limited ways. It can
1591 be used for allocation within a basic block, and may be spilled for
1592 reloading; that is all.
1594 If this macro is not defined, it means that none of the hard registers
1595 overlap each other. This is the usual situation.
1597 @findex INSN_CLOBBERS_REGNO_P
1598 @item INSN_CLOBBERS_REGNO_P (@var{insn}, @var{regno})
1599 If defined, this is a C expression whose value should be nonzero if
1600 the insn @var{insn} has the effect of mysteriously clobbering the
1601 contents of hard register number @var{regno}. By ``mysterious'' we
1602 mean that the insn's RTL expression doesn't describe such an effect.
1604 If this macro is not defined, it means that no insn clobbers registers
1605 mysteriously. This is the usual situation; all else being equal,
1606 it is best for the RTL expression to show all the activity.
1609 @findex PRESERVE_DEATH_INFO_REGNO_P
1610 @item PRESERVE_DEATH_INFO_REGNO_P (@var{regno})
1611 If defined, this is a C expression whose value is nonzero if correct
1612 @code{REG_DEAD} notes are needed for hard register number @var{regno}
1615 You would arrange to preserve death info for a register when some of the
1616 code in the machine description which is executed to write the assembler
1617 code looks at the death notes. This is necessary only when the actual
1618 hardware feature which GNU CC thinks of as a register is not actually a
1619 register of the usual sort. (It might, for example, be a hardware
1622 It is also useful for peepholes and linker relaxation.
1624 If this macro is not defined, it means that no death notes need to be
1625 preserved, and some may even be incorrect. This is the usual situation.
1628 @node Register Classes
1629 @section Register Classes
1630 @cindex register class definitions
1631 @cindex class definitions, register
1633 On many machines, the numbered registers are not all equivalent.
1634 For example, certain registers may not be allowed for indexed addressing;
1635 certain registers may not be allowed in some instructions. These machine
1636 restrictions are described to the compiler using @dfn{register classes}.
1638 You define a number of register classes, giving each one a name and saying
1639 which of the registers belong to it. Then you can specify register classes
1640 that are allowed as operands to particular instruction patterns.
1644 In general, each register will belong to several classes. In fact, one
1645 class must be named @code{ALL_REGS} and contain all the registers. Another
1646 class must be named @code{NO_REGS} and contain no registers. Often the
1647 union of two classes will be another class; however, this is not required.
1649 @findex GENERAL_REGS
1650 One of the classes must be named @code{GENERAL_REGS}. There is nothing
1651 terribly special about the name, but the operand constraint letters
1652 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
1653 the same as @code{ALL_REGS}, just define it as a macro which expands
1656 Order the classes so that if class @var{x} is contained in class @var{y}
1657 then @var{x} has a lower class number than @var{y}.
1659 The way classes other than @code{GENERAL_REGS} are specified in operand
1660 constraints is through machine-dependent operand constraint letters.
1661 You can define such letters to correspond to various classes, then use
1662 them in operand constraints.
1664 You should define a class for the union of two classes whenever some
1665 instruction allows both classes. For example, if an instruction allows
1666 either a floating point (coprocessor) register or a general register for a
1667 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
1668 which includes both of them. Otherwise you will get suboptimal code.
1670 You must also specify certain redundant information about the register
1671 classes: for each class, which classes contain it and which ones are
1672 contained in it; for each pair of classes, the largest class contained
1675 When a value occupying several consecutive registers is expected in a
1676 certain class, all the registers used must belong to that class.
1677 Therefore, register classes cannot be used to enforce a requirement for
1678 a register pair to start with an even-numbered register. The way to
1679 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
1681 Register classes used for input-operands of bitwise-and or shift
1682 instructions have a special requirement: each such class must have, for
1683 each fixed-point machine mode, a subclass whose registers can transfer that
1684 mode to or from memory. For example, on some machines, the operations for
1685 single-byte values (@code{QImode}) are limited to certain registers. When
1686 this is so, each register class that is used in a bitwise-and or shift
1687 instruction must have a subclass consisting of registers from which
1688 single-byte values can be loaded or stored. This is so that
1689 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
1692 @findex enum reg_class
1693 @item enum reg_class
1694 An enumeral type that must be defined with all the register class names
1695 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
1696 must be the last register class, followed by one more enumeral value,
1697 @code{LIM_REG_CLASSES}, which is not a register class but rather
1698 tells how many classes there are.
1700 Each register class has a number, which is the value of casting
1701 the class name to type @code{int}. The number serves as an index
1702 in many of the tables described below.
1704 @findex N_REG_CLASSES
1706 The number of distinct register classes, defined as follows:
1709 #define N_REG_CLASSES (int) LIM_REG_CLASSES
1712 @findex REG_CLASS_NAMES
1713 @item REG_CLASS_NAMES
1714 An initializer containing the names of the register classes as C string
1715 constants. These names are used in writing some of the debugging dumps.
1717 @findex REG_CLASS_CONTENTS
1718 @item REG_CLASS_CONTENTS
1719 An initializer containing the contents of the register classes, as integers
1720 which are bit masks. The @var{n}th integer specifies the contents of class
1721 @var{n}. The way the integer @var{mask} is interpreted is that
1722 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
1724 When the machine has more than 32 registers, an integer does not suffice.
1725 Then the integers are replaced by sub-initializers, braced groupings containing
1726 several integers. Each sub-initializer must be suitable as an initializer
1727 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
1729 @findex REGNO_REG_CLASS
1730 @item REGNO_REG_CLASS (@var{regno})
1731 A C expression whose value is a register class containing hard register
1732 @var{regno}. In general there is more than one such class; choose a class
1733 which is @dfn{minimal}, meaning that no smaller class also contains the
1736 @findex BASE_REG_CLASS
1737 @item BASE_REG_CLASS
1738 A macro whose definition is the name of the class to which a valid
1739 base register must belong. A base register is one used in an address
1740 which is the register value plus a displacement.
1742 @findex INDEX_REG_CLASS
1743 @item INDEX_REG_CLASS
1744 A macro whose definition is the name of the class to which a valid
1745 index register must belong. An index register is one used in an
1746 address where its value is either multiplied by a scale factor or
1747 added to another register (as well as added to a displacement).
1749 @findex REG_CLASS_FROM_LETTER
1750 @item REG_CLASS_FROM_LETTER (@var{char})
1751 A C expression which defines the machine-dependent operand constraint
1752 letters for register classes. If @var{char} is such a letter, the
1753 value should be the register class corresponding to it. Otherwise,
1754 the value should be @code{NO_REGS}. The register letter @samp{r},
1755 corresponding to class @code{GENERAL_REGS}, will not be passed
1756 to this macro; you do not need to handle it.
1758 @findex REGNO_OK_FOR_BASE_P
1759 @item REGNO_OK_FOR_BASE_P (@var{num})
1760 A C expression which is nonzero if register number @var{num} is
1761 suitable for use as a base register in operand addresses. It may be
1762 either a suitable hard register or a pseudo register that has been
1763 allocated such a hard register.
1765 @findex REGNO_MODE_OK_FOR_BASE_P
1766 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
1767 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
1768 that expression may examine the mode of the memory reference in
1769 @var{mode}. You should define this macro if the mode of the memory
1770 reference affects whether a register may be used as a base register. If
1771 you define this macro, the compiler will use it instead of
1772 @code{REGNO_OK_FOR_BASE_P}.
1774 @findex REGNO_OK_FOR_INDEX_P
1775 @item REGNO_OK_FOR_INDEX_P (@var{num})
1776 A C expression which is nonzero if register number @var{num} is
1777 suitable for use as an index register in operand addresses. It may be
1778 either a suitable hard register or a pseudo register that has been
1779 allocated such a hard register.
1781 The difference between an index register and a base register is that
1782 the index register may be scaled. If an address involves the sum of
1783 two registers, neither one of them scaled, then either one may be
1784 labeled the ``base'' and the other the ``index''; but whichever
1785 labeling is used must fit the machine's constraints of which registers
1786 may serve in each capacity. The compiler will try both labelings,
1787 looking for one that is valid, and will reload one or both registers
1788 only if neither labeling works.
1790 @findex PREFERRED_RELOAD_CLASS
1791 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
1792 A C expression that places additional restrictions on the register class
1793 to use when it is necessary to copy value @var{x} into a register in class
1794 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
1795 another, smaller class. On many machines, the following definition is
1799 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
1802 Sometimes returning a more restrictive class makes better code. For
1803 example, on the 68000, when @var{x} is an integer constant that is in range
1804 for a @samp{moveq} instruction, the value of this macro is always
1805 @code{DATA_REGS} as long as @var{class} includes the data registers.
1806 Requiring a data register guarantees that a @samp{moveq} will be used.
1808 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
1809 you can force @var{x} into a memory constant. This is useful on
1810 certain machines where immediate floating values cannot be loaded into
1811 certain kinds of registers.
1813 @findex PREFERRED_OUTPUT_RELOAD_CLASS
1814 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
1815 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
1816 input reloads. If you don't define this macro, the default is to use
1817 @var{class}, unchanged.
1819 @findex LIMIT_RELOAD_CLASS
1820 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
1821 A C expression that places additional restrictions on the register class
1822 to use when it is necessary to be able to hold a value of mode
1823 @var{mode} in a reload register for which class @var{class} would
1826 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
1827 there are certain modes that simply can't go in certain reload classes.
1829 The value is a register class; perhaps @var{class}, or perhaps another,
1832 Don't define this macro unless the target machine has limitations which
1833 require the macro to do something nontrivial.
1835 @findex SECONDARY_RELOAD_CLASS
1836 @findex SECONDARY_INPUT_RELOAD_CLASS
1837 @findex SECONDARY_OUTPUT_RELOAD_CLASS
1838 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1839 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1840 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1841 Many machines have some registers that cannot be copied directly to or
1842 from memory or even from other types of registers. An example is the
1843 @samp{MQ} register, which on most machines, can only be copied to or
1844 from general registers, but not memory. Some machines allow copying all
1845 registers to and from memory, but require a scratch register for stores
1846 to some memory locations (e.g., those with symbolic address on the RT,
1847 and those with certain symbolic address on the Sparc when compiling
1848 PIC). In some cases, both an intermediate and a scratch register are
1851 You should define these macros to indicate to the reload phase that it may
1852 need to allocate at least one register for a reload in addition to the
1853 register to contain the data. Specifically, if copying @var{x} to a
1854 register @var{class} in @var{mode} requires an intermediate register,
1855 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
1856 largest register class all of whose registers can be used as
1857 intermediate registers or scratch registers.
1859 If copying a register @var{class} in @var{mode} to @var{x} requires an
1860 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
1861 should be defined to return the largest register class required. If the
1862 requirements for input and output reloads are the same, the macro
1863 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
1866 The values returned by these macros are often @code{GENERAL_REGS}.
1867 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
1868 can be directly copied to or from a register of @var{class} in
1869 @var{mode} without requiring a scratch register. Do not define this
1870 macro if it would always return @code{NO_REGS}.
1872 If a scratch register is required (either with or without an
1873 intermediate register), you should define patterns for
1874 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
1875 (@pxref{Standard Names}. These patterns, which will normally be
1876 implemented with a @code{define_expand}, should be similar to the
1877 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
1880 Define constraints for the reload register and scratch register that
1881 contain a single register class. If the original reload register (whose
1882 class is @var{class}) can meet the constraint given in the pattern, the
1883 value returned by these macros is used for the class of the scratch
1884 register. Otherwise, two additional reload registers are required.
1885 Their classes are obtained from the constraints in the insn pattern.
1887 @var{x} might be a pseudo-register or a @code{subreg} of a
1888 pseudo-register, which could either be in a hard register or in memory.
1889 Use @code{true_regnum} to find out; it will return -1 if the pseudo is
1890 in memory and the hard register number if it is in a register.
1892 These macros should not be used in the case where a particular class of
1893 registers can only be copied to memory and not to another class of
1894 registers. In that case, secondary reload registers are not needed and
1895 would not be helpful. Instead, a stack location must be used to perform
1896 the copy and the @code{mov@var{m}} pattern should use memory as a
1897 intermediate storage. This case often occurs between floating-point and
1900 @findex SECONDARY_MEMORY_NEEDED
1901 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
1902 Certain machines have the property that some registers cannot be copied
1903 to some other registers without using memory. Define this macro on
1904 those machines to be a C expression that is non-zero if objects of mode
1905 @var{m} in registers of @var{class1} can only be copied to registers of
1906 class @var{class2} by storing a register of @var{class1} into memory
1907 and loading that memory location into a register of @var{class2}.
1909 Do not define this macro if its value would always be zero.
1911 @findex SECONDARY_MEMORY_NEEDED_RTX
1912 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
1913 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
1914 allocates a stack slot for a memory location needed for register copies.
1915 If this macro is defined, the compiler instead uses the memory location
1916 defined by this macro.
1918 Do not define this macro if you do not define
1919 @code{SECONDARY_MEMORY_NEEDED}.
1921 @findex SECONDARY_MEMORY_NEEDED_MODE
1922 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
1923 When the compiler needs a secondary memory location to copy between two
1924 registers of mode @var{mode}, it normally allocates sufficient memory to
1925 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
1926 load operations in a mode that many bits wide and whose class is the
1927 same as that of @var{mode}.
1929 This is right thing to do on most machines because it ensures that all
1930 bits of the register are copied and prevents accesses to the registers
1931 in a narrower mode, which some machines prohibit for floating-point
1934 However, this default behavior is not correct on some machines, such as
1935 the DEC Alpha, that store short integers in floating-point registers
1936 differently than in integer registers. On those machines, the default
1937 widening will not work correctly and you must define this macro to
1938 suppress that widening in some cases. See the file @file{alpha.h} for
1941 Do not define this macro if you do not define
1942 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
1943 is @code{BITS_PER_WORD} bits wide is correct for your machine.
1945 @findex SMALL_REGISTER_CLASSES
1946 @item SMALL_REGISTER_CLASSES
1947 Normally the compiler avoids choosing registers that have been
1948 explicitly mentioned in the rtl as spill registers (these registers are
1949 normally those used to pass parameters and return values). However,
1950 some machines have so few registers of certain classes that there
1951 would not be enough registers to use as spill registers if this were
1954 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a non-zero
1955 value on these machines. When this macro has a non-zero value, the
1956 compiler allows registers explicitly used in the rtl to be used as spill
1957 registers but avoids extending the lifetime of these registers.
1959 It is always safe to define this macro with a non-zero value, but if you
1960 unnecessarily define it, you will reduce the amount of optimizations
1961 that can be performed in some cases. If you do not define this macro
1962 with a non-zero value when it is required, the compiler will run out of
1963 spill registers and print a fatal error message. For most machines, you
1964 should not define this macro at all.
1966 @findex CLASS_LIKELY_SPILLED_P
1967 @item CLASS_LIKELY_SPILLED_P (@var{class})
1968 A C expression whose value is nonzero if pseudos that have been assigned
1969 to registers of class @var{class} would likely be spilled because
1970 registers of @var{class} are needed for spill registers.
1972 The default value of this macro returns 1 if @var{class} has exactly one
1973 register and zero otherwise. On most machines, this default should be
1974 used. Only define this macro to some other expression if pseudos
1975 allocated by @file{local-alloc.c} end up in memory because their hard
1976 registers were needed for spill registers. If this macro returns nonzero
1977 for those classes, those pseudos will only be allocated by
1978 @file{global.c}, which knows how to reallocate the pseudo to another
1979 register. If there would not be another register available for
1980 reallocation, you should not change the definition of this macro since
1981 the only effect of such a definition would be to slow down register
1984 @findex CLASS_MAX_NREGS
1985 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
1986 A C expression for the maximum number of consecutive registers
1987 of class @var{class} needed to hold a value of mode @var{mode}.
1989 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
1990 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
1991 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
1992 @var{mode})} for all @var{regno} values in the class @var{class}.
1994 This macro helps control the handling of multiple-word values
1997 @item CLASS_CANNOT_CHANGE_SIZE
1998 If defined, a C expression for a class that contains registers which the
1999 compiler must always access in a mode that is the same size as the mode
2000 in which it loaded the register.
2002 For the example, loading 32-bit integer or floating-point objects into
2003 floating-point registers on the Alpha extends them to 64-bits.
2004 Therefore loading a 64-bit object and then storing it as a 32-bit object
2005 does not store the low-order 32-bits, as would be the case for a normal
2006 register. Therefore, @file{alpha.h} defines this macro as
2010 Three other special macros describe which operands fit which constraint
2014 @findex CONST_OK_FOR_LETTER_P
2015 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2016 A C expression that defines the machine-dependent operand constraint
2017 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2018 particular ranges of integer values. If @var{c} is one of those
2019 letters, the expression should check that @var{value}, an integer, is in
2020 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2021 not one of those letters, the value should be 0 regardless of
2024 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2025 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2026 A C expression that defines the machine-dependent operand constraint
2027 letters that specify particular ranges of @code{const_double} values
2028 (@samp{G} or @samp{H}).
2030 If @var{c} is one of those letters, the expression should check that
2031 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2032 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2033 letters, the value should be 0 regardless of @var{value}.
2035 @code{const_double} is used for all floating-point constants and for
2036 @code{DImode} fixed-point constants. A given letter can accept either
2037 or both kinds of values. It can use @code{GET_MODE} to distinguish
2038 between these kinds.
2040 @findex EXTRA_CONSTRAINT
2041 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2042 A C expression that defines the optional machine-dependent constraint
2043 letters (@samp{Q}, @samp{R}, @samp{S}, @samp{T}, @samp{U}) that can
2044 be used to segregate specific types of operands, usually memory
2045 references, for the target machine. Normally this macro will not be
2046 defined. If it is required for a particular target machine, it should
2047 return 1 if @var{value} corresponds to the operand type represented by
2048 the constraint letter @var{c}. If @var{c} is not defined as an extra
2049 constraint, the value returned should be 0 regardless of @var{value}.
2051 For example, on the ROMP, load instructions cannot have their output in r0 if
2052 the memory reference contains a symbolic address. Constraint letter
2053 @samp{Q} is defined as representing a memory address that does
2054 @emph{not} contain a symbolic address. An alternative is specified with
2055 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2056 alternative specifies @samp{m} on the input and a register class that
2057 does not include r0 on the output.
2060 @node Stack and Calling
2061 @section Stack Layout and Calling Conventions
2062 @cindex calling conventions
2064 @c prevent bad page break with this line
2065 This describes the stack layout and calling conventions.
2073 * Register Arguments::
2075 * Aggregate Return::
2082 @subsection Basic Stack Layout
2083 @cindex stack frame layout
2084 @cindex frame layout
2086 @c prevent bad page break with this line
2087 Here is the basic stack layout.
2090 @findex STACK_GROWS_DOWNWARD
2091 @item STACK_GROWS_DOWNWARD
2092 Define this macro if pushing a word onto the stack moves the stack
2093 pointer to a smaller address.
2095 When we say, ``define this macro if @dots{},'' it means that the
2096 compiler checks this macro only with @code{#ifdef} so the precise
2097 definition used does not matter.
2099 @findex FRAME_GROWS_DOWNWARD
2100 @item FRAME_GROWS_DOWNWARD
2101 Define this macro if the addresses of local variable slots are at negative
2102 offsets from the frame pointer.
2104 @findex ARGS_GROW_DOWNWARD
2105 @item ARGS_GROW_DOWNWARD
2106 Define this macro if successive arguments to a function occupy decreasing
2107 addresses on the stack.
2109 @findex STARTING_FRAME_OFFSET
2110 @item STARTING_FRAME_OFFSET
2111 Offset from the frame pointer to the first local variable slot to be allocated.
2113 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2114 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2115 Otherwise, it is found by adding the length of the first slot to the
2116 value @code{STARTING_FRAME_OFFSET}.
2117 @c i'm not sure if the above is still correct.. had to change it to get
2118 @c rid of an overfull. --mew 2feb93
2120 @findex STACK_POINTER_OFFSET
2121 @item STACK_POINTER_OFFSET
2122 Offset from the stack pointer register to the first location at which
2123 outgoing arguments are placed. If not specified, the default value of
2124 zero is used. This is the proper value for most machines.
2126 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2127 the first location at which outgoing arguments are placed.
2129 @findex FIRST_PARM_OFFSET
2130 @item FIRST_PARM_OFFSET (@var{fundecl})
2131 Offset from the argument pointer register to the first argument's
2132 address. On some machines it may depend on the data type of the
2135 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2136 the first argument's address.
2138 @findex STACK_DYNAMIC_OFFSET
2139 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2140 Offset from the stack pointer register to an item dynamically allocated
2141 on the stack, e.g., by @code{alloca}.
2143 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2144 length of the outgoing arguments. The default is correct for most
2145 machines. See @file{function.c} for details.
2147 @findex DYNAMIC_CHAIN_ADDRESS
2148 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2149 A C expression whose value is RTL representing the address in a stack
2150 frame where the pointer to the caller's frame is stored. Assume that
2151 @var{frameaddr} is an RTL expression for the address of the stack frame
2154 If you don't define this macro, the default is to return the value
2155 of @var{frameaddr}---that is, the stack frame address is also the
2156 address of the stack word that points to the previous frame.
2158 @findex SETUP_FRAME_ADDRESSES
2159 @item SETUP_FRAME_ADDRESSES ()
2160 If defined, a C expression that produces the machine-specific code to
2161 setup the stack so that arbitrary frames can be accessed. For example,
2162 on the Sparc, we must flush all of the register windows to the stack
2163 before we can access arbitrary stack frames.
2164 This macro will seldom need to be defined.
2166 @findex RETURN_ADDR_RTX
2167 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2168 A C expression whose value is RTL representing the value of the return
2169 address for the frame @var{count} steps up from the current frame, after
2170 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2171 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2172 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2174 The value of the expression must always be the correct address when
2175 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2176 determine the return address of other frames.
2178 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2179 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2180 Define this if the return address of a particular stack frame is accessed
2181 from the frame pointer of the previous stack frame.
2183 @findex INCOMING_RETURN_ADDR_RTX
2184 @item INCOMING_RETURN_ADDR_RTX
2185 A C expression whose value is RTL representing the location of the
2186 incoming return address at the beginning of any function, before the
2187 prologue. This RTL is either a @code{REG}, indicating that the return
2188 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2191 You only need to define this macro if you want to support call frame
2192 debugging information like that provided by DWARF 2.
2194 @findex INCOMING_FRAME_SP_OFFSET
2195 @item INCOMING_FRAME_SP_OFFSET
2196 A C expression whose value is an integer giving the offset, in bytes,
2197 from the value of the stack pointer register to the top of the stack
2198 frame at the beginning of any function, before the prologue. The top of
2199 the frame is defined to be the value of the stack pointer in the
2200 previous frame, just before the call instruction.
2202 You only need to define this macro if you want to support call frame
2203 debugging information like that provided by DWARF 2.
2206 @node Stack Checking
2207 @subsection Specifying How Stack Checking is Done
2209 GNU CC will check that stack references are within the boundaries of
2210 the stack, if the @samp{-fstack-check} is specified, in one of three ways:
2214 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GNU CC
2215 will assume that you have arranged for stack checking to be done at
2216 appropriate places in the configuration files, e.g., in
2217 @code{FUNCTION_PROLOGUE}. GNU CC will do not other special processing.
2220 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
2221 called @code{check_stack} in your @file{md} file, GNU CC will call that
2222 pattern with one argument which is the address to compare the stack
2223 value against. You must arrange for this pattern to report an error if
2224 the stack pointer is out of range.
2227 If neither of the above are true, GNU CC will generate code to periodically
2228 ``probe'' the stack pointer using the values of the macros defined below.
2231 Normally, you will use the default values of these macros, so GNU CC
2232 will use the third approach.
2235 @findex STACK_CHECK_BUILTIN
2236 @item STACK_CHECK_BUILTIN
2237 A nonzero value if stack checking is done by the configuration files in a
2238 machine-dependent manner. You should define this macro if stack checking
2239 is require by the ABI of your machine or if you would like to have to stack
2240 checking in some more efficient way than GNU CC's portable approach.
2241 The default value of this macro is zero.
2243 @findex STACK_CHECK_PROBE_INTERVAL
2244 @item STACK_CHECK_PROBE_INTERVAL
2245 An integer representing the interval at which GNU CC must generate stack
2246 probe instructions. You will normally define this macro to be no larger
2247 than the size of the ``guard pages'' at the end of a stack area. The
2248 default value of 4096 is suitable for most systems.
2250 @findex STACK_CHECK_PROBE_LOAD
2251 @item STACK_CHECK_PROBE_LOAD
2252 A integer which is nonzero if GNU CC should perform the stack probe
2253 as a load instruction and zero if GNU CC should use a store instruction.
2254 The default is zero, which is the most efficient choice on most systems.
2256 @findex STACK_CHECK_PROTECT
2257 @item STACK_CHECK_PROTECT
2258 The number of bytes of stack needed to recover from a stack overflow,
2259 for languages where such a recovery is supported. The default value of
2260 75 words should be adequate for most machines.
2262 @findex STACK_CHECK_MAX_FRAME_SIZE
2263 @item STACK_CHECK_MAX_FRAME_SIZE
2264 The maximum size of a stack frame, in bytes. GNU CC will generate probe
2265 instructions in non-leaf functions to ensure at least this many bytes of
2266 stack are available. If a stack frame is larger than this size, stack
2267 checking will not be reliable and GNU CC will issue a warning. The
2268 default is chosen so that GNU CC only generates one instruction on most
2269 systems. You should normally not change the default value of this macro.
2271 @findex STACK_CHECK_FIXED_FRAME_SIZE
2272 @item STACK_CHECK_FIXED_FRAME_SIZE
2273 GNU CC uses this value to generate the above warning message. It
2274 represents the amount of fixed frame used by a function, not including
2275 space for any callee-saved registers, temporaries and user variables.
2276 You need only specify an upper bound for this amount and will normally
2277 use the default of four words.
2279 @findex STACK_CHECK_MAX_VAR_SIZE
2280 @item STACK_CHECK_MAX_VAR_SIZE
2281 The maximum size, in bytes, of an object that GNU CC will place in the
2282 fixed area of the stack frame when the user specifies
2283 @samp{-fstack-check}.
2284 GNU CC computed the default from the values of the above macros and you will
2285 normally not need to override that default.
2289 @node Frame Registers
2290 @subsection Registers That Address the Stack Frame
2292 @c prevent bad page break with this line
2293 This discusses registers that address the stack frame.
2296 @findex STACK_POINTER_REGNUM
2297 @item STACK_POINTER_REGNUM
2298 The register number of the stack pointer register, which must also be a
2299 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2300 the hardware determines which register this is.
2302 @findex FRAME_POINTER_REGNUM
2303 @item FRAME_POINTER_REGNUM
2304 The register number of the frame pointer register, which is used to
2305 access automatic variables in the stack frame. On some machines, the
2306 hardware determines which register this is. On other machines, you can
2307 choose any register you wish for this purpose.
2309 @findex HARD_FRAME_POINTER_REGNUM
2310 @item HARD_FRAME_POINTER_REGNUM
2311 On some machines the offset between the frame pointer and starting
2312 offset of the automatic variables is not known until after register
2313 allocation has been done (for example, because the saved registers are
2314 between these two locations). On those machines, define
2315 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
2316 be used internally until the offset is known, and define
2317 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
2318 used for the frame pointer.
