1 @c Copyright (C) 1988,89,92,93,94,96,97,98,1999 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
6 @chapter Target Description Macros
7 @cindex machine description macros
8 @cindex target description macros
9 @cindex macros, target description
10 @cindex @file{tm.h} macros
12 In addition to the file @file{@var{machine}.md}, a machine description
13 includes a C header file conventionally given the name
14 @file{@var{machine}.h}. This header file defines numerous macros
15 that convey the information about the target machine that does not fit
16 into the scheme of the @file{.md} file. The file @file{tm.h} should be
17 a link to @file{@var{machine}.h}. The header file @file{config.h}
18 includes @file{tm.h} and most compiler source files include
22 * Driver:: Controlling how the driver runs the compilation passes.
23 * Run-time Target:: Defining @samp{-m} options like @samp{-m68000} and @samp{-m68020}.
24 * Storage Layout:: Defining sizes and alignments of data.
25 * Type Layout:: Defining sizes and properties of basic user data types.
26 * Registers:: Naming and describing the hardware registers.
27 * Register Classes:: Defining the classes of hardware registers.
28 * Stack and Calling:: Defining which way the stack grows and by how much.
29 * Varargs:: Defining the varargs macros.
30 * Trampolines:: Code set up at run time to enter a nested function.
31 * Library Calls:: Controlling how library routines are implicitly called.
32 * Addressing Modes:: Defining addressing modes valid for memory operands.
33 * Condition Code:: Defining how insns update the condition code.
34 * Costs:: Defining relative costs of different operations.
35 * Sections:: Dividing storage into text, data, and other sections.
36 * PIC:: Macros for position independent code.
37 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
38 * Debugging Info:: Defining the format of debugging output.
39 * Cross-compilation:: Handling floating point for cross-compilers.
40 * Misc:: Everything else.
44 @section Controlling the Compilation Driver, @file{gcc}
46 @cindex controlling the compilation driver
48 @c prevent bad page break with this line
49 You can control the compilation driver.
52 @findex SWITCH_TAKES_ARG
53 @item SWITCH_TAKES_ARG (@var{char})
54 A C expression which determines whether the option @samp{-@var{char}}
55 takes arguments. The value should be the number of arguments that
56 option takes--zero, for many options.
58 By default, this macro is defined as
59 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
60 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
61 wish to add additional options which take arguments. Any redefinition
62 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
65 @findex WORD_SWITCH_TAKES_ARG
66 @item WORD_SWITCH_TAKES_ARG (@var{name})
67 A C expression which determines whether the option @samp{-@var{name}}
68 takes arguments. The value should be the number of arguments that
69 option takes--zero, for many options. This macro rather than
70 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
72 By default, this macro is defined as
73 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
74 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
75 wish to add additional options which take arguments. Any redefinition
76 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
79 @findex SWITCH_CURTAILS_COMPILATION
80 @item SWITCH_CURTAILS_COMPILATION (@var{char})
81 A C expression which determines whether the option @samp{-@var{char}}
82 stops compilation before the generation of an executable. The value is
83 boolean, non-zero if the option does stop an executable from being
84 generated, zero otherwise.
86 By default, this macro is defined as
87 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
88 options properly. You need not define
89 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
90 options which affect the generation of an executable. Any redefinition
91 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
92 for additional options.
94 @findex SWITCHES_NEED_SPACES
95 @item SWITCHES_NEED_SPACES
96 A string-valued C expression which enumerates the options for which
97 the linker needs a space between the option and its argument.
99 If this macro is not defined, the default value is @code{""}.
103 A C string constant that tells the GCC driver program options to
104 pass to CPP. It can also specify how to translate options you
105 give to GCC into options for GCC to pass to the CPP.
107 Do not define this macro if it does not need to do anything.
109 @findex NO_BUILTIN_SIZE_TYPE
110 @item NO_BUILTIN_SIZE_TYPE
111 If this macro is defined, the preprocessor will not define the builtin macro
112 @code{__SIZE_TYPE__}. The macro @code{__SIZE_TYPE__} must then be defined
113 by @code{CPP_SPEC} instead.
115 This should be defined if @code{SIZE_TYPE} depends on target dependent flags
116 which are not accessible to the preprocessor. Otherwise, it should not
119 @findex NO_BUILTIN_PTRDIFF_TYPE
120 @item NO_BUILTIN_PTRDIFF_TYPE
121 If this macro is defined, the preprocessor will not define the builtin macro
122 @code{__PTRDIFF_TYPE__}. The macro @code{__PTRDIFF_TYPE__} must then be
123 defined by @code{CPP_SPEC} instead.
125 This should be defined if @code{PTRDIFF_TYPE} depends on target dependent flags
126 which are not accessible to the preprocessor. Otherwise, it should not
129 @findex SIGNED_CHAR_SPEC
130 @item SIGNED_CHAR_SPEC
131 A C string constant that tells the GCC driver program options to
132 pass to CPP. By default, this macro is defined to pass the option
133 @samp{-D__CHAR_UNSIGNED__} to CPP if @code{char} will be treated as
134 @code{unsigned char} by @code{cc1}.
136 Do not define this macro unless you need to override the default
141 A C string constant that tells the GCC driver program options to
142 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
144 It can also specify how to translate options you give to GCC into options
145 for GCC to pass to front ends..
147 Do not define this macro if it does not need to do anything.
151 A C string constant that tells the GCC driver program options to
152 pass to @code{cc1plus}. It can also specify how to translate options you
153 give to GCC into options for GCC to pass to the @code{cc1plus}.
155 Do not define this macro if it does not need to do anything.
156 Note that everything defined in CC1_SPEC is already passed to
157 @code{cc1plus} so there is no need to duplicate the contents of
158 CC1_SPEC in CC1PLUS_SPEC.
162 A C string constant that tells the GCC driver program options to
163 pass to the assembler. It can also specify how to translate options
164 you give to GCC into options for GCC to pass to the assembler.
165 See the file @file{sun3.h} for an example of this.
167 Do not define this macro if it does not need to do anything.
169 @findex ASM_FINAL_SPEC
171 A C string constant that tells the GCC driver program how to
172 run any programs which cleanup after the normal assembler.
173 Normally, this is not needed. See the file @file{mips.h} for
176 Do not define this macro if it does not need to do anything.
180 A C string constant that tells the GCC driver program options to
181 pass to the linker. It can also specify how to translate options you
182 give to GCC into options for GCC to pass to the linker.
184 Do not define this macro if it does not need to do anything.
188 Another C string constant used much like @code{LINK_SPEC}. The difference
189 between the two is that @code{LIB_SPEC} is used at the end of the
190 command given to the linker.
192 If this macro is not defined, a default is provided that
193 loads the standard C library from the usual place. See @file{gcc.c}.
197 Another C string constant that tells the GCC driver program
198 how and when to place a reference to @file{libgcc.a} into the
199 linker command line. This constant is placed both before and after
200 the value of @code{LIB_SPEC}.
202 If this macro is not defined, the GCC driver provides a default that
203 passes the string @samp{-lgcc} to the linker unless the @samp{-shared}
206 @findex STARTFILE_SPEC
208 Another C string constant used much like @code{LINK_SPEC}. The
209 difference between the two is that @code{STARTFILE_SPEC} is used at
210 the very beginning of the command given to the linker.
212 If this macro is not defined, a default is provided that loads the
213 standard C startup file from the usual place. See @file{gcc.c}.
217 Another C string constant used much like @code{LINK_SPEC}. The
218 difference between the two is that @code{ENDFILE_SPEC} is used at
219 the very end of the command given to the linker.
221 Do not define this macro if it does not need to do anything.
225 Define this macro to provide additional specifications to put in the
226 @file{specs} file that can be used in various specifications like
229 The definition should be an initializer for an array of structures,
230 containing a string constant, that defines the specification name, and a
231 string constant that provides the specification.
233 Do not define this macro if it does not need to do anything.
235 @code{EXTRA_SPECS} is useful when an architecture contains several
236 related targets, which have various @code{..._SPECS} which are similar
237 to each other, and the maintainer would like one central place to keep
240 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
241 define either @code{_CALL_SYSV} when the System V calling sequence is
242 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
245 The @file{config/rs6000/rs6000.h} target file defines:
248 #define EXTRA_SPECS \
249 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
251 #define CPP_SYS_DEFAULT ""
254 The @file{config/rs6000/sysv.h} target file defines:
258 "%@{posix: -D_POSIX_SOURCE @} \
259 %@{mcall-sysv: -D_CALL_SYSV @} %@{mcall-aix: -D_CALL_AIX @} \
260 %@{!mcall-sysv: %@{!mcall-aix: %(cpp_sysv_default) @}@} \
261 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
263 #undef CPP_SYSV_DEFAULT
264 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
267 while the @file{config/rs6000/eabiaix.h} target file defines
268 @code{CPP_SYSV_DEFAULT} as:
271 #undef CPP_SYSV_DEFAULT
272 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
275 @findex LINK_LIBGCC_SPECIAL
276 @item LINK_LIBGCC_SPECIAL
277 Define this macro if the driver program should find the library
278 @file{libgcc.a} itself and should not pass @samp{-L} options to the
279 linker. If you do not define this macro, the driver program will pass
280 the argument @samp{-lgcc} to tell the linker to do the search and will
281 pass @samp{-L} options to it.
283 @findex LINK_LIBGCC_SPECIAL_1
284 @item LINK_LIBGCC_SPECIAL_1
285 Define this macro if the driver program should find the library
286 @file{libgcc.a}. If you do not define this macro, the driver program will pass
287 the argument @samp{-lgcc} to tell the linker to do the search.
288 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
289 not affect @samp{-L} options.
291 @findex LINK_COMMAND_SPEC
292 @item LINK_COMMAND_SPEC
293 A C string constant giving the complete command line need to execute the
294 linker. When you do this, you will need to update your port each time a
295 change is made to the link command line within @file{gcc.c}. Therefore,
296 define this macro only if you need to completely redefine the command
297 line for invoking the linker and there is no other way to accomplish
300 @findex MULTILIB_DEFAULTS
301 @item MULTILIB_DEFAULTS
302 Define this macro as a C expression for the initializer of an array of
303 string to tell the driver program which options are defaults for this
304 target and thus do not need to be handled specially when using
305 @code{MULTILIB_OPTIONS}.
307 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
308 the target makefile fragment or if none of the options listed in
309 @code{MULTILIB_OPTIONS} are set by default.
310 @xref{Target Fragment}.
312 @findex RELATIVE_PREFIX_NOT_LINKDIR
313 @item RELATIVE_PREFIX_NOT_LINKDIR
314 Define this macro to tell @code{gcc} that it should only translate
315 a @samp{-B} prefix into a @samp{-L} linker option if the prefix
316 indicates an absolute file name.
318 @findex STANDARD_EXEC_PREFIX
319 @item STANDARD_EXEC_PREFIX
320 Define this macro as a C string constant if you wish to override the
321 standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
322 try when searching for the executable files of the compiler.
324 @findex MD_EXEC_PREFIX
326 If defined, this macro is an additional prefix to try after
327 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
328 when the @samp{-b} option is used, or the compiler is built as a cross
329 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
330 to the list of directories used to find the assembler in @file{configure.in}.
332 @findex STANDARD_STARTFILE_PREFIX
333 @item STANDARD_STARTFILE_PREFIX
334 Define this macro as a C string constant if you wish to override the
335 standard choice of @file{/usr/local/lib/} as the default prefix to
336 try when searching for startup files such as @file{crt0.o}.
338 @findex MD_STARTFILE_PREFIX
339 @item MD_STARTFILE_PREFIX
340 If defined, this macro supplies an additional prefix to try after the
341 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
342 @samp{-b} option is used, or when the compiler is built as a cross
345 @findex MD_STARTFILE_PREFIX_1
346 @item MD_STARTFILE_PREFIX_1
347 If defined, this macro supplies yet another prefix to try after the
348 standard prefixes. It is not searched when the @samp{-b} option is
349 used, or when the compiler is built as a cross compiler.
351 @findex INIT_ENVIRONMENT
352 @item INIT_ENVIRONMENT
353 Define this macro as a C string constant if you wish to set environment
354 variables for programs called by the driver, such as the assembler and
355 loader. The driver passes the value of this macro to @code{putenv} to
356 initialize the necessary environment variables.
358 @findex LOCAL_INCLUDE_DIR
359 @item LOCAL_INCLUDE_DIR
360 Define this macro as a C string constant if you wish to override the
361 standard choice of @file{/usr/local/include} as the default prefix to
362 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
363 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
365 Cross compilers do not use this macro and do not search either
366 @file{/usr/local/include} or its replacement.
368 @findex SYSTEM_INCLUDE_DIR
369 @item SYSTEM_INCLUDE_DIR
370 Define this macro as a C string constant if you wish to specify a
371 system-specific directory to search for header files before the standard
372 directory. @code{SYSTEM_INCLUDE_DIR} comes before
373 @code{STANDARD_INCLUDE_DIR} in the search order.
375 Cross compilers do not use this macro and do not search the directory
378 @findex STANDARD_INCLUDE_DIR
379 @item STANDARD_INCLUDE_DIR
380 Define this macro as a C string constant if you wish to override the
381 standard choice of @file{/usr/include} as the default prefix to
382 try when searching for header files.
384 Cross compilers do not use this macro and do not search either
385 @file{/usr/include} or its replacement.
387 @findex STANDARD_INCLUDE_COMPONENT
388 @item STANDARD_INCLUDE_COMPONENT
389 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
390 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
391 If you do not define this macro, no component is used.
393 @findex INCLUDE_DEFAULTS
394 @item INCLUDE_DEFAULTS
395 Define this macro if you wish to override the entire default search path
396 for include files. For a native compiler, the default search path
397 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
398 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
399 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
400 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
401 and specify private search areas for GCC. The directory
402 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
404 The definition should be an initializer for an array of structures.
405 Each array element should have four elements: the directory name (a
406 string constant), the component name, and flag for C++-only directories,
407 and a flag showing that the includes in the directory don't need to be
408 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
409 the array with a null element.
411 The component name denotes what GNU package the include file is part of,
412 if any, in all upper-case letters. For example, it might be @samp{GCC}
413 or @samp{BINUTILS}. If the package is part of the a vendor-supplied
414 operating system, code the component name as @samp{0}.
417 For example, here is the definition used for VAX/VMS:
420 #define INCLUDE_DEFAULTS \
422 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
423 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
424 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
431 Here is the order of prefixes tried for exec files:
435 Any prefixes specified by the user with @samp{-B}.
438 The environment variable @code{GCC_EXEC_PREFIX}, if any.
441 The directories specified by the environment variable @code{COMPILER_PATH}.
444 The macro @code{STANDARD_EXEC_PREFIX}.
447 @file{/usr/lib/gcc/}.
450 The macro @code{MD_EXEC_PREFIX}, if any.
453 Here is the order of prefixes tried for startfiles:
457 Any prefixes specified by the user with @samp{-B}.
460 The environment variable @code{GCC_EXEC_PREFIX}, if any.
463 The directories specified by the environment variable @code{LIBRARY_PATH}
464 (native only, cross compilers do not use this).
467 The macro @code{STANDARD_EXEC_PREFIX}.
470 @file{/usr/lib/gcc/}.
473 The macro @code{MD_EXEC_PREFIX}, if any.
476 The macro @code{MD_STARTFILE_PREFIX}, if any.
479 The macro @code{STANDARD_STARTFILE_PREFIX}.
488 @node Run-time Target
489 @section Run-time Target Specification
490 @cindex run-time target specification
491 @cindex predefined macros
492 @cindex target specifications
494 @c prevent bad page break with this line
495 Here are run-time target specifications.
498 @findex CPP_PREDEFINES
500 Define this to be a string constant containing @samp{-D} options to
501 define the predefined macros that identify this machine and system.
502 These macros will be predefined unless the @samp{-ansi} option is
505 In addition, a parallel set of macros are predefined, whose names are
506 made by appending @samp{__} at the beginning and at the end. These
507 @samp{__} macros are permitted by the ANSI standard, so they are
508 predefined regardless of whether @samp{-ansi} is specified.
510 For example, on the Sun, one can use the following value:
513 "-Dmc68000 -Dsun -Dunix"
516 The result is to define the macros @code{__mc68000__}, @code{__sun__}
517 and @code{__unix__} unconditionally, and the macros @code{mc68000},
518 @code{sun} and @code{unix} provided @samp{-ansi} is not specified.
520 @findex extern int target_flags
521 @item extern int target_flags;
522 This declaration should be present.
524 @cindex optional hardware or system features
525 @cindex features, optional, in system conventions
527 This series of macros is to allow compiler command arguments to
528 enable or disable the use of optional features of the target machine.
529 For example, one machine description serves both the 68000 and
530 the 68020; a command argument tells the compiler whether it should
531 use 68020-only instructions or not. This command argument works
532 by means of a macro @code{TARGET_68020} that tests a bit in
535 Define a macro @code{TARGET_@var{featurename}} for each such option.
536 Its definition should test a bit in @code{target_flags}; for example:
539 #define TARGET_68020 (target_flags & 1)
542 One place where these macros are used is in the condition-expressions
543 of instruction patterns. Note how @code{TARGET_68020} appears
544 frequently in the 68000 machine description file, @file{m68k.md}.
545 Another place they are used is in the definitions of the other
546 macros in the @file{@var{machine}.h} file.
548 @findex TARGET_SWITCHES
549 @item TARGET_SWITCHES
550 This macro defines names of command options to set and clear
551 bits in @code{target_flags}. Its definition is an initializer
552 with a subgrouping for each command option.
554 Each subgrouping contains a string constant, that defines the option
555 name, a number, which contains the bits to set in
556 @code{target_flags}, and a second string which is the description
557 displayed by --help. If the number is negative then the bits specified
558 by the number are cleared instead of being set. If the description
559 string is present but empty, then no help information will be displayed
560 for that option, but it will not count as an undocumented option. The
561 actual option name is made by appending @samp{-m} to the specified name.
563 One of the subgroupings should have a null string. The number in
564 this grouping is the default value for @code{target_flags}. Any
565 target options act starting with that value.