2320 You should define this macro only in the very rare circumstances when it
2321 is not possible to calculate the offset between the frame pointer and
2322 the automatic variables until after register allocation has been
2323 completed. When this macro is defined, you must also indicate in your
2324 definition of @code{ELIMINABLE_REGS} how to eliminate
2325 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
2326 or @code{STACK_POINTER_REGNUM}.
2328 Do not define this macro if it would be the same as
2329 @code{FRAME_POINTER_REGNUM}.
2331 @findex ARG_POINTER_REGNUM
2332 @item ARG_POINTER_REGNUM
2333 The register number of the arg pointer register, which is used to access
2334 the function's argument list. On some machines, this is the same as the
2335 frame pointer register. On some machines, the hardware determines which
2336 register this is. On other machines, you can choose any register you
2337 wish for this purpose. If this is not the same register as the frame
2338 pointer register, then you must mark it as a fixed register according to
2339 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2340 (@pxref{Elimination}).
2342 @findex RETURN_ADDRESS_POINTER_REGNUM
2343 @item RETURN_ADDRESS_POINTER_REGNUM
2344 The register number of the return address pointer register, which is used to
2345 access the current function's return address from the stack. On some
2346 machines, the return address is not at a fixed offset from the frame
2347 pointer or stack pointer or argument pointer. This register can be defined
2348 to point to the return address on the stack, and then be converted by
2349 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
2351 Do not define this macro unless there is no other way to get the return
2352 address from the stack.
2354 @findex STATIC_CHAIN_REGNUM
2355 @findex STATIC_CHAIN_INCOMING_REGNUM
2356 @item STATIC_CHAIN_REGNUM
2357 @itemx STATIC_CHAIN_INCOMING_REGNUM
2358 Register numbers used for passing a function's static chain pointer. If
2359 register windows are used, the register number as seen by the called
2360 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
2361 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
2362 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
2363 not be defined.@refill
2365 The static chain register need not be a fixed register.
2367 If the static chain is passed in memory, these macros should not be
2368 defined; instead, the next two macros should be defined.
2370 @findex STATIC_CHAIN
2371 @findex STATIC_CHAIN_INCOMING
2373 @itemx STATIC_CHAIN_INCOMING
2374 If the static chain is passed in memory, these macros provide rtx giving
2375 @code{mem} expressions that denote where they are stored.
2376 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
2377 as seen by the calling and called functions, respectively. Often the former
2378 will be at an offset from the stack pointer and the latter at an offset from
2379 the frame pointer.@refill
2381 @findex stack_pointer_rtx
2382 @findex frame_pointer_rtx
2383 @findex arg_pointer_rtx
2384 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
2385 @code{arg_pointer_rtx} will have been initialized prior to the use of these
2386 macros and should be used to refer to those items.
2388 If the static chain is passed in a register, the two previous macros should
2393 @subsection Eliminating Frame Pointer and Arg Pointer
2395 @c prevent bad page break with this line
2396 This is about eliminating the frame pointer and arg pointer.
2399 @findex FRAME_POINTER_REQUIRED
2400 @item FRAME_POINTER_REQUIRED
2401 A C expression which is nonzero if a function must have and use a frame
2402 pointer. This expression is evaluated in the reload pass. If its value is
2403 nonzero the function will have a frame pointer.
2405 The expression can in principle examine the current function and decide
2406 according to the facts, but on most machines the constant 0 or the
2407 constant 1 suffices. Use 0 when the machine allows code to be generated
2408 with no frame pointer, and doing so saves some time or space. Use 1
2409 when there is no possible advantage to avoiding a frame pointer.
2411 In certain cases, the compiler does not know how to produce valid code
2412 without a frame pointer. The compiler recognizes those cases and
2413 automatically gives the function a frame pointer regardless of what
2414 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
2417 In a function that does not require a frame pointer, the frame pointer
2418 register can be allocated for ordinary usage, unless you mark it as a
2419 fixed register. See @code{FIXED_REGISTERS} for more information.
2421 @findex INITIAL_FRAME_POINTER_OFFSET
2422 @findex get_frame_size
2423 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
2424 A C statement to store in the variable @var{depth-var} the difference
2425 between the frame pointer and the stack pointer values immediately after
2426 the function prologue. The value would be computed from information
2427 such as the result of @code{get_frame_size ()} and the tables of
2428 registers @code{regs_ever_live} and @code{call_used_regs}.
2430 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
2431 need not be defined. Otherwise, it must be defined even if
2432 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
2433 case, you may set @var{depth-var} to anything.
2435 @findex ELIMINABLE_REGS
2436 @item ELIMINABLE_REGS
2437 If defined, this macro specifies a table of register pairs used to
2438 eliminate unneeded registers that point into the stack frame. If it is not
2439 defined, the only elimination attempted by the compiler is to replace
2440 references to the frame pointer with references to the stack pointer.
2442 The definition of this macro is a list of structure initializations, each
2443 of which specifies an original and replacement register.
2445 On some machines, the position of the argument pointer is not known until
2446 the compilation is completed. In such a case, a separate hard register
2447 must be used for the argument pointer. This register can be eliminated by
2448 replacing it with either the frame pointer or the argument pointer,
2449 depending on whether or not the frame pointer has been eliminated.
2451 In this case, you might specify:
2453 #define ELIMINABLE_REGS \
2454 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
2455 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
2456 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
2459 Note that the elimination of the argument pointer with the stack pointer is
2460 specified first since that is the preferred elimination.
2462 @findex CAN_ELIMINATE
2463 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
2464 A C expression that returns non-zero if the compiler is allowed to try
2465 to replace register number @var{from-reg} with register number
2466 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
2467 is defined, and will usually be the constant 1, since most of the cases
2468 preventing register elimination are things that the compiler already
2471 @findex INITIAL_ELIMINATION_OFFSET
2472 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
2473 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
2474 specifies the initial difference between the specified pair of
2475 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
2478 @findex LONGJMP_RESTORE_FROM_STACK
2479 @item LONGJMP_RESTORE_FROM_STACK
2480 Define this macro if the @code{longjmp} function restores registers from
2481 the stack frames, rather than from those saved specifically by
2482 @code{setjmp}. Certain quantities must not be kept in registers across
2483 a call to @code{setjmp} on such machines.
2486 @node Stack Arguments
2487 @subsection Passing Function Arguments on the Stack
2488 @cindex arguments on stack
2489 @cindex stack arguments
2491 The macros in this section control how arguments are passed
2492 on the stack. See the following section for other macros that
2493 control passing certain arguments in registers.
2496 @findex PROMOTE_PROTOTYPES
2497 @item PROMOTE_PROTOTYPES
2498 Define this macro if an argument declared in a prototype as an
2499 integral type smaller than @code{int} should actually be passed as an
2500 @code{int}. In addition to avoiding errors in certain cases of
2501 mismatch, it also makes for better code on certain machines.
2503 @findex PUSH_ROUNDING
2504 @item PUSH_ROUNDING (@var{npushed})
2505 A C expression that is the number of bytes actually pushed onto the
2506 stack when an instruction attempts to push @var{npushed} bytes.
2508 If the target machine does not have a push instruction, do not define
2509 this macro. That directs GNU CC to use an alternate strategy: to
2510 allocate the entire argument block and then store the arguments into
2513 On some machines, the definition
2516 #define PUSH_ROUNDING(BYTES) (BYTES)
2520 will suffice. But on other machines, instructions that appear
2521 to push one byte actually push two bytes in an attempt to maintain
2522 alignment. Then the definition should be
2525 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
2528 @findex ACCUMULATE_OUTGOING_ARGS
2529 @findex current_function_outgoing_args_size
2530 @item ACCUMULATE_OUTGOING_ARGS
2531 If defined, the maximum amount of space required for outgoing arguments
2532 will be computed and placed into the variable
2533 @code{current_function_outgoing_args_size}. No space will be pushed
2534 onto the stack for each call; instead, the function prologue should
2535 increase the stack frame size by this amount.
2537 Defining both @code{PUSH_ROUNDING} and @code{ACCUMULATE_OUTGOING_ARGS}
2540 @findex REG_PARM_STACK_SPACE
2541 @item REG_PARM_STACK_SPACE (@var{fndecl})
2542 Define this macro if functions should assume that stack space has been
2543 allocated for arguments even when their values are passed in
2546 The value of this macro is the size, in bytes, of the area reserved for
2547 arguments passed in registers for the function represented by @var{fndecl}.
2549 This space can be allocated by the caller, or be a part of the
2550 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
2552 @c above is overfull. not sure what to do. --mew 5feb93 did
2553 @c something, not sure if it looks good. --mew 10feb93
2555 @findex MAYBE_REG_PARM_STACK_SPACE
2556 @findex FINAL_REG_PARM_STACK_SPACE
2557 @item MAYBE_REG_PARM_STACK_SPACE
2558 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
2559 Define these macros in addition to the one above if functions might
2560 allocate stack space for arguments even when their values are passed
2561 in registers. These should be used when the stack space allocated
2562 for arguments in registers is not a simple constant independent of the
2563 function declaration.
2565 The value of the first macro is the size, in bytes, of the area that
2566 we should initially assume would be reserved for arguments passed in registers.
2568 The value of the second macro is the actual size, in bytes, of the area
2569 that will be reserved for arguments passed in registers. This takes two
2570 arguments: an integer representing the number of bytes of fixed sized
2571 arguments on the stack, and a tree representing the number of bytes of
2572 variable sized arguments on the stack.
2574 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
2575 called for libcall functions, the current function, or for a function
2576 being called when it is known that such stack space must be allocated.
2577 In each case this value can be easily computed.
2579 When deciding whether a called function needs such stack space, and how
2580 much space to reserve, GNU CC uses these two macros instead of
2581 @code{REG_PARM_STACK_SPACE}.
2583 @findex OUTGOING_REG_PARM_STACK_SPACE
2584 @item OUTGOING_REG_PARM_STACK_SPACE
2585 Define this if it is the responsibility of the caller to allocate the area
2586 reserved for arguments passed in registers.
2588 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
2589 whether the space for these arguments counts in the value of
2590 @code{current_function_outgoing_args_size}.
2592 @findex STACK_PARMS_IN_REG_PARM_AREA
2593 @item STACK_PARMS_IN_REG_PARM_AREA
2594 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
2595 stack parameters don't skip the area specified by it.
2596 @c i changed this, makes more sens and it should have taken care of the
2597 @c overfull.. not as specific, tho. --mew 5feb93
2599 Normally, when a parameter is not passed in registers, it is placed on the
2600 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
2601 suppresses this behavior and causes the parameter to be passed on the
2602 stack in its natural location.
2604 @findex RETURN_POPS_ARGS
2605 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
2606 A C expression that should indicate the number of bytes of its own
2607 arguments that a function pops on returning, or 0 if the
2608 function pops no arguments and the caller must therefore pop them all
2609 after the function returns.
2611 @var{fundecl} is a C variable whose value is a tree node that describes
2612 the function in question. Normally it is a node of type
2613 @code{FUNCTION_DECL} that describes the declaration of the function.
2614 From this you can obtain the DECL_MACHINE_ATTRIBUTES of the function.
2616 @var{funtype} is a C variable whose value is a tree node that
2617 describes the function in question. Normally it is a node of type
2618 @code{FUNCTION_TYPE} that describes the data type of the function.
2619 From this it is possible to obtain the data types of the value and
2620 arguments (if known).
2622 When a call to a library function is being considered, @var{fundecl}
2623 will contain an identifier node for the library function. Thus, if
2624 you need to distinguish among various library functions, you can do so
2625 by their names. Note that ``library function'' in this context means
2626 a function used to perform arithmetic, whose name is known specially
2627 in the compiler and was not mentioned in the C code being compiled.
2629 @var{stack-size} is the number of bytes of arguments passed on the
2630 stack. If a variable number of bytes is passed, it is zero, and
2631 argument popping will always be the responsibility of the calling function.
2633 On the Vax, all functions always pop their arguments, so the definition
2634 of this macro is @var{stack-size}. On the 68000, using the standard
2635 calling convention, no functions pop their arguments, so the value of
2636 the macro is always 0 in this case. But an alternative calling
2637 convention is available in which functions that take a fixed number of
2638 arguments pop them but other functions (such as @code{printf}) pop
2639 nothing (the caller pops all). When this convention is in use,
2640 @var{funtype} is examined to determine whether a function takes a fixed
2641 number of arguments.
2644 @node Register Arguments
2645 @subsection Passing Arguments in Registers
2646 @cindex arguments in registers
2647 @cindex registers arguments
2649 This section describes the macros which let you control how various
2650 types of arguments are passed in registers or how they are arranged in
2654 @findex FUNCTION_ARG
2655 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2656 A C expression that controls whether a function argument is passed
2657 in a register, and which register.
2659 The arguments are @var{cum}, which summarizes all the previous
2660 arguments; @var{mode}, the machine mode of the argument; @var{type},
2661 the data type of the argument as a tree node or 0 if that is not known
2662 (which happens for C support library functions); and @var{named},
2663 which is 1 for an ordinary argument and 0 for nameless arguments that
2664 correspond to @samp{@dots{}} in the called function's prototype.
2666 The value of the expression is usually either a @code{reg} RTX for the
2667 hard register in which to pass the argument, or zero to pass the
2668 argument on the stack.
2670 For machines like the Vax and 68000, where normally all arguments are
2671 pushed, zero suffices as a definition.
2673 The value of the expression can also be a @code{parallel} RTX. This is
2674 used when an argument is passed in multiple locations. The mode of the
2675 of the @code{parallel} should be the mode of the entire argument. The
2676 @code{parallel} holds any number of @code{expr_list} pairs; each one
2677 describes where part of the argument is passed. In each @code{expr_list},
2678 the first operand can be either a @code{reg} RTX for the hard register
2679 in which to pass this part of the argument, or zero to pass the argument
2680 on the stack. If this operand is a @code{reg}, then the mode indicates
2681 how large this part of the argument is. The second operand of the
2682 @code{expr_list} is a @code{const_int} which gives the offset in bytes
2683 into the entire argument where this part starts.
2685 @cindex @file{stdarg.h} and register arguments
2686 The usual way to make the ANSI library @file{stdarg.h} work on a machine
2687 where some arguments are usually passed in registers, is to cause
2688 nameless arguments to be passed on the stack instead. This is done
2689 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
2691 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
2692 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
2693 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
2694 in the definition of this macro to determine if this argument is of a
2695 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
2696 is not defined and @code{FUNCTION_ARG} returns non-zero for such an
2697 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
2698 defined, the argument will be computed in the stack and then loaded into
2701 @findex MUST_PASS_IN_STACK
2702 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
2703 Define as a C expression that evaluates to nonzero if we do not know how
2704 to pass TYPE solely in registers. The file @file{expr.h} defines a
2705 definition that is usually appropriate, refer to @file{expr.h} for additional
2708 @findex FUNCTION_INCOMING_ARG
2709 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2710 Define this macro if the target machine has ``register windows'', so
2711 that the register in which a function sees an arguments is not
2712 necessarily the same as the one in which the caller passed the
2715 For such machines, @code{FUNCTION_ARG} computes the register in which
2716 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
2717 be defined in a similar fashion to tell the function being called
2718 where the arguments will arrive.
2720 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
2721 serves both purposes.@refill
2723 @findex FUNCTION_ARG_PARTIAL_NREGS
2724 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
2725 A C expression for the number of words, at the beginning of an
2726 argument, must be put in registers. The value must be zero for
2727 arguments that are passed entirely in registers or that are entirely
2728 pushed on the stack.
2730 On some machines, certain arguments must be passed partially in
2731 registers and partially in memory. On these machines, typically the
2732 first @var{n} words of arguments are passed in registers, and the rest
2733 on the stack. If a multi-word argument (a @code{double} or a
2734 structure) crosses that boundary, its first few words must be passed
2735 in registers and the rest must be pushed. This macro tells the
2736 compiler when this occurs, and how many of the words should go in
2739 @code{FUNCTION_ARG} for these arguments should return the first
2740 register to be used by the caller for this argument; likewise
2741 @code{FUNCTION_INCOMING_ARG}, for the called function.
2743 @findex FUNCTION_ARG_PASS_BY_REFERENCE
2744 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2745 A C expression that indicates when an argument must be passed by reference.
2746 If nonzero for an argument, a copy of that argument is made in memory and a
2747 pointer to the argument is passed instead of the argument itself.
2748 The pointer is passed in whatever way is appropriate for passing a pointer
2751 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
2752 definition of this macro might be
2754 #define FUNCTION_ARG_PASS_BY_REFERENCE\
2755 (CUM, MODE, TYPE, NAMED) \
2756 MUST_PASS_IN_STACK (MODE, TYPE)
2758 @c this is *still* too long. --mew 5feb93
2760 @findex FUNCTION_ARG_CALLEE_COPIES
2761 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
2762 If defined, a C expression that indicates when it is the called function's
2763 responsibility to make a copy of arguments passed by invisible reference.
2764 Normally, the caller makes a copy and passes the address of the copy to the
2765 routine being called. When FUNCTION_ARG_CALLEE_COPIES is defined and is
2766 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
2767 ``live'' value. The called function must not modify this value. If it can be
2768 determined that the value won't be modified, it need not make a copy;
2769 otherwise a copy must be made.
2771 @findex CUMULATIVE_ARGS
2772 @item CUMULATIVE_ARGS
2773 A C type for declaring a variable that is used as the first argument of
2774 @code{FUNCTION_ARG} and other related values. For some target machines,
2775 the type @code{int} suffices and can hold the number of bytes of
2778 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
2779 arguments that have been passed on the stack. The compiler has other
2780 variables to keep track of that. For target machines on which all
2781 arguments are passed on the stack, there is no need to store anything in
2782 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
2783 should not be empty, so use @code{int}.
2785 @findex INIT_CUMULATIVE_ARGS
2786 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
2787 A C statement (sans semicolon) for initializing the variable @var{cum}
2788 for the state at the beginning of the argument list. The variable has
2789 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
2790 for the data type of the function which will receive the args, or 0
2791 if the args are to a compiler support library function. The value of
2792 @var{indirect} is nonzero when processing an indirect call, for example
2793 a call through a function pointer. The value of @var{indirect} is zero
2794 for a call to an explicitly named function, a library function call, or when
2795 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
2798 When processing a call to a compiler support library function,
2799 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
2800 contains the name of the function, as a string. @var{libname} is 0 when
2801 an ordinary C function call is being processed. Thus, each time this
2802 macro is called, either @var{libname} or @var{fntype} is nonzero, but
2803 never both of them at once.
2805 @findex INIT_CUMULATIVE_INCOMING_ARGS
2806 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
2807 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
2808 finding the arguments for the function being compiled. If this macro is
2809 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
2811 The value passed for @var{libname} is always 0, since library routines
2812 with special calling conventions are never compiled with GNU CC. The
2813 argument @var{libname} exists for symmetry with
2814 @code{INIT_CUMULATIVE_ARGS}.
2815 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
2816 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
2818 @findex FUNCTION_ARG_ADVANCE
2819 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2820 A C statement (sans semicolon) to update the summarizer variable
2821 @var{cum} to advance past an argument in the argument list. The
2822 values @var{mode}, @var{type} and @var{named} describe that argument.
2823 Once this is done, the variable @var{cum} is suitable for analyzing
2824 the @emph{following} argument with @code{FUNCTION_ARG}, etc.@refill
2826 This macro need not do anything if the argument in question was passed
2827 on the stack. The compiler knows how to track the amount of stack space
2828 used for arguments without any special help.
2830 @findex FUNCTION_ARG_PADDING
2831 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
2832 If defined, a C expression which determines whether, and in which direction,
2833 to pad out an argument with extra space. The value should be of type
2834 @code{enum direction}: either @code{upward} to pad above the argument,
2835 @code{downward} to pad below, or @code{none} to inhibit padding.
2837 The @emph{amount} of padding is always just enough to reach the next
2838 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
2841 This macro has a default definition which is right for most systems.
2842 For little-endian machines, the default is to pad upward. For
2843 big-endian machines, the default is to pad downward for an argument of
2844 constant size shorter than an @code{int}, and upward otherwise.
2846 @findex FUNCTION_ARG_BOUNDARY
2847 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
2848 If defined, a C expression that gives the alignment boundary, in bits,
2849 of an argument with the specified mode and type. If it is not defined,
2850 @code{PARM_BOUNDARY} is used for all arguments.
2852 @findex FUNCTION_ARG_REGNO_P
2853 @item FUNCTION_ARG_REGNO_P (@var{regno})
2854 A C expression that is nonzero if @var{regno} is the number of a hard
2855 register in which function arguments are sometimes passed. This does
2856 @emph{not} include implicit arguments such as the static chain and
2857 the structure-value address. On many machines, no registers can be
2858 used for this purpose since all function arguments are pushed on the
2861 @findex LOAD_ARGS_REVERSED
2862 @item LOAD_ARGS_REVERSED
2863 If defined, the order in which arguments are loaded into their
2864 respective argument registers is reversed so that the last
2865 argument is loaded first. This macro only effects arguments
2866 passed in registers.
2871 @subsection How Scalar Function Values Are Returned
2872 @cindex return values in registers
2873 @cindex values, returned by functions
2874 @cindex scalars, returned as values
2876 This section discusses the macros that control returning scalars as
2877 values---values that can fit in registers.
2880 @findex TRADITIONAL_RETURN_FLOAT
2881 @item TRADITIONAL_RETURN_FLOAT
2882 Define this macro if @samp{-traditional} should not cause functions
2883 declared to return @code{float} to convert the value to @code{double}.
2885 @findex FUNCTION_VALUE
2886 @item FUNCTION_VALUE (@var{valtype}, @var{func})
2887 A C expression to create an RTX representing the place where a
2888 function returns a value of data type @var{valtype}. @var{valtype} is
2889 a tree node representing a data type. Write @code{TYPE_MODE
2890 (@var{valtype})} to get the machine mode used to represent that type.
2891 On many machines, only the mode is relevant. (Actually, on most
2892 machines, scalar values are returned in the same place regardless of
2895 The value of the expression is usually a @code{reg} RTX for the hard
2896 register where the return value is stored. The value can also be a
2897 @code{parallel} RTX, if the return value is in multiple places. See
2898 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
2900 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
2901 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
2904 If the precise function being called is known, @var{func} is a tree
2905 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
2906 pointer. This makes it possible to use a different value-returning
2907 convention for specific functions when all their calls are
2910 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
2911 types, because these are returned in another way. See
2912 @code{STRUCT_VALUE_REGNUM} and related macros, below.
2914 @findex FUNCTION_OUTGOING_VALUE
2915 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
2916 Define this macro if the target machine has ``register windows''
2917 so that the register in which a function returns its value is not
2918 the same as the one in which the caller sees the value.
2920 For such machines, @code{FUNCTION_VALUE} computes the register in which
2921 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
2922 defined in a similar fashion to tell the function where to put the
2925 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
2926 @code{FUNCTION_VALUE} serves both purposes.@refill
2928 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
2929 aggregate data types, because these are returned in another way. See
2930 @code{STRUCT_VALUE_REGNUM} and related macros, below.
2932 @findex LIBCALL_VALUE
2933 @item LIBCALL_VALUE (@var{mode})
2934 A C expression to create an RTX representing the place where a library
2935 function returns a value of mode @var{mode}. If the precise function
2936 being called is known, @var{func} is a tree node
2937 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
2938 pointer. This makes it possible to use a different value-returning
2939 convention for specific functions when all their calls are
2942 Note that ``library function'' in this context means a compiler
2943 support routine, used to perform arithmetic, whose name is known
2944 specially by the compiler and was not mentioned in the C code being
2947 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
2948 data types, because none of the library functions returns such types.
2950 @findex FUNCTION_VALUE_REGNO_P
2951 @item FUNCTION_VALUE_REGNO_P (@var{regno})
2952 A C expression that is nonzero if @var{regno} is the number of a hard
2953 register in which the values of called function may come back.
2955 A register whose use for returning values is limited to serving as the
2956 second of a pair (for a value of type @code{double}, say) need not be
2957 recognized by this macro. So for most machines, this definition
2961 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
2964 If the machine has register windows, so that the caller and the called
2965 function use different registers for the return value, this macro
2966 should recognize only the caller's register numbers.
2968 @findex APPLY_RESULT_SIZE
2969 @item APPLY_RESULT_SIZE
2970 Define this macro if @samp{untyped_call} and @samp{untyped_return}
2971 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
2972 saving and restoring an arbitrary return value.
2975 @node Aggregate Return
2976 @subsection How Large Values Are Returned
2977 @cindex aggregates as return values
2978 @cindex large return values
2979 @cindex returning aggregate values
2980 @cindex structure value address
2982 When a function value's mode is @code{BLKmode} (and in some other
2983 cases), the value is not returned according to @code{FUNCTION_VALUE}
2984 (@pxref{Scalar Return}). Instead, the caller passes the address of a
2985 block of memory in which the value should be stored. This address
2986 is called the @dfn{structure value address}.
2988 This section describes how to control returning structure values in
2992 @findex RETURN_IN_MEMORY
2993 @item RETURN_IN_MEMORY (@var{type})
2994 A C expression which can inhibit the returning of certain function
2995 values in registers, based on the type of value. A nonzero value says
2996 to return the function value in memory, just as large structures are
2997 always returned. Here @var{type} will be a C expression of type
2998 @code{tree}, representing the data type of the value.
3000 Note that values of mode @code{BLKmode} must be explicitly handled
3001 by this macro. Also, the option @samp{-fpcc-struct-return}
3002 takes effect regardless of this macro. On most systems, it is
3003 possible to leave the macro undefined; this causes a default
3004 definition to be used, whose value is the constant 1 for @code{BLKmode}
3005 values, and 0 otherwise.
3007 Do not use this macro to indicate that structures and unions should always
3008 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3011 @findex DEFAULT_PCC_STRUCT_RETURN
3012 @item DEFAULT_PCC_STRUCT_RETURN
3013 Define this macro to be 1 if all structure and union return values must be
3014 in memory. Since this results in slower code, this should be defined
3015 only if needed for compatibility with other compilers or with an ABI.
3016 If you define this macro to be 0, then the conventions used for structure
3017 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3019 If not defined, this defaults to the value 1.
3021 @findex STRUCT_VALUE_REGNUM
3022 @item STRUCT_VALUE_REGNUM
3023 If the structure value address is passed in a register, then
3024 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3026 @findex STRUCT_VALUE
3028 If the structure value address is not passed in a register, define
3029 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3030 where the address is passed. If it returns 0, the address is passed as
3031 an ``invisible'' first argument.
3033 @findex STRUCT_VALUE_INCOMING_REGNUM
3034 @item STRUCT_VALUE_INCOMING_REGNUM
3035 On some architectures the place where the structure value address
3036 is found by the called function is not the same place that the
3037 caller put it. This can be due to register windows, or it could
3038 be because the function prologue moves it to a different place.
3040 If the incoming location of the structure value address is in a
3041 register, define this macro as the register number.
3043 @findex STRUCT_VALUE_INCOMING
3044 @item STRUCT_VALUE_INCOMING
3045 If the incoming location is not a register, then you should define
3046 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3047 called function should find the value. If it should find the value on
3048 the stack, define this to create a @code{mem} which refers to the frame
3049 pointer. A definition of 0 means that the address is passed as an
3050 ``invisible'' first argument.