567 Here is an example which defines @samp{-m68000} and @samp{-m68020}
568 with opposite meanings, and picks the latter as the default:
571 #define TARGET_SWITCHES \
572 @{ @{ "68020", 1, "" @}, \
573 @{ "68000", -1, "Compile for the 68000" @}, \
577 @findex TARGET_OPTIONS
579 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
580 options that have values. Its definition is an initializer with a
581 subgrouping for each command option.
583 Each subgrouping contains a string constant, that defines the fixed part
584 of the option name, the address of a variable, and a description string.
585 The variable, type @code{char *}, is set to the variable part of the
586 given option if the fixed part matches. The actual option name is made
587 by appending @samp{-m} to the specified name.
589 Here is an example which defines @samp{-mshort-data-@var{number}}. If the
590 given option is @samp{-mshort-data-512}, the variable @code{m88k_short_data}
591 will be set to the string @code{"512"}.
594 extern char *m88k_short_data;
595 #define TARGET_OPTIONS \
596 @{ @{ "short-data-", &m88k_short_data, "Specify the size of the short data section" @} @}
599 @findex TARGET_VERSION
601 This macro is a C statement to print on @code{stderr} a string
602 describing the particular machine description choice. Every machine
603 description should define @code{TARGET_VERSION}. For example:
607 #define TARGET_VERSION \
608 fprintf (stderr, " (68k, Motorola syntax)");
610 #define TARGET_VERSION \
611 fprintf (stderr, " (68k, MIT syntax)");
615 @findex OVERRIDE_OPTIONS
616 @item OVERRIDE_OPTIONS
617 Sometimes certain combinations of command options do not make sense on
618 a particular target machine. You can define a macro
619 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
620 defined, is executed once just after all the command options have been
623 Don't use this macro to turn on various extra optimizations for
624 @samp{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
626 @findex OPTIMIZATION_OPTIONS
627 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
628 Some machines may desire to change what optimizations are performed for
629 various optimization levels. This macro, if defined, is executed once
630 just after the optimization level is determined and before the remainder
631 of the command options have been parsed. Values set in this macro are
632 used as the default values for the other command line options.
634 @var{level} is the optimization level specified; 2 if @samp{-O2} is
635 specified, 1 if @samp{-O} is specified, and 0 if neither is specified.
637 @var{size} is non-zero if @samp{-Os} is specified and zero otherwise.
639 You should not use this macro to change options that are not
640 machine-specific. These should uniformly selected by the same
641 optimization level on all supported machines. Use this macro to enable
642 machine-specific optimizations.
644 @strong{Do not examine @code{write_symbols} in
645 this macro!} The debugging options are not supposed to alter the
648 @findex CAN_DEBUG_WITHOUT_FP
649 @item CAN_DEBUG_WITHOUT_FP
650 Define this macro if debugging can be performed even without a frame
651 pointer. If this macro is defined, GCC will turn on the
652 @samp{-fomit-frame-pointer} option whenever @samp{-O} is specified.
656 @section Storage Layout
657 @cindex storage layout
659 Note that the definitions of the macros in this table which are sizes or
660 alignments measured in bits do not need to be constant. They can be C
661 expressions that refer to static variables, such as the @code{target_flags}.
662 @xref{Run-time Target}.
665 @findex BITS_BIG_ENDIAN
666 @item BITS_BIG_ENDIAN
667 Define this macro to have the value 1 if the most significant bit in a
668 byte has the lowest number; otherwise define it to have the value zero.
669 This means that bit-field instructions count from the most significant
670 bit. If the machine has no bit-field instructions, then this must still
671 be defined, but it doesn't matter which value it is defined to. This
672 macro need not be a constant.
674 This macro does not affect the way structure fields are packed into
675 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
677 @findex BYTES_BIG_ENDIAN
678 @item BYTES_BIG_ENDIAN
679 Define this macro to have the value 1 if the most significant byte in a
680 word has the lowest number. This macro need not be a constant.
682 @findex WORDS_BIG_ENDIAN
683 @item WORDS_BIG_ENDIAN
684 Define this macro to have the value 1 if, in a multiword object, the
685 most significant word has the lowest number. This applies to both
686 memory locations and registers; GCC fundamentally assumes that the
687 order of words in memory is the same as the order in registers. This
688 macro need not be a constant.
690 @findex LIBGCC2_WORDS_BIG_ENDIAN
691 @item LIBGCC2_WORDS_BIG_ENDIAN
692 Define this macro if WORDS_BIG_ENDIAN is not constant. This must be a
693 constant value with the same meaning as WORDS_BIG_ENDIAN, which will be
694 used only when compiling libgcc2.c. Typically the value will be set
695 based on preprocessor defines.
697 @findex FLOAT_WORDS_BIG_ENDIAN
698 @item FLOAT_WORDS_BIG_ENDIAN
699 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
700 @code{TFmode} floating point numbers are stored in memory with the word
701 containing the sign bit at the lowest address; otherwise define it to
702 have the value 0. This macro need not be a constant.
704 You need not define this macro if the ordering is the same as for
707 @findex BITS_PER_UNIT
709 Define this macro to be the number of bits in an addressable storage
710 unit (byte); normally 8.
712 @findex BITS_PER_WORD
714 Number of bits in a word; normally 32.
716 @findex MAX_BITS_PER_WORD
717 @item MAX_BITS_PER_WORD
718 Maximum number of bits in a word. If this is undefined, the default is
719 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
720 largest value that @code{BITS_PER_WORD} can have at run-time.
722 @findex UNITS_PER_WORD
724 Number of storage units in a word; normally 4.
726 @findex MIN_UNITS_PER_WORD
727 @item MIN_UNITS_PER_WORD
728 Minimum number of units in a word. If this is undefined, the default is
729 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
730 smallest value that @code{UNITS_PER_WORD} can have at run-time.
734 Width of a pointer, in bits. You must specify a value no wider than the
735 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
736 you must define @code{POINTERS_EXTEND_UNSIGNED}.
738 @findex POINTERS_EXTEND_UNSIGNED
739 @item POINTERS_EXTEND_UNSIGNED
740 A C expression whose value is nonzero if pointers that need to be
741 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
742 be zero-extended and zero if they are to be sign-extended.
744 You need not define this macro if the @code{POINTER_SIZE} is equal
745 to the width of @code{Pmode}.
748 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
749 A macro to update @var{m} and @var{unsignedp} when an object whose type
750 is @var{type} and which has the specified mode and signedness is to be
751 stored in a register. This macro is only called when @var{type} is a
754 On most RISC machines, which only have operations that operate on a full
755 register, define this macro to set @var{m} to @code{word_mode} if
756 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
757 cases, only integer modes should be widened because wider-precision
758 floating-point operations are usually more expensive than their narrower
761 For most machines, the macro definition does not change @var{unsignedp}.
762 However, some machines, have instructions that preferentially handle
763 either signed or unsigned quantities of certain modes. For example, on
764 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
765 sign-extend the result to 64 bits. On such machines, set
766 @var{unsignedp} according to which kind of extension is more efficient.
768 Do not define this macro if it would never modify @var{m}.
770 @findex PROMOTE_FUNCTION_ARGS
771 @item PROMOTE_FUNCTION_ARGS
772 Define this macro if the promotion described by @code{PROMOTE_MODE}
773 should also be done for outgoing function arguments.
775 @findex PROMOTE_FUNCTION_RETURN
776 @item PROMOTE_FUNCTION_RETURN
777 Define this macro if the promotion described by @code{PROMOTE_MODE}
778 should also be done for the return value of functions.
780 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
781 promotions done by @code{PROMOTE_MODE}.
783 @findex PROMOTE_FOR_CALL_ONLY
784 @item PROMOTE_FOR_CALL_ONLY
785 Define this macro if the promotion described by @code{PROMOTE_MODE}
786 should @emph{only} be performed for outgoing function arguments or
787 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
788 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
790 @findex PARM_BOUNDARY
792 Normal alignment required for function parameters on the stack, in
793 bits. All stack parameters receive at least this much alignment
794 regardless of data type. On most machines, this is the same as the
797 @findex STACK_BOUNDARY
799 Define this macro if there is a guaranteed alignment for the stack
800 pointer on this machine. The definition is a C expression
801 for the desired alignment (measured in bits). This value is used as a
802 default if PREFERRED_STACK_BOUNDARY is not defined.
804 @findex PREFERRED_STACK_BOUNDARY
805 @item PREFERRED_STACK_BOUNDARY
806 Define this macro if you wish to preserve a certain alignment for
807 the stack pointer. The definition is a C expression
808 for the desired alignment (measured in bits). If STACK_BOUNDARY is
809 also defined, this macro must evaluate to a value equal to or larger
812 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
813 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
814 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
815 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
816 be momentarily unaligned while pushing arguments.
818 @findex FUNCTION_BOUNDARY
819 @item FUNCTION_BOUNDARY
820 Alignment required for a function entry point, in bits.
822 @findex BIGGEST_ALIGNMENT
823 @item BIGGEST_ALIGNMENT
824 Biggest alignment that any data type can require on this machine, in bits.
826 @findex MINIMUM_ATOMIC_ALIGNMENT
827 @item MINIMUM_ATOMIC_ALIGNMENT
828 If defined, the smallest alignment, in bits, that can be given to an
829 object that can be referenced in one operation, without disturbing any
830 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
831 on machines that don't have byte or half-word store operations.
833 @findex BIGGEST_FIELD_ALIGNMENT
834 @item BIGGEST_FIELD_ALIGNMENT
835 Biggest alignment that any structure field can require on this machine,
836 in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
837 structure fields only.
839 @findex ADJUST_FIELD_ALIGN
840 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
841 An expression for the alignment of a structure field @var{field} if the
842 alignment computed in the usual way is @var{computed}. GCC uses
843 this value instead of the value in @code{BIGGEST_ALIGNMENT} or
844 @code{BIGGEST_FIELD_ALIGNMENT}, if defined, for structure fields only.
846 @findex MAX_OFILE_ALIGNMENT
847 @item MAX_OFILE_ALIGNMENT
848 Biggest alignment supported by the object file format of this machine.
849 Use this macro to limit the alignment which can be specified using the
850 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
851 the default value is @code{BIGGEST_ALIGNMENT}.
853 @findex DATA_ALIGNMENT
854 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
855 If defined, a C expression to compute the alignment for a variables in
856 the static store. @var{type} is the data type, and @var{basic-align} is
857 the alignment that the object would ordinarily have. The value of this
858 macro is used instead of that alignment to align the object.
860 If this macro is not defined, then @var{basic-align} is used.
863 One use of this macro is to increase alignment of medium-size data to
864 make it all fit in fewer cache lines. Another is to cause character
865 arrays to be word-aligned so that @code{strcpy} calls that copy
866 constants to character arrays can be done inline.
868 @findex CONSTANT_ALIGNMENT
869 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
870 If defined, a C expression to compute the alignment given to a constant
871 that is being placed in memory. @var{constant} is the constant and
872 @var{basic-align} is the alignment that the object would ordinarily
873 have. The value of this macro is used instead of that alignment to
876 If this macro is not defined, then @var{basic-align} is used.
878 The typical use of this macro is to increase alignment for string
879 constants to be word aligned so that @code{strcpy} calls that copy
880 constants can be done inline.
882 @findex LOCAL_ALIGNMENT
883 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
884 If defined, a C expression to compute the alignment for a variables in
885 the local store. @var{type} is the data type, and @var{basic-align} is
886 the alignment that the object would ordinarily have. The value of this
887 macro is used instead of that alignment to align the object.
889 If this macro is not defined, then @var{basic-align} is used.
891 One use of this macro is to increase alignment of medium-size data to
892 make it all fit in fewer cache lines.
894 @findex EMPTY_FIELD_BOUNDARY
895 @item EMPTY_FIELD_BOUNDARY
896 Alignment in bits to be given to a structure bit field that follows an
897 empty field such as @code{int : 0;}.
899 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
900 that results from an empty field.
902 @findex STRUCTURE_SIZE_BOUNDARY
903 @item STRUCTURE_SIZE_BOUNDARY
904 Number of bits which any structure or union's size must be a multiple of.
905 Each structure or union's size is rounded up to a multiple of this.
907 If you do not define this macro, the default is the same as
908 @code{BITS_PER_UNIT}.
910 @findex STRICT_ALIGNMENT
911 @item STRICT_ALIGNMENT
912 Define this macro to be the value 1 if instructions will fail to work
913 if given data not on the nominal alignment. If instructions will merely
914 go slower in that case, define this macro as 0.
916 @findex PCC_BITFIELD_TYPE_MATTERS
917 @item PCC_BITFIELD_TYPE_MATTERS
918 Define this if you wish to imitate the way many other C compilers handle
919 alignment of bitfields and the structures that contain them.
921 The behavior is that the type written for a bitfield (@code{int},
922 @code{short}, or other integer type) imposes an alignment for the
923 entire structure, as if the structure really did contain an ordinary
924 field of that type. In addition, the bitfield is placed within the
925 structure so that it would fit within such a field, not crossing a
928 Thus, on most machines, a bitfield whose type is written as @code{int}
929 would not cross a four-byte boundary, and would force four-byte
930 alignment for the whole structure. (The alignment used may not be four
931 bytes; it is controlled by the other alignment parameters.)
933 If the macro is defined, its definition should be a C expression;
934 a nonzero value for the expression enables this behavior.
936 Note that if this macro is not defined, or its value is zero, some
937 bitfields may cross more than one alignment boundary. The compiler can
938 support such references if there are @samp{insv}, @samp{extv}, and
939 @samp{extzv} insns that can directly reference memory.
941 The other known way of making bitfields work is to define
942 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
943 Then every structure can be accessed with fullwords.
945 Unless the machine has bitfield instructions or you define
946 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
947 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
949 If your aim is to make GCC use the same conventions for laying out
950 bitfields as are used by another compiler, here is how to investigate
951 what the other compiler does. Compile and run this program:
970 printf ("Size of foo1 is %d\n",
971 sizeof (struct foo1));
972 printf ("Size of foo2 is %d\n",
973 sizeof (struct foo2));
978 If this prints 2 and 5, then the compiler's behavior is what you would
979 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
981 @findex BITFIELD_NBYTES_LIMITED
982 @item BITFIELD_NBYTES_LIMITED
983 Like PCC_BITFIELD_TYPE_MATTERS except that its effect is limited to
984 aligning a bitfield within the structure.
986 @findex ROUND_TYPE_SIZE
987 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
988 Define this macro as an expression for the overall size of a type
989 (given by @var{type} as a tree node) when the size computed in the
990 usual way is @var{computed} and the alignment is @var{specified}.
992 The default is to round @var{computed} up to a multiple of @var{specified}.
994 @findex ROUND_TYPE_ALIGN
995 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
996 Define this macro as an expression for the alignment of a type (given
997 by @var{type} as a tree node) if the alignment computed in the usual
998 way is @var{computed} and the alignment explicitly specified was
1001 The default is to use @var{specified} if it is larger; otherwise, use
1002 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1004 @findex MAX_FIXED_MODE_SIZE
1005 @item MAX_FIXED_MODE_SIZE
1006 An integer expression for the size in bits of the largest integer
1007 machine mode that should actually be used. All integer machine modes of
1008 this size or smaller can be used for structures and unions with the
1009 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1010 (DImode)} is assumed.
1012 @findex STACK_SAVEAREA_MODE
1013 @item STACK_SAVEAREA_MODE (@var{save_level})
1014 If defined, an expression of type @code{enum machine_mode} that
1015 specifies the mode of the save area operand of a
1016 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1017 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1018 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1019 having its mode specified.
1021 You need not define this macro if it always returns @code{Pmode}. You
1022 would most commonly define this macro if the
1023 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1026 @findex STACK_SIZE_MODE
1027 @item STACK_SIZE_MODE
1028 If defined, an expression of type @code{enum machine_mode} that
1029 specifies the mode of the size increment operand of an
1030 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1032 You need not define this macro if it always returns @code{word_mode}.
1033 You would most commonly define this macro if the @code{allocate_stack}
1034 pattern needs to support both a 32- and a 64-bit mode.
1036 @findex CHECK_FLOAT_VALUE
1037 @item CHECK_FLOAT_VALUE (@var{mode}, @var{value}, @var{overflow})
1038 A C statement to validate the value @var{value} (of type
1039 @code{double}) for mode @var{mode}. This means that you check whether
1040 @var{value} fits within the possible range of values for mode
1041 @var{mode} on this target machine. The mode @var{mode} is always
1042 a mode of class @code{MODE_FLOAT}. @var{overflow} is nonzero if
1043 the value is already known to be out of range.
1045 If @var{value} is not valid or if @var{overflow} is nonzero, you should
1046 set @var{overflow} to 1 and then assign some valid value to @var{value}.
1047 Allowing an invalid value to go through the compiler can produce
1048 incorrect assembler code which may even cause Unix assemblers to crash.
1050 This macro need not be defined if there is no work for it to do.
1052 @findex TARGET_FLOAT_FORMAT
1053 @item TARGET_FLOAT_FORMAT
1054 A code distinguishing the floating point format of the target machine.
1055 There are three defined values:
1058 @findex IEEE_FLOAT_FORMAT
1059 @item IEEE_FLOAT_FORMAT
1060 This code indicates IEEE floating point. It is the default; there is no
1061 need to define this macro when the format is IEEE.
1063 @findex VAX_FLOAT_FORMAT
1064 @item VAX_FLOAT_FORMAT
1065 This code indicates the peculiar format used on the Vax.
1067 @findex UNKNOWN_FLOAT_FORMAT
1068 @item UNKNOWN_FLOAT_FORMAT
1069 This code indicates any other format.
1072 The value of this macro is compared with @code{HOST_FLOAT_FORMAT}
1073 (@pxref{Config}) to determine whether the target machine has the same
1074 format as the host machine. If any other formats are actually in use on
1075 supported machines, new codes should be defined for them.
1077 The ordering of the component words of floating point values stored in
1078 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN} for the target
1079 machine and @code{HOST_FLOAT_WORDS_BIG_ENDIAN} for the host.
1081 @findex DEFAULT_VTABLE_THUNKS
1082 @item DEFAULT_VTABLE_THUNKS
1083 GCC supports two ways of implementing C++ vtables: traditional or with
1084 so-called ``thunks''. The flag @samp{-fvtable-thunk} chooses between them.