3052 @findex PCC_STATIC_STRUCT_RETURN
3053 @item PCC_STATIC_STRUCT_RETURN
3054 Define this macro if the usual system convention on the target machine
3055 for returning structures and unions is for the called function to return
3056 the address of a static variable containing the value.
3058 Do not define this if the usual system convention is for the caller to
3059 pass an address to the subroutine.
3061 This macro has effect in @samp{-fpcc-struct-return} mode, but it does
3062 nothing when you use @samp{-freg-struct-return} mode.
3066 @subsection Caller-Saves Register Allocation
3068 If you enable it, GNU CC can save registers around function calls. This
3069 makes it possible to use call-clobbered registers to hold variables that
3070 must live across calls.
3073 @findex DEFAULT_CALLER_SAVES
3074 @item DEFAULT_CALLER_SAVES
3075 Define this macro if function calls on the target machine do not preserve
3076 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3077 for all registers. This macro enables @samp{-fcaller-saves} by default.
3078 Eventually that option will be enabled by default on all machines and both
3079 the option and this macro will be eliminated.
3081 @findex CALLER_SAVE_PROFITABLE
3082 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3083 A C expression to determine whether it is worthwhile to consider placing
3084 a pseudo-register in a call-clobbered hard register and saving and
3085 restoring it around each function call. The expression should be 1 when
3086 this is worth doing, and 0 otherwise.
3088 If you don't define this macro, a default is used which is good on most
3089 machines: @code{4 * @var{calls} < @var{refs}}.
3092 @node Function Entry
3093 @subsection Function Entry and Exit
3094 @cindex function entry and exit
3098 This section describes the macros that output function entry
3099 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3102 @findex FUNCTION_PROLOGUE
3103 @item FUNCTION_PROLOGUE (@var{file}, @var{size})
3104 A C compound statement that outputs the assembler code for entry to a
3105 function. The prologue is responsible for setting up the stack frame,
3106 initializing the frame pointer register, saving registers that must be
3107 saved, and allocating @var{size} additional bytes of storage for the
3108 local variables. @var{size} is an integer. @var{file} is a stdio
3109 stream to which the assembler code should be output.
3111 The label for the beginning of the function need not be output by this
3112 macro. That has already been done when the macro is run.
3114 @findex regs_ever_live
3115 To determine which registers to save, the macro can refer to the array
3116 @code{regs_ever_live}: element @var{r} is nonzero if hard register
3117 @var{r} is used anywhere within the function. This implies the function
3118 prologue should save register @var{r}, provided it is not one of the
3119 call-used registers. (@code{FUNCTION_EPILOGUE} must likewise use
3120 @code{regs_ever_live}.)
3122 On machines that have ``register windows'', the function entry code does
3123 not save on the stack the registers that are in the windows, even if
3124 they are supposed to be preserved by function calls; instead it takes
3125 appropriate steps to ``push'' the register stack, if any non-call-used
3126 registers are used in the function.
3128 @findex frame_pointer_needed
3129 On machines where functions may or may not have frame-pointers, the
3130 function entry code must vary accordingly; it must set up the frame
3131 pointer if one is wanted, and not otherwise. To determine whether a
3132 frame pointer is in wanted, the macro can refer to the variable
3133 @code{frame_pointer_needed}. The variable's value will be 1 at run
3134 time in a function that needs a frame pointer. @xref{Elimination}.
3136 The function entry code is responsible for allocating any stack space
3137 required for the function. This stack space consists of the regions
3138 listed below. In most cases, these regions are allocated in the
3139 order listed, with the last listed region closest to the top of the
3140 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
3141 the highest address if it is not defined). You can use a different order
3142 for a machine if doing so is more convenient or required for
3143 compatibility reasons. Except in cases where required by standard
3144 or by a debugger, there is no reason why the stack layout used by GCC
3145 need agree with that used by other compilers for a machine.
3149 @findex current_function_pretend_args_size
3150 A region of @code{current_function_pretend_args_size} bytes of
3151 uninitialized space just underneath the first argument arriving on the
3152 stack. (This may not be at the very start of the allocated stack region
3153 if the calling sequence has pushed anything else since pushing the stack
3154 arguments. But usually, on such machines, nothing else has been pushed
3155 yet, because the function prologue itself does all the pushing.) This
3156 region is used on machines where an argument may be passed partly in
3157 registers and partly in memory, and, in some cases to support the
3158 features in @file{varargs.h} and @file{stdargs.h}.
3161 An area of memory used to save certain registers used by the function.
3162 The size of this area, which may also include space for such things as
3163 the return address and pointers to previous stack frames, is
3164 machine-specific and usually depends on which registers have been used
3165 in the function. Machines with register windows often do not require
3169 A region of at least @var{size} bytes, possibly rounded up to an allocation
3170 boundary, to contain the local variables of the function. On some machines,
3171 this region and the save area may occur in the opposite order, with the
3172 save area closer to the top of the stack.
3175 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3176 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3177 @code{current_function_outgoing_args_size} bytes to be used for outgoing
3178 argument lists of the function. @xref{Stack Arguments}.
3181 Normally, it is necessary for the macros @code{FUNCTION_PROLOGUE} and
3182 @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C
3183 variable @code{leaf_function} is nonzero for such a function.
3185 @findex EXIT_IGNORE_STACK
3186 @item EXIT_IGNORE_STACK
3187 Define this macro as a C expression that is nonzero if the return
3188 instruction or the function epilogue ignores the value of the stack
3189 pointer; in other words, if it is safe to delete an instruction to
3190 adjust the stack pointer before a return from the function.
3192 Note that this macro's value is relevant only for functions for which
3193 frame pointers are maintained. It is never safe to delete a final
3194 stack adjustment in a function that has no frame pointer, and the
3195 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3197 @findex EPILOGUE_USES
3198 @item EPILOGUE_USES (@var{regno})
3199 Define this macro as a C expression that is nonzero for registers are
3200 used by the epilogue or the @samp{return} pattern. The stack and frame
3201 pointer registers are already be assumed to be used as needed.
3203 @findex FUNCTION_EPILOGUE
3204 @item FUNCTION_EPILOGUE (@var{file}, @var{size})
3205 A C compound statement that outputs the assembler code for exit from a
3206 function. The epilogue is responsible for restoring the saved
3207 registers and stack pointer to their values when the function was
3208 called, and returning control to the caller. This macro takes the
3209 same arguments as the macro @code{FUNCTION_PROLOGUE}, and the
3210 registers to restore are determined from @code{regs_ever_live} and
3211 @code{CALL_USED_REGISTERS} in the same way.
3213 On some machines, there is a single instruction that does all the work
3214 of returning from the function. On these machines, give that
3215 instruction the name @samp{return} and do not define the macro
3216 @code{FUNCTION_EPILOGUE} at all.
3218 Do not define a pattern named @samp{return} if you want the
3219 @code{FUNCTION_EPILOGUE} to be used. If you want the target switches
3220 to control whether return instructions or epilogues are used, define a
3221 @samp{return} pattern with a validity condition that tests the target
3222 switches appropriately. If the @samp{return} pattern's validity
3223 condition is false, epilogues will be used.
3225 On machines where functions may or may not have frame-pointers, the
3226 function exit code must vary accordingly. Sometimes the code for these
3227 two cases is completely different. To determine whether a frame pointer
3228 is wanted, the macro can refer to the variable
3229 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
3230 a function that needs a frame pointer.
3232 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
3233 treat leaf functions specially. The C variable @code{leaf_function} is
3234 nonzero for such a function. @xref{Leaf Functions}.
3236 On some machines, some functions pop their arguments on exit while
3237 others leave that for the caller to do. For example, the 68020 when
3238 given @samp{-mrtd} pops arguments in functions that take a fixed
3239 number of arguments.
3241 @findex current_function_pops_args
3242 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
3243 functions pop their own arguments. @code{FUNCTION_EPILOGUE} needs to
3244 know what was decided. The variable that is called
3245 @code{current_function_pops_args} is the number of bytes of its
3246 arguments that a function should pop. @xref{Scalar Return}.
3247 @c what is the "its arguments" in the above sentence referring to, pray
3248 @c tell? --mew 5feb93
3250 @findex DELAY_SLOTS_FOR_EPILOGUE
3251 @item DELAY_SLOTS_FOR_EPILOGUE
3252 Define this macro if the function epilogue contains delay slots to which
3253 instructions from the rest of the function can be ``moved''. The
3254 definition should be a C expression whose value is an integer
3255 representing the number of delay slots there.
3257 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
3258 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
3259 A C expression that returns 1 if @var{insn} can be placed in delay
3260 slot number @var{n} of the epilogue.
3262 The argument @var{n} is an integer which identifies the delay slot now
3263 being considered (since different slots may have different rules of
3264 eligibility). It is never negative and is always less than the number
3265 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
3266 If you reject a particular insn for a given delay slot, in principle, it
3267 may be reconsidered for a subsequent delay slot. Also, other insns may
3268 (at least in principle) be considered for the so far unfilled delay
3271 @findex current_function_epilogue_delay_list
3272 @findex final_scan_insn
3273 The insns accepted to fill the epilogue delay slots are put in an RTL
3274 list made with @code{insn_list} objects, stored in the variable
3275 @code{current_function_epilogue_delay_list}. The insn for the first
3276 delay slot comes first in the list. Your definition of the macro
3277 @code{FUNCTION_EPILOGUE} should fill the delay slots by outputting the
3278 insns in this list, usually by calling @code{final_scan_insn}.
3280 You need not define this macro if you did not define
3281 @code{DELAY_SLOTS_FOR_EPILOGUE}.
3283 @findex ASM_OUTPUT_MI_THUNK
3284 @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function})
3285 A C compound statement that outputs the assembler code for a thunk
3286 function, used to implement C++ virtual function calls with multiple
3287 inheritance. The thunk acts as a wrapper around a virtual function,
3288 adjusting the implicit object parameter before handing control off to
3291 First, emit code to add the integer @var{delta} to the location that
3292 contains the incoming first argument. Assume that this argument
3293 contains a pointer, and is the one used to pass the @code{this} pointer
3294 in C++. This is the incoming argument @emph{before} the function prologue,
3295 e.g. @samp{%o0} on a sparc. The addition must preserve the values of
3296 all other incoming arguments.
3298 After the addition, emit code to jump to @var{function}, which is a
3299 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
3300 not touch the return address. Hence returning from @var{FUNCTION} will
3301 return to whoever called the current @samp{thunk}.
3303 The effect must be as if @var{function} had been called directly with
3304 the adjusted first argument. This macro is responsible for emitting all
3305 of the code for a thunk function; @code{FUNCTION_PROLOGUE} and
3306 @code{FUNCTION_EPILOGUE} are not invoked.
3308 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
3309 have already been extracted from it.) It might possibly be useful on
3310 some targets, but probably not.
3312 If you do not define this macro, the target-independent code in the C++
3313 frontend will generate a less efficient heavyweight thunk that calls
3314 @var{function} instead of jumping to it. The generic approach does
3315 not support varargs.
3319 @subsection Generating Code for Profiling
3320 @cindex profiling, code generation
3322 These macros will help you generate code for profiling.
3325 @findex FUNCTION_PROFILER
3326 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
3327 A C statement or compound statement to output to @var{file} some
3328 assembler code to call the profiling subroutine @code{mcount}.
3329 Before calling, the assembler code must load the address of a
3330 counter variable into a register where @code{mcount} expects to
3331 find the address. The name of this variable is @samp{LP} followed
3332 by the number @var{labelno}, so you would generate the name using
3333 @samp{LP%d} in a @code{fprintf}.
3336 The details of how the address should be passed to @code{mcount} are
3337 determined by your operating system environment, not by GNU CC. To
3338 figure them out, compile a small program for profiling using the
3339 system's installed C compiler and look at the assembler code that
3342 @findex PROFILE_BEFORE_PROLOGUE
3343 @item PROFILE_BEFORE_PROLOGUE
3344 Define this macro if the code for function profiling should come before
3345 the function prologue. Normally, the profiling code comes after.
3347 @findex FUNCTION_BLOCK_PROFILER
3348 @vindex profile_block_flag
3349 @item FUNCTION_BLOCK_PROFILER (@var{file}, @var{labelno})
3350 A C statement or compound statement to output to @var{file} some
3351 assembler code to initialize basic-block profiling for the current
3352 object module. The global compile flag @code{profile_block_flag}
3353 distinguishes two profile modes.
3356 @findex __bb_init_func
3357 @item profile_block_flag != 2
3358 Output code to call the subroutine @code{__bb_init_func} once per
3359 object module, passing it as its sole argument the address of a block
3360 allocated in the object module.
3362 The name of the block is a local symbol made with this statement:
3365 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3368 Of course, since you are writing the definition of
3369 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3370 can take a short cut in the definition of this macro and use the name
3371 that you know will result.
3373 The first word of this block is a flag which will be nonzero if the
3374 object module has already been initialized. So test this word first,
3375 and do not call @code{__bb_init_func} if the flag is
3376 nonzero. BLOCK_OR_LABEL contains a unique number which may be used to
3377 generate a label as a branch destination when @code{__bb_init_func}
3380 Described in assembler language, the code to be output looks like:
3390 @findex __bb_init_trace_func
3391 @item profile_block_flag == 2
3392 Output code to call the subroutine @code{__bb_init_trace_func}
3393 and pass two parameters to it. The first parameter is the same as
3394 for @code{__bb_init_func}. The second parameter is the number of the
3395 first basic block of the function as given by BLOCK_OR_LABEL. Note
3396 that @code{__bb_init_trace_func} has to be called, even if the object
3397 module has been initialized already.
3399 Described in assembler language, the code to be output looks like:
3402 parameter2 <- BLOCK_OR_LABEL
3403 call __bb_init_trace_func
3407 @findex BLOCK_PROFILER
3408 @vindex profile_block_flag
3409 @item BLOCK_PROFILER (@var{file}, @var{blockno})
3410 A C statement or compound statement to output to @var{file} some
3411 assembler code to increment the count associated with the basic
3412 block number @var{blockno}. The global compile flag
3413 @code{profile_block_flag} distinguishes two profile modes.
3416 @item profile_block_flag != 2
3417 Output code to increment the counter directly. Basic blocks are
3418 numbered separately from zero within each compilation. The count
3419 associated with block number @var{blockno} is at index
3420 @var{blockno} in a vector of words; the name of this array is a local
3421 symbol made with this statement:
3424 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 2);
3427 @c This paragraph is the same as one a few paragraphs up.
3428 @c That is not an error.
3429 Of course, since you are writing the definition of
3430 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3431 can take a short cut in the definition of this macro and use the name
3432 that you know will result.
3434 Described in assembler language, the code to be output looks like:
3437 inc (LPBX2+4*BLOCKNO)
3441 @findex __bb_trace_func
3442 @item profile_block_flag == 2
3443 Output code to initialize the global structure @code{__bb} and
3444 call the function @code{__bb_trace_func}, which will increment the
3447 @code{__bb} consists of two words. In the first word, the current
3448 basic block number, as given by BLOCKNO, has to be stored. In
3449 the second word, the address of a block allocated in the object
3450 module has to be stored. The address is given by the label created
3451 with this statement:
3454 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3457 Described in assembler language, the code to be output looks like:
3459 move BLOCKNO -> (__bb)
3460 move LPBX0 -> (__bb+4)
3461 call __bb_trace_func
3465 @findex FUNCTION_BLOCK_PROFILER_EXIT
3466 @findex __bb_trace_ret
3467 @vindex profile_block_flag
3468 @item FUNCTION_BLOCK_PROFILER_EXIT (@var{file})
3469 A C statement or compound statement to output to @var{file}
3470 assembler code to call function @code{__bb_trace_ret}. The
3471 assembler code should only be output
3472 if the global compile flag @code{profile_block_flag} == 2. This
3473 macro has to be used at every place where code for returning from
3474 a function is generated (e.g. @code{FUNCTION_EPILOGUE}). Although
3475 you have to write the definition of @code{FUNCTION_EPILOGUE}
3476 as well, you have to define this macro to tell the compiler, that
3477 the proper call to @code{__bb_trace_ret} is produced.
3479 @findex MACHINE_STATE_SAVE
3480 @findex __bb_init_trace_func
3481 @findex __bb_trace_func
3482 @findex __bb_trace_ret
3483 @item MACHINE_STATE_SAVE (@var{id})
3484 A C statement or compound statement to save all registers, which may
3485 be clobbered by a function call, including condition codes. The
3486 @code{asm} statement will be mostly likely needed to handle this
3487 task. Local labels in the assembler code can be concatenated with the
3488 string @var{id}, to obtain a unique lable name.
3490 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3491 @code{FUNCTION_EPILOGUE} must be saved in the macros
3492 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3493 @code{BLOCK_PROFILER} prior calling @code{__bb_init_trace_func},
3494 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3496 @findex MACHINE_STATE_RESTORE
3497 @findex __bb_init_trace_func
3498 @findex __bb_trace_func
3499 @findex __bb_trace_ret
3500 @item MACHINE_STATE_RESTORE (@var{id})
3501 A C statement or compound statement to restore all registers, including
3502 condition codes, saved by @code{MACHINE_STATE_SAVE}.
3504 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3505 @code{FUNCTION_EPILOGUE} must be restored in the macros
3506 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3507 @code{BLOCK_PROFILER} after calling @code{__bb_init_trace_func},
3508 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3510 @findex BLOCK_PROFILER_CODE
3511 @item BLOCK_PROFILER_CODE
3512 A C function or functions which are needed in the library to
3513 support block profiling.
3517 @section Implementing the Varargs Macros
3518 @cindex varargs implementation
3520 GNU CC comes with an implementation of @file{varargs.h} and
3521 @file{stdarg.h} that work without change on machines that pass arguments
3522 on the stack. Other machines require their own implementations of
3523 varargs, and the two machine independent header files must have
3524 conditionals to include it.
3526 ANSI @file{stdarg.h} differs from traditional @file{varargs.h} mainly in
3527 the calling convention for @code{va_start}. The traditional
3528 implementation takes just one argument, which is the variable in which
3529 to store the argument pointer. The ANSI implementation of
3530 @code{va_start} takes an additional second argument. The user is
3531 supposed to write the last named argument of the function here.
3533 However, @code{va_start} should not use this argument. The way to find
3534 the end of the named arguments is with the built-in functions described
3538 @findex __builtin_saveregs
3539 @item __builtin_saveregs ()
3540 Use this built-in function to save the argument registers in memory so
3541 that the varargs mechanism can access them. Both ANSI and traditional
3542 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3543 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
3545 On some machines, @code{__builtin_saveregs} is open-coded under the
3546 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
3547 it calls a routine written in assembler language, found in
3550 Code generated for the call to @code{__builtin_saveregs} appears at the
3551 beginning of the function, as opposed to where the call to
3552 @code{__builtin_saveregs} is written, regardless of what the code is.
3553 This is because the registers must be saved before the function starts
3554 to use them for its own purposes.
3555 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3558 @findex __builtin_args_info
3559 @item __builtin_args_info (@var{category})
3560 Use this built-in function to find the first anonymous arguments in
3563 In general, a machine may have several categories of registers used for
3564 arguments, each for a particular category of data types. (For example,
3565 on some machines, floating-point registers are used for floating-point
3566 arguments while other arguments are passed in the general registers.)
3567 To make non-varargs functions use the proper calling convention, you
3568 have defined the @code{CUMULATIVE_ARGS} data type to record how many
3569 registers in each category have been used so far
3571 @code{__builtin_args_info} accesses the same data structure of type
3572 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
3573 with it, with @var{category} specifying which word to access. Thus, the
3574 value indicates the first unused register in a given category.
3576 Normally, you would use @code{__builtin_args_info} in the implementation
3577 of @code{va_start}, accessing each category just once and storing the
3578 value in the @code{va_list} object. This is because @code{va_list} will
3579 have to update the values, and there is no way to alter the
3580 values accessed by @code{__builtin_args_info}.
3582 @findex __builtin_next_arg
3583 @item __builtin_next_arg (@var{lastarg})
3584 This is the equivalent of @code{__builtin_args_info}, for stack
3585 arguments. It returns the address of the first anonymous stack
3586 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
3587 returns the address of the location above the first anonymous stack
3588 argument. Use it in @code{va_start} to initialize the pointer for
3589 fetching arguments from the stack. Also use it in @code{va_start} to
3590 verify that the second parameter @var{lastarg} is the last named argument
3591 of the current function.
3593 @findex __builtin_classify_type
3594 @item __builtin_classify_type (@var{object})
3595 Since each machine has its own conventions for which data types are
3596 passed in which kind of register, your implementation of @code{va_arg}
3597 has to embody these conventions. The easiest way to categorize the
3598 specified data type is to use @code{__builtin_classify_type} together
3599 with @code{sizeof} and @code{__alignof__}.
3601 @code{__builtin_classify_type} ignores the value of @var{object},
3602 considering only its data type. It returns an integer describing what
3603 kind of type that is---integer, floating, pointer, structure, and so on.
3605 The file @file{typeclass.h} defines an enumeration that you can use to
3606 interpret the values of @code{__builtin_classify_type}.
3609 These machine description macros help implement varargs:
3612 @findex EXPAND_BUILTIN_SAVEREGS
3613 @item EXPAND_BUILTIN_SAVEREGS (@var{args})
3614 If defined, is a C expression that produces the machine-specific code
3615 for a call to @code{__builtin_saveregs}. This code will be moved to the
3616 very beginning of the function, before any parameter access are made.
3617 The return value of this function should be an RTX that contains the
3618 value to use as the return of @code{__builtin_saveregs}.
3620 The argument @var{args} is a @code{tree_list} containing the arguments
3621 that were passed to @code{__builtin_saveregs}.
3623 If this macro is not defined, the compiler will output an ordinary
3624 call to the library function @samp{__builtin_saveregs}.
3626 @c !!! a bug in texinfo; how to make the entry on the @item line allow
3627 @c more than one line of text... help... --mew 10feb93
3628 @findex SETUP_INCOMING_VARARGS
3629 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type},
3630 @var{pretend_args_size}, @var{second_time})
3631 This macro offers an alternative to using @code{__builtin_saveregs} and
3632 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
3633 anonymous register arguments into the stack so that all the arguments
3634 appear to have been passed consecutively on the stack. Once this is
3635 done, you can use the standard implementation of varargs that works for
3636 machines that pass all their arguments on the stack.
3638 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
3639 structure, containing the values that obtain after processing of the
3640 named arguments. The arguments @var{mode} and @var{type} describe the
3641 last named argument---its machine mode and its data type as a tree node.
3643 The macro implementation should do two things: first, push onto the
3644 stack all the argument registers @emph{not} used for the named
3645 arguments, and second, store the size of the data thus pushed into the
3646 @code{int}-valued variable whose name is supplied as the argument
3647 @var{pretend_args_size}. The value that you store here will serve as
3648 additional offset for setting up the stack frame.
3650 Because you must generate code to push the anonymous arguments at
3651 compile time without knowing their data types,
3652 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
3653 a single category of argument register and use it uniformly for all data
3656 If the argument @var{second_time} is nonzero, it means that the
3657 arguments of the function are being analyzed for the second time. This
3658 happens for an inline function, which is not actually compiled until the
3659 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
3660 not generate any instructions in this case.
3662 @findex STRICT_ARGUMENT_NAMING
3663 @item STRICT_ARGUMENT_NAMING
3664 Define this macro if the location where a function argument is passed
3665 depends on whether or not it is a named argument.
3667 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
3668 is set for varargs and stdarg functions. With this macro defined,
3669 the @var{named} argument is always true for named arguments, and false for
3670 unnamed arguments. If this is not defined, but @code{SETUP_INCOMING_VARARGS}
3671 is defined, then all arguments are treated as named. Otherwise, all named
3672 arguments except the last are treated as named.
3676 @section Trampolines for Nested Functions
3677 @cindex trampolines for nested functions
3678 @cindex nested functions, trampolines for
3680 A @dfn{trampoline} is a small piece of code that is created at run time
3681 when the address of a nested function is taken. It normally resides on
3682 the stack, in the stack frame of the containing function. These macros
3683 tell GNU CC how to generate code to allocate and initialize a
3686 The instructions in the trampoline must do two things: load a constant
3687 address into the static chain register, and jump to the real address of
3688 the nested function. On CISC machines such as the m68k, this requires
3689 two instructions, a move immediate and a jump. Then the two addresses
3690 exist in the trampoline as word-long immediate operands. On RISC
3691 machines, it is often necessary to load each address into a register in
3692 two parts. Then pieces of each address form separate immediate
3695 The code generated to initialize the trampoline must store the variable
3696 parts---the static chain value and the function address---into the
3697 immediate operands of the instructions. On a CISC machine, this is
3698 simply a matter of copying each address to a memory reference at the
3699 proper offset from the start of the trampoline. On a RISC machine, it
3700 may be necessary to take out pieces of the address and store them
3704 @findex TRAMPOLINE_TEMPLATE
3705 @item TRAMPOLINE_TEMPLATE (@var{file})
3706 A C statement to output, on the stream @var{file}, assembler code for a
3707 block of data that contains the constant parts of a trampoline. This
3708 code should not include a label---the label is taken care of
3711 If you do not define this macro, it means no template is needed
3712 for the target. Do not define this macro on systems where the block move
3713 code to copy the trampoline into place would be larger than the code
3714 to generate it on the spot.
3716 @findex TRAMPOLINE_SECTION
3717 @item TRAMPOLINE_SECTION
3718 The name of a subroutine to switch to the section in which the
3719 trampoline template is to be placed (@pxref{Sections}). The default is
3720 a value of @samp{readonly_data_section}, which places the trampoline in
3721 the section containing read-only data.
3723 @findex TRAMPOLINE_SIZE
3724 @item TRAMPOLINE_SIZE
3725 A C expression for the size in bytes of the trampoline, as an integer.
3727 @findex TRAMPOLINE_ALIGNMENT
3728 @item TRAMPOLINE_ALIGNMENT
3729 Alignment required for trampolines, in bits.
3731 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
3732 is used for aligning trampolines.
3734 @findex INITIALIZE_TRAMPOLINE
3735 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
3736 A C statement to initialize the variable parts of a trampoline.
3737 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
3738 an RTX for the address of the nested function; @var{static_chain} is an
3739 RTX for the static chain value that should be passed to the function
3742 @findex ALLOCATE_TRAMPOLINE
3743 @item ALLOCATE_TRAMPOLINE (@var{fp})
3744 A C expression to allocate run-time space for a trampoline. The
3745 expression value should be an RTX representing a memory reference to the
3746 space for the trampoline.
3748 @cindex @code{FUNCTION_EPILOGUE} and trampolines
3749 @cindex @code{FUNCTION_PROLOGUE} and trampolines
3750 If this macro is not defined, by default the trampoline is allocated as
3751 a stack slot. This default is right for most machines. The exceptions
3752 are machines where it is impossible to execute instructions in the stack
3753 area. On such machines, you may have to implement a separate stack,
3754 using this macro in conjunction with @code{FUNCTION_PROLOGUE} and
3755 @code{FUNCTION_EPILOGUE}.