1085 Define this macro to be a C expression for the default value of that flag.
1086 If @code{DEFAULT_VTABLE_THUNKS} is 0, GCC uses the traditional
1087 implementation by default. The ``thunk'' implementation is more efficient
1088 (especially if you have provided an implementation of
1089 @code{ASM_OUTPUT_MI_THUNK}, see @ref{Function Entry}), but is not binary
1090 compatible with code compiled using the traditional implementation.
1091 If you are writing a new ports, define @code{DEFAULT_VTABLE_THUNKS} to 1.
1093 If you do not define this macro, the default for @samp{-fvtable-thunk} is 0.
1097 @section Layout of Source Language Data Types
1099 These macros define the sizes and other characteristics of the standard
1100 basic data types used in programs being compiled. Unlike the macros in
1101 the previous section, these apply to specific features of C and related
1102 languages, rather than to fundamental aspects of storage layout.
1105 @findex INT_TYPE_SIZE
1107 A C expression for the size in bits of the type @code{int} on the
1108 target machine. If you don't define this, the default is one word.
1110 @findex MAX_INT_TYPE_SIZE
1111 @item MAX_INT_TYPE_SIZE
1112 Maximum number for the size in bits of the type @code{int} on the target
1113 machine. If this is undefined, the default is @code{INT_TYPE_SIZE}.
1114 Otherwise, it is the constant value that is the largest value that
1115 @code{INT_TYPE_SIZE} can have at run-time. This is used in @code{cpp}.
1117 @findex SHORT_TYPE_SIZE
1118 @item SHORT_TYPE_SIZE
1119 A C expression for the size in bits of the type @code{short} on the
1120 target machine. If you don't define this, the default is half a word.
1121 (If this would be less than one storage unit, it is rounded up to one
1124 @findex LONG_TYPE_SIZE
1125 @item LONG_TYPE_SIZE
1126 A C expression for the size in bits of the type @code{long} on the
1127 target machine. If you don't define this, the default is one word.
1129 @findex MAX_LONG_TYPE_SIZE
1130 @item MAX_LONG_TYPE_SIZE
1131 Maximum number for the size in bits of the type @code{long} on the
1132 target machine. If this is undefined, the default is
1133 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1134 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1137 @findex LONG_LONG_TYPE_SIZE
1138 @item LONG_LONG_TYPE_SIZE
1139 A C expression for the size in bits of the type @code{long long} on the
1140 target machine. If you don't define this, the default is two
1141 words. If you want to support GNU Ada on your machine, the value of
1142 macro must be at least 64.
1144 @findex CHAR_TYPE_SIZE
1145 @item CHAR_TYPE_SIZE
1146 A C expression for the size in bits of the type @code{char} on the
1147 target machine. If you don't define this, the default is one quarter
1148 of a word. (If this would be less than one storage unit, it is rounded up
1151 @findex MAX_CHAR_TYPE_SIZE
1152 @item MAX_CHAR_TYPE_SIZE
1153 Maximum number for the size in bits of the type @code{char} on the
1154 target machine. If this is undefined, the default is
1155 @code{CHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1156 largest value that @code{CHAR_TYPE_SIZE} can have at run-time. This is
1159 @findex FLOAT_TYPE_SIZE
1160 @item FLOAT_TYPE_SIZE
1161 A C expression for the size in bits of the type @code{float} on the
1162 target machine. If you don't define this, the default is one word.
1164 @findex DOUBLE_TYPE_SIZE
1165 @item DOUBLE_TYPE_SIZE
1166 A C expression for the size in bits of the type @code{double} on the
1167 target machine. If you don't define this, the default is two
1170 @findex LONG_DOUBLE_TYPE_SIZE
1171 @item LONG_DOUBLE_TYPE_SIZE
1172 A C expression for the size in bits of the type @code{long double} on
1173 the target machine. If you don't define this, the default is two
1176 @findex WIDEST_HARDWARE_FP_SIZE
1177 @item WIDEST_HARDWARE_FP_SIZE
1178 A C expression for the size in bits of the widest floating-point format
1179 supported by the hardware. If you define this macro, you must specify a
1180 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1181 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1184 @findex DEFAULT_SIGNED_CHAR
1185 @item DEFAULT_SIGNED_CHAR
1186 An expression whose value is 1 or 0, according to whether the type
1187 @code{char} should be signed or unsigned by default. The user can
1188 always override this default with the options @samp{-fsigned-char}
1189 and @samp{-funsigned-char}.
1191 @findex DEFAULT_SHORT_ENUMS
1192 @item DEFAULT_SHORT_ENUMS
1193 A C expression to determine whether to give an @code{enum} type
1194 only as many bytes as it takes to represent the range of possible values
1195 of that type. A nonzero value means to do that; a zero value means all
1196 @code{enum} types should be allocated like @code{int}.
1198 If you don't define the macro, the default is 0.
1202 A C expression for a string describing the name of the data type to use
1203 for size values. The typedef name @code{size_t} is defined using the
1204 contents of the string.
1206 The string can contain more than one keyword. If so, separate them with
1207 spaces, and write first any length keyword, then @code{unsigned} if
1208 appropriate, and finally @code{int}. The string must exactly match one
1209 of the data type names defined in the function
1210 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1211 omit @code{int} or change the order---that would cause the compiler to
1214 If you don't define this macro, the default is @code{"long unsigned
1217 @findex PTRDIFF_TYPE
1219 A C expression for a string describing the name of the data type to use
1220 for the result of subtracting two pointers. The typedef name
1221 @code{ptrdiff_t} is defined using the contents of the string. See
1222 @code{SIZE_TYPE} above for more information.
1224 If you don't define this macro, the default is @code{"long int"}.
1228 A C expression for a string describing the name of the data type to use
1229 for wide characters. The typedef name @code{wchar_t} is defined using
1230 the contents of the string. See @code{SIZE_TYPE} above for more
1233 If you don't define this macro, the default is @code{"int"}.
1235 @findex WCHAR_TYPE_SIZE
1236 @item WCHAR_TYPE_SIZE
1237 A C expression for the size in bits of the data type for wide
1238 characters. This is used in @code{cpp}, which cannot make use of
1241 @findex MAX_WCHAR_TYPE_SIZE
1242 @item MAX_WCHAR_TYPE_SIZE
1243 Maximum number for the size in bits of the data type for wide
1244 characters. If this is undefined, the default is
1245 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1246 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1249 @findex OBJC_INT_SELECTORS
1250 @item OBJC_INT_SELECTORS
1251 Define this macro if the type of Objective C selectors should be
1254 If this macro is not defined, then selectors should have the type
1255 @code{struct objc_selector *}.
1257 @findex OBJC_SELECTORS_WITHOUT_LABELS
1258 @item OBJC_SELECTORS_WITHOUT_LABELS
1259 Define this macro if the compiler can group all the selectors together
1260 into a vector and use just one label at the beginning of the vector.
1261 Otherwise, the compiler must give each selector its own assembler
1264 On certain machines, it is important to have a separate label for each
1265 selector because this enables the linker to eliminate duplicate selectors.
1269 A C constant expression for the integer value for escape sequence
1274 @findex TARGET_NEWLINE
1277 @itemx TARGET_NEWLINE
1278 C constant expressions for the integer values for escape sequences
1279 @samp{\b}, @samp{\t} and @samp{\n}.
1287 C constant expressions for the integer values for escape sequences
1288 @samp{\v}, @samp{\f} and @samp{\r}.
1292 @section Register Usage
1293 @cindex register usage
1295 This section explains how to describe what registers the target machine
1296 has, and how (in general) they can be used.
1298 The description of which registers a specific instruction can use is
1299 done with register classes; see @ref{Register Classes}. For information
1300 on using registers to access a stack frame, see @ref{Frame Registers}.
1301 For passing values in registers, see @ref{Register Arguments}.
1302 For returning values in registers, see @ref{Scalar Return}.
1305 * Register Basics:: Number and kinds of registers.
1306 * Allocation Order:: Order in which registers are allocated.
1307 * Values in Registers:: What kinds of values each reg can hold.
1308 * Leaf Functions:: Renumbering registers for leaf functions.
1309 * Stack Registers:: Handling a register stack such as 80387.
1312 @node Register Basics
1313 @subsection Basic Characteristics of Registers
1315 @c prevent bad page break with this line
1316 Registers have various characteristics.
1319 @findex FIRST_PSEUDO_REGISTER
1320 @item FIRST_PSEUDO_REGISTER
1321 Number of hardware registers known to the compiler. They receive
1322 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1323 pseudo register's number really is assigned the number
1324 @code{FIRST_PSEUDO_REGISTER}.
1326 @item FIXED_REGISTERS
1327 @findex FIXED_REGISTERS
1328 @cindex fixed register
1329 An initializer that says which registers are used for fixed purposes
1330 all throughout the compiled code and are therefore not available for
1331 general allocation. These would include the stack pointer, the frame
1332 pointer (except on machines where that can be used as a general
1333 register when no frame pointer is needed), the program counter on
1334 machines where that is considered one of the addressable registers,
1335 and any other numbered register with a standard use.
1337 This information is expressed as a sequence of numbers, separated by
1338 commas and surrounded by braces. The @var{n}th number is 1 if
1339 register @var{n} is fixed, 0 otherwise.
1341 The table initialized from this macro, and the table initialized by
1342 the following one, may be overridden at run time either automatically,
1343 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1344 the user with the command options @samp{-ffixed-@var{reg}},
1345 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}}.
1347 @findex CALL_USED_REGISTERS
1348 @item CALL_USED_REGISTERS
1349 @cindex call-used register
1350 @cindex call-clobbered register
1351 @cindex call-saved register
1352 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1353 clobbered (in general) by function calls as well as for fixed
1354 registers. This macro therefore identifies the registers that are not
1355 available for general allocation of values that must live across
1358 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1359 automatically saves it on function entry and restores it on function
1360 exit, if the register is used within the function.
1362 @findex HARD_REGNO_CALL_PART_CLOBBERED
1363 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1364 @cindex call-used register
1365 @cindex call-clobbered register
1366 @cindex call-saved register
1367 A C expression that is non-zero if it is not permissible to store a
1368 value of mode @var{mode} in hard register number @var{regno} across a
1369 call without some part of it being clobbered. For most machines this
1370 macro need not be defined. It is only required for machines that do not
1371 preserve the entire contents of a register across a call.
1373 @findex CONDITIONAL_REGISTER_USAGE
1375 @findex call_used_regs
1376 @item CONDITIONAL_REGISTER_USAGE
1377 Zero or more C statements that may conditionally modify four variables
1378 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs}
1379 (these three are of type @code{char []}) and @code{reg_class_contents}
1380 (of type @code{HARD_REG_SET}).
1381 Before the macro is called @code{fixed_regs}, @code{call_used_regs}
1382 and @code{reg_class_contents} have been initialized from
1383 @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS} and
1384 @code{REG_CLASS_CONTENTS}, respectively,
1385 @code{global_regs} has been cleared, and any @samp{-ffixed-@var{reg}},
1386 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}} command
1387 options have been applied.
1389 This is necessary in case the fixed or call-clobbered registers depend
1392 You need not define this macro if it has no work to do.
1394 @cindex disabling certain registers
1395 @cindex controlling register usage
1396 If the usage of an entire class of registers depends on the target
1397 flags, you may indicate this to GCC by using this macro to modify
1398 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1399 registers in the classes which should not be used by GCC. Also define
1400 the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1401 is called with a letter for a class that shouldn't be used.
1403 (However, if this class is not included in @code{GENERAL_REGS} and all
1404 of the insn patterns whose constraints permit this class are
1405 controlled by target switches, then GCC will automatically avoid using
1406 these registers when the target switches are opposed to them.)
1408 @findex NON_SAVING_SETJMP
1409 @item NON_SAVING_SETJMP
1410 If this macro is defined and has a nonzero value, it means that
1411 @code{setjmp} and related functions fail to save the registers, or that
1412 @code{longjmp} fails to restore them. To compensate, the compiler
1413 avoids putting variables in registers in functions that use
1416 @findex INCOMING_REGNO
1417 @item INCOMING_REGNO (@var{out})
1418 Define this macro if the target machine has register windows. This C
1419 expression returns the register number as seen by the called function
1420 corresponding to the register number @var{out} as seen by the calling
1421 function. Return @var{out} if register number @var{out} is not an
1424 @findex OUTGOING_REGNO
1425 @item OUTGOING_REGNO (@var{in})
1426 Define this macro if the target machine has register windows. This C
1427 expression returns the register number as seen by the calling function
1428 corresponding to the register number @var{in} as seen by the called
1429 function. Return @var{in} if register number @var{in} is not an inbound
1435 If the program counter has a register number, define this as that
1436 register number. Otherwise, do not define it.
1440 @node Allocation Order
1441 @subsection Order of Allocation of Registers
1442 @cindex order of register allocation
1443 @cindex register allocation order
1445 @c prevent bad page break with this line
1446 Registers are allocated in order.
1449 @findex REG_ALLOC_ORDER
1450 @item REG_ALLOC_ORDER
1451 If defined, an initializer for a vector of integers, containing the
1452 numbers of hard registers in the order in which GCC should prefer
1453 to use them (from most preferred to least).
1455 If this macro is not defined, registers are used lowest numbered first
1456 (all else being equal).
1458 One use of this macro is on machines where the highest numbered
1459 registers must always be saved and the save-multiple-registers
1460 instruction supports only sequences of consecutive registers. On such
1461 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1462 the highest numbered allocable register first.
1464 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1465 @item ORDER_REGS_FOR_LOCAL_ALLOC
1466 A C statement (sans semicolon) to choose the order in which to allocate
1467 hard registers for pseudo-registers local to a basic block.
1469 Store the desired register order in the array @code{reg_alloc_order}.
1470 Element 0 should be the register to allocate first; element 1, the next
1471 register; and so on.
1473 The macro body should not assume anything about the contents of
1474 @code{reg_alloc_order} before execution of the macro.
1476 On most machines, it is not necessary to define this macro.
1479 @node Values in Registers
1480 @subsection How Values Fit in Registers
1482 This section discusses the macros that describe which kinds of values
1483 (specifically, which machine modes) each register can hold, and how many
1484 consecutive registers are needed for a given mode.
1487 @findex HARD_REGNO_NREGS
1488 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1489 A C expression for the number of consecutive hard registers, starting
1490 at register number @var{regno}, required to hold a value of mode
1493 On a machine where all registers are exactly one word, a suitable
1494 definition of this macro is
1497 #define HARD_REGNO_NREGS(REGNO, MODE) \
1498 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1502 @findex ALTER_HARD_SUBREG
1503 @item ALTER_HARD_SUBREG (@var{tgt_mode}, @var{word}, @var{src_mode}, @var{regno})
1504 A C expression that returns an adjusted hard register number for
1507 (subreg:@var{tgt_mode} (reg:@var{src_mode} @var{regno}) @var{word})
1510 This may be needed if the target machine has mixed sized big-endian
1511 registers, like Sparc v9.
1513 @findex HARD_REGNO_MODE_OK
1514 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1515 A C expression that is nonzero if it is permissible to store a value
1516 of mode @var{mode} in hard register number @var{regno} (or in several
1517 registers starting with that one). For a machine where all registers
1518 are equivalent, a suitable definition is
1521 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1524 You need not include code to check for the numbers of fixed registers,
1525 because the allocation mechanism considers them to be always occupied.
1527 @cindex register pairs
1528 On some machines, double-precision values must be kept in even/odd
1529 register pairs. You can implement that by defining this macro to reject
1530 odd register numbers for such modes.
1532 The minimum requirement for a mode to be OK in a register is that the
1533 @samp{mov@var{mode}} instruction pattern support moves between the
1534 register and other hard register in the same class and that moving a
1535 value into the register and back out not alter it.
1537 Since the same instruction used to move @code{word_mode} will work for
1538 all narrower integer modes, it is not necessary on any machine for
1539 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1540 you define patterns @samp{movhi}, etc., to take advantage of this. This
1541 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1542 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1545 Many machines have special registers for floating point arithmetic.
1546 Often people assume that floating point machine modes are allowed only
1547 in floating point registers. This is not true. Any registers that
1548 can hold integers can safely @emph{hold} a floating point machine
1549 mode, whether or not floating arithmetic can be done on it in those
1550 registers. Integer move instructions can be used to move the values.
1552 On some machines, though, the converse is true: fixed-point machine
1553 modes may not go in floating registers. This is true if the floating
1554 registers normalize any value stored in them, because storing a
1555 non-floating value there would garble it. In this case,
1556 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1557 floating registers. But if the floating registers do not automatically
1558 normalize, if you can store any bit pattern in one and retrieve it
1559 unchanged without a trap, then any machine mode may go in a floating
1560 register, so you can define this macro to say so.
1562 The primary significance of special floating registers is rather that
1563 they are the registers acceptable in floating point arithmetic
1564 instructions. However, this is of no concern to
1565 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1566 constraints for those instructions.
1568 On some machines, the floating registers are especially slow to access,
1569 so that it is better to store a value in a stack frame than in such a
1570 register if floating point arithmetic is not being done. As long as the
1571 floating registers are not in class @code{GENERAL_REGS}, they will not
1572 be used unless some pattern's constraint asks for one.
1574 @findex MODES_TIEABLE_P
1575 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
1576 A C expression that is nonzero if a value of mode
1577 @var{mode1} is accessible in mode @var{mode2} without copying.
1579 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
1580 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
1581 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
1582 should be nonzero. If they differ for any @var{r}, you should define
1583 this macro to return zero unless some other mechanism ensures the
1584 accessibility of the value in a narrower mode.
1586 You should define this macro to return nonzero in as many cases as
1587 possible since doing so will allow GCC to perform better register
1590 @findex AVOID_CCMODE_COPIES
1591 @item AVOID_CCMODE_COPIES
1592 Define this macro if the compiler should avoid copies to/from @code{CCmode}
1593 registers. You should only define this macro if support fo copying to/from
1594 @code{CCmode} is incomplete.