3757 @var{fp} points to a data structure, a @code{struct function}, which
3758 describes the compilation status of the immediate containing function of
3759 the function which the trampoline is for. Normally (when
3760 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
3761 trampoline is in the stack frame of this containing function. Other
3762 allocation strategies probably must do something analogous with this
3766 Implementing trampolines is difficult on many machines because they have
3767 separate instruction and data caches. Writing into a stack location
3768 fails to clear the memory in the instruction cache, so when the program
3769 jumps to that location, it executes the old contents.
3771 Here are two possible solutions. One is to clear the relevant parts of
3772 the instruction cache whenever a trampoline is set up. The other is to
3773 make all trampolines identical, by having them jump to a standard
3774 subroutine. The former technique makes trampoline execution faster; the
3775 latter makes initialization faster.
3777 To clear the instruction cache when a trampoline is initialized, define
3778 the following macros which describe the shape of the cache.
3781 @findex INSN_CACHE_SIZE
3782 @item INSN_CACHE_SIZE
3783 The total size in bytes of the cache.
3785 @findex INSN_CACHE_LINE_WIDTH
3786 @item INSN_CACHE_LINE_WIDTH
3787 The length in bytes of each cache line. The cache is divided into cache
3788 lines which are disjoint slots, each holding a contiguous chunk of data
3789 fetched from memory. Each time data is brought into the cache, an
3790 entire line is read at once. The data loaded into a cache line is
3791 always aligned on a boundary equal to the line size.
3793 @findex INSN_CACHE_DEPTH
3794 @item INSN_CACHE_DEPTH
3795 The number of alternative cache lines that can hold any particular memory
3799 Alternatively, if the machine has system calls or instructions to clear
3800 the instruction cache directly, you can define the following macro.
3803 @findex CLEAR_INSN_CACHE
3804 @item CLEAR_INSN_CACHE (@var{BEG}, @var{END})
3805 If defined, expands to a C expression clearing the @emph{instruction
3806 cache} in the specified interval. If it is not defined, and the macro
3807 INSN_CACHE_SIZE is defined, some generic code is generated to clear the
3808 cache. The definition of this macro would typically be a series of
3809 @code{asm} statements. Both @var{BEG} and @var{END} are both pointer
3813 To use a standard subroutine, define the following macro. In addition,
3814 you must make sure that the instructions in a trampoline fill an entire
3815 cache line with identical instructions, or else ensure that the
3816 beginning of the trampoline code is always aligned at the same point in
3817 its cache line. Look in @file{m68k.h} as a guide.
3820 @findex TRANSFER_FROM_TRAMPOLINE
3821 @item TRANSFER_FROM_TRAMPOLINE
3822 Define this macro if trampolines need a special subroutine to do their
3823 work. The macro should expand to a series of @code{asm} statements
3824 which will be compiled with GNU CC. They go in a library function named
3825 @code{__transfer_from_trampoline}.
3827 If you need to avoid executing the ordinary prologue code of a compiled
3828 C function when you jump to the subroutine, you can do so by placing a
3829 special label of your own in the assembler code. Use one @code{asm}
3830 statement to generate an assembler label, and another to make the label
3831 global. Then trampolines can use that label to jump directly to your
3832 special assembler code.
3836 @section Implicit Calls to Library Routines
3837 @cindex library subroutine names
3838 @cindex @file{libgcc.a}
3840 @c prevent bad page break with this line
3841 Here is an explanation of implicit calls to library routines.
3844 @findex MULSI3_LIBCALL
3845 @item MULSI3_LIBCALL
3846 A C string constant giving the name of the function to call for
3847 multiplication of one signed full-word by another. If you do not
3848 define this macro, the default name is used, which is @code{__mulsi3},
3849 a function defined in @file{libgcc.a}.
3851 @findex DIVSI3_LIBCALL
3852 @item DIVSI3_LIBCALL
3853 A C string constant giving the name of the function to call for
3854 division of one signed full-word by another. If you do not define
3855 this macro, the default name is used, which is @code{__divsi3}, a
3856 function defined in @file{libgcc.a}.
3858 @findex UDIVSI3_LIBCALL
3859 @item UDIVSI3_LIBCALL
3860 A C string constant giving the name of the function to call for
3861 division of one unsigned full-word by another. If you do not define
3862 this macro, the default name is used, which is @code{__udivsi3}, a
3863 function defined in @file{libgcc.a}.
3865 @findex MODSI3_LIBCALL
3866 @item MODSI3_LIBCALL
3867 A C string constant giving the name of the function to call for the
3868 remainder in division of one signed full-word by another. If you do
3869 not define this macro, the default name is used, which is
3870 @code{__modsi3}, a function defined in @file{libgcc.a}.
3872 @findex UMODSI3_LIBCALL
3873 @item UMODSI3_LIBCALL
3874 A C string constant giving the name of the function to call for the
3875 remainder in division of one unsigned full-word by another. If you do
3876 not define this macro, the default name is used, which is
3877 @code{__umodsi3}, a function defined in @file{libgcc.a}.
3879 @findex MULDI3_LIBCALL
3880 @item MULDI3_LIBCALL
3881 A C string constant giving the name of the function to call for
3882 multiplication of one signed double-word by another. If you do not
3883 define this macro, the default name is used, which is @code{__muldi3},
3884 a function defined in @file{libgcc.a}.
3886 @findex DIVDI3_LIBCALL
3887 @item DIVDI3_LIBCALL
3888 A C string constant giving the name of the function to call for
3889 division of one signed double-word by another. If you do not define
3890 this macro, the default name is used, which is @code{__divdi3}, a
3891 function defined in @file{libgcc.a}.
3893 @findex UDIVDI3_LIBCALL
3894 @item UDIVDI3_LIBCALL
3895 A C string constant giving the name of the function to call for
3896 division of one unsigned full-word by another. If you do not define
3897 this macro, the default name is used, which is @code{__udivdi3}, a
3898 function defined in @file{libgcc.a}.
3900 @findex MODDI3_LIBCALL
3901 @item MODDI3_LIBCALL
3902 A C string constant giving the name of the function to call for the
3903 remainder in division of one signed double-word by another. If you do
3904 not define this macro, the default name is used, which is
3905 @code{__moddi3}, a function defined in @file{libgcc.a}.
3907 @findex UMODDI3_LIBCALL
3908 @item UMODDI3_LIBCALL
3909 A C string constant giving the name of the function to call for the
3910 remainder in division of one unsigned full-word by another. If you do
3911 not define this macro, the default name is used, which is
3912 @code{__umoddi3}, a function defined in @file{libgcc.a}.
3914 @findex INIT_TARGET_OPTABS
3915 @item INIT_TARGET_OPTABS
3916 Define this macro as a C statement that declares additional library
3917 routines renames existing ones. @code{init_optabs} calls this macro after
3918 initializing all the normal library routines.
3921 @cindex @code{EDOM}, implicit usage
3923 The value of @code{EDOM} on the target machine, as a C integer constant
3924 expression. If you don't define this macro, GNU CC does not attempt to
3925 deposit the value of @code{EDOM} into @code{errno} directly. Look in
3926 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
3929 If you do not define @code{TARGET_EDOM}, then compiled code reports
3930 domain errors by calling the library function and letting it report the
3931 error. If mathematical functions on your system use @code{matherr} when
3932 there is an error, then you should leave @code{TARGET_EDOM} undefined so
3933 that @code{matherr} is used normally.
3935 @findex GEN_ERRNO_RTX
3936 @cindex @code{errno}, implicit usage
3938 Define this macro as a C expression to create an rtl expression that
3939 refers to the global ``variable'' @code{errno}. (On certain systems,
3940 @code{errno} may not actually be a variable.) If you don't define this
3941 macro, a reasonable default is used.
3943 @findex TARGET_MEM_FUNCTIONS
3944 @cindex @code{bcopy}, implicit usage
3945 @cindex @code{memcpy}, implicit usage
3946 @cindex @code{bzero}, implicit usage
3947 @cindex @code{memset}, implicit usage
3948 @item TARGET_MEM_FUNCTIONS
3949 Define this macro if GNU CC should generate calls to the System V
3950 (and ANSI C) library functions @code{memcpy} and @code{memset}
3951 rather than the BSD functions @code{bcopy} and @code{bzero}.
3953 @findex LIBGCC_NEEDS_DOUBLE
3954 @item LIBGCC_NEEDS_DOUBLE
3955 Define this macro if only @code{float} arguments cannot be passed to
3956 library routines (so they must be converted to @code{double}). This
3957 macro affects both how library calls are generated and how the library
3958 routines in @file{libgcc1.c} accept their arguments. It is useful on
3959 machines where floating and fixed point arguments are passed
3960 differently, such as the i860.
3962 @findex FLOAT_ARG_TYPE
3963 @item FLOAT_ARG_TYPE
3964 Define this macro to override the type used by the library routines to
3965 pick up arguments of type @code{float}. (By default, they use a union
3966 of @code{float} and @code{int}.)
3968 The obvious choice would be @code{float}---but that won't work with
3969 traditional C compilers that expect all arguments declared as @code{float}
3970 to arrive as @code{double}. To avoid this conversion, the library routines
3971 ask for the value as some other type and then treat it as a @code{float}.
3973 On some systems, no other type will work for this. For these systems,
3974 you must use @code{LIBGCC_NEEDS_DOUBLE} instead, to force conversion of
3975 the values @code{double} before they are passed.
3978 @item FLOATIFY (@var{passed-value})
3979 Define this macro to override the way library routines redesignate a
3980 @code{float} argument as a @code{float} instead of the type it was
3981 passed as. The default is an expression which takes the @code{float}
3984 @findex FLOAT_VALUE_TYPE
3985 @item FLOAT_VALUE_TYPE
3986 Define this macro to override the type used by the library routines to
3987 return values that ought to have type @code{float}. (By default, they
3990 The obvious choice would be @code{float}---but that won't work with
3991 traditional C compilers gratuitously convert values declared as
3992 @code{float} into @code{double}.
3995 @item INTIFY (@var{float-value})
3996 Define this macro to override the way the value of a
3997 @code{float}-returning library routine should be packaged in order to
3998 return it. These functions are actually declared to return type
3999 @code{FLOAT_VALUE_TYPE} (normally @code{int}).
4001 These values can't be returned as type @code{float} because traditional
4002 C compilers would gratuitously convert the value to a @code{double}.
4004 A local variable named @code{intify} is always available when the macro
4005 @code{INTIFY} is used. It is a union of a @code{float} field named
4006 @code{f} and a field named @code{i} whose type is
4007 @code{FLOAT_VALUE_TYPE} or @code{int}.
4009 If you don't define this macro, the default definition works by copying
4010 the value through that union.
4012 @findex nongcc_SI_type
4013 @item nongcc_SI_type
4014 Define this macro as the name of the data type corresponding to
4015 @code{SImode} in the system's own C compiler.
4017 You need not define this macro if that type is @code{long int}, as it usually
4020 @findex nongcc_word_type
4021 @item nongcc_word_type
4022 Define this macro as the name of the data type corresponding to the
4023 word_mode in the system's own C compiler.
4025 You need not define this macro if that type is @code{long int}, as it usually
4028 @findex perform_@dots{}
4029 @item perform_@dots{}
4030 Define these macros to supply explicit C statements to carry out various
4031 arithmetic operations on types @code{float} and @code{double} in the
4032 library routines in @file{libgcc1.c}. See that file for a full list
4033 of these macros and their arguments.
4035 On most machines, you don't need to define any of these macros, because
4036 the C compiler that comes with the system takes care of doing them.
4038 @findex NEXT_OBJC_RUNTIME
4039 @item NEXT_OBJC_RUNTIME
4040 Define this macro to generate code for Objective C message sending using
4041 the calling convention of the NeXT system. This calling convention
4042 involves passing the object, the selector and the method arguments all
4043 at once to the method-lookup library function.
4045 The default calling convention passes just the object and the selector
4046 to the lookup function, which returns a pointer to the method.
4049 @node Addressing Modes
4050 @section Addressing Modes
4051 @cindex addressing modes
4053 @c prevent bad page break with this line
4054 This is about addressing modes.
4057 @findex HAVE_POST_INCREMENT
4058 @item HAVE_POST_INCREMENT
4059 Define this macro if the machine supports post-increment addressing.
4061 @findex HAVE_PRE_INCREMENT
4062 @findex HAVE_POST_DECREMENT
4063 @findex HAVE_PRE_DECREMENT
4064 @item HAVE_PRE_INCREMENT
4065 @itemx HAVE_POST_DECREMENT
4066 @itemx HAVE_PRE_DECREMENT
4067 Similar for other kinds of addressing.
4069 @findex CONSTANT_ADDRESS_P
4070 @item CONSTANT_ADDRESS_P (@var{x})
4071 A C expression that is 1 if the RTX @var{x} is a constant which
4072 is a valid address. On most machines, this can be defined as
4073 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4074 in which constant addresses are supported.
4077 @code{CONSTANT_P} accepts integer-values expressions whose values are
4078 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4079 @code{high} expressions and @code{const} arithmetic expressions, in
4080 addition to @code{const_int} and @code{const_double} expressions.
4082 @findex MAX_REGS_PER_ADDRESS
4083 @item MAX_REGS_PER_ADDRESS
4084 A number, the maximum number of registers that can appear in a valid
4085 memory address. Note that it is up to you to specify a value equal to
4086 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4089 @findex GO_IF_LEGITIMATE_ADDRESS
4090 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4091 A C compound statement with a conditional @code{goto @var{label};}
4092 executed if @var{x} (an RTX) is a legitimate memory address on the
4093 target machine for a memory operand of mode @var{mode}.
4095 It usually pays to define several simpler macros to serve as
4096 subroutines for this one. Otherwise it may be too complicated to
4099 This macro must exist in two variants: a strict variant and a
4100 non-strict one. The strict variant is used in the reload pass. It
4101 must be defined so that any pseudo-register that has not been
4102 allocated a hard register is considered a memory reference. In
4103 contexts where some kind of register is required, a pseudo-register
4104 with no hard register must be rejected.
4106 The non-strict variant is used in other passes. It must be defined to
4107 accept all pseudo-registers in every context where some kind of
4108 register is required.
4110 @findex REG_OK_STRICT
4111 Compiler source files that want to use the strict variant of this
4112 macro define the macro @code{REG_OK_STRICT}. You should use an
4113 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4114 in that case and the non-strict variant otherwise.
4116 Subroutines to check for acceptable registers for various purposes (one
4117 for base registers, one for index registers, and so on) are typically
4118 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4119 Then only these subroutine macros need have two variants; the higher
4120 levels of macros may be the same whether strict or not.@refill
4122 Normally, constant addresses which are the sum of a @code{symbol_ref}
4123 and an integer are stored inside a @code{const} RTX to mark them as
4124 constant. Therefore, there is no need to recognize such sums
4125 specifically as legitimate addresses. Normally you would simply
4126 recognize any @code{const} as legitimate.
4128 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4129 sums that are not marked with @code{const}. It assumes that a naked
4130 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4131 naked constant sums as illegitimate addresses, so that none of them will
4132 be given to @code{PRINT_OPERAND_ADDRESS}.
4134 @cindex @code{ENCODE_SECTION_INFO} and address validation
4135 On some machines, whether a symbolic address is legitimate depends on
4136 the section that the address refers to. On these machines, define the
4137 macro @code{ENCODE_SECTION_INFO} to store the information into the
4138 @code{symbol_ref}, and then check for it here. When you see a
4139 @code{const}, you will have to look inside it to find the
4140 @code{symbol_ref} in order to determine the section. @xref{Assembler
4143 @findex saveable_obstack
4144 The best way to modify the name string is by adding text to the
4145 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4146 the new name in @code{saveable_obstack}. You will have to modify
4147 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4148 output the name accordingly, and define @code{STRIP_NAME_ENCODING} to
4149 access the original name string.
4151 You can check the information stored here into the @code{symbol_ref} in
4152 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4153 @code{PRINT_OPERAND_ADDRESS}.
4155 @findex REG_OK_FOR_BASE_P
4156 @item REG_OK_FOR_BASE_P (@var{x})
4157 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4158 RTX) is valid for use as a base register. For hard registers, it
4159 should always accept those which the hardware permits and reject the
4160 others. Whether the macro accepts or rejects pseudo registers must be
4161 controlled by @code{REG_OK_STRICT} as described above. This usually
4162 requires two variant definitions, of which @code{REG_OK_STRICT}
4163 controls the one actually used.
4165 @findex REG_MODE_OK_FOR_BASE_P
4166 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4167 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4168 that expression may examine the mode of the memory reference in
4169 @var{mode}. You should define this macro if the mode of the memory
4170 reference affects whether a register may be used as a base register. If
4171 you define this macro, the compiler will use it instead of
4172 @code{REG_OK_FOR_BASE_P}.
4174 @findex REG_OK_FOR_INDEX_P
4175 @item REG_OK_FOR_INDEX_P (@var{x})
4176 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4177 RTX) is valid for use as an index register.
4179 The difference between an index register and a base register is that
4180 the index register may be scaled. If an address involves the sum of
4181 two registers, neither one of them scaled, then either one may be
4182 labeled the ``base'' and the other the ``index''; but whichever
4183 labeling is used must fit the machine's constraints of which registers
4184 may serve in each capacity. The compiler will try both labelings,
4185 looking for one that is valid, and will reload one or both registers
4186 only if neither labeling works.
4188 @findex LEGITIMIZE_ADDRESS
4189 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4190 A C compound statement that attempts to replace @var{x} with a valid
4191 memory address for an operand of mode @var{mode}. @var{win} will be a
4192 C statement label elsewhere in the code; the macro definition may use
4195 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4199 to avoid further processing if the address has become legitimate.
4201 @findex break_out_memory_refs
4202 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4203 and @var{oldx} will be the operand that was given to that function to produce
4206 The code generated by this macro should not alter the substructure of
4207 @var{x}. If it transforms @var{x} into a more legitimate form, it
4208 should assign @var{x} (which will always be a C variable) a new value.
4210 It is not necessary for this macro to come up with a legitimate
4211 address. The compiler has standard ways of doing so in all cases. In
4212 fact, it is safe for this macro to do nothing. But often a
4213 machine-dependent strategy can generate better code.
4215 @findex GO_IF_MODE_DEPENDENT_ADDRESS
4216 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4217 A C statement or compound statement with a conditional @code{goto
4218 @var{label};} executed if memory address @var{x} (an RTX) can have
4219 different meanings depending on the machine mode of the memory
4220 reference it is used for or if the address is valid for some modes
4223 Autoincrement and autodecrement addresses typically have mode-dependent
4224 effects because the amount of the increment or decrement is the size
4225 of the operand being addressed. Some machines have other mode-dependent
4226 addresses. Many RISC machines have no mode-dependent addresses.
4228 You may assume that @var{addr} is a valid address for the machine.
4230 @findex LEGITIMATE_CONSTANT_P
4231 @item LEGITIMATE_CONSTANT_P (@var{x})
4232 A C expression that is nonzero if @var{x} is a legitimate constant for
4233 an immediate operand on the target machine. You can assume that
4234 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4235 @samp{1} is a suitable definition for this macro on machines where
4236 anything @code{CONSTANT_P} is valid.@refill
4238 @findex DONT_RECORD_EQUIVALENCE
4239 @item DONT_RECORD_EQUIVALENCE (@var{note})
4240 A C expression that is nonzero if the @code{REG_EQUAL} note @var{x} should not
4241 be promoted to a @code{REG_EQUIV} note.
4243 Define this macro if @var{note} refers to a constant that must be accepted
4244 by @code{LEGITIMATE_CONSTANT_P}, but must not appear as an immediate operand.
4246 Most machine descriptions do not need to define this macro.
4249 @node Condition Code
4250 @section Condition Code Status
4251 @cindex condition code status
4253 @c prevent bad page break with this line
4254 This describes the condition code status.
4257 The file @file{conditions.h} defines a variable @code{cc_status} to
4258 describe how the condition code was computed (in case the interpretation of
4259 the condition code depends on the instruction that it was set by). This
4260 variable contains the RTL expressions on which the condition code is
4261 currently based, and several standard flags.
4263 Sometimes additional machine-specific flags must be defined in the machine
4264 description header file. It can also add additional machine-specific
4265 information by defining @code{CC_STATUS_MDEP}.
4268 @findex CC_STATUS_MDEP
4269 @item CC_STATUS_MDEP
4270 C code for a data type which is used for declaring the @code{mdep}
4271 component of @code{cc_status}. It defaults to @code{int}.
4273 This macro is not used on machines that do not use @code{cc0}.
4275 @findex CC_STATUS_MDEP_INIT
4276 @item CC_STATUS_MDEP_INIT
4277 A C expression to initialize the @code{mdep} field to ``empty''.
4278 The default definition does nothing, since most machines don't use
4279 the field anyway. If you want to use the field, you should probably
4280 define this macro to initialize it.
4282 This macro is not used on machines that do not use @code{cc0}.
4284 @findex NOTICE_UPDATE_CC
4285 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4286 A C compound statement to set the components of @code{cc_status}
4287 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4288 this macro's responsibility to recognize insns that set the condition
4289 code as a byproduct of other activity as well as those that explicitly
4292 This macro is not used on machines that do not use @code{cc0}.
4294 If there are insns that do not set the condition code but do alter
4295 other machine registers, this macro must check to see whether they
4296 invalidate the expressions that the condition code is recorded as
4297 reflecting. For example, on the 68000, insns that store in address
4298 registers do not set the condition code, which means that usually
4299 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4300 insns. But suppose that the previous insn set the condition code
4301 based on location @samp{a4@@(102)} and the current insn stores a new
4302 value in @samp{a4}. Although the condition code is not changed by
4303 this, it will no longer be true that it reflects the contents of
4304 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4305 @code{cc_status} in this case to say that nothing is known about the
4306 condition code value.
4308 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4309 with the results of peephole optimization: insns whose patterns are
4310 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4311 constants which are just the operands. The RTL structure of these
4312 insns is not sufficient to indicate what the insns actually do. What
4313 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4314 @code{CC_STATUS_INIT}.
4316 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4317 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4318 @samp{cc}. This avoids having detailed information about patterns in
4319 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4321 @findex EXTRA_CC_MODES
4322 @item EXTRA_CC_MODES
4323 A list of names to be used for additional modes for condition code
4324 values in registers (@pxref{Jump Patterns}). These names are added
4325 to @code{enum machine_mode} and all have class @code{MODE_CC}. By
4326 convention, they should start with @samp{CC} and end with @samp{mode}.
4328 You should only define this macro if your machine does not use @code{cc0}
4329 and only if additional modes are required.
4331 @findex EXTRA_CC_NAMES
4332 @item EXTRA_CC_NAMES
4333 A list of C strings giving the names for the modes listed in
4334 @code{EXTRA_CC_MODES}. For example, the Sparc defines this macro and
4335 @code{EXTRA_CC_MODES} as
4338 #define EXTRA_CC_MODES CC_NOOVmode, CCFPmode, CCFPEmode
4339 #define EXTRA_CC_NAMES "CC_NOOV", "CCFP", "CCFPE"
4342 This macro is not required if @code{EXTRA_CC_MODES} is not defined.
4344 @findex SELECT_CC_MODE
4345 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4346 Returns a mode from class @code{MODE_CC} to be used when comparison
4347 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
4348 example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
4349 @pxref{Jump Patterns} for a description of the reason for this
4353 #define SELECT_CC_MODE(OP,X,Y) \
4354 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4355 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
4356 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4357 || GET_CODE (X) == NEG) \
4358 ? CC_NOOVmode : CCmode))
4361 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
4363 @findex CANONICALIZE_COMPARISON
4364 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
4365 One some machines not all possible comparisons are defined, but you can
4366 convert an invalid comparison into a valid one. For example, the Alpha
4367 does not have a @code{GT} comparison, but you can use an @code{LT}
4368 comparison instead and swap the order of the operands.
4370 On such machines, define this macro to be a C statement to do any
4371 required conversions. @var{code} is the initial comparison code
4372 and @var{op0} and @var{op1} are the left and right operands of the
4373 comparison, respectively. You should modify @var{code}, @var{op0}, and
4374 @var{op1} as required.
4376 GNU CC will not assume that the comparison resulting from this macro is
4377 valid but will see if the resulting insn matches a pattern in the
4380 You need not define this macro if it would never change the comparison
4383 @findex REVERSIBLE_CC_MODE
4384 @item REVERSIBLE_CC_MODE (@var{mode})
4385 A C expression whose value is one if it is always safe to reverse a
4386 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4387 can ever return @var{mode} for a floating-point inequality comparison,
4388 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4390 You need not define this macro if it would always returns zero or if the
4391 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4392 For example, here is the definition used on the Sparc, where floating-point
4393 inequality comparisons are always given @code{CCFPEmode}:
4396 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
4402 @section Describing Relative Costs of Operations
4403 @cindex costs of instructions
4404 @cindex relative costs
4405 @cindex speed of instructions
4407 These macros let you describe the relative speed of various operations
4408 on the target machine.
4412 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
4413 A part of a C @code{switch} statement that describes the relative costs
4414 of constant RTL expressions. It must contain @code{case} labels for
4415 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
4416 @code{label_ref} and @code{const_double}. Each case must ultimately
4417 reach a @code{return} statement to return the relative cost of the use
4418 of that kind of constant value in an expression. The cost may depend on
4419 the precise value of the constant, which is available for examination in
4420 @var{x}, and the rtx code of the expression in which it is contained,
4421 found in @var{outer_code}.
4423 @var{code} is the expression code---redundant, since it can be
4424 obtained with @code{GET_CODE (@var{x})}.
4427 @findex COSTS_N_INSNS
4428 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4429 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
4430 This can be used, for example, to indicate how costly a multiply
4431 instruction is. In writing this macro, you can use the construct
4432 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
4433 instructions. @var{outer_code} is the code of the expression in which
4434 @var{x} is contained.
4436 This macro is optional; do not define it if the default cost assumptions
4437 are adequate for the target machine.
4439 @findex DEFAULT_RTX_COSTS
4440 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4441 This macro, if defined, is called for any case not handled by the
4442 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
4443 to put case labels into the macro, but the code, or any functions it
4444 calls, must assume that the RTL in @var{x} could be of any type that has
4445 not already been handled. The arguments are the same as for
4446 @code{RTX_COSTS}, and the macro should execute a return statement giving
4447 the cost of any RTL expressions that it can handle. The default cost
4448 calculation is used for any RTL for which this macro does not return a
4451 This macro is optional; do not define it if the default cost assumptions
4452 are adequate for the target machine.
4454 @findex ADDRESS_COST
4455 @item ADDRESS_COST (@var{address})
4456 An expression giving the cost of an addressing mode that contains
4457 @var{address}. If not defined, the cost is computed from
4458 the @var{address} expression and the @code{CONST_COSTS} values.
4460 For most CISC machines, the default cost is a good approximation of the
4461 true cost of the addressing mode. However, on RISC machines, all
4462 instructions normally have the same length and execution time. Hence
4463 all addresses will have equal costs.
4465 In cases where more than one form of an address is known, the form with
4466 the lowest cost will be used. If multiple forms have the same, lowest,
4467 cost, the one that is the most complex will be used.
4469 For example, suppose an address that is equal to the sum of a register
4470 and a constant is used twice in the same basic block. When this macro
4471 is not defined, the address will be computed in a register and memory
4472 references will be indirect through that register. On machines where
4473 the cost of the addressing mode containing the sum is no higher than
4474 that of a simple indirect reference, this will produce an additional
4475 instruction and possibly require an additional register. Proper
4476 specification of this macro eliminates this overhead for such machines.
4478 Similar use of this macro is made in strength reduction of loops.
4480 @var{address} need not be valid as an address. In such a case, the cost
4481 is not relevant and can be any value; invalid addresses need not be
4482 assigned a different cost.
4484 On machines where an address involving more than one register is as
4485 cheap as an address computation involving only one register, defining
4486 @code{ADDRESS_COST} to reflect this can cause two registers to be live
4487 over a region of code where only one would have been if
4488 @code{ADDRESS_COST} were not defined in that manner. This effect should
4489 be considered in the definition of this macro. Equivalent costs should
4490 probably only be given to addresses with different numbers of registers
4491 on machines with lots of registers.
4493 This macro will normally either not be defined or be defined as a
4496 @findex REGISTER_MOVE_COST
4497 @item REGISTER_MOVE_COST (@var{from}, @var{to})
4498 A C expression for the cost of moving data from a register in class
4499 @var{from} to one in class @var{to}. The classes are expressed using
4500 the enumeration values such as @code{GENERAL_REGS}. A value of 2 is the
4501 default; other values are interpreted relative to that.
4503 It is not required that the cost always equal 2 when @var{from} is the
4504 same as @var{to}; on some machines it is expensive to move between
4505 registers if they are not general registers.
4507 If reload sees an insn consisting of a single @code{set} between two
4508 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4509 classes returns a value of 2, reload does not check to ensure that the
4510 constraints of the insn are met. Setting a cost of other than 2 will
4511 allow reload to verify that the constraints are met. You should do this
4512 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4514 @findex MEMORY_MOVE_COST
4515 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4516 A C expression for the cost of moving data of mode @var{mode} between a
4517 register of class @var{class} and memory; @var{in} is zero if the value
4518 is to be written to memory, non-zero if it is to be read in. If this
4519 macro is not defined, the default cost is assumed to be 4, plus any costs
4520 that would be incurred copying via a secondary reload register, if
4521 needed. This cost is relative to those in @code{REGISTER_MOVE_COST}.
4523 If moving between registers and memory is more expensive than between
4524 two registers, you should define this macro to express the relative cost.
4526 If a secondary reload register would be required for @var{class}, but the
4527 reload mechanism is more complex than copying via an intermediate, this
4528 macro should be defined to reflect the actual cost of the move.
4530 The function @code{memory_move_secondary_cost}, which is defined if
4531 secondary reloads are needed, will compute the costs due to copying; you
4532 can use this function if you need to take other factors into account as
4533 well, or if the default base value of 4 is not correct for your machine.
4537 A C expression for the cost of a branch instruction. A value of 1 is
4538 the default; other values are interpreted relative to that.
4541 Here are additional macros which do not specify precise relative costs,
4542 but only that certain actions are more expensive than GNU CC would
4546 @findex SLOW_BYTE_ACCESS
4547 @item SLOW_BYTE_ACCESS
4548 Define this macro as a C expression which is nonzero if accessing less
4549 than a word of memory (i.e. a @code{char} or a @code{short}) is no
4550 faster than accessing a word of memory, i.e., if such access
4551 require more than one instruction or if there is no difference in cost
4552 between byte and (aligned) word loads.
4554 When this macro is not defined, the compiler will access a field by
4555 finding the smallest containing object; when it is defined, a fullword
4556 load will be used if alignment permits. Unless bytes accesses are
4557 faster than word accesses, using word accesses is preferable since it
4558 may eliminate subsequent memory access if subsequent accesses occur to
4559 other fields in the same word of the structure, but to different bytes.
4561 @findex SLOW_ZERO_EXTEND
4562 @item SLOW_ZERO_EXTEND
4563 Define this macro if zero-extension (of a @code{char} or @code{short}
4564 to an @code{int}) can be done faster if the destination is a register
4565 that is known to be zero.
4567 If you define this macro, you must have instruction patterns that
4568 recognize RTL structures like this:
4571 (set (strict_low_part (subreg:QI (reg:SI @dots{}) 0)) @dots{})
4575 and likewise for @code{HImode}.
4577 @findex SLOW_UNALIGNED_ACCESS
4578 @item SLOW_UNALIGNED_ACCESS
4579 Define this macro to be the value 1 if unaligned accesses have a cost
4580 many times greater than aligned accesses, for example if they are
4581 emulated in a trap handler.
4583 When this macro is non-zero, the compiler will act as if
4584 @code{STRICT_ALIGNMENT} were non-zero when generating code for block
4585 moves. This can cause significantly more instructions to be produced.
4586 Therefore, do not set this macro non-zero if unaligned accesses only add a
4587 cycle or two to the time for a memory access.
4589 If the value of this macro is always zero, it need not be defined.
4591 @findex DONT_REDUCE_ADDR
4592 @item DONT_REDUCE_ADDR
4593 Define this macro to inhibit strength reduction of memory addresses.
4594 (On some machines, such strength reduction seems to do harm rather
4599 The number of scalar move insns which should be generated instead of a
4600 string move insn or a library call. Increasing the value will always
4601 make code faster, but eventually incurs high cost in increased code size.
4603 If you don't define this, a reasonable default is used.
4605 @findex NO_FUNCTION_CSE
4606 @item NO_FUNCTION_CSE
4607 Define this macro if it is as good or better to call a constant
4608 function address than to call an address kept in a register.
4610 @findex NO_RECURSIVE_FUNCTION_CSE
4611 @item NO_RECURSIVE_FUNCTION_CSE
4612 Define this macro if it is as good or better for a function to call
4613 itself with an explicit address than to call an address kept in a
4617 @item ADJUST_COST (@var{insn}, @var{link}, @var{dep_insn}, @var{cost})
4618 A C statement (sans semicolon) to update the integer variable @var{cost}
4619 based on the relationship between @var{insn} that is dependent on
4620 @var{dep_insn} through the dependence @var{link}. The default is to
4621 make no adjustment to @var{cost}. This can be used for example to
4622 specify to the scheduler that an output- or anti-dependence does not
4623 incur the same cost as a data-dependence.
4625 @findex ADJUST_PRIORITY
4626 @item ADJUST_PRIORITY (@var{insn})
4627 A C statement (sans semicolon) to update the integer scheduling
4628 priority @code{INSN_PRIORITY(@var{insn})}. Reduce the priority
4629 to execute the @var{insn} earlier, increase the priority to execute
4630 @var{insn} later. Do not define this macro if you do not need to
4631 adjust the scheduling priorities of insns.
4635 @section Dividing the Output into Sections (Texts, Data, @dots{})
4636 @c the above section title is WAY too long. maybe cut the part between
4637 @c the (...)? --mew 10feb93
4639 An object file is divided into sections containing different types of
4640 data. In the most common case, there are three sections: the @dfn{text
4641 section}, which holds instructions and read-only data; the @dfn{data
4642 section}, which holds initialized writable data; and the @dfn{bss
4643 section}, which holds uninitialized data. Some systems have other kinds
4646 The compiler must tell the assembler when to switch sections. These
4647 macros control what commands to output to tell the assembler this. You
4648 can also define additional sections.
4651 @findex TEXT_SECTION_ASM_OP
4652 @item TEXT_SECTION_ASM_OP
4653 A C expression whose value is a string containing the assembler
4654 operation that should precede instructions and read-only data. Normally
4655 @code{".text"} is right.
4657 @findex DATA_SECTION_ASM_OP
4658 @item DATA_SECTION_ASM_OP
4659 A C expression whose value is a string containing the assembler
4660 operation to identify the following data as writable initialized data.
4661 Normally @code{".data"} is right.
4663 @findex SHARED_SECTION_ASM_OP
4664 @item SHARED_SECTION_ASM_OP
4665 If defined, a C expression whose value is a string containing the
4666 assembler operation to identify the following data as shared data. If
4667 not defined, @code{DATA_SECTION_ASM_OP} will be used.
4669 @findex BSS_SECTION_ASM_OP
4670 @item BSS_SECTION_ASM_OP
4671 If defined, a C expression whose value is a string containing the
4672 assembler operation to identify the following data as uninitialized global
4673 data. If not defined, and neither @code{ASM_OUTPUT_BSS} nor
4674 @code{ASM_OUTPUT_ALIGNED_BSS} are defined, uninitialized global data will be
4675 output in the data section if @samp{-fno-common} is passed, otherwise
4676 @code{ASM_OUTPUT_COMMON} will be used.
4678 @findex SHARED_BSS_SECTION_ASM_OP
4679 @item SHARED_BSS_SECTION_ASM_OP
4680 If defined, a C expression whose value is a string containing the
4681 assembler operation to identify the following data as uninitialized global
4682 shared data. If not defined, and @code{BSS_SECTION_ASM_OP} is, the latter
4685 @findex INIT_SECTION_ASM_OP
4686 @item INIT_SECTION_ASM_OP
4687 If defined, a C expression whose value is a string containing the
4688 assembler operation to identify the following data as initialization
4689 code. If not defined, GNU CC will assume such a section does not
4692 @findex EXTRA_SECTIONS
4695 @item EXTRA_SECTIONS
4696 A list of names for sections other than the standard two, which are
4697 @code{in_text} and @code{in_data}. You need not define this macro
4698 on a system with no other sections (that GCC needs to use).
4700 @findex EXTRA_SECTION_FUNCTIONS
4701 @findex text_section
4702 @findex data_section
4703 @item EXTRA_SECTION_FUNCTIONS
4704 One or more functions to be defined in @file{varasm.c}. These
4705 functions should do jobs analogous to those of @code{text_section} and
4706 @code{data_section}, for your additional sections. Do not define this
4707 macro if you do not define @code{EXTRA_SECTIONS}.
4709 @findex READONLY_DATA_SECTION
4710 @item READONLY_DATA_SECTION
4711 On most machines, read-only variables, constants, and jump tables are
4712 placed in the text section. If this is not the case on your machine,
4713 this macro should be defined to be the name of a function (either
4714 @code{data_section} or a function defined in @code{EXTRA_SECTIONS}) that
4715 switches to the section to be used for read-only items.
4717 If these items should be placed in the text section, this macro should
4720 @findex SELECT_SECTION
4721 @item SELECT_SECTION (@var{exp}, @var{reloc})
4722 A C statement or statements to switch to the appropriate section for
4723 output of @var{exp}. You can assume that @var{exp} is either a
4724 @code{VAR_DECL} node or a constant of some sort. @var{reloc}
4725 indicates whether the initial value of @var{exp} requires link-time
4726 relocations. Select the section by calling @code{text_section} or one
4727 of the alternatives for other sections.
4729 Do not define this macro if you put all read-only variables and
4730 constants in the read-only data section (usually the text section).
4732 @findex SELECT_RTX_SECTION
4733 @item SELECT_RTX_SECTION (@var{mode}, @var{rtx})
4734 A C statement or statements to switch to the appropriate section for
4735 output of @var{rtx} in mode @var{mode}. You can assume that @var{rtx}
4736 is some kind of constant in RTL. The argument @var{mode} is redundant
4737 except in the case of a @code{const_int} rtx. Select the section by
4738 calling @code{text_section} or one of the alternatives for other
4741 Do not define this macro if you put all constants in the read-only
4744 @findex JUMP_TABLES_IN_TEXT_SECTION
4745 @item JUMP_TABLES_IN_TEXT_SECTION
4746 Define this macro if jump tables (for @code{tablejump} insns) should be
4747 output in the text section, along with the assembler instructions.
4748 Otherwise, the readonly data section is used.
4750 This macro is irrelevant if there is no separate readonly data section.
4752 @findex ENCODE_SECTION_INFO
4753 @item ENCODE_SECTION_INFO (@var{decl})
4754 Define this macro if references to a symbol must be treated differently
4755 depending on something about the variable or function named by the
4756 symbol (such as what section it is in).
4758 The macro definition, if any, is executed immediately after the rtl for
4759 @var{decl} has been created and stored in @code{DECL_RTL (@var{decl})}.
4760 The value of the rtl will be a @code{mem} whose address is a
4763 @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO}
4764 The usual thing for this macro to do is to record a flag in the
4765 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
4766 modified name string in the @code{symbol_ref} (if one bit is not enough
4769 @findex STRIP_NAME_ENCODING
4770 @item STRIP_NAME_ENCODING (@var{var}, @var{sym_name})
4771 Decode @var{sym_name} and store the real name part in @var{var}, sans
4772 the characters that encode section info. Define this macro if
4773 @code{ENCODE_SECTION_INFO} alters the symbol's name string.
4775 @findex UNIQUE_SECTION_P
4776 @item UNIQUE_SECTION_P (@var{decl})
4777 A C expression which evaluates to true if @var{decl} should be placed
4778 into a unique section for some target-specific reason. If you do not
4779 define this macro, the default is @samp{0}. Note that the flag
4780 @samp{-ffunction-sections} will also cause functions to be placed into
4783 @findex UNIQUE_SECTION
4784 @item UNIQUE_SECTION (@var{decl}, @var{reloc})
4785 A C statement to build up a unique section name, expressed as a
4786 STRING_CST node, and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
4787 @var{reloc} indicates whether the initial value of @var{exp} requires
4788 link-time relocations. If you do not define this macro, GNU CC will use
4789 the symbol name prefixed by @samp{.} as the section name.
4793 @section Position Independent Code
4794 @cindex position independent code
4797 This section describes macros that help implement generation of position
4798 independent code. Simply defining these macros is not enough to
4799 generate valid PIC; you must also add support to the macros
4800 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
4801 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
4802 @samp{movsi} to do something appropriate when the source operand
4803 contains a symbolic address. You may also need to alter the handling of
4804 switch statements so that they use relative addresses.
4805 @c i rearranged the order of the macros above to try to force one of
4806 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
4809 @findex PIC_OFFSET_TABLE_REGNUM
4810 @item PIC_OFFSET_TABLE_REGNUM
4811 The register number of the register used to address a table of static
4812 data addresses in memory. In some cases this register is defined by a
4813 processor's ``application binary interface'' (ABI). When this macro
4814 is defined, RTL is generated for this register once, as with the stack
4815 pointer and frame pointer registers. If this macro is not defined, it
4816 is up to the machine-dependent files to allocate such a register (if
4819 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
4820 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
4821 Define this macro if the register defined by
4822 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
4823 this macro if @code{PPIC_OFFSET_TABLE_REGNUM} is not defined.
4825 @findex FINALIZE_PIC
4827 By generating position-independent code, when two different programs (A
4828 and B) share a common library (libC.a), the text of the library can be
4829 shared whether or not the library is linked at the same address for both
4830 programs. In some of these environments, position-independent code
4831 requires not only the use of different addressing modes, but also
4832 special code to enable the use of these addressing modes.
4834 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
4835 codes once the function is being compiled into assembly code, but not
4836 before. (It is not done before, because in the case of compiling an
4837 inline function, it would lead to multiple PIC prologues being
4838 included in functions which used inline functions and were compiled to
4841 @findex LEGITIMATE_PIC_OPERAND_P
4842 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
4843 A C expression that is nonzero if @var{x} is a legitimate immediate
4844 operand on the target machine when generating position independent code.
4845 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
4846 check this. You can also assume @var{flag_pic} is true, so you need not
4847 check it either. You need not define this macro if all constants
4848 (including @code{SYMBOL_REF}) can be immediate operands when generating
4849 position independent code.
4852 @node Assembler Format
4853 @section Defining the Output Assembler Language
4855 This section describes macros whose principal purpose is to describe how
4856 to write instructions in assembler language--rather than what the
4860 * File Framework:: Structural information for the assembler file.
4861 * Data Output:: Output of constants (numbers, strings, addresses).
4862 * Uninitialized Data:: Output of uninitialized variables.
4863 * Label Output:: Output and generation of labels.
4864 * Initialization:: General principles of initialization
4865 and termination routines.
4866 * Macros for Initialization::
4867 Specific macros that control the handling of
4868 initialization and termination routines.
4869 * Instruction Output:: Output of actual instructions.
4870 * Dispatch Tables:: Output of jump tables.
4871 * Exception Region Output:: Output of exception region code.
4872 * Alignment Output:: Pseudo ops for alignment and skipping data.
4875 @node File Framework
4876 @subsection The Overall Framework of an Assembler File
4877 @cindex assembler format
4878 @cindex output of assembler code
4880 @c prevent bad page break with this line
4881 This describes the overall framework of an assembler file.
4884 @findex ASM_FILE_START
4885 @item ASM_FILE_START (@var{stream})
4886 A C expression which outputs to the stdio stream @var{stream}
4887 some appropriate text to go at the start of an assembler file.
4889 Normally this macro is defined to output a line containing
4890 @samp{#NO_APP}, which is a comment that has no effect on most
4891 assemblers but tells the GNU assembler that it can save time by not
4892 checking for certain assembler constructs.
4894 On systems that use SDB, it is necessary to output certain commands;
4895 see @file{attasm.h}.
4897 @findex ASM_FILE_END
4898 @item ASM_FILE_END (@var{stream})
4899 A C expression which outputs to the stdio stream @var{stream}
4900 some appropriate text to go at the end of an assembler file.
4902 If this macro is not defined, the default is to output nothing
4903 special at the end of the file. Most systems don't require any
4906 On systems that use SDB, it is necessary to output certain commands;
4907 see @file{attasm.h}.
4909 @findex ASM_IDENTIFY_GCC
4910 @item ASM_IDENTIFY_GCC (@var{file})
4911 A C statement to output assembler commands which will identify
4912 the object file as having been compiled with GNU CC (or another
4915 If you don't define this macro, the string @samp{gcc_compiled.:}
4916 is output. This string is calculated to define a symbol which,
4917 on BSD systems, will never be defined for any other reason.
4918 GDB checks for the presence of this symbol when reading the
4919 symbol table of an executable.
4921 On non-BSD systems, you must arrange communication with GDB in
4922 some other fashion. If GDB is not used on your system, you can
4923 define this macro with an empty body.
4925 @findex ASM_COMMENT_START
4926 @item ASM_COMMENT_START
4927 A C string constant describing how to begin a comment in the target
4928 assembler language. The compiler assumes that the comment will end at
4929 the end of the line.
4933 A C string constant for text to be output before each @code{asm}
4934 statement or group of consecutive ones. Normally this is
4935 @code{"#APP"}, which is a comment that has no effect on most
4936 assemblers but tells the GNU assembler that it must check the lines
4937 that follow for all valid assembler constructs.
4941 A C string constant for text to be output after each @code{asm}
4942 statement or group of consecutive ones. Normally this is
4943 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
4944 time-saving assumptions that are valid for ordinary compiler output.
4946 @findex ASM_OUTPUT_SOURCE_FILENAME
4947 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
4948 A C statement to output COFF information or DWARF debugging information
4949 which indicates that filename @var{name} is the current source file to
4950 the stdio stream @var{stream}.
4952 This macro need not be defined if the standard form of output
4953 for the file format in use is appropriate.
4955 @findex OUTPUT_QUOTED_STRING
4956 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{name})
4957 A C statement to output the string @var{string} to the stdio stream
4958 @var{stream}. If you do not call the function @code{output_quoted_string}
4959 in your config files, GNU CC will only call it to output filenames to
4960 the assembler source. So you can use it to canonicalize the format
4961 of the filename using this macro.
4963 @findex ASM_OUTPUT_SOURCE_LINE
4964 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
4965 A C statement to output DBX or SDB debugging information before code
4966 for line number @var{line} of the current source file to the
4967 stdio stream @var{stream}.
4969 This macro need not be defined if the standard form of debugging
4970 information for the debugger in use is appropriate.
4972 @findex ASM_OUTPUT_IDENT
4973 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
4974 A C statement to output something to the assembler file to handle a
4975 @samp{#ident} directive containing the text @var{string}. If this
4976 macro is not defined, nothing is output for a @samp{#ident} directive.
4978 @findex ASM_OUTPUT_SECTION_NAME
4979 @item ASM_OUTPUT_SECTION_NAME (@var{stream}, @var{decl}, @var{name}, @var{reloc})
4980 A C statement to output something to the assembler file to switch to section
4981 @var{name} for object @var{decl} which is either a @code{FUNCTION_DECL}, a
4982 @code{VAR_DECL} or @code{NULL_TREE}. @var{reloc}
4983 indicates whether the initial value of @var{exp} requires link-time
4984 relocations. Some target formats do not support
4985 arbitrary sections. Do not define this macro in such cases.
4987 At present this macro is only used to support section attributes.
4988 When this macro is undefined, section attributes are disabled.
4990 @findex OBJC_PROLOGUE
4992 A C statement to output any assembler statements which are required to
4993 precede any Objective C object definitions or message sending. The
4994 statement is executed only when compiling an Objective C program.
4999 @subsection Output of Data
5001 @c prevent bad page break with this line
5002 This describes data output.
5005 @findex ASM_OUTPUT_LONG_DOUBLE
5006 @findex ASM_OUTPUT_DOUBLE
5007 @findex ASM_OUTPUT_FLOAT
5008 @item ASM_OUTPUT_LONG_DOUBLE (@var{stream}, @var{value})
5009 @itemx ASM_OUTPUT_DOUBLE (@var{stream}, @var{value})
5010 @itemx ASM_OUTPUT_FLOAT (@var{stream}, @var{value})
5011 @itemx ASM_OUTPUT_THREE_QUARTER_FLOAT (@var{stream}, @var{value})
5012 @itemx ASM_OUTPUT_SHORT_FLOAT (@var{stream}, @var{value})
5013 @itemx ASM_OUTPUT_BYTE_FLOAT (@var{stream}, @var{value})
5014 A C statement to output to the stdio stream @var{stream} an assembler
5015 instruction to assemble a floating-point constant of @code{TFmode},
5016 @code{DFmode}, @code{SFmode}, @code{TQFmode}, @code{HFmode}, or
5017 @code{QFmode}, respectively, whose value is @var{value}. @var{value}
5018 will be a C expression of type @code{REAL_VALUE_TYPE}. Macros such as
5019 @code{REAL_VALUE_TO_TARGET_DOUBLE} are useful for writing these
5022 @findex ASM_OUTPUT_QUADRUPLE_INT
5023 @findex ASM_OUTPUT_DOUBLE_INT
5024 @findex ASM_OUTPUT_INT
5025 @findex ASM_OUTPUT_SHORT
5026 @findex ASM_OUTPUT_CHAR
5027 @findex output_addr_const
5028 @item ASM_OUTPUT_QUADRUPLE_INT (@var{stream}, @var{exp})
5029 @itemx ASM_OUTPUT_DOUBLE_INT (@var{stream}, @var{exp})
5030 @itemx ASM_OUTPUT_INT (@var{stream}, @var{exp})
5031 @itemx ASM_OUTPUT_SHORT (@var{stream}, @var{exp})
5032 @itemx ASM_OUTPUT_CHAR (@var{stream}, @var{exp})
5033 A C statement to output to the stdio stream @var{stream} an assembler
5034 instruction to assemble an integer of 16, 8, 4, 2 or 1 bytes,
5035 respectively, whose value is @var{value}. The argument @var{exp} will
5036 be an RTL expression which represents a constant value. Use
5037 @samp{output_addr_const (@var{stream}, @var{exp})} to output this value
5038 as an assembler expression.@refill
5040 For sizes larger than @code{UNITS_PER_WORD}, if the action of a macro
5041 would be identical to repeatedly calling the macro corresponding to
5042 a size of @code{UNITS_PER_WORD}, once for each word, you need not define
5045 @findex ASM_OUTPUT_BYTE
5046 @item ASM_OUTPUT_BYTE (@var{stream}, @var{value})
5047 A C statement to output to the stdio stream @var{stream} an assembler
5048 instruction to assemble a single byte containing the number @var{value}.
5052 A C string constant giving the pseudo-op to use for a sequence of
5053 single-byte constants. If this macro is not defined, the default is
5056 @findex ASM_OUTPUT_ASCII
5057 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5058 A C statement to output to the stdio stream @var{stream} an assembler
5059 instruction to assemble a string constant containing the @var{len}
5060 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5061 @code{char *} and @var{len} a C expression of type @code{int}.
5063 If the assembler has a @code{.ascii} pseudo-op as found in the
5064 Berkeley Unix assembler, do not define the macro
5065 @code{ASM_OUTPUT_ASCII}.
5067 @findex CONSTANT_POOL_BEFORE_FUNCTION
5068 @item CONSTANT_POOL_BEFORE_FUNCTION
5069 You may define this macro as a C expression. You should define the
5070 expression to have a non-zero value if GNU CC should output the constant
5071 pool for a function before the code for the function, or a zero value if
5072 GNU CC should output the constant pool after the function. If you do
5073 not define this macro, the usual case, GNU CC will output the constant
5074 pool before the function.
5076 @findex ASM_OUTPUT_POOL_PROLOGUE
5077 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5078 A C statement to output assembler commands to define the start of the
5079 constant pool for a function. @var{funname} is a string giving
5080 the name of the function. Should the return type of the function
5081 be required, it can be obtained via @var{fundecl}. @var{size}
5082 is the size, in bytes, of the constant pool that will be written
5083 immediately after this call.
5085 If no constant-pool prefix is required, the usual case, this macro need
5088 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
5089 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5090 A C statement (with or without semicolon) to output a constant in the
5091 constant pool, if it needs special treatment. (This macro need not do
5092 anything for RTL expressions that can be output normally.)
5094 The argument @var{file} is the standard I/O stream to output the
5095 assembler code on. @var{x} is the RTL expression for the constant to
5096 output, and @var{mode} is the machine mode (in case @var{x} is a
5097 @samp{const_int}). @var{align} is the required alignment for the value
5098 @var{x}; you should output an assembler directive to force this much
5101 The argument @var{labelno} is a number to use in an internal label for
5102 the address of this pool entry. The definition of this macro is
5103 responsible for outputting the label definition at the proper place.
5104 Here is how to do this:
5107 ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
5110 When you output a pool entry specially, you should end with a
5111 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5112 entry from being output a second time in the usual manner.
5114 You need not define this macro if it would do nothing.
5116 @findex CONSTANT_AFTER_FUNCTION_P
5117 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
5118 Define this macro as a C expression which is nonzero if the constant
5119 @var{exp}, of type @code{tree}, should be output after the code for a
5120 function. The compiler will normally output all constants before the
5121 function; you need not define this macro if this is OK.
5123 @findex ASM_OUTPUT_POOL_EPILOGUE
5124 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5125 A C statement to output assembler commands to at the end of the constant
5126 pool for a function. @var{funname} is a string giving the name of the
5127 function. Should the return type of the function be required, you can
5128 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5129 constant pool that GNU CC wrote immediately before this call.
5131 If no constant-pool epilogue is required, the usual case, you need not
5134 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
5135 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
5136 Define this macro as a C expression which is nonzero if @var{C} is
5137 used as a logical line separator by the assembler.
5139 If you do not define this macro, the default is that only
5140 the character @samp{;} is treated as a logical line separator.