1597 @node Leaf Functions
1598 @subsection Handling Leaf Functions
1600 @cindex leaf functions
1601 @cindex functions, leaf
1602 On some machines, a leaf function (i.e., one which makes no calls) can run
1603 more efficiently if it does not make its own register window. Often this
1604 means it is required to receive its arguments in the registers where they
1605 are passed by the caller, instead of the registers where they would
1608 The special treatment for leaf functions generally applies only when
1609 other conditions are met; for example, often they may use only those
1610 registers for its own variables and temporaries. We use the term ``leaf
1611 function'' to mean a function that is suitable for this special
1612 handling, so that functions with no calls are not necessarily ``leaf
1615 GCC assigns register numbers before it knows whether the function is
1616 suitable for leaf function treatment. So it needs to renumber the
1617 registers in order to output a leaf function. The following macros
1621 @findex LEAF_REGISTERS
1622 @item LEAF_REGISTERS
1623 A C initializer for a vector, indexed by hard register number, which
1624 contains 1 for a register that is allowable in a candidate for leaf
1627 If leaf function treatment involves renumbering the registers, then the
1628 registers marked here should be the ones before renumbering---those that
1629 GCC would ordinarily allocate. The registers which will actually be
1630 used in the assembler code, after renumbering, should not be marked with 1
1633 Define this macro only if the target machine offers a way to optimize
1634 the treatment of leaf functions.
1636 @findex LEAF_REG_REMAP
1637 @item LEAF_REG_REMAP (@var{regno})
1638 A C expression whose value is the register number to which @var{regno}
1639 should be renumbered, when a function is treated as a leaf function.
1641 If @var{regno} is a register number which should not appear in a leaf
1642 function before renumbering, then the expression should yield -1, which
1643 will cause the compiler to abort.
1645 Define this macro only if the target machine offers a way to optimize the
1646 treatment of leaf functions, and registers need to be renumbered to do
1650 @findex current_function_is_leaf
1651 @findex current_function_uses_only_leaf_regs
1652 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
1653 treat leaf functions specially. They can test the C variable
1654 @code{current_function_is_leaf} which is nonzero for leaf functions.
1655 @code{current_function_is_leaf} is set prior to local register allocation
1656 and is valid for the remaining compiler passes. They can also test the C
1657 variable @code{current_function_uses_only_leaf_regs} which is nonzero for
1658 leaf functions which only use leaf registers.
1659 @code{current_function_uses_only_leaf_regs} is valid after reload and is
1660 only useful if @code{LEAF_REGISTERS} is defined.
1661 @c changed this to fix overfull. ALSO: why the "it" at the beginning
1662 @c of the next paragraph?! --mew 2feb93
1664 @node Stack Registers
1665 @subsection Registers That Form a Stack
1667 There are special features to handle computers where some of the
1668 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
1669 Stack registers are normally written by pushing onto the stack, and are
1670 numbered relative to the top of the stack.
1672 Currently, GCC can only handle one group of stack-like registers, and
1673 they must be consecutively numbered.
1678 Define this if the machine has any stack-like registers.
1680 @findex FIRST_STACK_REG
1681 @item FIRST_STACK_REG
1682 The number of the first stack-like register. This one is the top
1685 @findex LAST_STACK_REG
1686 @item LAST_STACK_REG
1687 The number of the last stack-like register. This one is the bottom of
1691 @node Register Classes
1692 @section Register Classes
1693 @cindex register class definitions
1694 @cindex class definitions, register
1696 On many machines, the numbered registers are not all equivalent.
1697 For example, certain registers may not be allowed for indexed addressing;
1698 certain registers may not be allowed in some instructions. These machine
1699 restrictions are described to the compiler using @dfn{register classes}.
1701 You define a number of register classes, giving each one a name and saying
1702 which of the registers belong to it. Then you can specify register classes
1703 that are allowed as operands to particular instruction patterns.
1707 In general, each register will belong to several classes. In fact, one
1708 class must be named @code{ALL_REGS} and contain all the registers. Another
1709 class must be named @code{NO_REGS} and contain no registers. Often the
1710 union of two classes will be another class; however, this is not required.
1712 @findex GENERAL_REGS
1713 One of the classes must be named @code{GENERAL_REGS}. There is nothing
1714 terribly special about the name, but the operand constraint letters
1715 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
1716 the same as @code{ALL_REGS}, just define it as a macro which expands
1719 Order the classes so that if class @var{x} is contained in class @var{y}
1720 then @var{x} has a lower class number than @var{y}.
1722 The way classes other than @code{GENERAL_REGS} are specified in operand
1723 constraints is through machine-dependent operand constraint letters.
1724 You can define such letters to correspond to various classes, then use
1725 them in operand constraints.
1727 You should define a class for the union of two classes whenever some
1728 instruction allows both classes. For example, if an instruction allows
1729 either a floating point (coprocessor) register or a general register for a
1730 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
1731 which includes both of them. Otherwise you will get suboptimal code.
1733 You must also specify certain redundant information about the register
1734 classes: for each class, which classes contain it and which ones are
1735 contained in it; for each pair of classes, the largest class contained
1738 When a value occupying several consecutive registers is expected in a
1739 certain class, all the registers used must belong to that class.
1740 Therefore, register classes cannot be used to enforce a requirement for
1741 a register pair to start with an even-numbered register. The way to
1742 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
1744 Register classes used for input-operands of bitwise-and or shift
1745 instructions have a special requirement: each such class must have, for
1746 each fixed-point machine mode, a subclass whose registers can transfer that
1747 mode to or from memory. For example, on some machines, the operations for
1748 single-byte values (@code{QImode}) are limited to certain registers. When
1749 this is so, each register class that is used in a bitwise-and or shift
1750 instruction must have a subclass consisting of registers from which
1751 single-byte values can be loaded or stored. This is so that
1752 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
1755 @findex enum reg_class
1756 @item enum reg_class
1757 An enumeral type that must be defined with all the register class names
1758 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
1759 must be the last register class, followed by one more enumeral value,
1760 @code{LIM_REG_CLASSES}, which is not a register class but rather
1761 tells how many classes there are.
1763 Each register class has a number, which is the value of casting
1764 the class name to type @code{int}. The number serves as an index
1765 in many of the tables described below.
1767 @findex N_REG_CLASSES
1769 The number of distinct register classes, defined as follows:
1772 #define N_REG_CLASSES (int) LIM_REG_CLASSES
1775 @findex REG_CLASS_NAMES
1776 @item REG_CLASS_NAMES
1777 An initializer containing the names of the register classes as C string
1778 constants. These names are used in writing some of the debugging dumps.
1780 @findex REG_CLASS_CONTENTS
1781 @item REG_CLASS_CONTENTS
1782 An initializer containing the contents of the register classes, as integers
1783 which are bit masks. The @var{n}th integer specifies the contents of class
1784 @var{n}. The way the integer @var{mask} is interpreted is that
1785 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
1787 When the machine has more than 32 registers, an integer does not suffice.
1788 Then the integers are replaced by sub-initializers, braced groupings containing
1789 several integers. Each sub-initializer must be suitable as an initializer
1790 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
1792 @findex REGNO_REG_CLASS
1793 @item REGNO_REG_CLASS (@var{regno})
1794 A C expression whose value is a register class containing hard register
1795 @var{regno}. In general there is more than one such class; choose a class
1796 which is @dfn{minimal}, meaning that no smaller class also contains the
1799 @findex BASE_REG_CLASS
1800 @item BASE_REG_CLASS
1801 A macro whose definition is the name of the class to which a valid
1802 base register must belong. A base register is one used in an address
1803 which is the register value plus a displacement.
1805 @findex INDEX_REG_CLASS
1806 @item INDEX_REG_CLASS
1807 A macro whose definition is the name of the class to which a valid
1808 index register must belong. An index register is one used in an
1809 address where its value is either multiplied by a scale factor or
1810 added to another register (as well as added to a displacement).
1812 @findex REG_CLASS_FROM_LETTER
1813 @item REG_CLASS_FROM_LETTER (@var{char})
1814 A C expression which defines the machine-dependent operand constraint
1815 letters for register classes. If @var{char} is such a letter, the
1816 value should be the register class corresponding to it. Otherwise,
1817 the value should be @code{NO_REGS}. The register letter @samp{r},
1818 corresponding to class @code{GENERAL_REGS}, will not be passed
1819 to this macro; you do not need to handle it.
1821 @findex REGNO_OK_FOR_BASE_P
1822 @item REGNO_OK_FOR_BASE_P (@var{num})
1823 A C expression which is nonzero if register number @var{num} is
1824 suitable for use as a base register in operand addresses. It may be
1825 either a suitable hard register or a pseudo register that has been
1826 allocated such a hard register.
1828 @findex REGNO_MODE_OK_FOR_BASE_P
1829 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
1830 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
1831 that expression may examine the mode of the memory reference in
1832 @var{mode}. You should define this macro if the mode of the memory
1833 reference affects whether a register may be used as a base register. If
1834 you define this macro, the compiler will use it instead of
1835 @code{REGNO_OK_FOR_BASE_P}.
1837 @findex REGNO_OK_FOR_INDEX_P
1838 @item REGNO_OK_FOR_INDEX_P (@var{num})
1839 A C expression which is nonzero if register number @var{num} is
1840 suitable for use as an index register in operand addresses. It may be
1841 either a suitable hard register or a pseudo register that has been
1842 allocated such a hard register.
1844 The difference between an index register and a base register is that
1845 the index register may be scaled. If an address involves the sum of
1846 two registers, neither one of them scaled, then either one may be
1847 labeled the ``base'' and the other the ``index''; but whichever
1848 labeling is used must fit the machine's constraints of which registers
1849 may serve in each capacity. The compiler will try both labelings,
1850 looking for one that is valid, and will reload one or both registers
1851 only if neither labeling works.
1853 @findex PREFERRED_RELOAD_CLASS
1854 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
1855 A C expression that places additional restrictions on the register class
1856 to use when it is necessary to copy value @var{x} into a register in class
1857 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
1858 another, smaller class. On many machines, the following definition is
1862 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
1865 Sometimes returning a more restrictive class makes better code. For
1866 example, on the 68000, when @var{x} is an integer constant that is in range
1867 for a @samp{moveq} instruction, the value of this macro is always
1868 @code{DATA_REGS} as long as @var{class} includes the data registers.
1869 Requiring a data register guarantees that a @samp{moveq} will be used.
1871 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
1872 you can force @var{x} into a memory constant. This is useful on
1873 certain machines where immediate floating values cannot be loaded into
1874 certain kinds of registers.
1876 @findex PREFERRED_OUTPUT_RELOAD_CLASS
1877 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
1878 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
1879 input reloads. If you don't define this macro, the default is to use
1880 @var{class}, unchanged.
1882 @findex LIMIT_RELOAD_CLASS
1883 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
1884 A C expression that places additional restrictions on the register class
1885 to use when it is necessary to be able to hold a value of mode
1886 @var{mode} in a reload register for which class @var{class} would
1889 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
1890 there are certain modes that simply can't go in certain reload classes.
1892 The value is a register class; perhaps @var{class}, or perhaps another,
1895 Don't define this macro unless the target machine has limitations which
1896 require the macro to do something nontrivial.
1898 @findex SECONDARY_RELOAD_CLASS
1899 @findex SECONDARY_INPUT_RELOAD_CLASS
1900 @findex SECONDARY_OUTPUT_RELOAD_CLASS
1901 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1902 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1903 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1904 Many machines have some registers that cannot be copied directly to or
1905 from memory or even from other types of registers. An example is the
1906 @samp{MQ} register, which on most machines, can only be copied to or
1907 from general registers, but not memory. Some machines allow copying all
1908 registers to and from memory, but require a scratch register for stores
1909 to some memory locations (e.g., those with symbolic address on the RT,
1910 and those with certain symbolic address on the Sparc when compiling
1911 PIC). In some cases, both an intermediate and a scratch register are
1914 You should define these macros to indicate to the reload phase that it may
1915 need to allocate at least one register for a reload in addition to the
1916 register to contain the data. Specifically, if copying @var{x} to a
1917 register @var{class} in @var{mode} requires an intermediate register,
1918 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
1919 largest register class all of whose registers can be used as
1920 intermediate registers or scratch registers.
1922 If copying a register @var{class} in @var{mode} to @var{x} requires an
1923 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
1924 should be defined to return the largest register class required. If the
1925 requirements for input and output reloads are the same, the macro
1926 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
1929 The values returned by these macros are often @code{GENERAL_REGS}.
1930 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
1931 can be directly copied to or from a register of @var{class} in
1932 @var{mode} without requiring a scratch register. Do not define this
1933 macro if it would always return @code{NO_REGS}.
1935 If a scratch register is required (either with or without an
1936 intermediate register), you should define patterns for
1937 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
1938 (@pxref{Standard Names}. These patterns, which will normally be
1939 implemented with a @code{define_expand}, should be similar to the
1940 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
1943 Define constraints for the reload register and scratch register that
1944 contain a single register class. If the original reload register (whose
1945 class is @var{class}) can meet the constraint given in the pattern, the
1946 value returned by these macros is used for the class of the scratch
1947 register. Otherwise, two additional reload registers are required.
1948 Their classes are obtained from the constraints in the insn pattern.
1950 @var{x} might be a pseudo-register or a @code{subreg} of a
1951 pseudo-register, which could either be in a hard register or in memory.
1952 Use @code{true_regnum} to find out; it will return -1 if the pseudo is
1953 in memory and the hard register number if it is in a register.
1955 These macros should not be used in the case where a particular class of
1956 registers can only be copied to memory and not to another class of
1957 registers. In that case, secondary reload registers are not needed and
1958 would not be helpful. Instead, a stack location must be used to perform
1959 the copy and the @code{mov@var{m}} pattern should use memory as a
1960 intermediate storage. This case often occurs between floating-point and
1963 @findex SECONDARY_MEMORY_NEEDED
1964 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
1965 Certain machines have the property that some registers cannot be copied
1966 to some other registers without using memory. Define this macro on
1967 those machines to be a C expression that is non-zero if objects of mode
1968 @var{m} in registers of @var{class1} can only be copied to registers of
1969 class @var{class2} by storing a register of @var{class1} into memory
1970 and loading that memory location into a register of @var{class2}.
1972 Do not define this macro if its value would always be zero.
1974 @findex SECONDARY_MEMORY_NEEDED_RTX
1975 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
1976 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
1977 allocates a stack slot for a memory location needed for register copies.
1978 If this macro is defined, the compiler instead uses the memory location
1979 defined by this macro.
1981 Do not define this macro if you do not define
1982 @code{SECONDARY_MEMORY_NEEDED}.
1984 @findex SECONDARY_MEMORY_NEEDED_MODE
1985 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
1986 When the compiler needs a secondary memory location to copy between two
1987 registers of mode @var{mode}, it normally allocates sufficient memory to
1988 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
1989 load operations in a mode that many bits wide and whose class is the
1990 same as that of @var{mode}.
1992 This is right thing to do on most machines because it ensures that all
1993 bits of the register are copied and prevents accesses to the registers
1994 in a narrower mode, which some machines prohibit for floating-point
1997 However, this default behavior is not correct on some machines, such as
1998 the DEC Alpha, that store short integers in floating-point registers
1999 differently than in integer registers. On those machines, the default
2000 widening will not work correctly and you must define this macro to
2001 suppress that widening in some cases. See the file @file{alpha.h} for
2004 Do not define this macro if you do not define
2005 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2006 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2008 @findex SMALL_REGISTER_CLASSES
2009 @item SMALL_REGISTER_CLASSES
2010 On some machines, it is risky to let hard registers live across arbitrary
2011 insns. Typically, these machines have instructions that require values
2012 to be in specific registers (like an accumulator), and reload will fail
2013 if the required hard register is used for another purpose across such an
2016 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a non-zero
2017 value on these machines. When this macro has a non-zero value, the
2018 compiler will try to minimize the lifetime of hard registers.
2020 It is always safe to define this macro with a non-zero value, but if you
2021 unnecessarily define it, you will reduce the amount of optimizations
2022 that can be performed in some cases. If you do not define this macro
2023 with a non-zero value when it is required, the compiler will run out of
2024 spill registers and print a fatal error message. For most machines, you
2025 should not define this macro at all.
2027 @findex CLASS_LIKELY_SPILLED_P
2028 @item CLASS_LIKELY_SPILLED_P (@var{class})
2029 A C expression whose value is nonzero if pseudos that have been assigned
2030 to registers of class @var{class} would likely be spilled because
2031 registers of @var{class} are needed for spill registers.
2033 The default value of this macro returns 1 if @var{class} has exactly one
2034 register and zero otherwise. On most machines, this default should be
2035 used. Only define this macro to some other expression if pseudos
2036 allocated by @file{local-alloc.c} end up in memory because their hard
2037 registers were needed for spill registers. If this macro returns nonzero
2038 for those classes, those pseudos will only be allocated by
2039 @file{global.c}, which knows how to reallocate the pseudo to another
2040 register. If there would not be another register available for
2041 reallocation, you should not change the definition of this macro since
2042 the only effect of such a definition would be to slow down register
2045 @findex CLASS_MAX_NREGS
2046 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2047 A C expression for the maximum number of consecutive registers
2048 of class @var{class} needed to hold a value of mode @var{mode}.
2050 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2051 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2052 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2053 @var{mode})} for all @var{regno} values in the class @var{class}.
2055 This macro helps control the handling of multiple-word values
2058 @item CLASS_CANNOT_CHANGE_SIZE
2059 If defined, a C expression for a class that contains registers which the
2060 compiler must always access in a mode that is the same size as the mode
2061 in which it loaded the register.
2063 For the example, loading 32-bit integer or floating-point objects into
2064 floating-point registers on the Alpha extends them to 64-bits.
2065 Therefore loading a 64-bit object and then storing it as a 32-bit object
2066 does not store the low-order 32-bits, as would be the case for a normal
2067 register. Therefore, @file{alpha.h} defines this macro as
2071 Three other special macros describe which operands fit which constraint
2075 @findex CONST_OK_FOR_LETTER_P
2076 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2077 A C expression that defines the machine-dependent operand constraint
2078 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2079 particular ranges of integer values. If @var{c} is one of those
2080 letters, the expression should check that @var{value}, an integer, is in
2081 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2082 not one of those letters, the value should be 0 regardless of
2085 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2086 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2087 A C expression that defines the machine-dependent operand constraint
2088 letters that specify particular ranges of @code{const_double} values
2089 (@samp{G} or @samp{H}).