5143 @findex ASM_OPEN_PAREN
5144 @findex ASM_CLOSE_PAREN
5145 @item ASM_OPEN_PAREN
5146 @itemx ASM_CLOSE_PAREN
5147 These macros are defined as C string constant, describing the syntax
5148 in the assembler for grouping arithmetic expressions. The following
5149 definitions are correct for most assemblers:
5152 #define ASM_OPEN_PAREN "("
5153 #define ASM_CLOSE_PAREN ")"
5157 These macros are provided by @file{real.h} for writing the definitions
5158 of @code{ASM_OUTPUT_DOUBLE} and the like:
5161 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5162 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5163 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5164 @findex REAL_VALUE_TO_TARGET_SINGLE
5165 @findex REAL_VALUE_TO_TARGET_DOUBLE
5166 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
5167 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
5168 floating point representation, and store its bit pattern in the array of
5169 @code{long int} whose address is @var{l}. The number of elements in the
5170 output array is determined by the size of the desired target floating
5171 point data type: 32 bits of it go in each @code{long int} array
5172 element. Each array element holds 32 bits of the result, even if
5173 @code{long int} is wider than 32 bits on the host machine.
5175 The array element values are designed so that you can print them out
5176 using @code{fprintf} in the order they should appear in the target
5179 @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
5180 @findex REAL_VALUE_TO_DECIMAL
5181 This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
5182 decimal number and stores it as a string into @var{string}.
5183 You must pass, as @var{string}, the address of a long enough block
5184 of space to hold the result.
5186 The argument @var{format} is a @code{printf}-specification that serves
5187 as a suggestion for how to format the output string.
5190 @node Uninitialized Data
5191 @subsection Output of Uninitialized Variables
5193 Each of the macros in this section is used to do the whole job of
5194 outputting a single uninitialized variable.
5197 @findex ASM_OUTPUT_COMMON
5198 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5199 A C statement (sans semicolon) to output to the stdio stream
5200 @var{stream} the assembler definition of a common-label named
5201 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5202 is the size rounded up to whatever alignment the caller wants.
5204 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5205 output the name itself; before and after that, output the additional
5206 assembler syntax for defining the name, and a newline.
5208 This macro controls how the assembler definitions of uninitialized
5209 common global variables are output.
5211 @findex ASM_OUTPUT_ALIGNED_COMMON
5212 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5213 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5214 separate, explicit argument. If you define this macro, it is used in
5215 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5216 handling the required alignment of the variable. The alignment is specified
5217 as the number of bits.
5219 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
5220 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5221 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5222 variable to be output, if there is one, or @code{NULL_TREE} if there
5223 is not corresponding variable. If you define this macro, GNU CC wil use it
5224 in place of both @code{ASM_OUTPUT_COMMON} and
5225 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5226 the variable's decl in order to chose what to output.
5228 @findex ASM_OUTPUT_SHARED_COMMON
5229 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5230 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
5231 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
5234 @findex ASM_OUTPUT_BSS
5235 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5236 A C statement (sans semicolon) to output to the stdio stream
5237 @var{stream} the assembler definition of uninitialized global @var{decl} named
5238 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5239 is the size rounded up to whatever alignment the caller wants.
5241 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
5242 defining this macro. If unable, use the expression
5243 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5244 before and after that, output the additional assembler syntax for defining
5245 the name, and a newline.
5247 This macro controls how the assembler definitions of uninitialized global
5248 variables are output. This macro exists to properly support languages like
5249 @code{c++} which do not have @code{common} data. However, this macro currently
5250 is not defined for all targets. If this macro and
5251 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
5252 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
5253 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
5255 @findex ASM_OUTPUT_ALIGNED_BSS
5256 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5257 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
5258 separate, explicit argument. If you define this macro, it is used in
5259 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
5260 handling the required alignment of the variable. The alignment is specified
5261 as the number of bits.
5263 Try to use function @code{asm_output_aligned_bss} defined in file
5264 @file{varasm.c} when defining this macro.
5266 @findex ASM_OUTPUT_SHARED_BSS
5267 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5268 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
5269 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
5272 @findex ASM_OUTPUT_LOCAL
5273 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5274 A C statement (sans semicolon) to output to the stdio stream
5275 @var{stream} the assembler definition of a local-common-label named
5276 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5277 is the size rounded up to whatever alignment the caller wants.
5279 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5280 output the name itself; before and after that, output the additional
5281 assembler syntax for defining the name, and a newline.
5283 This macro controls how the assembler definitions of uninitialized
5284 static variables are output.
5286 @findex ASM_OUTPUT_ALIGNED_LOCAL
5287 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5288 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5289 separate, explicit argument. If you define this macro, it is used in
5290 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5291 handling the required alignment of the variable. The alignment is specified
5292 as the number of bits.
5294 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
5295 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5296 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5297 variable to be output, if there is one, or @code{NULL_TREE} if there
5298 is not corresponding variable. If you define this macro, GNU CC wil use it
5299 in place of both @code{ASM_OUTPUT_DECL} and
5300 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5301 the variable's decl in order to chose what to output.
5304 @findex ASM_OUTPUT_SHARED_LOCAL
5305 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5306 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
5307 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
5312 @subsection Output and Generation of Labels
5314 @c prevent bad page break with this line
5315 This is about outputting labels.
5318 @findex ASM_OUTPUT_LABEL
5319 @findex assemble_name
5320 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5321 A C statement (sans semicolon) to output to the stdio stream
5322 @var{stream} the assembler definition of a label named @var{name}.
5323 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5324 output the name itself; before and after that, output the additional
5325 assembler syntax for defining the name, and a newline.
5327 @findex ASM_DECLARE_FUNCTION_NAME
5328 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5329 A C statement (sans semicolon) to output to the stdio stream
5330 @var{stream} any text necessary for declaring the name @var{name} of a
5331 function which is being defined. This macro is responsible for
5332 outputting the label definition (perhaps using
5333 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
5334 @code{FUNCTION_DECL} tree node representing the function.
5336 If this macro is not defined, then the function name is defined in the
5337 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5339 @findex ASM_DECLARE_FUNCTION_SIZE
5340 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5341 A C statement (sans semicolon) to output to the stdio stream
5342 @var{stream} any text necessary for declaring the size of a function
5343 which is being defined. The argument @var{name} is the name of the
5344 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5345 representing the function.
5347 If this macro is not defined, then the function size is not defined.
5349 @findex ASM_DECLARE_OBJECT_NAME
5350 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5351 A C statement (sans semicolon) to output to the stdio stream
5352 @var{stream} any text necessary for declaring the name @var{name} of an
5353 initialized variable which is being defined. This macro must output the
5354 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
5355 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
5357 If this macro is not defined, then the variable name is defined in the
5358 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5360 @findex ASM_FINISH_DECLARE_OBJECT
5361 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5362 A C statement (sans semicolon) to finish up declaring a variable name
5363 once the compiler has processed its initializer fully and thus has had a
5364 chance to determine the size of an array when controlled by an
5365 initializer. This is used on systems where it's necessary to declare
5366 something about the size of the object.
5368 If you don't define this macro, that is equivalent to defining it to do
5371 @findex ASM_GLOBALIZE_LABEL
5372 @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
5373 A C statement (sans semicolon) to output to the stdio stream
5374 @var{stream} some commands that will make the label @var{name} global;
5375 that is, available for reference from other files. Use the expression
5376 @code{assemble_name (@var{stream}, @var{name})} to output the name
5377 itself; before and after that, output the additional assembler syntax
5378 for making that name global, and a newline.
5380 @findex ASM_WEAKEN_LABEL
5381 @item ASM_WEAKEN_LABEL
5382 A C statement (sans semicolon) to output to the stdio stream
5383 @var{stream} some commands that will make the label @var{name} weak;
5384 that is, available for reference from other files but only used if
5385 no other definition is available. Use the expression
5386 @code{assemble_name (@var{stream}, @var{name})} to output the name
5387 itself; before and after that, output the additional assembler syntax
5388 for making that name weak, and a newline.
5390 If you don't define this macro, GNU CC will not support weak
5391 symbols and you should not define the @code{SUPPORTS_WEAK} macro.
5393 @findex SUPPORTS_WEAK
5395 A C expression which evaluates to true if the target supports weak symbols.
5397 If you don't define this macro, @file{defaults.h} provides a default
5398 definition. If @code{ASM_WEAKEN_LABEL} is defined, the default
5399 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5400 you want to control weak symbol support with a compiler flag such as
5403 @findex MAKE_DECL_ONE_ONLY (@var{decl})
5404 @item MAKE_DECL_ONE_ONLY
5405 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5406 public symbol such that extra copies in multiple translation units will
5407 be discarded by the linker. Define this macro if your object file
5408 format provides support for this concept, such as the @samp{COMDAT}
5409 section flags in the Microsoft Windows PE/COFF format, and this support
5410 requires changes to @var{decl}, such as putting it in a separate section.
5412 @findex SUPPORTS_ONE_ONLY
5413 @item SUPPORTS_ONE_ONLY
5414 A C expression which evaluates to true if the target supports one-only
5417 If you don't define this macro, @file{varasm.c} provides a default
5418 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
5419 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5420 you want to control one-only symbol support with a compiler flag, or if
5421 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
5422 be emitted as one-only.
5424 @findex ASM_OUTPUT_EXTERNAL
5425 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
5426 A C statement (sans semicolon) to output to the stdio stream
5427 @var{stream} any text necessary for declaring the name of an external
5428 symbol named @var{name} which is referenced in this compilation but
5429 not defined. The value of @var{decl} is the tree node for the
5432 This macro need not be defined if it does not need to output anything.
5433 The GNU assembler and most Unix assemblers don't require anything.
5435 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
5436 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
5437 A C statement (sans semicolon) to output on @var{stream} an assembler
5438 pseudo-op to declare a library function name external. The name of the
5439 library function is given by @var{symref}, which has type @code{rtx} and
5440 is a @code{symbol_ref}.
5442 This macro need not be defined if it does not need to output anything.
5443 The GNU assembler and most Unix assemblers don't require anything.
5445 @findex ASM_OUTPUT_LABELREF
5446 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
5447 A C statement (sans semicolon) to output to the stdio stream
5448 @var{stream} a reference in assembler syntax to a label named
5449 @var{name}. This should add @samp{_} to the front of the name, if that
5450 is customary on your operating system, as it is in most Berkeley Unix
5451 systems. This macro is used in @code{assemble_name}.
5453 @ignore @c Seems not to exist anymore.
5454 @findex ASM_OUTPUT_LABELREF_AS_INT
5455 @item ASM_OUTPUT_LABELREF_AS_INT (@var{file}, @var{label})
5456 Define this macro for systems that use the program @code{collect2}.
5457 The definition should be a C statement to output a word containing
5458 a reference to the label @var{label}.
5461 @findex ASM_OUTPUT_INTERNAL_LABEL
5462 @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
5463 A C statement to output to the stdio stream @var{stream} a label whose
5464 name is made from the string @var{prefix} and the number @var{num}.
5466 It is absolutely essential that these labels be distinct from the labels
5467 used for user-level functions and variables. Otherwise, certain programs
5468 will have name conflicts with internal labels.
5470 It is desirable to exclude internal labels from the symbol table of the
5471 object file. Most assemblers have a naming convention for labels that
5472 should be excluded; on many systems, the letter @samp{L} at the
5473 beginning of a label has this effect. You should find out what
5474 convention your system uses, and follow it.
5476 The usual definition of this macro is as follows:
5479 fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
5482 @findex ASM_GENERATE_INTERNAL_LABEL
5483 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
5484 A C statement to store into the string @var{string} a label whose name
5485 is made from the string @var{prefix} and the number @var{num}.
5487 This string, when output subsequently by @code{assemble_name}, should
5488 produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
5489 with the same @var{prefix} and @var{num}.
5491 If the string begins with @samp{*}, then @code{assemble_name} will
5492 output the rest of the string unchanged. It is often convenient for
5493 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
5494 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
5495 to output the string, and may change it. (Of course,
5496 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
5497 you should know what it does on your machine.)
5499 @findex ASM_FORMAT_PRIVATE_NAME
5500 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
5501 A C expression to assign to @var{outvar} (which is a variable of type
5502 @code{char *}) a newly allocated string made from the string
5503 @var{name} and the number @var{number}, with some suitable punctuation
5504 added. Use @code{alloca} to get space for the string.
5506 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
5507 produce an assembler label for an internal static variable whose name is
5508 @var{name}. Therefore, the string must be such as to result in valid
5509 assembler code. The argument @var{number} is different each time this
5510 macro is executed; it prevents conflicts between similarly-named
5511 internal static variables in different scopes.
5513 Ideally this string should not be a valid C identifier, to prevent any
5514 conflict with the user's own symbols. Most assemblers allow periods
5515 or percent signs in assembler symbols; putting at least one of these
5516 between the name and the number will suffice.
5518 @findex ASM_OUTPUT_DEF
5519 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
5520 A C statement to output to the stdio stream @var{stream} assembler code
5521 which defines (equates) the symbol @var{name} to have the value @var{value}.
5523 If SET_ASM_OP is defined, a default definition is provided which is
5524 correct for most systems.
5526 @findex ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL
5527 @item ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL (@var{stream}, @var{symbol}, @var{high}, @var{low})
5528 A C statement to output to the stdio stream @var{stream} assembler code
5529 which defines (equates) the symbol @var{symbol} to have a value equal to
5530 the difference of the two symbols @var{high} and @var{low}, i.e.
5531 @var{high} minus @var{low}. GNU CC guarantees that the symbols @var{high}
5532 and @var{low} are already known by the assembler so that the difference
5533 resolves into a constant.
5535 If SET_ASM_OP is defined, a default definition is provided which is
5536 correct for most systems.
5538 @findex ASM_OUTPUT_WEAK_ALIAS
5539 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
5540 A C statement to output to the stdio stream @var{stream} assembler code
5541 which defines (equates) the weak symbol @var{name} to have the value
5544 Define this macro if the target only supports weak aliases; define
5545 ASM_OUTPUT_DEF instead if possible.
5547 @findex OBJC_GEN_METHOD_LABEL
5548 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
5549 Define this macro to override the default assembler names used for
5550 Objective C methods.
5552 The default name is a unique method number followed by the name of the
5553 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
5554 the category is also included in the assembler name (e.g.@:
5557 These names are safe on most systems, but make debugging difficult since
5558 the method's selector is not present in the name. Therefore, particular
5559 systems define other ways of computing names.
5561 @var{buf} is an expression of type @code{char *} which gives you a
5562 buffer in which to store the name; its length is as long as
5563 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
5564 50 characters extra.
5566 The argument @var{is_inst} specifies whether the method is an instance
5567 method or a class method; @var{class_name} is the name of the class;
5568 @var{cat_name} is the name of the category (or NULL if the method is not
5569 in a category); and @var{sel_name} is the name of the selector.
5571 On systems where the assembler can handle quoted names, you can use this
5572 macro to provide more human-readable names.
5575 @node Initialization
5576 @subsection How Initialization Functions Are Handled
5577 @cindex initialization routines
5578 @cindex termination routines
5579 @cindex constructors, output of
5580 @cindex destructors, output of
5582 The compiled code for certain languages includes @dfn{constructors}
5583 (also called @dfn{initialization routines})---functions to initialize
5584 data in the program when the program is started. These functions need
5585 to be called before the program is ``started''---that is to say, before
5586 @code{main} is called.
5588 Compiling some languages generates @dfn{destructors} (also called
5589 @dfn{termination routines}) that should be called when the program
5592 To make the initialization and termination functions work, the compiler
5593 must output something in the assembler code to cause those functions to
5594 be called at the appropriate time. When you port the compiler to a new
5595 system, you need to specify how to do this.
5597 There are two major ways that GCC currently supports the execution of
5598 initialization and termination functions. Each way has two variants.
5599 Much of the structure is common to all four variations.
5601 @findex __CTOR_LIST__
5602 @findex __DTOR_LIST__
5603 The linker must build two lists of these functions---a list of
5604 initialization functions, called @code{__CTOR_LIST__}, and a list of
5605 termination functions, called @code{__DTOR_LIST__}.
5607 Each list always begins with an ignored function pointer (which may hold
5608 0, @minus{}1, or a count of the function pointers after it, depending on
5609 the environment). This is followed by a series of zero or more function
5610 pointers to constructors (or destructors), followed by a function
5611 pointer containing zero.
5613 Depending on the operating system and its executable file format, either
5614 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
5615 time and exit time. Constructors are called in reverse order of the
5616 list; destructors in forward order.
5618 The best way to handle static constructors works only for object file
5619 formats which provide arbitrarily-named sections. A section is set
5620 aside for a list of constructors, and another for a list of destructors.
5621 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
5622 object file that defines an initialization function also puts a word in
5623 the constructor section to point to that function. The linker
5624 accumulates all these words into one contiguous @samp{.ctors} section.
5625 Termination functions are handled similarly.
5627 To use this method, you need appropriate definitions of the macros
5628 @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}. Usually
5629 you can get them by including @file{svr4.h}.
5631 When arbitrary sections are available, there are two variants, depending
5632 upon how the code in @file{crtstuff.c} is called. On systems that
5633 support an @dfn{init} section which is executed at program startup,
5634 parts of @file{crtstuff.c} are compiled into that section. The
5635 program is linked by the @code{gcc} driver like this:
5638 ld -o @var{output_file} crtbegin.o @dots{} crtend.o -lgcc
5641 The head of a function (@code{__do_global_ctors}) appears in the init
5642 section of @file{crtbegin.o}; the remainder of the function appears in
5643 the init section of @file{crtend.o}. The linker will pull these two
5644 parts of the section together, making a whole function. If any of the
5645 user's object files linked into the middle of it contribute code, then that
5646 code will be executed as part of the body of @code{__do_global_ctors}.
5648 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
5651 If no init section is available, do not define
5652 @code{INIT_SECTION_ASM_OP}. Then @code{__do_global_ctors} is built into
5653 the text section like all other functions, and resides in
5654 @file{libgcc.a}. When GCC compiles any function called @code{main}, it
5655 inserts a procedure call to @code{__main} as the first executable code
5656 after the function prologue. The @code{__main} function, also defined
5657 in @file{libgcc2.c}, simply calls @file{__do_global_ctors}.
5659 In file formats that don't support arbitrary sections, there are again
5660 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
5661 and an `a.out' format must be used. In this case,
5662 @code{ASM_OUTPUT_CONSTRUCTOR} is defined to produce a @code{.stabs}
5663 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
5664 and with the address of the void function containing the initialization
5665 code as its value. The GNU linker recognizes this as a request to add
5666 the value to a ``set''; the values are accumulated, and are eventually
5667 placed in the executable as a vector in the format described above, with
5668 a leading (ignored) count and a trailing zero element.
5669 @code{ASM_OUTPUT_DESTRUCTOR} is handled similarly. Since no init
5670 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
5671 the compilation of @code{main} to call @code{__main} as above, starting
5672 the initialization process.
5674 The last variant uses neither arbitrary sections nor the GNU linker.
5675 This is preferable when you want to do dynamic linking and when using
5676 file formats which the GNU linker does not support, such as `ECOFF'. In
5677 this case, @code{ASM_OUTPUT_CONSTRUCTOR} does not produce an
5678 @code{N_SETT} symbol; initialization and termination functions are
5679 recognized simply by their names. This requires an extra program in the
5680 linkage step, called @code{collect2}. This program pretends to be the
5681 linker, for use with GNU CC; it does its job by running the ordinary
5682 linker, but also arranges to include the vectors of initialization and
5683 termination functions. These functions are called via @code{__main} as
5686 Choosing among these configuration options has been simplified by a set
5687 of operating-system-dependent files in the @file{config} subdirectory.
5688 These files define all of the relevant parameters. Usually it is
5689 sufficient to include one into your specific machine-dependent
5690 configuration file. These files are:
5694 For operating systems using the `a.out' format.
5697 For operating systems using the `MachO' format.
5700 For System V Release 3 and similar systems using `COFF' format.
5703 For System V Release 4 and similar systems using `ELF' format.
5706 For the VMS operating system.
5710 The following section describes the specific macros that control and
5711 customize the handling of initialization and termination functions.
5714 @node Macros for Initialization
5715 @subsection Macros Controlling Initialization Routines
5717 Here are the macros that control how the compiler handles initialization
5718 and termination functions:
5721 @findex INIT_SECTION_ASM_OP
5722 @item INIT_SECTION_ASM_OP
5723 If defined, a C string constant for the assembler operation to identify
5724 the following data as initialization code. If not defined, GNU CC will
5725 assume such a section does not exist. When you are using special
5726 sections for initialization and termination functions, this macro also
5727 controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to run the
5728 initialization functions.
5730 @item HAS_INIT_SECTION
5731 @findex HAS_INIT_SECTION
5732 If defined, @code{main} will not call @code{__main} as described above.
5733 This macro should be defined for systems that control the contents of the
5734 init section on a symbol-by-symbol basis, such as OSF/1, and should not
5735 be defined explicitly for systems that support
5736 @code{INIT_SECTION_ASM_OP}.
5738 @item LD_INIT_SWITCH
5739 @findex LD_INIT_SWITCH
5740 If defined, a C string constant for a switch that tells the linker that
5741 the following symbol is an initialization routine.
5743 @item LD_FINI_SWITCH
5744 @findex LD_FINI_SWITCH
5745 If defined, a C string constant for a switch that tells the linker that
5746 the following symbol is a finalization routine.
5749 @findex INVOKE__main
5750 If defined, @code{main} will call @code{__main} despite the presence of
5751 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
5752 where the init section is not actually run automatically, but is still
5753 useful for collecting the lists of constructors and destructors.
5755 @item ASM_OUTPUT_CONSTRUCTOR (@var{stream}, @var{name})
5756 @findex ASM_OUTPUT_CONSTRUCTOR
5757 Define this macro as a C statement to output on the stream @var{stream}
5758 the assembler code to arrange to call the function named @var{name} at
5759 initialization time.
5761 Assume that @var{name} is the name of a C function generated
5762 automatically by the compiler. This function takes no arguments. Use
5763 the function @code{assemble_name} to output the name @var{name}; this
5764 performs any system-specific syntactic transformations such as adding an
5767 If you don't define this macro, nothing special is output to arrange to
5768 call the function. This is correct when the function will be called in
5769 some other manner---for example, by means of the @code{collect2} program,
5770 which looks through the symbol table to find these functions by their
5773 @item ASM_OUTPUT_DESTRUCTOR (@var{stream}, @var{name})
5774 @findex ASM_OUTPUT_DESTRUCTOR
5775 This is like @code{ASM_OUTPUT_CONSTRUCTOR} but used for termination
5776 functions rather than initialization functions.
5779 If your system uses @code{collect2} as the means of processing
5780 constructors, then that program normally uses @code{nm} to scan an
5781 object file for constructor functions to be called. On certain kinds of
5782 systems, you can define these macros to make @code{collect2} work faster
5783 (and, in some cases, make it work at all):
5786 @findex OBJECT_FORMAT_COFF
5787 @item OBJECT_FORMAT_COFF
5788 Define this macro if the system uses COFF (Common Object File Format)
5789 object files, so that @code{collect2} can assume this format and scan
5790 object files directly for dynamic constructor/destructor functions.
5792 @findex OBJECT_FORMAT_ROSE
5793 @item OBJECT_FORMAT_ROSE
5794 Define this macro if the system uses ROSE format object files, so that
5795 @code{collect2} can assume this format and scan object files directly
5796 for dynamic constructor/destructor functions.
5798 These macros are effective only in a native compiler; @code{collect2} as
5799 part of a cross compiler always uses @code{nm} for the target machine.
5801 @findex REAL_NM_FILE_NAME
5802 @item REAL_NM_FILE_NAME
5803 Define this macro as a C string constant containing the file name to use
5804 to execute @code{nm}. The default is to search the path normally for
5807 If your system supports shared libraries and has a program to list the
5808 dynamic dependencies of a given library or executable, you can define
5809 these macros to enable support for running initialization and
5810 termination functions in shared libraries:
5814 Define this macro to a C string constant containing the name of the
5815 program which lists dynamic dependencies, like @code{"ldd"} under SunOS 4.
5817 @findex PARSE_LDD_OUTPUT
5818 @item PARSE_LDD_OUTPUT (@var{PTR})
5819 Define this macro to be C code that extracts filenames from the output
5820 of the program denoted by @code{LDD_SUFFIX}. @var{PTR} is a variable
5821 of type @code{char *} that points to the beginning of a line of output
5822 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
5823 code must advance @var{PTR} to the beginning of the filename on that
5824 line. Otherwise, it must set @var{PTR} to @code{NULL}.
5828 @node Instruction Output
5829 @subsection Output of Assembler Instructions
5831 @c prevent bad page break with this line
5832 This describes assembler instruction output.
5835 @findex REGISTER_NAMES
5836 @item REGISTER_NAMES
5837 A C initializer containing the assembler's names for the machine
5838 registers, each one as a C string constant. This is what translates
5839 register numbers in the compiler into assembler language.
5841 @findex ADDITIONAL_REGISTER_NAMES
5842 @item ADDITIONAL_REGISTER_NAMES
5843 If defined, a C initializer for an array of structures containing a name
5844 and a register number. This macro defines additional names for hard
5845 registers, thus allowing the @code{asm} option in declarations to refer
5846 to registers using alternate names.
5848 @findex ASM_OUTPUT_OPCODE
5849 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
5850 Define this macro if you are using an unusual assembler that
5851 requires different names for the machine instructions.
5853 The definition is a C statement or statements which output an
5854 assembler instruction opcode to the stdio stream @var{stream}. The
5855 macro-operand @var{ptr} is a variable of type @code{char *} which
5856 points to the opcode name in its ``internal'' form---the form that is
5857 written in the machine description. The definition should output the
5858 opcode name to @var{stream}, performing any translation you desire, and
5859 increment the variable @var{ptr} to point at the end of the opcode
5860 so that it will not be output twice.
5862 In fact, your macro definition may process less than the entire opcode
5863 name, or more than the opcode name; but if you want to process text
5864 that includes @samp{%}-sequences to substitute operands, you must take
5865 care of the substitution yourself. Just be sure to increment
5866 @var{ptr} over whatever text should not be output normally.
5868 @findex recog_operand
5869 If you need to look at the operand values, they can be found as the
5870 elements of @code{recog_operand}.
5872 If the macro definition does nothing, the instruction is output
5875 @findex FINAL_PRESCAN_INSN
5876 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
5877 If defined, a C statement to be executed just prior to the output of
5878 assembler code for @var{insn}, to modify the extracted operands so
5879 they will be output differently.
5881 Here the argument @var{opvec} is the vector containing the operands
5882 extracted from @var{insn}, and @var{noperands} is the number of
5883 elements of the vector which contain meaningful data for this insn.
5884 The contents of this vector are what will be used to convert the insn
5885 template into assembler code, so you can change the assembler output
5886 by changing the contents of the vector.
5888 This macro is useful when various assembler syntaxes share a single
5889 file of instruction patterns; by defining this macro differently, you
5890 can cause a large class of instructions to be output differently (such
5891 as with rearranged operands). Naturally, variations in assembler
5892 syntax affecting individual insn patterns ought to be handled by
5893 writing conditional output routines in those patterns.
5895 If this macro is not defined, it is equivalent to a null statement.
5897 @findex FINAL_PRESCAN_LABEL
5898 @item FINAL_PRESCAN_LABEL
5899 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
5900 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
5901 @var{noperands} will be zero.