2091 If @var{c} is one of those letters, the expression should check that
2092 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2093 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2094 letters, the value should be 0 regardless of @var{value}.
2096 @code{const_double} is used for all floating-point constants and for
2097 @code{DImode} fixed-point constants. A given letter can accept either
2098 or both kinds of values. It can use @code{GET_MODE} to distinguish
2099 between these kinds.
2101 @findex EXTRA_CONSTRAINT
2102 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2103 A C expression that defines the optional machine-dependent constraint
2104 letters (@samp{Q}, @samp{R}, @samp{S}, @samp{T}, @samp{U}) that can
2105 be used to segregate specific types of operands, usually memory
2106 references, for the target machine. Normally this macro will not be
2107 defined. If it is required for a particular target machine, it should
2108 return 1 if @var{value} corresponds to the operand type represented by
2109 the constraint letter @var{c}. If @var{c} is not defined as an extra
2110 constraint, the value returned should be 0 regardless of @var{value}.
2112 For example, on the ROMP, load instructions cannot have their output in r0 if
2113 the memory reference contains a symbolic address. Constraint letter
2114 @samp{Q} is defined as representing a memory address that does
2115 @emph{not} contain a symbolic address. An alternative is specified with
2116 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2117 alternative specifies @samp{m} on the input and a register class that
2118 does not include r0 on the output.
2121 @node Stack and Calling
2122 @section Stack Layout and Calling Conventions
2123 @cindex calling conventions
2125 @c prevent bad page break with this line
2126 This describes the stack layout and calling conventions.
2134 * Register Arguments::
2136 * Aggregate Return::
2143 @subsection Basic Stack Layout
2144 @cindex stack frame layout
2145 @cindex frame layout
2147 @c prevent bad page break with this line
2148 Here is the basic stack layout.
2151 @findex STACK_GROWS_DOWNWARD
2152 @item STACK_GROWS_DOWNWARD
2153 Define this macro if pushing a word onto the stack moves the stack
2154 pointer to a smaller address.
2156 When we say, ``define this macro if @dots{},'' it means that the
2157 compiler checks this macro only with @code{#ifdef} so the precise
2158 definition used does not matter.
2160 @findex FRAME_GROWS_DOWNWARD
2161 @item FRAME_GROWS_DOWNWARD
2162 Define this macro if the addresses of local variable slots are at negative
2163 offsets from the frame pointer.
2165 @findex ARGS_GROW_DOWNWARD
2166 @item ARGS_GROW_DOWNWARD
2167 Define this macro if successive arguments to a function occupy decreasing
2168 addresses on the stack.
2170 @findex STARTING_FRAME_OFFSET
2171 @item STARTING_FRAME_OFFSET
2172 Offset from the frame pointer to the first local variable slot to be allocated.
2174 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2175 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2176 Otherwise, it is found by adding the length of the first slot to the
2177 value @code{STARTING_FRAME_OFFSET}.
2178 @c i'm not sure if the above is still correct.. had to change it to get
2179 @c rid of an overfull. --mew 2feb93
2181 @findex STACK_POINTER_OFFSET
2182 @item STACK_POINTER_OFFSET
2183 Offset from the stack pointer register to the first location at which
2184 outgoing arguments are placed. If not specified, the default value of
2185 zero is used. This is the proper value for most machines.
2187 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2188 the first location at which outgoing arguments are placed.
2190 @findex FIRST_PARM_OFFSET
2191 @item FIRST_PARM_OFFSET (@var{fundecl})
2192 Offset from the argument pointer register to the first argument's
2193 address. On some machines it may depend on the data type of the
2196 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2197 the first argument's address.
2199 @findex STACK_DYNAMIC_OFFSET
2200 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2201 Offset from the stack pointer register to an item dynamically allocated
2202 on the stack, e.g., by @code{alloca}.
2204 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2205 length of the outgoing arguments. The default is correct for most
2206 machines. See @file{function.c} for details.
2208 @findex DYNAMIC_CHAIN_ADDRESS
2209 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2210 A C expression whose value is RTL representing the address in a stack
2211 frame where the pointer to the caller's frame is stored. Assume that
2212 @var{frameaddr} is an RTL expression for the address of the stack frame
2215 If you don't define this macro, the default is to return the value
2216 of @var{frameaddr}---that is, the stack frame address is also the
2217 address of the stack word that points to the previous frame.
2219 @findex SETUP_FRAME_ADDRESSES
2220 @item SETUP_FRAME_ADDRESSES
2221 If defined, a C expression that produces the machine-specific code to
2222 setup the stack so that arbitrary frames can be accessed. For example,
2223 on the Sparc, we must flush all of the register windows to the stack
2224 before we can access arbitrary stack frames. You will seldom need to
2227 @findex BUILTIN_SETJMP_FRAME_VALUE
2228 @item BUILTIN_SETJMP_FRAME_VALUE
2229 If defined, a C expression that contains an rtx that is used to store
2230 the address of the current frame into the built in @code{setjmp} buffer.
2231 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2232 machines. One reason you may need to define this macro is if
2233 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2235 @findex RETURN_ADDR_RTX
2236 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2237 A C expression whose value is RTL representing the value of the return
2238 address for the frame @var{count} steps up from the current frame, after
2239 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2240 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2241 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2243 The value of the expression must always be the correct address when
2244 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2245 determine the return address of other frames.
2247 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2248 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2249 Define this if the return address of a particular stack frame is accessed
2250 from the frame pointer of the previous stack frame.
2252 @findex INCOMING_RETURN_ADDR_RTX
2253 @item INCOMING_RETURN_ADDR_RTX
2254 A C expression whose value is RTL representing the location of the
2255 incoming return address at the beginning of any function, before the
2256 prologue. This RTL is either a @code{REG}, indicating that the return
2257 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2260 You only need to define this macro if you want to support call frame
2261 debugging information like that provided by DWARF 2.
2263 @findex INCOMING_FRAME_SP_OFFSET
2264 @item INCOMING_FRAME_SP_OFFSET
2265 A C expression whose value is an integer giving the offset, in bytes,
2266 from the value of the stack pointer register to the top of the stack
2267 frame at the beginning of any function, before the prologue. The top of
2268 the frame is defined to be the value of the stack pointer in the
2269 previous frame, just before the call instruction.
2271 You only need to define this macro if you want to support call frame
2272 debugging information like that provided by DWARF 2.
2274 @findex ARG_POINTER_CFA_OFFSET
2275 @item ARG_POINTER_CFA_OFFSET
2276 A C expression whose value is an integer giving the offset, in bytes,
2277 from the argument pointer to the canonical frame address (cfa). The
2278 final value should coincide with that calculated by
2279 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2280 during virtual register instantiation.
2282 You only need to define this macro if you want to support call frame
2283 debugging information like that provided by DWARF 2.
2286 @node Stack Checking
2287 @subsection Specifying How Stack Checking is Done
2289 GCC will check that stack references are within the boundaries of
2290 the stack, if the @samp{-fstack-check} is specified, in one of three ways:
2294 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2295 will assume that you have arranged for stack checking to be done at
2296 appropriate places in the configuration files, e.g., in
2297 @code{FUNCTION_PROLOGUE}. GCC will do not other special processing.
2300 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
2301 called @code{check_stack} in your @file{md} file, GCC will call that
2302 pattern with one argument which is the address to compare the stack
2303 value against. You must arrange for this pattern to report an error if
2304 the stack pointer is out of range.
2307 If neither of the above are true, GCC will generate code to periodically
2308 ``probe'' the stack pointer using the values of the macros defined below.
2311 Normally, you will use the default values of these macros, so GCC
2312 will use the third approach.
2315 @findex STACK_CHECK_BUILTIN
2316 @item STACK_CHECK_BUILTIN
2317 A nonzero value if stack checking is done by the configuration files in a
2318 machine-dependent manner. You should define this macro if stack checking
2319 is require by the ABI of your machine or if you would like to have to stack
2320 checking in some more efficient way than GCC's portable approach.
2321 The default value of this macro is zero.
2323 @findex STACK_CHECK_PROBE_INTERVAL
2324 @item STACK_CHECK_PROBE_INTERVAL
2325 An integer representing the interval at which GCC must generate stack
2326 probe instructions. You will normally define this macro to be no larger
2327 than the size of the ``guard pages'' at the end of a stack area. The
2328 default value of 4096 is suitable for most systems.
2330 @findex STACK_CHECK_PROBE_LOAD
2331 @item STACK_CHECK_PROBE_LOAD
2332 A integer which is nonzero if GCC should perform the stack probe
2333 as a load instruction and zero if GCC should use a store instruction.
2334 The default is zero, which is the most efficient choice on most systems.
2336 @findex STACK_CHECK_PROTECT
2337 @item STACK_CHECK_PROTECT
2338 The number of bytes of stack needed to recover from a stack overflow,
2339 for languages where such a recovery is supported. The default value of
2340 75 words should be adequate for most machines.
2342 @findex STACK_CHECK_MAX_FRAME_SIZE
2343 @item STACK_CHECK_MAX_FRAME_SIZE
2344 The maximum size of a stack frame, in bytes. GCC will generate probe
2345 instructions in non-leaf functions to ensure at least this many bytes of
2346 stack are available. If a stack frame is larger than this size, stack
2347 checking will not be reliable and GCC will issue a warning. The
2348 default is chosen so that GCC only generates one instruction on most
2349 systems. You should normally not change the default value of this macro.
2351 @findex STACK_CHECK_FIXED_FRAME_SIZE
2352 @item STACK_CHECK_FIXED_FRAME_SIZE
2353 GCC uses this value to generate the above warning message. It
2354 represents the amount of fixed frame used by a function, not including
2355 space for any callee-saved registers, temporaries and user variables.
2356 You need only specify an upper bound for this amount and will normally
2357 use the default of four words.
2359 @findex STACK_CHECK_MAX_VAR_SIZE
2360 @item STACK_CHECK_MAX_VAR_SIZE
2361 The maximum size, in bytes, of an object that GCC will place in the
2362 fixed area of the stack frame when the user specifies
2363 @samp{-fstack-check}.
2364 GCC computed the default from the values of the above macros and you will
2365 normally not need to override that default.
2369 @node Frame Registers
2370 @subsection Registers That Address the Stack Frame
2372 @c prevent bad page break with this line
2373 This discusses registers that address the stack frame.
2376 @findex STACK_POINTER_REGNUM
2377 @item STACK_POINTER_REGNUM
2378 The register number of the stack pointer register, which must also be a
2379 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2380 the hardware determines which register this is.
2382 @findex FRAME_POINTER_REGNUM
2383 @item FRAME_POINTER_REGNUM
2384 The register number of the frame pointer register, which is used to
2385 access automatic variables in the stack frame. On some machines, the
2386 hardware determines which register this is. On other machines, you can
2387 choose any register you wish for this purpose.
2389 @findex HARD_FRAME_POINTER_REGNUM
2390 @item HARD_FRAME_POINTER_REGNUM
2391 On some machines the offset between the frame pointer and starting
2392 offset of the automatic variables is not known until after register
2393 allocation has been done (for example, because the saved registers are
2394 between these two locations). On those machines, define
2395 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
2396 be used internally until the offset is known, and define
2397 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
2398 used for the frame pointer.
2400 You should define this macro only in the very rare circumstances when it
2401 is not possible to calculate the offset between the frame pointer and
2402 the automatic variables until after register allocation has been
2403 completed. When this macro is defined, you must also indicate in your
2404 definition of @code{ELIMINABLE_REGS} how to eliminate
2405 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
2406 or @code{STACK_POINTER_REGNUM}.
2408 Do not define this macro if it would be the same as
2409 @code{FRAME_POINTER_REGNUM}.
2411 @findex ARG_POINTER_REGNUM
2412 @item ARG_POINTER_REGNUM
2413 The register number of the arg pointer register, which is used to access
2414 the function's argument list. On some machines, this is the same as the
2415 frame pointer register. On some machines, the hardware determines which
2416 register this is. On other machines, you can choose any register you
2417 wish for this purpose. If this is not the same register as the frame
2418 pointer register, then you must mark it as a fixed register according to
2419 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2420 (@pxref{Elimination}).
2422 @findex RETURN_ADDRESS_POINTER_REGNUM
2423 @item RETURN_ADDRESS_POINTER_REGNUM
2424 The register number of the return address pointer register, which is used to
2425 access the current function's return address from the stack. On some
2426 machines, the return address is not at a fixed offset from the frame
2427 pointer or stack pointer or argument pointer. This register can be defined
2428 to point to the return address on the stack, and then be converted by
2429 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
2431 Do not define this macro unless there is no other way to get the return
2432 address from the stack.
2434 @findex STATIC_CHAIN_REGNUM
2435 @findex STATIC_CHAIN_INCOMING_REGNUM
2436 @item STATIC_CHAIN_REGNUM
2437 @itemx STATIC_CHAIN_INCOMING_REGNUM
2438 Register numbers used for passing a function's static chain pointer. If
2439 register windows are used, the register number as seen by the called
2440 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
2441 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
2442 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
2443 not be defined.@refill
2445 The static chain register need not be a fixed register.
2447 If the static chain is passed in memory, these macros should not be
2448 defined; instead, the next two macros should be defined.
2450 @findex STATIC_CHAIN
2451 @findex STATIC_CHAIN_INCOMING
2453 @itemx STATIC_CHAIN_INCOMING
2454 If the static chain is passed in memory, these macros provide rtx giving
2455 @code{mem} expressions that denote where they are stored.
2456 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
2457 as seen by the calling and called functions, respectively. Often the former
2458 will be at an offset from the stack pointer and the latter at an offset from
2459 the frame pointer.@refill
2461 @findex stack_pointer_rtx
2462 @findex frame_pointer_rtx
2463 @findex arg_pointer_rtx
2464 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
2465 @code{arg_pointer_rtx} will have been initialized prior to the use of these
2466 macros and should be used to refer to those items.
2468 If the static chain is passed in a register, the two previous macros should
2473 @subsection Eliminating Frame Pointer and Arg Pointer
2475 @c prevent bad page break with this line
2476 This is about eliminating the frame pointer and arg pointer.
2479 @findex FRAME_POINTER_REQUIRED
2480 @item FRAME_POINTER_REQUIRED
2481 A C expression which is nonzero if a function must have and use a frame
2482 pointer. This expression is evaluated in the reload pass. If its value is
2483 nonzero the function will have a frame pointer.
2485 The expression can in principle examine the current function and decide
2486 according to the facts, but on most machines the constant 0 or the
2487 constant 1 suffices. Use 0 when the machine allows code to be generated
2488 with no frame pointer, and doing so saves some time or space. Use 1
2489 when there is no possible advantage to avoiding a frame pointer.
2491 In certain cases, the compiler does not know how to produce valid code
2492 without a frame pointer. The compiler recognizes those cases and
2493 automatically gives the function a frame pointer regardless of what
2494 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
2497 In a function that does not require a frame pointer, the frame pointer
2498 register can be allocated for ordinary usage, unless you mark it as a
2499 fixed register. See @code{FIXED_REGISTERS} for more information.
2501 @findex INITIAL_FRAME_POINTER_OFFSET
2502 @findex get_frame_size
2503 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
2504 A C statement to store in the variable @var{depth-var} the difference
2505 between the frame pointer and the stack pointer values immediately after
2506 the function prologue. The value would be computed from information
2507 such as the result of @code{get_frame_size ()} and the tables of
2508 registers @code{regs_ever_live} and @code{call_used_regs}.
2510 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
2511 need not be defined. Otherwise, it must be defined even if
2512 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
2513 case, you may set @var{depth-var} to anything.
2515 @findex ELIMINABLE_REGS
2516 @item ELIMINABLE_REGS
2517 If defined, this macro specifies a table of register pairs used to
2518 eliminate unneeded registers that point into the stack frame. If it is not
2519 defined, the only elimination attempted by the compiler is to replace
2520 references to the frame pointer with references to the stack pointer.
2522 The definition of this macro is a list of structure initializations, each
2523 of which specifies an original and replacement register.
2525 On some machines, the position of the argument pointer is not known until
2526 the compilation is completed. In such a case, a separate hard register
2527 must be used for the argument pointer. This register can be eliminated by
2528 replacing it with either the frame pointer or the argument pointer,
2529 depending on whether or not the frame pointer has been eliminated.
2531 In this case, you might specify:
2533 #define ELIMINABLE_REGS \
2534 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
2535 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
2536 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
2539 Note that the elimination of the argument pointer with the stack pointer is
2540 specified first since that is the preferred elimination.
2542 @findex CAN_ELIMINATE
2543 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
2544 A C expression that returns non-zero if the compiler is allowed to try
2545 to replace register number @var{from-reg} with register number
2546 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
2547 is defined, and will usually be the constant 1, since most of the cases
2548 preventing register elimination are things that the compiler already
2551 @findex INITIAL_ELIMINATION_OFFSET
2552 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
2553 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
2554 specifies the initial difference between the specified pair of
2555 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
2558 @findex LONGJMP_RESTORE_FROM_STACK
2559 @item LONGJMP_RESTORE_FROM_STACK
2560 Define this macro if the @code{longjmp} function restores registers from
2561 the stack frames, rather than from those saved specifically by
2562 @code{setjmp}. Certain quantities must not be kept in registers across
2563 a call to @code{setjmp} on such machines.
2566 @node Stack Arguments
2567 @subsection Passing Function Arguments on the Stack
2568 @cindex arguments on stack
2569 @cindex stack arguments
2571 The macros in this section control how arguments are passed
2572 on the stack. See the following section for other macros that
2573 control passing certain arguments in registers.
2576 @findex PROMOTE_PROTOTYPES
2577 @item PROMOTE_PROTOTYPES
2578 A C expression whose value is nonzero if an argument declared in
2579 a prototype as an integral type smaller than @code{int} should
2580 actually be passed as an @code{int}. In addition to avoiding
2581 errors in certain cases of mismatch, it also makes for better
2582 code on certain machines. If the macro is not defined in target
2583 header files, it defaults to 0.
2585 @findex PUSH_ROUNDING
2586 @item PUSH_ROUNDING (@var{npushed})
2587 A C expression that is the number of bytes actually pushed onto the
2588 stack when an instruction attempts to push @var{npushed} bytes.
2590 If the target machine does not have a push instruction, do not define
2591 this macro. That directs GCC to use an alternate strategy: to
2592 allocate the entire argument block and then store the arguments into
2595 On some machines, the definition
2598 #define PUSH_ROUNDING(BYTES) (BYTES)
2602 will suffice. But on other machines, instructions that appear
2603 to push one byte actually push two bytes in an attempt to maintain
2604 alignment. Then the definition should be
2607 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
2610 @findex ACCUMULATE_OUTGOING_ARGS
2611 @findex current_function_outgoing_args_size
2612 @item ACCUMULATE_OUTGOING_ARGS
2613 If defined, the maximum amount of space required for outgoing arguments
2614 will be computed and placed into the variable
2615 @code{current_function_outgoing_args_size}. No space will be pushed
2616 onto the stack for each call; instead, the function prologue should
2617 increase the stack frame size by this amount.
2619 Defining both @code{PUSH_ROUNDING} and @code{ACCUMULATE_OUTGOING_ARGS}
2622 @findex REG_PARM_STACK_SPACE
2623 @item REG_PARM_STACK_SPACE (@var{fndecl})
2624 Define this macro if functions should assume that stack space has been
2625 allocated for arguments even when their values are passed in
2628 The value of this macro is the size, in bytes, of the area reserved for
2629 arguments passed in registers for the function represented by @var{fndecl},
2630 which can be zero if GCC is calling a library function.
2632 This space can be allocated by the caller, or be a part of the
2633 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
2635 @c above is overfull. not sure what to do. --mew 5feb93 did
2636 @c something, not sure if it looks good. --mew 10feb93
2638 @findex MAYBE_REG_PARM_STACK_SPACE
2639 @findex FINAL_REG_PARM_STACK_SPACE
2640 @item MAYBE_REG_PARM_STACK_SPACE
2641 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
2642 Define these macros in addition to the one above if functions might
2643 allocate stack space for arguments even when their values are passed
2644 in registers. These should be used when the stack space allocated
2645 for arguments in registers is not a simple constant independent of the
2646 function declaration.
2648 The value of the first macro is the size, in bytes, of the area that
2649 we should initially assume would be reserved for arguments passed in registers.
2651 The value of the second macro is the actual size, in bytes, of the area
2652 that will be reserved for arguments passed in registers. This takes two
2653 arguments: an integer representing the number of bytes of fixed sized
2654 arguments on the stack, and a tree representing the number of bytes of
2655 variable sized arguments on the stack.
2657 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
2658 called for libcall functions, the current function, or for a function
2659 being called when it is known that such stack space must be allocated.
2660 In each case this value can be easily computed.
2662 When deciding whether a called function needs such stack space, and how
2663 much space to reserve, GCC uses these two macros instead of
2664 @code{REG_PARM_STACK_SPACE}.
2666 @findex OUTGOING_REG_PARM_STACK_SPACE
2667 @item OUTGOING_REG_PARM_STACK_SPACE
2668 Define this if it is the responsibility of the caller to allocate the area
2669 reserved for arguments passed in registers.
2671 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
2672 whether the space for these arguments counts in the value of
2673 @code{current_function_outgoing_args_size}.
2675 @findex STACK_PARMS_IN_REG_PARM_AREA
2676 @item STACK_PARMS_IN_REG_PARM_AREA
2677 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
2678 stack parameters don't skip the area specified by it.
2679 @c i changed this, makes more sens and it should have taken care of the
2680 @c overfull.. not as specific, tho. --mew 5feb93
2682 Normally, when a parameter is not passed in registers, it is placed on the
2683 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
2684 suppresses this behavior and causes the parameter to be passed on the
2685 stack in its natural location.
2687 @findex RETURN_POPS_ARGS
2688 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
2689 A C expression that should indicate the number of bytes of its own
2690 arguments that a function pops on returning, or 0 if the
2691 function pops no arguments and the caller must therefore pop them all
2692 after the function returns.
2694 @var{fundecl} is a C variable whose value is a tree node that describes
2695 the function in question. Normally it is a node of type
2696 @code{FUNCTION_DECL} that describes the declaration of the function.
2697 From this you can obtain the DECL_MACHINE_ATTRIBUTES of the function.
2699 @var{funtype} is a C variable whose value is a tree node that
2700 describes the function in question. Normally it is a node of type
2701 @code{FUNCTION_TYPE} that describes the data type of the function.
2702 From this it is possible to obtain the data types of the value and
2703 arguments (if known).
2705 When a call to a library function is being considered, @var{fundecl}
2706 will contain an identifier node for the library function. Thus, if
2707 you need to distinguish among various library functions, you can do so
2708 by their names. Note that ``library function'' in this context means
2709 a function used to perform arithmetic, whose name is known specially
2710 in the compiler and was not mentioned in the C code being compiled.
2712 @var{stack-size} is the number of bytes of arguments passed on the
2713 stack. If a variable number of bytes is passed, it is zero, and
2714 argument popping will always be the responsibility of the calling function.
2716 On the Vax, all functions always pop their arguments, so the definition
2717 of this macro is @var{stack-size}. On the 68000, using the standard
2718 calling convention, no functions pop their arguments, so the value of
2719 the macro is always 0 in this case. But an alternative calling
2720 convention is available in which functions that take a fixed number of
2721 arguments pop them but other functions (such as @code{printf}) pop
2722 nothing (the caller pops all). When this convention is in use,
2723 @var{funtype} is examined to determine whether a function takes a fixed
2724 number of arguments.
2727 @node Register Arguments
2728 @subsection Passing Arguments in Registers
2729 @cindex arguments in registers
2730 @cindex registers arguments
2732 This section describes the macros which let you control how various
2733 types of arguments are passed in registers or how they are arranged in
2737 @findex FUNCTION_ARG
2738 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2739 A C expression that controls whether a function argument is passed
2740 in a register, and which register.
2742 The arguments are @var{cum}, which summarizes all the previous
2743 arguments; @var{mode}, the machine mode of the argument; @var{type},
2744 the data type of the argument as a tree node or 0 if that is not known
2745 (which happens for C support library functions); and @var{named},
2746 which is 1 for an ordinary argument and 0 for nameless arguments that
2747 correspond to @samp{@dots{}} in the called function's prototype.
2749 The value of the expression is usually either a @code{reg} RTX for the
2750 hard register in which to pass the argument, or zero to pass the
2751 argument on the stack.
2753 For machines like the Vax and 68000, where normally all arguments are
2754 pushed, zero suffices as a definition.
2756 The value of the expression can also be a @code{parallel} RTX. This is
2757 used when an argument is passed in multiple locations. The mode of the
2758 of the @code{parallel} should be the mode of the entire argument. The
2759 @code{parallel} holds any number of @code{expr_list} pairs; each one
2760 describes where part of the argument is passed. In each
2761 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
2762 register in which to pass this part of the argument, and the mode of the
2763 register RTX indicates how large this part of the argument is. The
2764 second operand of the @code{expr_list} is a @code{const_int} which gives
2765 the offset in bytes into the entire argument of where this part starts.
2766 As a special exception the first @code{expr_list} in the @code{parallel}
2767 RTX may have a first operand of zero. This indicates that the entire
2768 argument is also stored on the stack.
2770 @cindex @file{stdarg.h} and register arguments
2771 The usual way to make the ANSI library @file{stdarg.h} work on a machine
2772 where some arguments are usually passed in registers, is to cause
2773 nameless arguments to be passed on the stack instead. This is done
2774 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
2776 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
2777 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
2778 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
2779 in the definition of this macro to determine if this argument is of a
2780 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
2781 is not defined and @code{FUNCTION_ARG} returns non-zero for such an
2782 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
2783 defined, the argument will be computed in the stack and then loaded into
2786 @findex MUST_PASS_IN_STACK
2787 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
2788 Define as a C expression that evaluates to nonzero if we do not know how
2789 to pass TYPE solely in registers. The file @file{expr.h} defines a
2790 definition that is usually appropriate, refer to @file{expr.h} for additional
2793 @findex FUNCTION_INCOMING_ARG
2794 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2795 Define this macro if the target machine has ``register windows'', so
2796 that the register in which a function sees an arguments is not
2797 necessarily the same as the one in which the caller passed the
2800 For such machines, @code{FUNCTION_ARG} computes the register in which
2801 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
2802 be defined in a similar fashion to tell the function being called
2803 where the arguments will arrive.
2805 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
2806 serves both purposes.@refill
2808 @findex FUNCTION_ARG_PARTIAL_NREGS
2809 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
2810 A C expression for the number of words, at the beginning of an
2811 argument, must be put in registers. The value must be zero for
2812 arguments that are passed entirely in registers or that are entirely
2813 pushed on the stack.
2815 On some machines, certain arguments must be passed partially in
2816 registers and partially in memory. On these machines, typically the
2817 first @var{n} words of arguments are passed in registers, and the rest
2818 on the stack. If a multi-word argument (a @code{double} or a
2819 structure) crosses that boundary, its first few words must be passed
2820 in registers and the rest must be pushed. This macro tells the
2821 compiler when this occurs, and how many of the words should go in
2824 @code{FUNCTION_ARG} for these arguments should return the first
2825 register to be used by the caller for this argument; likewise
2826 @code{FUNCTION_INCOMING_ARG}, for the called function.
2828 @findex FUNCTION_ARG_PASS_BY_REFERENCE
2829 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2830 A C expression that indicates when an argument must be passed by reference.
2831 If nonzero for an argument, a copy of that argument is made in memory and a
2832 pointer to the argument is passed instead of the argument itself.
2833 The pointer is passed in whatever way is appropriate for passing a pointer
2836 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
2837 definition of this macro might be
2839 #define FUNCTION_ARG_PASS_BY_REFERENCE\
2840 (CUM, MODE, TYPE, NAMED) \
2841 MUST_PASS_IN_STACK (MODE, TYPE)
2843 @c this is *still* too long. --mew 5feb93
2845 @findex FUNCTION_ARG_CALLEE_COPIES
2846 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
2847 If defined, a C expression that indicates when it is the called function's
2848 responsibility to make a copy of arguments passed by invisible reference.
2849 Normally, the caller makes a copy and passes the address of the copy to the
2850 routine being called. When FUNCTION_ARG_CALLEE_COPIES is defined and is
2851 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
2852 ``live'' value. The called function must not modify this value. If it can be
2853 determined that the value won't be modified, it need not make a copy;
2854 otherwise a copy must be made.
2856 @findex CUMULATIVE_ARGS
2857 @item CUMULATIVE_ARGS
2858 A C type for declaring a variable that is used as the first argument of
2859 @code{FUNCTION_ARG} and other related values. For some target machines,
2860 the type @code{int} suffices and can hold the number of bytes of
2863 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
2864 arguments that have been passed on the stack. The compiler has other
2865 variables to keep track of that. For target machines on which all
2866 arguments are passed on the stack, there is no need to store anything in
2867 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
2868 should not be empty, so use @code{int}.
2870 @findex INIT_CUMULATIVE_ARGS
2871 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
2872 A C statement (sans semicolon) for initializing the variable @var{cum}
2873 for the state at the beginning of the argument list. The variable has
2874 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
2875 for the data type of the function which will receive the args, or 0
2876 if the args are to a compiler support library function. The value of
2877 @var{indirect} is nonzero when processing an indirect call, for example
2878 a call through a function pointer. The value of @var{indirect} is zero
2879 for a call to an explicitly named function, a library function call, or when
2880 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
2883 When processing a call to a compiler support library function,
2884 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
2885 contains the name of the function, as a string. @var{libname} is 0 when
2886 an ordinary C function call is being processed. Thus, each time this
2887 macro is called, either @var{libname} or @var{fntype} is nonzero, but
2888 never both of them at once.
2890 @findex INIT_CUMULATIVE_INCOMING_ARGS
2891 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
2892 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
2893 finding the arguments for the function being compiled. If this macro is
2894 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
2896 The value passed for @var{libname} is always 0, since library routines
2897 with special calling conventions are never compiled with GCC. The
2898 argument @var{libname} exists for symmetry with
2899 @code{INIT_CUMULATIVE_ARGS}.
2900 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
2901 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
2903 @findex FUNCTION_ARG_ADVANCE
2904 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2905 A C statement (sans semicolon) to update the summarizer variable
2906 @var{cum} to advance past an argument in the argument list. The
2907 values @var{mode}, @var{type} and @var{named} describe that argument.
2908 Once this is done, the variable @var{cum} is suitable for analyzing
2909 the @emph{following} argument with @code{FUNCTION_ARG}, etc.@refill
2911 This macro need not do anything if the argument in question was passed
2912 on the stack. The compiler knows how to track the amount of stack space
2913 used for arguments without any special help.
2915 @findex FUNCTION_ARG_PADDING
2916 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
2917 If defined, a C expression which determines whether, and in which direction,
2918 to pad out an argument with extra space. The value should be of type
2919 @code{enum direction}: either @code{upward} to pad above the argument,
2920 @code{downward} to pad below, or @code{none} to inhibit padding.
2922 The @emph{amount} of padding is always just enough to reach the next
2923 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
2926 This macro has a default definition which is right for most systems.
2927 For little-endian machines, the default is to pad upward. For
2928 big-endian machines, the default is to pad downward for an argument of
2929 constant size shorter than an @code{int}, and upward otherwise.
2931 @findex FUNCTION_ARG_BOUNDARY
2932 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
2933 If defined, a C expression that gives the alignment boundary, in bits,
2934 of an argument with the specified mode and type. If it is not defined,
2935 @code{PARM_BOUNDARY} is used for all arguments.
2937 @findex FUNCTION_ARG_REGNO_P
2938 @item FUNCTION_ARG_REGNO_P (@var{regno})
2939 A C expression that is nonzero if @var{regno} is the number of a hard
2940 register in which function arguments are sometimes passed. This does
2941 @emph{not} include implicit arguments such as the static chain and
2942 the structure-value address. On many machines, no registers can be
2943 used for this purpose since all function arguments are pushed on the
2946 @findex LOAD_ARGS_REVERSED
2947 @item LOAD_ARGS_REVERSED
2948 If defined, the order in which arguments are loaded into their
2949 respective argument registers is reversed so that the last
2950 argument is loaded first. This macro only affects arguments
2951 passed in registers.
2956 @subsection How Scalar Function Values Are Returned
2957 @cindex return values in registers
2958 @cindex values, returned by functions
2959 @cindex scalars, returned as values
2961 This section discusses the macros that control returning scalars as
2962 values---values that can fit in registers.
2965 @findex TRADITIONAL_RETURN_FLOAT
2966 @item TRADITIONAL_RETURN_FLOAT
2967 Define this macro if @samp{-traditional} should not cause functions
2968 declared to return @code{float} to convert the value to @code{double}.
2970 @findex FUNCTION_VALUE
2971 @item FUNCTION_VALUE (@var{valtype}, @var{func})
2972 A C expression to create an RTX representing the place where a
2973 function returns a value of data type @var{valtype}. @var{valtype} is
2974 a tree node representing a data type. Write @code{TYPE_MODE
2975 (@var{valtype})} to get the machine mode used to represent that type.
2976 On many machines, only the mode is relevant. (Actually, on most
2977 machines, scalar values are returned in the same place regardless of
2980 The value of the expression is usually a @code{reg} RTX for the hard
2981 register where the return value is stored. The value can also be a
2982 @code{parallel} RTX, if the return value is in multiple places. See
2983 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
2985 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
2986 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
2989 If the precise function being called is known, @var{func} is a tree
2990 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
2991 pointer. This makes it possible to use a different value-returning
2992 convention for specific functions when all their calls are
2995 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
2996 types, because these are returned in another way. See
2997 @code{STRUCT_VALUE_REGNUM} and related macros, below.
2999 @findex FUNCTION_OUTGOING_VALUE
3000 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3001 Define this macro if the target machine has ``register windows''
3002 so that the register in which a function returns its value is not
3003 the same as the one in which the caller sees the value.
3005 For such machines, @code{FUNCTION_VALUE} computes the register in which
3006 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3007 defined in a similar fashion to tell the function where to put the
3010 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3011 @code{FUNCTION_VALUE} serves both purposes.@refill
3013 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3014 aggregate data types, because these are returned in another way. See
3015 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3017 @findex LIBCALL_VALUE
3018 @item LIBCALL_VALUE (@var{mode})
3019 A C expression to create an RTX representing the place where a library
3020 function returns a value of mode @var{mode}. If the precise function
3021 being called is known, @var{func} is a tree node
3022 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3023 pointer. This makes it possible to use a different value-returning
3024 convention for specific functions when all their calls are
3027 Note that ``library function'' in this context means a compiler
3028 support routine, used to perform arithmetic, whose name is known
3029 specially by the compiler and was not mentioned in the C code being
3032 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3033 data types, because none of the library functions returns such types.
3035 @findex FUNCTION_VALUE_REGNO_P
3036 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3037 A C expression that is nonzero if @var{regno} is the number of a hard
3038 register in which the values of called function may come back.
3040 A register whose use for returning values is limited to serving as the
3041 second of a pair (for a value of type @code{double}, say) need not be
3042 recognized by this macro. So for most machines, this definition
3046 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3049 If the machine has register windows, so that the caller and the called
3050 function use different registers for the return value, this macro
3051 should recognize only the caller's register numbers.
3053 @findex APPLY_RESULT_SIZE
3054 @item APPLY_RESULT_SIZE
3055 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3056 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3057 saving and restoring an arbitrary return value.
3060 @node Aggregate Return
3061 @subsection How Large Values Are Returned
3062 @cindex aggregates as return values
3063 @cindex large return values
3064 @cindex returning aggregate values
3065 @cindex structure value address
3067 When a function value's mode is @code{BLKmode} (and in some other
3068 cases), the value is not returned according to @code{FUNCTION_VALUE}
3069 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3070 block of memory in which the value should be stored. This address
3071 is called the @dfn{structure value address}.
3073 This section describes how to control returning structure values in
3077 @findex RETURN_IN_MEMORY
3078 @item RETURN_IN_MEMORY (@var{type})
3079 A C expression which can inhibit the returning of certain function
3080 values in registers, based on the type of value. A nonzero value says
3081 to return the function value in memory, just as large structures are
3082 always returned. Here @var{type} will be a C expression of type
3083 @code{tree}, representing the data type of the value.