5903 @findex PRINT_OPERAND
5904 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
5905 A C compound statement to output to stdio stream @var{stream} the
5906 assembler syntax for an instruction operand @var{x}. @var{x} is an
5909 @var{code} is a value that can be used to specify one of several ways
5910 of printing the operand. It is used when identical operands must be
5911 printed differently depending on the context. @var{code} comes from
5912 the @samp{%} specification that was used to request printing of the
5913 operand. If the specification was just @samp{%@var{digit}} then
5914 @var{code} is 0; if the specification was @samp{%@var{ltr}
5915 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
5918 If @var{x} is a register, this macro should print the register's name.
5919 The names can be found in an array @code{reg_names} whose type is
5920 @code{char *[]}. @code{reg_names} is initialized from
5921 @code{REGISTER_NAMES}.
5923 When the machine description has a specification @samp{%@var{punct}}
5924 (a @samp{%} followed by a punctuation character), this macro is called
5925 with a null pointer for @var{x} and the punctuation character for
5928 @findex PRINT_OPERAND_PUNCT_VALID_P
5929 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
5930 A C expression which evaluates to true if @var{code} is a valid
5931 punctuation character for use in the @code{PRINT_OPERAND} macro. If
5932 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
5933 punctuation characters (except for the standard one, @samp{%}) are used
5936 @findex PRINT_OPERAND_ADDRESS
5937 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
5938 A C compound statement to output to stdio stream @var{stream} the
5939 assembler syntax for an instruction operand that is a memory reference
5940 whose address is @var{x}. @var{x} is an RTL expression.
5942 @cindex @code{ENCODE_SECTION_INFO} usage
5943 On some machines, the syntax for a symbolic address depends on the
5944 section that the address refers to. On these machines, define the macro
5945 @code{ENCODE_SECTION_INFO} to store the information into the
5946 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
5948 @findex DBR_OUTPUT_SEQEND
5949 @findex dbr_sequence_length
5950 @item DBR_OUTPUT_SEQEND(@var{file})
5951 A C statement, to be executed after all slot-filler instructions have
5952 been output. If necessary, call @code{dbr_sequence_length} to
5953 determine the number of slots filled in a sequence (zero if not
5954 currently outputting a sequence), to decide how many no-ops to output,
5957 Don't define this macro if it has nothing to do, but it is helpful in
5958 reading assembly output if the extent of the delay sequence is made
5959 explicit (e.g. with white space).
5961 @findex final_sequence
5962 Note that output routines for instructions with delay slots must be
5963 prepared to deal with not being output as part of a sequence (i.e.
5964 when the scheduling pass is not run, or when no slot fillers could be
5965 found.) The variable @code{final_sequence} is null when not
5966 processing a sequence, otherwise it contains the @code{sequence} rtx
5969 @findex REGISTER_PREFIX
5970 @findex LOCAL_LABEL_PREFIX
5971 @findex USER_LABEL_PREFIX
5972 @findex IMMEDIATE_PREFIX
5974 @item REGISTER_PREFIX
5975 @itemx LOCAL_LABEL_PREFIX
5976 @itemx USER_LABEL_PREFIX
5977 @itemx IMMEDIATE_PREFIX
5978 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
5979 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
5980 @file{final.c}). These are useful when a single @file{md} file must
5981 support multiple assembler formats. In that case, the various @file{tm.h}
5982 files can define these macros differently.
5984 @findex ASSEMBLER_DIALECT
5985 @item ASSEMBLER_DIALECT
5986 If your target supports multiple dialects of assembler language (such as
5987 different opcodes), define this macro as a C expression that gives the
5988 numeric index of the assembler language dialect to use, with zero as the
5991 If this macro is defined, you may use constructs of the form
5992 @samp{@{option0|option1|option2@dots{}@}} in the output
5993 templates of patterns (@pxref{Output Template}) or in the first argument
5994 of @code{asm_fprintf}. This construct outputs @samp{option0},
5995 @samp{option1} or @samp{option2}, etc., if the value of
5996 @code{ASSEMBLER_DIALECT} is zero, one or two, etc. Any special
5997 characters within these strings retain their usual meaning.
5999 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6000 @samp{@}} do not have any special meaning when used in templates or
6001 operands to @code{asm_fprintf}.
6003 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6004 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6005 the variations in assembly language syntax with that mechanism. Define
6006 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6007 if the syntax variant are larger and involve such things as different
6008 opcodes or operand order.
6010 @findex ASM_OUTPUT_REG_PUSH
6011 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6012 A C expression to output to @var{stream} some assembler code
6013 which will push hard register number @var{regno} onto the stack.
6014 The code need not be optimal, since this macro is used only when
6017 @findex ASM_OUTPUT_REG_POP
6018 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6019 A C expression to output to @var{stream} some assembler code
6020 which will pop hard register number @var{regno} off of the stack.
6021 The code need not be optimal, since this macro is used only when
6025 @node Dispatch Tables
6026 @subsection Output of Dispatch Tables
6028 @c prevent bad page break with this line
6029 This concerns dispatch tables.
6032 @cindex dispatch table
6033 @findex ASM_OUTPUT_ADDR_DIFF_ELT
6034 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6035 A C statement to output to the stdio stream @var{stream} an assembler
6036 pseudo-instruction to generate a difference between two labels.
6037 @var{value} and @var{rel} are the numbers of two internal labels. The
6038 definitions of these labels are output using
6039 @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
6040 way here. For example,
6043 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6044 @var{value}, @var{rel})
6047 You must provide this macro on machines where the addresses in a
6048 dispatch table are relative to the table's own address. If defined, GNU
6049 CC will also use this macro on all machines when producing PIC.
6050 @var{body} is the body of the ADDR_DIFF_VEC; it is provided so that the
6051 mode and flags can be read.
6053 @findex ASM_OUTPUT_ADDR_VEC_ELT
6054 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6055 This macro should be provided on machines where the addresses
6056 in a dispatch table are absolute.
6058 The definition should be a C statement to output to the stdio stream
6059 @var{stream} an assembler pseudo-instruction to generate a reference to
6060 a label. @var{value} is the number of an internal label whose
6061 definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
6065 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6068 @findex ASM_OUTPUT_CASE_LABEL
6069 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6070 Define this if the label before a jump-table needs to be output
6071 specially. The first three arguments are the same as for
6072 @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
6073 jump-table which follows (a @code{jump_insn} containing an
6074 @code{addr_vec} or @code{addr_diff_vec}).
6076 This feature is used on system V to output a @code{swbeg} statement
6079 If this macro is not defined, these labels are output with
6080 @code{ASM_OUTPUT_INTERNAL_LABEL}.
6082 @findex ASM_OUTPUT_CASE_END
6083 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6084 Define this if something special must be output at the end of a
6085 jump-table. The definition should be a C statement to be executed
6086 after the assembler code for the table is written. It should write
6087 the appropriate code to stdio stream @var{stream}. The argument
6088 @var{table} is the jump-table insn, and @var{num} is the label-number
6089 of the preceding label.
6091 If this macro is not defined, nothing special is output at the end of
6095 @node Exception Region Output
6096 @subsection Assembler Commands for Exception Regions
6098 @c prevent bad page break with this line
6100 This describes commands marking the start and the end of an exception
6104 @findex ASM_OUTPUT_EH_REGION_BEG
6105 @item ASM_OUTPUT_EH_REGION_BEG ()
6106 A C expression to output text to mark the start of an exception region.
6108 This macro need not be defined on most platforms.
6110 @findex ASM_OUTPUT_EH_REGION_END
6111 @item ASM_OUTPUT_EH_REGION_END ()
6112 A C expression to output text to mark the end of an exception region.
6114 This macro need not be defined on most platforms.
6116 @findex EXCEPTION_SECTION
6117 @item EXCEPTION_SECTION ()
6118 A C expression to switch to the section in which the main
6119 exception table is to be placed (@pxref{Sections}). The default is a
6120 section named @code{.gcc_except_table} on machines that support named
6121 sections via @code{ASM_OUTPUT_SECTION_NAME}, otherwise if @samp{-fpic}
6122 or @samp{-fPIC} is in effect, the @code{data_section}, otherwise the
6123 @code{readonly_data_section}.
6125 @findex EH_FRAME_SECTION_ASM_OP
6126 @item EH_FRAME_SECTION_ASM_OP
6127 If defined, a C string constant for the assembler operation to switch to
6128 the section for exception handling frame unwind information. If not
6129 defined, GNU CC will provide a default definition if the target supports
6130 named sections. @file{crtstuff.c} uses this macro to switch to the
6131 appropriate section.
6133 You should define this symbol if your target supports DWARF 2 frame
6134 unwind information and the default definition does not work.
6136 @findex OMIT_EH_TABLE
6137 @item OMIT_EH_TABLE ()
6138 A C expression that is nonzero if the normal exception table output
6141 This macro need not be defined on most platforms.
6143 @findex EH_TABLE_LOOKUP
6144 @item EH_TABLE_LOOKUP ()
6145 Alternate runtime support for looking up an exception at runtime and
6146 finding the associated handler, if the default method won't work.
6148 This macro need not be defined on most platforms.
6150 @findex DOESNT_NEED_UNWINDER
6151 @item DOESNT_NEED_UNWINDER
6152 A C expression that decides whether or not the current function needs to
6153 have a function unwinder generated for it. See the file @code{except.c}
6154 for details on when to define this, and how.
6156 @findex MASK_RETURN_ADDR
6157 @item MASK_RETURN_ADDR
6158 An rtx used to mask the return address found via RETURN_ADDR_RTX, so
6159 that it does not contain any extraneous set bits in it.
6161 @findex DWARF2_UNWIND_INFO
6162 @item DWARF2_UNWIND_INFO
6163 Define this macro to 0 if your target supports DWARF 2 frame unwind
6164 information, but it does not yet work with exception handling.
6165 Otherwise, if your target supports this information (if it defines
6166 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
6167 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
6170 If this macro is defined to 1, the DWARF 2 unwinder will be the default
6171 exception handling mechanism; otherwise, setjmp/longjmp will be used by
6174 If this macro is defined to anything, the DWARF 2 unwinder will be used
6175 instead of inline unwinders and __unwind_function in the non-setjmp case.
6179 @node Alignment Output
6180 @subsection Assembler Commands for Alignment
6182 @c prevent bad page break with this line
6183 This describes commands for alignment.
6186 @findex LABEL_ALIGN_AFTER_BARRIER
6187 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
6188 The alignment (log base 2) to put in front of @var{label}, which follows
6191 This macro need not be defined if you don't want any special alignment
6192 to be done at such a time. Most machine descriptions do not currently
6196 @item LOOP_ALIGN (@var{label})
6197 The alignment (log base 2) to put in front of @var{label}, which follows
6198 a NOTE_INSN_LOOP_BEG note.
6200 This macro need not be defined if you don't want any special alignment
6201 to be done at such a time. Most machine descriptions do not currently
6205 @item LABEL_ALIGN (@var{label})
6206 The alignment (log base 2) to put in front of @var{label}.
6207 If LABEL_ALIGN_AFTER_BARRIER / LOOP_ALIGN specify a different alignment,
6208 the maximum of the specified values is used.
6210 @findex ASM_OUTPUT_SKIP
6211 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6212 A C statement to output to the stdio stream @var{stream} an assembler
6213 instruction to advance the location counter by @var{nbytes} bytes.
6214 Those bytes should be zero when loaded. @var{nbytes} will be a C
6215 expression of type @code{int}.
6217 @findex ASM_NO_SKIP_IN_TEXT
6218 @item ASM_NO_SKIP_IN_TEXT
6219 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6220 text section because it fails to put zeros in the bytes that are skipped.
6221 This is true on many Unix systems, where the pseudo--op to skip bytes
6222 produces no-op instructions rather than zeros when used in the text
6225 @findex ASM_OUTPUT_ALIGN
6226 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6227 A C statement to output to the stdio stream @var{stream} an assembler
6228 command to advance the location counter to a multiple of 2 to the
6229 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
6233 @node Debugging Info
6234 @section Controlling Debugging Information Format
6236 @c prevent bad page break with this line
6237 This describes how to specify debugging information.
6240 * All Debuggers:: Macros that affect all debugging formats uniformly.
6241 * DBX Options:: Macros enabling specific options in DBX format.
6242 * DBX Hooks:: Hook macros for varying DBX format.
6243 * File Names and DBX:: Macros controlling output of file names in DBX format.
6244 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
6248 @subsection Macros Affecting All Debugging Formats
6250 @c prevent bad page break with this line
6251 These macros affect all debugging formats.
6254 @findex DBX_REGISTER_NUMBER
6255 @item DBX_REGISTER_NUMBER (@var{regno})
6256 A C expression that returns the DBX register number for the compiler
6257 register number @var{regno}. In simple cases, the value of this
6258 expression may be @var{regno} itself. But sometimes there are some
6259 registers that the compiler knows about and DBX does not, or vice
6260 versa. In such cases, some register may need to have one number in
6261 the compiler and another for DBX.
6263 If two registers have consecutive numbers inside GNU CC, and they can be
6264 used as a pair to hold a multiword value, then they @emph{must} have
6265 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
6266 Otherwise, debuggers will be unable to access such a pair, because they
6267 expect register pairs to be consecutive in their own numbering scheme.
6269 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
6270 does not preserve register pairs, then what you must do instead is
6271 redefine the actual register numbering scheme.
6273 @findex DEBUGGER_AUTO_OFFSET
6274 @item DEBUGGER_AUTO_OFFSET (@var{x})
6275 A C expression that returns the integer offset value for an automatic
6276 variable having address @var{x} (an RTL expression). The default
6277 computation assumes that @var{x} is based on the frame-pointer and
6278 gives the offset from the frame-pointer. This is required for targets
6279 that produce debugging output for DBX or COFF-style debugging output
6280 for SDB and allow the frame-pointer to be eliminated when the
6281 @samp{-g} options is used.
6283 @findex DEBUGGER_ARG_OFFSET
6284 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6285 A C expression that returns the integer offset value for an argument
6286 having address @var{x} (an RTL expression). The nominal offset is
6289 @findex PREFERRED_DEBUGGING_TYPE
6290 @item PREFERRED_DEBUGGING_TYPE
6291 A C expression that returns the type of debugging output GNU CC produces
6292 when the user specifies @samp{-g} or @samp{-ggdb}. Define this if you
6293 have arranged for GNU CC to support more than one format of debugging
6294 output. Currently, the allowable values are @code{DBX_DEBUG},
6295 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG}, and
6298 If the user specifies @samp{-ggdb}, then there are two special cases
6299 where this macro's value is ignored and another value is substituted.
6300 If @code{DWARF2_DEBUGGING_INFO} is defined and
6301 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, then
6302 @code{DWARF2_DEBUG} is substituted. Otherwise, if
6303 @code{DBX_DEBUGGING_INFO} is defined, then @code{DBX_DEBUG} is
6306 The value of this macro only affects the default debugging output; the
6307 user can always get a specific type of output by using @samp{-gstabs},
6308 @samp{-gcoff}, @samp{-gdwarf-1}, @samp{-gdwarf-2}, or @samp{-gxcoff}.
6312 @subsection Specific Options for DBX Output
6314 @c prevent bad page break with this line
6315 These are specific options for DBX output.
6318 @findex DBX_DEBUGGING_INFO
6319 @item DBX_DEBUGGING_INFO
6320 Define this macro if GNU CC should produce debugging output for DBX
6321 in response to the @samp{-g} option.
6323 @findex XCOFF_DEBUGGING_INFO
6324 @item XCOFF_DEBUGGING_INFO
6325 Define this macro if GNU CC should produce XCOFF format debugging output
6326 in response to the @samp{-g} option. This is a variant of DBX format.
6328 @findex DEFAULT_GDB_EXTENSIONS
6329 @item DEFAULT_GDB_EXTENSIONS
6330 Define this macro to control whether GNU CC should by default generate
6331 GDB's extended version of DBX debugging information (assuming DBX-format
6332 debugging information is enabled at all). If you don't define the
6333 macro, the default is 1: always generate the extended information
6334 if there is any occasion to.
6336 @findex DEBUG_SYMS_TEXT
6337 @item DEBUG_SYMS_TEXT
6338 Define this macro if all @code{.stabs} commands should be output while
6339 in the text section.
6341 @findex ASM_STABS_OP
6343 A C string constant naming the assembler pseudo op to use instead of
6344 @code{.stabs} to define an ordinary debugging symbol. If you don't
6345 define this macro, @code{.stabs} is used. This macro applies only to
6346 DBX debugging information format.
6348 @findex ASM_STABD_OP
6350 A C string constant naming the assembler pseudo op to use instead of
6351 @code{.stabd} to define a debugging symbol whose value is the current
6352 location. If you don't define this macro, @code{.stabd} is used.
6353 This macro applies only to DBX debugging information format.
6355 @findex ASM_STABN_OP
6357 A C string constant naming the assembler pseudo op to use instead of
6358 @code{.stabn} to define a debugging symbol with no name. If you don't
6359 define this macro, @code{.stabn} is used. This macro applies only to
6360 DBX debugging information format.
6362 @findex DBX_NO_XREFS
6364 Define this macro if DBX on your system does not support the construct
6365 @samp{xs@var{tagname}}. On some systems, this construct is used to
6366 describe a forward reference to a structure named @var{tagname}.
6367 On other systems, this construct is not supported at all.
6369 @findex DBX_CONTIN_LENGTH
6370 @item DBX_CONTIN_LENGTH
6371 A symbol name in DBX-format debugging information is normally
6372 continued (split into two separate @code{.stabs} directives) when it
6373 exceeds a certain length (by default, 80 characters). On some
6374 operating systems, DBX requires this splitting; on others, splitting
6375 must not be done. You can inhibit splitting by defining this macro
6376 with the value zero. You can override the default splitting-length by
6377 defining this macro as an expression for the length you desire.
6379 @findex DBX_CONTIN_CHAR
6380 @item DBX_CONTIN_CHAR
6381 Normally continuation is indicated by adding a @samp{\} character to
6382 the end of a @code{.stabs} string when a continuation follows. To use
6383 a different character instead, define this macro as a character
6384 constant for the character you want to use. Do not define this macro
6385 if backslash is correct for your system.
6387 @findex DBX_STATIC_STAB_DATA_SECTION
6388 @item DBX_STATIC_STAB_DATA_SECTION
6389 Define this macro if it is necessary to go to the data section before
6390 outputting the @samp{.stabs} pseudo-op for a non-global static
6393 @findex DBX_TYPE_DECL_STABS_CODE
6394 @item DBX_TYPE_DECL_STABS_CODE
6395 The value to use in the ``code'' field of the @code{.stabs} directive
6396 for a typedef. The default is @code{N_LSYM}.
6398 @findex DBX_STATIC_CONST_VAR_CODE
6399 @item DBX_STATIC_CONST_VAR_CODE
6400 The value to use in the ``code'' field of the @code{.stabs} directive
6401 for a static variable located in the text section. DBX format does not
6402 provide any ``right'' way to do this. The default is @code{N_FUN}.
6404 @findex DBX_REGPARM_STABS_CODE
6405 @item DBX_REGPARM_STABS_CODE
6406 The value to use in the ``code'' field of the @code{.stabs} directive
6407 for a parameter passed in registers. DBX format does not provide any
6408 ``right'' way to do this. The default is @code{N_RSYM}.
6410 @findex DBX_REGPARM_STABS_LETTER
6411 @item DBX_REGPARM_STABS_LETTER
6412 The letter to use in DBX symbol data to identify a symbol as a parameter
6413 passed in registers. DBX format does not customarily provide any way to
6414 do this. The default is @code{'P'}.
6416 @findex DBX_MEMPARM_STABS_LETTER
6417 @item DBX_MEMPARM_STABS_LETTER
6418 The letter to use in DBX symbol data to identify a symbol as a stack
6419 parameter. The default is @code{'p'}.
6421 @findex DBX_FUNCTION_FIRST
6422 @item DBX_FUNCTION_FIRST
6423 Define this macro if the DBX information for a function and its
6424 arguments should precede the assembler code for the function. Normally,
6425 in DBX format, the debugging information entirely follows the assembler
6428 @findex DBX_LBRAC_FIRST
6429 @item DBX_LBRAC_FIRST
6430 Define this macro if the @code{N_LBRAC} symbol for a block should
6431 precede the debugging information for variables and functions defined in
6432 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
6435 @findex DBX_BLOCKS_FUNCTION_RELATIVE
6436 @item DBX_BLOCKS_FUNCTION_RELATIVE
6437 Define this macro if the value of a symbol describing the scope of a
6438 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
6439 of the enclosing function. Normally, GNU C uses an absolute address.
6441 @findex DBX_USE_BINCL
6443 Define this macro if GNU C should generate @code{N_BINCL} and
6444 @code{N_EINCL} stabs for included header files, as on Sun systems. This
6445 macro also directs GNU C to output a type number as a pair of a file
6446 number and a type number within the file. Normally, GNU C does not
6447 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
6448 number for a type number.
6452 @subsection Open-Ended Hooks for DBX Format
6454 @c prevent bad page break with this line
6455 These are hooks for DBX format.
6458 @findex DBX_OUTPUT_LBRAC
6459 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
6460 Define this macro to say how to output to @var{stream} the debugging
6461 information for the start of a scope level for variable names. The
6462 argument @var{name} is the name of an assembler symbol (for use with
6463 @code{assemble_name}) whose value is the address where the scope begins.
6465 @findex DBX_OUTPUT_RBRAC
6466 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
6467 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
6469 @findex DBX_OUTPUT_ENUM
6470 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
6471 Define this macro if the target machine requires special handling to
6472 output an enumeration type. The definition should be a C statement
6473 (sans semicolon) to output the appropriate information to @var{stream}
6474 for the type @var{type}.
6476 @findex DBX_OUTPUT_FUNCTION_END
6477 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
6478 Define this macro if the target machine requires special output at the
6479 end of the debugging information for a function. The definition should
6480 be a C statement (sans semicolon) to output the appropriate information
6481 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
6484 @findex DBX_OUTPUT_STANDARD_TYPES
6485 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
6486 Define this macro if you need to control the order of output of the
6487 standard data types at the beginning of compilation. The argument
6488 @var{syms} is a @code{tree} which is a chain of all the predefined
6489 global symbols, including names of data types.
6491 Normally, DBX output starts with definitions of the types for integers
6492 and characters, followed by all the other predefined types of the
6493 particular language in no particular order.
6495 On some machines, it is necessary to output different particular types
6496 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
6497 those symbols in the necessary order. Any predefined types that you
6498 don't explicitly output will be output afterward in no particular order.
6500 Be careful not to define this macro so that it works only for C. There
6501 are no global variables to access most of the built-in types, because
6502 another language may have another set of types. The way to output a
6503 particular type is to look through @var{syms} to see if you can find it.
6509 for (decl = syms; decl; decl = TREE_CHAIN (decl))
6510 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
6512 dbxout_symbol (decl);
6518 This does nothing if the expected type does not exist.
6520 See the function @code{init_decl_processing} in @file{c-decl.c} to find
6521 the names to use for all the built-in C types.
6523 Here is another way of finding a particular type:
6525 @c this is still overfull. --mew 10feb93
6529 for (decl = syms; decl; decl = TREE_CHAIN (decl))
6530 if (TREE_CODE (decl) == TYPE_DECL
6531 && (TREE_CODE (TREE_TYPE (decl))
6533 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
6534 && TYPE_UNSIGNED (TREE_TYPE (decl)))
6536 /* @r{This must be @code{unsigned short}.} */
6537 dbxout_symbol (decl);
6543 @findex NO_DBX_FUNCTION_END
6544 @item NO_DBX_FUNCTION_END
6545 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
6546 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extention construct.
6547 On those machines, define this macro to turn this feature off without
6548 disturbing the rest of the gdb extensions.
6552 @node File Names and DBX
6553 @subsection File Names in DBX Format
6555 @c prevent bad page break with this line
6556 This describes file names in DBX format.
6559 @findex DBX_WORKING_DIRECTORY
6560 @item DBX_WORKING_DIRECTORY
6561 Define this if DBX wants to have the current directory recorded in each
6564 Note that the working directory is always recorded if GDB extensions are
6567 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
6568 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
6569 A C statement to output DBX debugging information to the stdio stream
6570 @var{stream} which indicates that file @var{name} is the main source
6571 file---the file specified as the input file for compilation.
6572 This macro is called only once, at the beginning of compilation.
6574 This macro need not be defined if the standard form of output
6575 for DBX debugging information is appropriate.
6577 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
6578 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
6579 A C statement to output DBX debugging information to the stdio stream
6580 @var{stream} which indicates that the current directory during
6581 compilation is named @var{name}.
6583 This macro need not be defined if the standard form of output
6584 for DBX debugging information is appropriate.
6586 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
6587 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
6588 A C statement to output DBX debugging information at the end of
6589 compilation of the main source file @var{name}.
6591 If you don't define this macro, nothing special is output at the end
6592 of compilation, which is correct for most machines.
6594 @findex DBX_OUTPUT_SOURCE_FILENAME
6595 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6596 A C statement to output DBX debugging information to the stdio stream
6597 @var{stream} which indicates that file @var{name} is the current source
6598 file. This output is generated each time input shifts to a different
6599 source file as a result of @samp{#include}, the end of an included file,
6600 or a @samp{#line} command.
6602 This macro need not be defined if the standard form of output
6603 for DBX debugging information is appropriate.
6608 @subsection Macros for SDB and DWARF Output
6610 @c prevent bad page break with this line
6611 Here are macros for SDB and DWARF output.
6614 @findex SDB_DEBUGGING_INFO
6615 @item SDB_DEBUGGING_INFO
6616 Define this macro if GNU CC should produce COFF-style debugging output
6617 for SDB in response to the @samp{-g} option.
6619 @findex DWARF_DEBUGGING_INFO
6620 @item DWARF_DEBUGGING_INFO
6621 Define this macro if GNU CC should produce dwarf format debugging output
6622 in response to the @samp{-g} option.
6624 @findex DWARF2_DEBUGGING_INFO
6625 @item DWARF2_DEBUGGING_INFO
6626 Define this macro if GNU CC should produce dwarf version 2 format
6627 debugging output in response to the @samp{-g} option.
6629 To support optional call frame debugging information, you must also
6630 define @code{INCOMING_RETURN_ADDR_RTX} and either set
6631 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
6632 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
6633 as appropriate from @code{FUNCTION_PROLOGUE} if you don't.
6635 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
6636 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
6637 Define this macro if the linker does not work with dwarf version 2.
6638 Normally, if the user specifies only @samp{-ggdb}, then GNU CC will use
6639 dwarf version 2 if available; this macro causes GNU CC to use the format
6640 specified by @code{PREFERRED_DEBUGGING_TYPE} instead.
6642 @findex PUT_SDB_@dots{}
6643 @item PUT_SDB_@dots{}
6644 Define these macros to override the assembler syntax for the special
6645 SDB assembler directives. See @file{sdbout.c} for a list of these
6646 macros and their arguments. If the standard syntax is used, you need
6647 not define them yourself.