3085 Note that values of mode @code{BLKmode} must be explicitly handled
3086 by this macro. Also, the option @samp{-fpcc-struct-return}
3087 takes effect regardless of this macro. On most systems, it is
3088 possible to leave the macro undefined; this causes a default
3089 definition to be used, whose value is the constant 1 for @code{BLKmode}
3090 values, and 0 otherwise.
3092 Do not use this macro to indicate that structures and unions should always
3093 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3096 @findex DEFAULT_PCC_STRUCT_RETURN
3097 @item DEFAULT_PCC_STRUCT_RETURN
3098 Define this macro to be 1 if all structure and union return values must be
3099 in memory. Since this results in slower code, this should be defined
3100 only if needed for compatibility with other compilers or with an ABI.
3101 If you define this macro to be 0, then the conventions used for structure
3102 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3104 If not defined, this defaults to the value 1.
3106 @findex STRUCT_VALUE_REGNUM
3107 @item STRUCT_VALUE_REGNUM
3108 If the structure value address is passed in a register, then
3109 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3111 @findex STRUCT_VALUE
3113 If the structure value address is not passed in a register, define
3114 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3115 where the address is passed. If it returns 0, the address is passed as
3116 an ``invisible'' first argument.
3118 @findex STRUCT_VALUE_INCOMING_REGNUM
3119 @item STRUCT_VALUE_INCOMING_REGNUM
3120 On some architectures the place where the structure value address
3121 is found by the called function is not the same place that the
3122 caller put it. This can be due to register windows, or it could
3123 be because the function prologue moves it to a different place.
3125 If the incoming location of the structure value address is in a
3126 register, define this macro as the register number.
3128 @findex STRUCT_VALUE_INCOMING
3129 @item STRUCT_VALUE_INCOMING
3130 If the incoming location is not a register, then you should define
3131 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3132 called function should find the value. If it should find the value on
3133 the stack, define this to create a @code{mem} which refers to the frame
3134 pointer. A definition of 0 means that the address is passed as an
3135 ``invisible'' first argument.
3137 @findex PCC_STATIC_STRUCT_RETURN
3138 @item PCC_STATIC_STRUCT_RETURN
3139 Define this macro if the usual system convention on the target machine
3140 for returning structures and unions is for the called function to return
3141 the address of a static variable containing the value.
3143 Do not define this if the usual system convention is for the caller to
3144 pass an address to the subroutine.
3146 This macro has effect in @samp{-fpcc-struct-return} mode, but it does
3147 nothing when you use @samp{-freg-struct-return} mode.
3151 @subsection Caller-Saves Register Allocation
3153 If you enable it, GCC can save registers around function calls. This
3154 makes it possible to use call-clobbered registers to hold variables that
3155 must live across calls.
3158 @findex DEFAULT_CALLER_SAVES
3159 @item DEFAULT_CALLER_SAVES
3160 Define this macro if function calls on the target machine do not preserve
3161 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3162 for all registers. When defined, this macro enables @samp{-fcaller-saves}
3163 by default for all optimization levels. It has no effect for optimization
3164 levels 2 and higher, where @samp{-fcaller-saves} is the default.
3166 @findex CALLER_SAVE_PROFITABLE
3167 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3168 A C expression to determine whether it is worthwhile to consider placing
3169 a pseudo-register in a call-clobbered hard register and saving and
3170 restoring it around each function call. The expression should be 1 when
3171 this is worth doing, and 0 otherwise.
3173 If you don't define this macro, a default is used which is good on most
3174 machines: @code{4 * @var{calls} < @var{refs}}.
3176 @findex HARD_REGNO_CALLER_SAVE_MODE
3177 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3178 A C expression specifying which mode is required for saving @var{nregs}
3179 of a pseudo-register in call-clobbered hard register @var{regno}. If
3180 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3181 returned. For most machines this macro need not be defined since GCC
3182 will select the smallest suitable mode.
3185 @node Function Entry
3186 @subsection Function Entry and Exit
3187 @cindex function entry and exit
3191 This section describes the macros that output function entry
3192 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3195 @findex FUNCTION_PROLOGUE
3196 @item FUNCTION_PROLOGUE (@var{file}, @var{size})
3197 A C compound statement that outputs the assembler code for entry to a
3198 function. The prologue is responsible for setting up the stack frame,
3199 initializing the frame pointer register, saving registers that must be
3200 saved, and allocating @var{size} additional bytes of storage for the
3201 local variables. @var{size} is an integer. @var{file} is a stdio
3202 stream to which the assembler code should be output.
3204 The label for the beginning of the function need not be output by this
3205 macro. That has already been done when the macro is run.
3207 @findex regs_ever_live
3208 To determine which registers to save, the macro can refer to the array
3209 @code{regs_ever_live}: element @var{r} is nonzero if hard register
3210 @var{r} is used anywhere within the function. This implies the function
3211 prologue should save register @var{r}, provided it is not one of the
3212 call-used registers. (@code{FUNCTION_EPILOGUE} must likewise use
3213 @code{regs_ever_live}.)
3215 On machines that have ``register windows'', the function entry code does
3216 not save on the stack the registers that are in the windows, even if
3217 they are supposed to be preserved by function calls; instead it takes
3218 appropriate steps to ``push'' the register stack, if any non-call-used
3219 registers are used in the function.
3221 @findex frame_pointer_needed
3222 On machines where functions may or may not have frame-pointers, the
3223 function entry code must vary accordingly; it must set up the frame
3224 pointer if one is wanted, and not otherwise. To determine whether a
3225 frame pointer is in wanted, the macro can refer to the variable
3226 @code{frame_pointer_needed}. The variable's value will be 1 at run
3227 time in a function that needs a frame pointer. @xref{Elimination}.
3229 The function entry code is responsible for allocating any stack space
3230 required for the function. This stack space consists of the regions
3231 listed below. In most cases, these regions are allocated in the
3232 order listed, with the last listed region closest to the top of the
3233 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
3234 the highest address if it is not defined). You can use a different order
3235 for a machine if doing so is more convenient or required for
3236 compatibility reasons. Except in cases where required by standard
3237 or by a debugger, there is no reason why the stack layout used by GCC
3238 need agree with that used by other compilers for a machine.
3242 @findex current_function_pretend_args_size
3243 A region of @code{current_function_pretend_args_size} bytes of
3244 uninitialized space just underneath the first argument arriving on the
3245 stack. (This may not be at the very start of the allocated stack region
3246 if the calling sequence has pushed anything else since pushing the stack
3247 arguments. But usually, on such machines, nothing else has been pushed
3248 yet, because the function prologue itself does all the pushing.) This
3249 region is used on machines where an argument may be passed partly in
3250 registers and partly in memory, and, in some cases to support the
3251 features in @file{varargs.h} and @file{stdargs.h}.
3254 An area of memory used to save certain registers used by the function.
3255 The size of this area, which may also include space for such things as
3256 the return address and pointers to previous stack frames, is
3257 machine-specific and usually depends on which registers have been used
3258 in the function. Machines with register windows often do not require
3262 A region of at least @var{size} bytes, possibly rounded up to an allocation
3263 boundary, to contain the local variables of the function. On some machines,
3264 this region and the save area may occur in the opposite order, with the
3265 save area closer to the top of the stack.
3268 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3269 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3270 @code{current_function_outgoing_args_size} bytes to be used for outgoing
3271 argument lists of the function. @xref{Stack Arguments}.
3274 Normally, it is necessary for the macros @code{FUNCTION_PROLOGUE} and
3275 @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C
3276 variable @code{current_function_is_leaf} is nonzero for such a function.
3278 @findex EXIT_IGNORE_STACK
3279 @item EXIT_IGNORE_STACK
3280 Define this macro as a C expression that is nonzero if the return
3281 instruction or the function epilogue ignores the value of the stack
3282 pointer; in other words, if it is safe to delete an instruction to
3283 adjust the stack pointer before a return from the function.
3285 Note that this macro's value is relevant only for functions for which
3286 frame pointers are maintained. It is never safe to delete a final
3287 stack adjustment in a function that has no frame pointer, and the
3288 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3290 @findex EPILOGUE_USES
3291 @item EPILOGUE_USES (@var{regno})
3292 Define this macro as a C expression that is nonzero for registers are
3293 used by the epilogue or the @samp{return} pattern. The stack and frame
3294 pointer registers are already be assumed to be used as needed.
3296 @findex FUNCTION_EPILOGUE
3297 @item FUNCTION_EPILOGUE (@var{file}, @var{size})
3298 A C compound statement that outputs the assembler code for exit from a
3299 function. The epilogue is responsible for restoring the saved
3300 registers and stack pointer to their values when the function was
3301 called, and returning control to the caller. This macro takes the
3302 same arguments as the macro @code{FUNCTION_PROLOGUE}, and the
3303 registers to restore are determined from @code{regs_ever_live} and
3304 @code{CALL_USED_REGISTERS} in the same way.
3306 On some machines, there is a single instruction that does all the work
3307 of returning from the function. On these machines, give that
3308 instruction the name @samp{return} and do not define the macro
3309 @code{FUNCTION_EPILOGUE} at all.
3311 Do not define a pattern named @samp{return} if you want the
3312 @code{FUNCTION_EPILOGUE} to be used. If you want the target switches
3313 to control whether return instructions or epilogues are used, define a
3314 @samp{return} pattern with a validity condition that tests the target
3315 switches appropriately. If the @samp{return} pattern's validity
3316 condition is false, epilogues will be used.
3318 On machines where functions may or may not have frame-pointers, the
3319 function exit code must vary accordingly. Sometimes the code for these
3320 two cases is completely different. To determine whether a frame pointer
3321 is wanted, the macro can refer to the variable
3322 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
3323 a function that needs a frame pointer.
3325 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
3326 treat leaf functions specially. The C variable @code{current_function_is_leaf}
3327 is nonzero for such a function. @xref{Leaf Functions}.
3329 On some machines, some functions pop their arguments on exit while
3330 others leave that for the caller to do. For example, the 68020 when
3331 given @samp{-mrtd} pops arguments in functions that take a fixed
3332 number of arguments.
3334 @findex current_function_pops_args
3335 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
3336 functions pop their own arguments. @code{FUNCTION_EPILOGUE} needs to
3337 know what was decided. The variable that is called
3338 @code{current_function_pops_args} is the number of bytes of its
3339 arguments that a function should pop. @xref{Scalar Return}.
3340 @c what is the "its arguments" in the above sentence referring to, pray
3341 @c tell? --mew 5feb93
3343 @findex DELAY_SLOTS_FOR_EPILOGUE
3344 @item DELAY_SLOTS_FOR_EPILOGUE
3345 Define this macro if the function epilogue contains delay slots to which
3346 instructions from the rest of the function can be ``moved''. The
3347 definition should be a C expression whose value is an integer
3348 representing the number of delay slots there.
3350 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
3351 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
3352 A C expression that returns 1 if @var{insn} can be placed in delay
3353 slot number @var{n} of the epilogue.
3355 The argument @var{n} is an integer which identifies the delay slot now
3356 being considered (since different slots may have different rules of
3357 eligibility). It is never negative and is always less than the number
3358 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
3359 If you reject a particular insn for a given delay slot, in principle, it
3360 may be reconsidered for a subsequent delay slot. Also, other insns may
3361 (at least in principle) be considered for the so far unfilled delay
3364 @findex current_function_epilogue_delay_list
3365 @findex final_scan_insn
3366 The insns accepted to fill the epilogue delay slots are put in an RTL
3367 list made with @code{insn_list} objects, stored in the variable
3368 @code{current_function_epilogue_delay_list}. The insn for the first
3369 delay slot comes first in the list. Your definition of the macro
3370 @code{FUNCTION_EPILOGUE} should fill the delay slots by outputting the
3371 insns in this list, usually by calling @code{final_scan_insn}.
3373 You need not define this macro if you did not define
3374 @code{DELAY_SLOTS_FOR_EPILOGUE}.
3376 @findex ASM_OUTPUT_MI_THUNK
3377 @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function})
3378 A C compound statement that outputs the assembler code for a thunk
3379 function, used to implement C++ virtual function calls with multiple
3380 inheritance. The thunk acts as a wrapper around a virtual function,
3381 adjusting the implicit object parameter before handing control off to
3384 First, emit code to add the integer @var{delta} to the location that
3385 contains the incoming first argument. Assume that this argument
3386 contains a pointer, and is the one used to pass the @code{this} pointer
3387 in C++. This is the incoming argument @emph{before} the function prologue,
3388 e.g. @samp{%o0} on a sparc. The addition must preserve the values of
3389 all other incoming arguments.
3391 After the addition, emit code to jump to @var{function}, which is a
3392 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
3393 not touch the return address. Hence returning from @var{FUNCTION} will
3394 return to whoever called the current @samp{thunk}.
3396 The effect must be as if @var{function} had been called directly with
3397 the adjusted first argument. This macro is responsible for emitting all
3398 of the code for a thunk function; @code{FUNCTION_PROLOGUE} and
3399 @code{FUNCTION_EPILOGUE} are not invoked.
3401 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
3402 have already been extracted from it.) It might possibly be useful on
3403 some targets, but probably not.
3405 If you do not define this macro, the target-independent code in the C++
3406 frontend will generate a less efficient heavyweight thunk that calls
3407 @var{function} instead of jumping to it. The generic approach does
3408 not support varargs.
3412 @subsection Generating Code for Profiling
3413 @cindex profiling, code generation
3415 These macros will help you generate code for profiling.
3418 @findex FUNCTION_PROFILER
3419 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
3420 A C statement or compound statement to output to @var{file} some
3421 assembler code to call the profiling subroutine @code{mcount}.
3422 Before calling, the assembler code must load the address of a
3423 counter variable into a register where @code{mcount} expects to
3424 find the address. The name of this variable is @samp{LP} followed
3425 by the number @var{labelno}, so you would generate the name using
3426 @samp{LP%d} in a @code{fprintf}.
3429 The details of how the address should be passed to @code{mcount} are
3430 determined by your operating system environment, not by GCC. To
3431 figure them out, compile a small program for profiling using the
3432 system's installed C compiler and look at the assembler code that
3435 @findex PROFILE_BEFORE_PROLOGUE
3436 @item PROFILE_BEFORE_PROLOGUE
3437 Define this macro if the code for function profiling should come before
3438 the function prologue. Normally, the profiling code comes after.
3440 @findex FUNCTION_BLOCK_PROFILER
3441 @vindex profile_block_flag
3442 @item FUNCTION_BLOCK_PROFILER (@var{file}, @var{labelno})
3443 A C statement or compound statement to output to @var{file} some
3444 assembler code to initialize basic-block profiling for the current
3445 object module. The global compile flag @code{profile_block_flag}
3446 distinguishes two profile modes.
3449 @findex __bb_init_func
3450 @item profile_block_flag != 2
3451 Output code to call the subroutine @code{__bb_init_func} once per
3452 object module, passing it as its sole argument the address of a block
3453 allocated in the object module.
3455 The name of the block is a local symbol made with this statement:
3458 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3461 Of course, since you are writing the definition of
3462 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3463 can take a short cut in the definition of this macro and use the name
3464 that you know will result.
3466 The first word of this block is a flag which will be nonzero if the
3467 object module has already been initialized. So test this word first,
3468 and do not call @code{__bb_init_func} if the flag is
3469 nonzero. BLOCK_OR_LABEL contains a unique number which may be used to
3470 generate a label as a branch destination when @code{__bb_init_func}
3473 Described in assembler language, the code to be output looks like:
3483 @findex __bb_init_trace_func
3484 @item profile_block_flag == 2
3485 Output code to call the subroutine @code{__bb_init_trace_func}
3486 and pass two parameters to it. The first parameter is the same as
3487 for @code{__bb_init_func}. The second parameter is the number of the
3488 first basic block of the function as given by BLOCK_OR_LABEL. Note
3489 that @code{__bb_init_trace_func} has to be called, even if the object
3490 module has been initialized already.
3492 Described in assembler language, the code to be output looks like:
3495 parameter2 <- BLOCK_OR_LABEL
3496 call __bb_init_trace_func
3500 @findex BLOCK_PROFILER
3501 @vindex profile_block_flag
3502 @item BLOCK_PROFILER (@var{file}, @var{blockno})
3503 A C statement or compound statement to output to @var{file} some
3504 assembler code to increment the count associated with the basic
3505 block number @var{blockno}. The global compile flag
3506 @code{profile_block_flag} distinguishes two profile modes.
3509 @item profile_block_flag != 2
3510 Output code to increment the counter directly. Basic blocks are
3511 numbered separately from zero within each compilation. The count
3512 associated with block number @var{blockno} is at index
3513 @var{blockno} in a vector of words; the name of this array is a local
3514 symbol made with this statement:
3517 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 2);
3520 @c This paragraph is the same as one a few paragraphs up.
3521 @c That is not an error.
3522 Of course, since you are writing the definition of
3523 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3524 can take a short cut in the definition of this macro and use the name
3525 that you know will result.
3527 Described in assembler language, the code to be output looks like:
3530 inc (LPBX2+4*BLOCKNO)
3534 @findex __bb_trace_func
3535 @item profile_block_flag == 2
3536 Output code to initialize the global structure @code{__bb} and
3537 call the function @code{__bb_trace_func}, which will increment the
3540 @code{__bb} consists of two words. In the first word, the current
3541 basic block number, as given by BLOCKNO, has to be stored. In
3542 the second word, the address of a block allocated in the object
3543 module has to be stored. The address is given by the label created
3544 with this statement:
3547 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3550 Described in assembler language, the code to be output looks like:
3552 move BLOCKNO -> (__bb)
3553 move LPBX0 -> (__bb+4)
3554 call __bb_trace_func
3558 @findex FUNCTION_BLOCK_PROFILER_EXIT
3559 @findex __bb_trace_ret
3560 @vindex profile_block_flag
3561 @item FUNCTION_BLOCK_PROFILER_EXIT (@var{file})
3562 A C statement or compound statement to output to @var{file}
3563 assembler code to call function @code{__bb_trace_ret}. The
3564 assembler code should only be output
3565 if the global compile flag @code{profile_block_flag} == 2. This
3566 macro has to be used at every place where code for returning from
3567 a function is generated (e.g. @code{FUNCTION_EPILOGUE}). Although
3568 you have to write the definition of @code{FUNCTION_EPILOGUE}
3569 as well, you have to define this macro to tell the compiler, that
3570 the proper call to @code{__bb_trace_ret} is produced.