6651 Some assemblers do not support a semicolon as a delimiter, even between
6652 SDB assembler directives. In that case, define this macro to be the
6653 delimiter to use (usually @samp{\n}). It is not necessary to define
6654 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
6657 @findex SDB_GENERATE_FAKE
6658 @item SDB_GENERATE_FAKE
6659 Define this macro to override the usual method of constructing a dummy
6660 name for anonymous structure and union types. See @file{sdbout.c} for
6663 @findex SDB_ALLOW_UNKNOWN_REFERENCES
6664 @item SDB_ALLOW_UNKNOWN_REFERENCES
6665 Define this macro to allow references to unknown structure,
6666 union, or enumeration tags to be emitted. Standard COFF does not
6667 allow handling of unknown references, MIPS ECOFF has support for
6670 @findex SDB_ALLOW_FORWARD_REFERENCES
6671 @item SDB_ALLOW_FORWARD_REFERENCES
6672 Define this macro to allow references to structure, union, or
6673 enumeration tags that have not yet been seen to be handled. Some
6674 assemblers choke if forward tags are used, while some require it.
6677 @node Cross-compilation
6678 @section Cross Compilation and Floating Point
6679 @cindex cross compilation and floating point
6680 @cindex floating point and cross compilation
6682 While all modern machines use 2's complement representation for integers,
6683 there are a variety of representations for floating point numbers. This
6684 means that in a cross-compiler the representation of floating point numbers
6685 in the compiled program may be different from that used in the machine
6686 doing the compilation.
6689 Because different representation systems may offer different amounts of
6690 range and precision, the cross compiler cannot safely use the host
6691 machine's floating point arithmetic. Therefore, floating point constants
6692 must be represented in the target machine's format. This means that the
6693 cross compiler cannot use @code{atof} to parse a floating point constant;
6694 it must have its own special routine to use instead. Also, constant
6695 folding must emulate the target machine's arithmetic (or must not be done
6698 The macros in the following table should be defined only if you are cross
6699 compiling between different floating point formats.
6701 Otherwise, don't define them. Then default definitions will be set up which
6702 use @code{double} as the data type, @code{==} to test for equality, etc.
6704 You don't need to worry about how many times you use an operand of any
6705 of these macros. The compiler never uses operands which have side effects.
6708 @findex REAL_VALUE_TYPE
6709 @item REAL_VALUE_TYPE
6710 A macro for the C data type to be used to hold a floating point value
6711 in the target machine's format. Typically this would be a
6712 @code{struct} containing an array of @code{int}.
6714 @findex REAL_VALUES_EQUAL
6715 @item REAL_VALUES_EQUAL (@var{x}, @var{y})
6716 A macro for a C expression which compares for equality the two values,
6717 @var{x} and @var{y}, both of type @code{REAL_VALUE_TYPE}.
6719 @findex REAL_VALUES_LESS
6720 @item REAL_VALUES_LESS (@var{x}, @var{y})
6721 A macro for a C expression which tests whether @var{x} is less than
6722 @var{y}, both values being of type @code{REAL_VALUE_TYPE} and
6723 interpreted as floating point numbers in the target machine's
6726 @findex REAL_VALUE_LDEXP
6728 @item REAL_VALUE_LDEXP (@var{x}, @var{scale})
6729 A macro for a C expression which performs the standard library
6730 function @code{ldexp}, but using the target machine's floating point
6731 representation. Both @var{x} and the value of the expression have
6732 type @code{REAL_VALUE_TYPE}. The second argument, @var{scale}, is an
6735 @findex REAL_VALUE_FIX
6736 @item REAL_VALUE_FIX (@var{x})
6737 A macro whose definition is a C expression to convert the target-machine
6738 floating point value @var{x} to a signed integer. @var{x} has type
6739 @code{REAL_VALUE_TYPE}.
6741 @findex REAL_VALUE_UNSIGNED_FIX
6742 @item REAL_VALUE_UNSIGNED_FIX (@var{x})
6743 A macro whose definition is a C expression to convert the target-machine
6744 floating point value @var{x} to an unsigned integer. @var{x} has type
6745 @code{REAL_VALUE_TYPE}.
6747 @findex REAL_VALUE_RNDZINT
6748 @item REAL_VALUE_RNDZINT (@var{x})
6749 A macro whose definition is a C expression to round the target-machine
6750 floating point value @var{x} towards zero to an integer value (but still
6751 as a floating point number). @var{x} has type @code{REAL_VALUE_TYPE},
6752 and so does the value.
6754 @findex REAL_VALUE_UNSIGNED_RNDZINT
6755 @item REAL_VALUE_UNSIGNED_RNDZINT (@var{x})
6756 A macro whose definition is a C expression to round the target-machine
6757 floating point value @var{x} towards zero to an unsigned integer value
6758 (but still represented as a floating point number). @var{x} has type
6759 @code{REAL_VALUE_TYPE}, and so does the value.
6761 @findex REAL_VALUE_ATOF
6762 @item REAL_VALUE_ATOF (@var{string}, @var{mode})
6763 A macro for a C expression which converts @var{string}, an expression of
6764 type @code{char *}, into a floating point number in the target machine's
6765 representation for mode @var{mode}. The value has type
6766 @code{REAL_VALUE_TYPE}.
6768 @findex REAL_INFINITY
6770 Define this macro if infinity is a possible floating point value, and
6771 therefore division by 0 is legitimate.
6773 @findex REAL_VALUE_ISINF
6775 @item REAL_VALUE_ISINF (@var{x})
6776 A macro for a C expression which determines whether @var{x}, a floating
6777 point value, is infinity. The value has type @code{int}.
6778 By default, this is defined to call @code{isinf}.
6780 @findex REAL_VALUE_ISNAN
6782 @item REAL_VALUE_ISNAN (@var{x})
6783 A macro for a C expression which determines whether @var{x}, a floating
6784 point value, is a ``nan'' (not-a-number). The value has type
6785 @code{int}. By default, this is defined to call @code{isnan}.
6788 @cindex constant folding and floating point
6789 Define the following additional macros if you want to make floating
6790 point constant folding work while cross compiling. If you don't
6791 define them, cross compilation is still possible, but constant folding
6792 will not happen for floating point values.
6795 @findex REAL_ARITHMETIC
6796 @item REAL_ARITHMETIC (@var{output}, @var{code}, @var{x}, @var{y})
6797 A macro for a C statement which calculates an arithmetic operation of
6798 the two floating point values @var{x} and @var{y}, both of type
6799 @code{REAL_VALUE_TYPE} in the target machine's representation, to
6800 produce a result of the same type and representation which is stored
6801 in @var{output} (which will be a variable).
6803 The operation to be performed is specified by @var{code}, a tree code
6804 which will always be one of the following: @code{PLUS_EXPR},
6805 @code{MINUS_EXPR}, @code{MULT_EXPR}, @code{RDIV_EXPR},
6806 @code{MAX_EXPR}, @code{MIN_EXPR}.@refill
6808 @cindex overflow while constant folding
6809 The expansion of this macro is responsible for checking for overflow.
6810 If overflow happens, the macro expansion should execute the statement
6811 @code{return 0;}, which indicates the inability to perform the
6812 arithmetic operation requested.
6814 @findex REAL_VALUE_NEGATE
6815 @item REAL_VALUE_NEGATE (@var{x})
6816 A macro for a C expression which returns the negative of the floating
6817 point value @var{x}. Both @var{x} and the value of the expression
6818 have type @code{REAL_VALUE_TYPE} and are in the target machine's
6819 floating point representation.
6821 There is no way for this macro to report overflow, since overflow
6822 can't happen in the negation operation.
6824 @findex REAL_VALUE_TRUNCATE
6825 @item REAL_VALUE_TRUNCATE (@var{mode}, @var{x})
6826 A macro for a C expression which converts the floating point value
6827 @var{x} to mode @var{mode}.
6829 Both @var{x} and the value of the expression are in the target machine's
6830 floating point representation and have type @code{REAL_VALUE_TYPE}.
6831 However, the value should have an appropriate bit pattern to be output
6832 properly as a floating constant whose precision accords with mode
6835 There is no way for this macro to report overflow.
6837 @findex REAL_VALUE_TO_INT
6838 @item REAL_VALUE_TO_INT (@var{low}, @var{high}, @var{x})
6839 A macro for a C expression which converts a floating point value
6840 @var{x} into a double-precision integer which is then stored into
6841 @var{low} and @var{high}, two variables of type @var{int}.
6843 @item REAL_VALUE_FROM_INT (@var{x}, @var{low}, @var{high}, @var{mode})
6844 @findex REAL_VALUE_FROM_INT
6845 A macro for a C expression which converts a double-precision integer
6846 found in @var{low} and @var{high}, two variables of type @var{int},
6847 into a floating point value which is then stored into @var{x}.
6848 The value is in the target machine's representation for mode @var{mode}
6849 and has the type @code{REAL_VALUE_TYPE}.
6853 @section Miscellaneous Parameters
6854 @cindex parameters, miscellaneous
6856 @c prevent bad page break with this line
6857 Here are several miscellaneous parameters.
6860 @item PREDICATE_CODES
6861 @findex PREDICATE_CODES
6862 Define this if you have defined special-purpose predicates in the file
6863 @file{@var{machine}.c}. This macro is called within an initializer of an
6864 array of structures. The first field in the structure is the name of a
6865 predicate and the second field is an array of rtl codes. For each
6866 predicate, list all rtl codes that can be in expressions matched by the
6867 predicate. The list should have a trailing comma. Here is an example
6868 of two entries in the list for a typical RISC machine:
6871 #define PREDICATE_CODES \
6872 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
6873 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
6876 Defining this macro does not affect the generated code (however,
6877 incorrect definitions that omit an rtl code that may be matched by the
6878 predicate can cause the compiler to malfunction). Instead, it allows
6879 the table built by @file{genrecog} to be more compact and efficient,
6880 thus speeding up the compiler. The most important predicates to include
6881 in the list specified by this macro are those used in the most insn
6884 @findex CASE_VECTOR_MODE
6885 @item CASE_VECTOR_MODE
6886 An alias for a machine mode name. This is the machine mode that
6887 elements of a jump-table should have.
6889 @findex CASE_VECTOR_SHORTEN_MODE
6890 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
6891 Optional: return the preferred mode for an @code{addr_diff_vec}
6892 when the minimum and maximum offset are known. If you define this,
6893 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
6894 To make this work, you also have to define INSN_ALIGN and
6895 make the alignment for @code{addr_diff_vec} explicit.
6896 The @var{body} argument is provided so that teh offset_unsigned and scale
6897 flags can be updated.
6899 @findex CASE_VECTOR_PC_RELATIVE
6900 @item CASE_VECTOR_PC_RELATIVE
6901 Define this macro to be a C expression to indicate when jump-tables
6902 should contain relative addresses. If jump-tables never contain
6903 relative addresses, then you need not define this macro.
6905 @findex CASE_DROPS_THROUGH
6906 @item CASE_DROPS_THROUGH
6907 Define this if control falls through a @code{case} insn when the index
6908 value is out of range. This means the specified default-label is
6909 actually ignored by the @code{case} insn proper.
6911 @findex CASE_VALUES_THRESHOLD
6912 @item CASE_VALUES_THRESHOLD
6913 Define this to be the smallest number of different values for which it
6914 is best to use a jump-table instead of a tree of conditional branches.
6915 The default is four for machines with a @code{casesi} instruction and
6916 five otherwise. This is best for most machines.
6918 @findex WORD_REGISTER_OPERATIONS
6919 @item WORD_REGISTER_OPERATIONS
6920 Define this macro if operations between registers with integral mode
6921 smaller than a word are always performed on the entire register.
6922 Most RISC machines have this property and most CISC machines do not.
6924 @findex LOAD_EXTEND_OP
6925 @item LOAD_EXTEND_OP (@var{mode})
6926 Define this macro to be a C expression indicating when insns that read
6927 memory in @var{mode}, an integral mode narrower than a word, set the
6928 bits outside of @var{mode} to be either the sign-extension or the
6929 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
6930 of @var{mode} for which the
6931 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
6932 @code{NIL} for other modes.
6934 This macro is not called with @var{mode} non-integral or with a width
6935 greater than or equal to @code{BITS_PER_WORD}, so you may return any
6936 value in this case. Do not define this macro if it would always return
6937 @code{NIL}. On machines where this macro is defined, you will normally
6938 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
6940 @findex SHORT_IMMEDIATES_SIGN_EXTEND
6941 @item SHORT_IMMEDIATES_SIGN_EXTEND
6942 Define this macro if loading short immediate values into registers sign
6945 @findex IMPLICIT_FIX_EXPR
6946 @item IMPLICIT_FIX_EXPR
6947 An alias for a tree code that should be used by default for conversion
6948 of floating point values to fixed point. Normally,
6949 @code{FIX_ROUND_EXPR} is used.@refill
6951 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
6952 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
6953 Define this macro if the same instructions that convert a floating
6954 point number to a signed fixed point number also convert validly to an
6957 @findex EASY_DIV_EXPR
6959 An alias for a tree code that is the easiest kind of division to
6960 compile code for in the general case. It may be
6961 @code{TRUNC_DIV_EXPR}, @code{FLOOR_DIV_EXPR}, @code{CEIL_DIV_EXPR} or
6962 @code{ROUND_DIV_EXPR}. These four division operators differ in how
6963 they round the result to an integer. @code{EASY_DIV_EXPR} is used
6964 when it is permissible to use any of those kinds of division and the
6965 choice should be made on the basis of efficiency.@refill
6969 The maximum number of bytes that a single instruction can move quickly
6970 between memory and registers or between two memory locations.
6972 @findex MAX_MOVE_MAX
6974 The maximum number of bytes that a single instruction can move quickly
6975 between memory and registers or between two memory locations. If this
6976 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
6977 constant value that is the largest value that @code{MOVE_MAX} can have
6980 @findex SHIFT_COUNT_TRUNCATED
6981 @item SHIFT_COUNT_TRUNCATED
6982 A C expression that is nonzero if on this machine the number of bits
6983 actually used for the count of a shift operation is equal to the number
6984 of bits needed to represent the size of the object being shifted. When
6985 this macro is non-zero, the compiler will assume that it is safe to omit
6986 a sign-extend, zero-extend, and certain bitwise `and' instructions that
6987 truncates the count of a shift operation. On machines that have
6988 instructions that act on bitfields at variable positions, which may
6989 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
6990 also enables deletion of truncations of the values that serve as
6991 arguments to bitfield instructions.
6993 If both types of instructions truncate the count (for shifts) and
6994 position (for bitfield operations), or if no variable-position bitfield
6995 instructions exist, you should define this macro.
6997 However, on some machines, such as the 80386 and the 680x0, truncation
6998 only applies to shift operations and not the (real or pretended)
6999 bitfield operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7000 such machines. Instead, add patterns to the @file{md} file that include
7001 the implied truncation of the shift instructions.
7003 You need not define this macro if it would always have the value of zero.
7005 @findex TRULY_NOOP_TRUNCATION
7006 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
7007 A C expression which is nonzero if on this machine it is safe to
7008 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
7009 bits (where @var{outprec} is smaller than @var{inprec}) by merely
7010 operating on it as if it had only @var{outprec} bits.
7012 On many machines, this expression can be 1.
7014 @c rearranged this, removed the phrase "it is reported that". this was
7015 @c to fix an overfull hbox. --mew 10feb93
7016 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
7017 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
7018 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
7019 such cases may improve things.
7021 @findex STORE_FLAG_VALUE
7022 @item STORE_FLAG_VALUE
7023 A C expression describing the value returned by a comparison operator
7024 with an integral mode and stored by a store-flag instruction
7025 (@samp{s@var{cond}}) when the condition is true. This description must
7026 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
7027 comparison operators whose results have a @code{MODE_INT} mode.
7029 A value of 1 or -1 means that the instruction implementing the
7030 comparison operator returns exactly 1 or -1 when the comparison is true
7031 and 0 when the comparison is false. Otherwise, the value indicates
7032 which bits of the result are guaranteed to be 1 when the comparison is
7033 true. This value is interpreted in the mode of the comparison
7034 operation, which is given by the mode of the first operand in the
7035 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
7036 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7039 If @code{STORE_FLAG_VALUE} is neither 1 or -1, the compiler will
7040 generate code that depends only on the specified bits. It can also
7041 replace comparison operators with equivalent operations if they cause
7042 the required bits to be set, even if the remaining bits are undefined.
7043 For example, on a machine whose comparison operators return an
7044 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7045 @samp{0x80000000}, saying that just the sign bit is relevant, the
7049 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7056 (ashift:SI @var{x} (const_int @var{n}))
7060 where @var{n} is the appropriate shift count to move the bit being
7061 tested into the sign bit.
7063 There is no way to describe a machine that always sets the low-order bit
7064 for a true value, but does not guarantee the value of any other bits,
7065 but we do not know of any machine that has such an instruction. If you
7066 are trying to port GNU CC to such a machine, include an instruction to
7067 perform a logical-and of the result with 1 in the pattern for the
7068 comparison operators and let us know
7070 (@pxref{Bug Reporting,,How to Report Bugs}).
7073 (@pxref{Bug Reporting,,How to Report Bugs,gcc.info,Using GCC}).
7076 Often, a machine will have multiple instructions that obtain a value
7077 from a comparison (or the condition codes). Here are rules to guide the
7078 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7083 Use the shortest sequence that yields a valid definition for
7084 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7085 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7086 comparison operators to do so because there may be opportunities to
7087 combine the normalization with other operations.
7090 For equal-length sequences, use a value of 1 or -1, with -1 being
7091 slightly preferred on machines with expensive jumps and 1 preferred on
7095 As a second choice, choose a value of @samp{0x80000001} if instructions
7096 exist that set both the sign and low-order bits but do not define the
7100 Otherwise, use a value of @samp{0x80000000}.
7103 Many machines can produce both the value chosen for
7104 @code{STORE_FLAG_VALUE} and its negation in the same number of
7105 instructions. On those machines, you should also define a pattern for
7106 those cases, e.g., one matching
7109 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7112 Some machines can also perform @code{and} or @code{plus} operations on
7113 condition code values with less instructions than the corresponding
7114 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
7115 machines, define the appropriate patterns. Use the names @code{incscc}
7116 and @code{decscc}, respectively, for the patterns which perform
7117 @code{plus} or @code{minus} operations on condition code values. See
7118 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
7119 find such instruction sequences on other machines.
7121 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7124 @findex FLOAT_STORE_FLAG_VALUE
7125 @item FLOAT_STORE_FLAG_VALUE
7126 A C expression that gives a non-zero floating point value that is
7127 returned when comparison operators with floating-point results are true.
7128 Define this macro on machine that have comparison operations that return
7129 floating-point values. If there are no such operations, do not define
7134 An alias for the machine mode for pointers. On most machines, define
7135 this to be the integer mode corresponding to the width of a hardware
7136 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7137 On some machines you must define this to be one of the partial integer
7138 modes, such as @code{PSImode}.
7140 The width of @code{Pmode} must be at least as large as the value of
7141 @code{POINTER_SIZE}. If it is not equal, you must define the macro
7142 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7145 @findex FUNCTION_MODE
7147 An alias for the machine mode used for memory references to functions
7148 being called, in @code{call} RTL expressions. On most machines this
7149 should be @code{QImode}.
7151 @findex INTEGRATE_THRESHOLD
7152 @item INTEGRATE_THRESHOLD (@var{decl})
7153 A C expression for the maximum number of instructions above which the
7154 function @var{decl} should not be inlined. @var{decl} is a
7155 @code{FUNCTION_DECL} node.
7157 The default definition of this macro is 64 plus 8 times the number of
7158 arguments that the function accepts. Some people think a larger
7159 threshold should be used on RISC machines.
7161 @findex SCCS_DIRECTIVE
7162 @item SCCS_DIRECTIVE
7163 Define this if the preprocessor should ignore @code{#sccs} directives
7164 and print no error message.
7166 @findex NO_IMPLICIT_EXTERN_C
7167 @item NO_IMPLICIT_EXTERN_C
7168 Define this macro if the system header files support C++ as well as C.
7169 This macro inhibits the usual method of using system header files in
7170 C++, which is to pretend that the file's contents are enclosed in
7171 @samp{extern "C" @{@dots{}@}}.
7173 @findex HANDLE_PRAGMA
7176 @item HANDLE_PRAGMA (@var{stream}, @var{node})
7177 Define this macro if you want to implement any pragmas. If defined, it
7178 is a C expression whose value is 1 if the pragma was handled by the function.
7179 The argument @var{stream} is the stdio input stream from which the source text
7180 can be read. @var{node} is the tree node for the identifier after the
7183 It is generally a bad idea to implement new uses of @code{#pragma}. The
7184 only reason to define this macro is for compatibility with other
7185 compilers that do support @code{#pragma} for the sake of any user
7186 programs which already use it.
7188 @findex VALID_MACHINE_DECL_ATTRIBUTE
7189 @item VALID_MACHINE_DECL_ATTRIBUTE (@var{decl}, @var{attributes}, @var{identifier}, @var{args})
7190 If defined, a C expression whose value is nonzero if @var{identifier} with
7191 arguments @var{args} is a valid machine specific attribute for @var{decl}.
7192 The attributes in @var{attributes} have previously been assigned to @var{decl}.
7194 @findex VALID_MACHINE_TYPE_ATTRIBUTE
7195 @item VALID_MACHINE_TYPE_ATTRIBUTE (@var{type}, @var{attributes}, @var{identifier}, @var{args})
7196 If defined, a C expression whose value is nonzero if @var{identifier} with
7197 arguments @var{args} is a valid machine specific attribute for @var{type}.
7198 The attributes in @var{attributes} have previously been assigned to @var{type}.
7200 @findex COMP_TYPE_ATTRIBUTES
7201 @item COMP_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
7202 If defined, a C expression whose value is zero if the attributes on
7203 @var{type1} and @var{type2} are incompatible, one if they are compatible,
7204 and two if they are nearly compatible (which causes a warning to be
7207 @findex SET_DEFAULT_TYPE_ATTRIBUTES
7208 @item SET_DEFAULT_TYPE_ATTRIBUTES (@var{type})
7209 If defined, a C statement that assigns default attributes to
7210 newly defined @var{type}.
7212 @findex MERGE_MACHINE_TYPE_ATTRIBUTES
7213 @item MERGE_MACHINE_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
7214 Define this macro if the merging of type attributes needs special handling.
7215 If defined, the result is a list of the combined TYPE_ATTRIBUTES of
7216 @var{type1} and @var{type2}. It is assumed that comptypes has already been
7217 called and returned 1.
7219 @findex MERGE_MACHINE_DECL_ATTRIBUTES
7220 @item MERGE_MACHINE_DECL_ATTRIBUTES (@var{olddecl}, @var{newdecl})
7221 Define this macro if the merging of decl attributes needs special handling.
7222 If defined, the result is a list of the combined DECL_MACHINE_ATTRIBUTES of
7223 @var{olddecl} and @var{newdecl}. @var{newdecl} is a duplicate declaration
7224 of @var{olddecl}. Examples of when this is needed are when one attribute
7225 overrides another, or when an attribute is nullified by a subsequent
7228 @findex DOLLARS_IN_IDENTIFIERS
7229 @item DOLLARS_IN_IDENTIFIERS
7230 Define this macro to control use of the character @samp{$} in identifier
7231 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
7232 1 is the default; there is no need to define this macro in that case.
7233 This macro controls the compiler proper; it does not affect the preprocessor.
7235 @findex NO_DOLLAR_IN_LABEL
7236 @item NO_DOLLAR_IN_LABEL
7237 Define this macro if the assembler does not accept the character
7238 @samp{$} in label names. By default constructors and destructors in
7239 G++ have @samp{$} in the identifiers. If this macro is defined,
7240 @samp{.} is used instead.
7242 @findex NO_DOT_IN_LABEL
7243 @item NO_DOT_IN_LABEL
7244 Define this macro if the assembler does not accept the character
7245 @samp{.} in label names. By default constructors and destructors in G++
7246 have names that use @samp{.}. If this macro is defined, these names
7247 are rewritten to avoid @samp{.}.
7249 @findex DEFAULT_MAIN_RETURN
7250 @item DEFAULT_MAIN_RETURN
7251 Define this macro if the target system expects every program's @code{main}
7252 function to return a standard ``success'' value by default (if no other
7253 value is explicitly returned).
7255 The definition should be a C statement (sans semicolon) to generate the
7256 appropriate rtl instructions. It is used only when compiling the end of
7261 Define this if the target system supports the function
7262 @code{atexit} from the ANSI C standard. If this is not defined,
7263 and @code{INIT_SECTION_ASM_OP} is not defined, a default
7264 @code{exit} function will be provided to support C++.
7268 Define this if your @code{exit} function needs to do something
7269 besides calling an external function @code{_cleanup} before
7270 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
7271 only needed if neither @code{HAVE_ATEXIT} nor
7272 @code{INIT_SECTION_ASM_OP} are defined.
7274 @findex INSN_SETS_ARE_DELAYED
7275 @item INSN_SETS_ARE_DELAYED (@var{insn})
7276 Define this macro as a C expression that is nonzero if it is safe for the
7277 delay slot scheduler to place instructions in the delay slot of @var{insn},
7278 even if they appear to use a resource set or clobbered in @var{insn}.
7279 @var{insn} is always a @code{jump_insn} or an @code{insn}; GNU CC knows that
7280 every @code{call_insn} has this behavior. On machines where some @code{insn}
7281 or @code{jump_insn} is really a function call and hence has this behavior,
7282 you should define this macro.
7284 You need not define this macro if it would always return zero.
7286 @findex INSN_REFERENCES_ARE_DELAYED
7287 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
7288 Define this macro as a C expression that is nonzero if it is safe for the
7289 delay slot scheduler to place instructions in the delay slot of @var{insn},
7290 even if they appear to set or clobber a resource referenced in @var{insn}.
7291 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
7292 some @code{insn} or @code{jump_insn} is really a function call and its operands
7293 are registers whose use is actually in the subroutine it calls, you should
7294 define this macro. Doing so allows the delay slot scheduler to move
7295 instructions which copy arguments into the argument registers into the delay
7298 You need not define this macro if it would always return zero.
7300 @findex MACHINE_DEPENDENT_REORG
7301 @item MACHINE_DEPENDENT_REORG (@var{insn})
7302 In rare cases, correct code generation requires extra machine
7303 dependent processing between the second jump optimization pass and
7304 delayed branch scheduling. On those machines, define this macro as a C
7305 statement to act on the code starting at @var{insn}.
7307 @findex MULTIPLE_SYMBOL_SPACES
7308 @item MULTIPLE_SYMBOL_SPACES
7309 Define this macro if in some cases global symbols from one translation
7310 unit may not be bound to undefined symbols in another translation unit
7311 without user intervention. For instance, under Microsoft Windows
7312 symbols must be explicitly imported from shared libraries (DLLs).
7314 @findex GIV_SORT_CRITERION
7315 @item GIV_SORT_CRITERION (@var{giv1}, @var{giv2})
7316 In some cases, the strength reduction optimization pass can produce better
7317 code if this is defined. This macro controls the order that induction
7318 variables are combined. This macro is particularly useful if the target has
7319 limited addressing modes. For instance, the SH target has only positive
7320 offsets in addresses. Thus sorting to put the smallest address first
7321 allows the most combinations to be found.
7325 A C expression that returns how many instructions can be issued at the
7326 same time if the machine is a superscalar machine. This is only used by
7327 the @samp{Haifa} scheduler, and not the traditional scheduler.