3572 @findex MACHINE_STATE_SAVE
3573 @findex __bb_init_trace_func
3574 @findex __bb_trace_func
3575 @findex __bb_trace_ret
3576 @item MACHINE_STATE_SAVE (@var{id})
3577 A C statement or compound statement to save all registers, which may
3578 be clobbered by a function call, including condition codes. The
3579 @code{asm} statement will be mostly likely needed to handle this
3580 task. Local labels in the assembler code can be concatenated with the
3581 string @var{id}, to obtain a unique lable name.
3583 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3584 @code{FUNCTION_EPILOGUE} must be saved in the macros
3585 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3586 @code{BLOCK_PROFILER} prior calling @code{__bb_init_trace_func},
3587 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3589 @findex MACHINE_STATE_RESTORE
3590 @findex __bb_init_trace_func
3591 @findex __bb_trace_func
3592 @findex __bb_trace_ret
3593 @item MACHINE_STATE_RESTORE (@var{id})
3594 A C statement or compound statement to restore all registers, including
3595 condition codes, saved by @code{MACHINE_STATE_SAVE}.
3597 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3598 @code{FUNCTION_EPILOGUE} must be restored in the macros
3599 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3600 @code{BLOCK_PROFILER} after calling @code{__bb_init_trace_func},
3601 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3603 @findex BLOCK_PROFILER_CODE
3604 @item BLOCK_PROFILER_CODE
3605 A C function or functions which are needed in the library to
3606 support block profiling.
3610 @section Implementing the Varargs Macros
3611 @cindex varargs implementation
3613 GCC comes with an implementation of @file{varargs.h} and
3614 @file{stdarg.h} that work without change on machines that pass arguments
3615 on the stack. Other machines require their own implementations of
3616 varargs, and the two machine independent header files must have
3617 conditionals to include it.
3619 ANSI @file{stdarg.h} differs from traditional @file{varargs.h} mainly in
3620 the calling convention for @code{va_start}. The traditional
3621 implementation takes just one argument, which is the variable in which
3622 to store the argument pointer. The ANSI implementation of
3623 @code{va_start} takes an additional second argument. The user is
3624 supposed to write the last named argument of the function here.
3626 However, @code{va_start} should not use this argument. The way to find
3627 the end of the named arguments is with the built-in functions described
3631 @findex __builtin_saveregs
3632 @item __builtin_saveregs ()
3633 Use this built-in function to save the argument registers in memory so
3634 that the varargs mechanism can access them. Both ANSI and traditional
3635 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3636 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
3638 On some machines, @code{__builtin_saveregs} is open-coded under the
3639 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
3640 it calls a routine written in assembler language, found in
3643 Code generated for the call to @code{__builtin_saveregs} appears at the
3644 beginning of the function, as opposed to where the call to
3645 @code{__builtin_saveregs} is written, regardless of what the code is.
3646 This is because the registers must be saved before the function starts
3647 to use them for its own purposes.
3648 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3651 @findex __builtin_args_info
3652 @item __builtin_args_info (@var{category})
3653 Use this built-in function to find the first anonymous arguments in
3656 In general, a machine may have several categories of registers used for
3657 arguments, each for a particular category of data types. (For example,
3658 on some machines, floating-point registers are used for floating-point
3659 arguments while other arguments are passed in the general registers.)
3660 To make non-varargs functions use the proper calling convention, you
3661 have defined the @code{CUMULATIVE_ARGS} data type to record how many
3662 registers in each category have been used so far
3664 @code{__builtin_args_info} accesses the same data structure of type
3665 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
3666 with it, with @var{category} specifying which word to access. Thus, the
3667 value indicates the first unused register in a given category.
3669 Normally, you would use @code{__builtin_args_info} in the implementation
3670 of @code{va_start}, accessing each category just once and storing the
3671 value in the @code{va_list} object. This is because @code{va_list} will
3672 have to update the values, and there is no way to alter the
3673 values accessed by @code{__builtin_args_info}.
3675 @findex __builtin_next_arg
3676 @item __builtin_next_arg (@var{lastarg})
3677 This is the equivalent of @code{__builtin_args_info}, for stack
3678 arguments. It returns the address of the first anonymous stack
3679 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
3680 returns the address of the location above the first anonymous stack
3681 argument. Use it in @code{va_start} to initialize the pointer for
3682 fetching arguments from the stack. Also use it in @code{va_start} to
3683 verify that the second parameter @var{lastarg} is the last named argument
3684 of the current function.
3686 @findex __builtin_classify_type
3687 @item __builtin_classify_type (@var{object})
3688 Since each machine has its own conventions for which data types are
3689 passed in which kind of register, your implementation of @code{va_arg}
3690 has to embody these conventions. The easiest way to categorize the
3691 specified data type is to use @code{__builtin_classify_type} together
3692 with @code{sizeof} and @code{__alignof__}.
3694 @code{__builtin_classify_type} ignores the value of @var{object},
3695 considering only its data type. It returns an integer describing what
3696 kind of type that is---integer, floating, pointer, structure, and so on.
3698 The file @file{typeclass.h} defines an enumeration that you can use to
3699 interpret the values of @code{__builtin_classify_type}.
3702 These machine description macros help implement varargs:
3705 @findex EXPAND_BUILTIN_SAVEREGS
3706 @item EXPAND_BUILTIN_SAVEREGS ()
3707 If defined, is a C expression that produces the machine-specific code
3708 for a call to @code{__builtin_saveregs}. This code will be moved to the
3709 very beginning of the function, before any parameter access are made.
3710 The return value of this function should be an RTX that contains the
3711 value to use as the return of @code{__builtin_saveregs}.
3713 @findex SETUP_INCOMING_VARARGS
3714 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
3715 This macro offers an alternative to using @code{__builtin_saveregs} and
3716 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
3717 anonymous register arguments into the stack so that all the arguments
3718 appear to have been passed consecutively on the stack. Once this is
3719 done, you can use the standard implementation of varargs that works for
3720 machines that pass all their arguments on the stack.
3722 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
3723 structure, containing the values that obtain after processing of the
3724 named arguments. The arguments @var{mode} and @var{type} describe the
3725 last named argument---its machine mode and its data type as a tree node.
3727 The macro implementation should do two things: first, push onto the
3728 stack all the argument registers @emph{not} used for the named
3729 arguments, and second, store the size of the data thus pushed into the
3730 @code{int}-valued variable whose name is supplied as the argument
3731 @var{pretend_args_size}. The value that you store here will serve as
3732 additional offset for setting up the stack frame.
3734 Because you must generate code to push the anonymous arguments at
3735 compile time without knowing their data types,
3736 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
3737 a single category of argument register and use it uniformly for all data
3740 If the argument @var{second_time} is nonzero, it means that the
3741 arguments of the function are being analyzed for the second time. This
3742 happens for an inline function, which is not actually compiled until the
3743 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
3744 not generate any instructions in this case.
3746 @findex STRICT_ARGUMENT_NAMING
3747 @item STRICT_ARGUMENT_NAMING
3748 Define this macro to be a nonzero value if the location where a function
3749 argument is passed depends on whether or not it is a named argument.
3751 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
3752 is set for varargs and stdarg functions. If this macro returns a
3753 nonzero value, the @var{named} argument is always true for named
3754 arguments, and false for unnamed arguments. If it returns a value of
3755 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
3756 are treated as named. Otherwise, all named arguments except the last
3757 are treated as named.
3759 You need not define this macro if it always returns zero.
3761 @findex PRETEND_OUTGOING_VARARGS_NAMED
3762 @item PRETEND_OUTGOING_VARARGS_NAMED
3763 If you need to conditionally change ABIs so that one works with
3764 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
3765 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
3766 defined, then define this macro to return nonzero if
3767 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
3768 Otherwise, you should not define this macro.
3772 @section Trampolines for Nested Functions
3773 @cindex trampolines for nested functions
3774 @cindex nested functions, trampolines for
3776 A @dfn{trampoline} is a small piece of code that is created at run time
3777 when the address of a nested function is taken. It normally resides on
3778 the stack, in the stack frame of the containing function. These macros
3779 tell GCC how to generate code to allocate and initialize a
3782 The instructions in the trampoline must do two things: load a constant
3783 address into the static chain register, and jump to the real address of
3784 the nested function. On CISC machines such as the m68k, this requires
3785 two instructions, a move immediate and a jump. Then the two addresses
3786 exist in the trampoline as word-long immediate operands. On RISC
3787 machines, it is often necessary to load each address into a register in
3788 two parts. Then pieces of each address form separate immediate
3791 The code generated to initialize the trampoline must store the variable
3792 parts---the static chain value and the function address---into the
3793 immediate operands of the instructions. On a CISC machine, this is
3794 simply a matter of copying each address to a memory reference at the
3795 proper offset from the start of the trampoline. On a RISC machine, it
3796 may be necessary to take out pieces of the address and store them
3800 @findex TRAMPOLINE_TEMPLATE
3801 @item TRAMPOLINE_TEMPLATE (@var{file})
3802 A C statement to output, on the stream @var{file}, assembler code for a
3803 block of data that contains the constant parts of a trampoline. This
3804 code should not include a label---the label is taken care of
3807 If you do not define this macro, it means no template is needed
3808 for the target. Do not define this macro on systems where the block move
3809 code to copy the trampoline into place would be larger than the code
3810 to generate it on the spot.
3812 @findex TRAMPOLINE_SECTION
3813 @item TRAMPOLINE_SECTION
3814 The name of a subroutine to switch to the section in which the
3815 trampoline template is to be placed (@pxref{Sections}). The default is
3816 a value of @samp{readonly_data_section}, which places the trampoline in
3817 the section containing read-only data.
3819 @findex TRAMPOLINE_SIZE
3820 @item TRAMPOLINE_SIZE
3821 A C expression for the size in bytes of the trampoline, as an integer.
3823 @findex TRAMPOLINE_ALIGNMENT
3824 @item TRAMPOLINE_ALIGNMENT
3825 Alignment required for trampolines, in bits.
3827 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
3828 is used for aligning trampolines.
3830 @findex INITIALIZE_TRAMPOLINE
3831 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
3832 A C statement to initialize the variable parts of a trampoline.
3833 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
3834 an RTX for the address of the nested function; @var{static_chain} is an
3835 RTX for the static chain value that should be passed to the function
3838 @findex ALLOCATE_TRAMPOLINE
3839 @item ALLOCATE_TRAMPOLINE (@var{fp})
3840 A C expression to allocate run-time space for a trampoline. The
3841 expression value should be an RTX representing a memory reference to the
3842 space for the trampoline.
3844 @cindex @code{FUNCTION_EPILOGUE} and trampolines
3845 @cindex @code{FUNCTION_PROLOGUE} and trampolines
3846 If this macro is not defined, by default the trampoline is allocated as
3847 a stack slot. This default is right for most machines. The exceptions
3848 are machines where it is impossible to execute instructions in the stack
3849 area. On such machines, you may have to implement a separate stack,
3850 using this macro in conjunction with @code{FUNCTION_PROLOGUE} and
3851 @code{FUNCTION_EPILOGUE}.
3853 @var{fp} points to a data structure, a @code{struct function}, which
3854 describes the compilation status of the immediate containing function of
3855 the function which the trampoline is for. Normally (when
3856 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
3857 trampoline is in the stack frame of this containing function. Other
3858 allocation strategies probably must do something analogous with this
3862 Implementing trampolines is difficult on many machines because they have
3863 separate instruction and data caches. Writing into a stack location
3864 fails to clear the memory in the instruction cache, so when the program
3865 jumps to that location, it executes the old contents.
3867 Here are two possible solutions. One is to clear the relevant parts of
3868 the instruction cache whenever a trampoline is set up. The other is to
3869 make all trampolines identical, by having them jump to a standard
3870 subroutine. The former technique makes trampoline execution faster; the
3871 latter makes initialization faster.
3873 To clear the instruction cache when a trampoline is initialized, define
3874 the following macros which describe the shape of the cache.
3877 @findex INSN_CACHE_SIZE
3878 @item INSN_CACHE_SIZE
3879 The total size in bytes of the cache.
3881 @findex INSN_CACHE_LINE_WIDTH
3882 @item INSN_CACHE_LINE_WIDTH
3883 The length in bytes of each cache line. The cache is divided into cache
3884 lines which are disjoint slots, each holding a contiguous chunk of data
3885 fetched from memory. Each time data is brought into the cache, an
3886 entire line is read at once. The data loaded into a cache line is
3887 always aligned on a boundary equal to the line size.
3889 @findex INSN_CACHE_DEPTH
3890 @item INSN_CACHE_DEPTH
3891 The number of alternative cache lines that can hold any particular memory
3895 Alternatively, if the machine has system calls or instructions to clear
3896 the instruction cache directly, you can define the following macro.
3899 @findex CLEAR_INSN_CACHE
3900 @item CLEAR_INSN_CACHE (@var{BEG}, @var{END})
3901 If defined, expands to a C expression clearing the @emph{instruction
3902 cache} in the specified interval. If it is not defined, and the macro
3903 INSN_CACHE_SIZE is defined, some generic code is generated to clear the
3904 cache. The definition of this macro would typically be a series of
3905 @code{asm} statements. Both @var{BEG} and @var{END} are both pointer
3909 To use a standard subroutine, define the following macro. In addition,
3910 you must make sure that the instructions in a trampoline fill an entire
3911 cache line with identical instructions, or else ensure that the
3912 beginning of the trampoline code is always aligned at the same point in
3913 its cache line. Look in @file{m68k.h} as a guide.
3916 @findex TRANSFER_FROM_TRAMPOLINE
3917 @item TRANSFER_FROM_TRAMPOLINE
3918 Define this macro if trampolines need a special subroutine to do their
3919 work. The macro should expand to a series of @code{asm} statements
3920 which will be compiled with GCC. They go in a library function named
3921 @code{__transfer_from_trampoline}.
3923 If you need to avoid executing the ordinary prologue code of a compiled
3924 C function when you jump to the subroutine, you can do so by placing a
3925 special label of your own in the assembler code. Use one @code{asm}
3926 statement to generate an assembler label, and another to make the label
3927 global. Then trampolines can use that label to jump directly to your
3928 special assembler code.
3932 @section Implicit Calls to Library Routines
3933 @cindex library subroutine names
3934 @cindex @file{libgcc.a}
3936 @c prevent bad page break with this line
3937 Here is an explanation of implicit calls to library routines.
3940 @findex MULSI3_LIBCALL
3941 @item MULSI3_LIBCALL
3942 A C string constant giving the name of the function to call for
3943 multiplication of one signed full-word by another. If you do not
3944 define this macro, the default name is used, which is @code{__mulsi3},
3945 a function defined in @file{libgcc.a}.
3947 @findex DIVSI3_LIBCALL
3948 @item DIVSI3_LIBCALL
3949 A C string constant giving the name of the function to call for
3950 division of one signed full-word by another. If you do not define
3951 this macro, the default name is used, which is @code{__divsi3}, a
3952 function defined in @file{libgcc.a}.
3954 @findex UDIVSI3_LIBCALL
3955 @item UDIVSI3_LIBCALL
3956 A C string constant giving the name of the function to call for
3957 division of one unsigned full-word by another. If you do not define
3958 this macro, the default name is used, which is @code{__udivsi3}, a
3959 function defined in @file{libgcc.a}.
3961 @findex MODSI3_LIBCALL
3962 @item MODSI3_LIBCALL
3963 A C string constant giving the name of the function to call for the
3964 remainder in division of one signed full-word by another. If you do
3965 not define this macro, the default name is used, which is
3966 @code{__modsi3}, a function defined in @file{libgcc.a}.
3968 @findex UMODSI3_LIBCALL
3969 @item UMODSI3_LIBCALL
3970 A C string constant giving the name of the function to call for the
3971 remainder in division of one unsigned full-word by another. If you do
3972 not define this macro, the default name is used, which is
3973 @code{__umodsi3}, a function defined in @file{libgcc.a}.
3975 @findex MULDI3_LIBCALL
3976 @item MULDI3_LIBCALL
3977 A C string constant giving the name of the function to call for
3978 multiplication of one signed double-word by another. If you do not
3979 define this macro, the default name is used, which is @code{__muldi3},
3980 a function defined in @file{libgcc.a}.
3982 @findex DIVDI3_LIBCALL
3983 @item DIVDI3_LIBCALL
3984 A C string constant giving the name of the function to call for
3985 division of one signed double-word by another. If you do not define
3986 this macro, the default name is used, which is @code{__divdi3}, a
3987 function defined in @file{libgcc.a}.
3989 @findex UDIVDI3_LIBCALL
3990 @item UDIVDI3_LIBCALL
3991 A C string constant giving the name of the function to call for
3992 division of one unsigned full-word by another. If you do not define
3993 this macro, the default name is used, which is @code{__udivdi3}, a
3994 function defined in @file{libgcc.a}.
3996 @findex MODDI3_LIBCALL
3997 @item MODDI3_LIBCALL
3998 A C string constant giving the name of the function to call for the
3999 remainder in division of one signed double-word by another. If you do
4000 not define this macro, the default name is used, which is
4001 @code{__moddi3}, a function defined in @file{libgcc.a}.
4003 @findex UMODDI3_LIBCALL
4004 @item UMODDI3_LIBCALL
4005 A C string constant giving the name of the function to call for the
4006 remainder in division of one unsigned full-word by another. If you do
4007 not define this macro, the default name is used, which is
4008 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4010 @findex INIT_TARGET_OPTABS
4011 @item INIT_TARGET_OPTABS
4012 Define this macro as a C statement that declares additional library
4013 routines renames existing ones. @code{init_optabs} calls this macro after
4014 initializing all the normal library routines.
4016 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4017 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4018 Define this macro as a C statement that returns nonzero if a call to
4019 the floating point comparison library function will return a boolean
4020 value that indicates the result of the comparison. It should return
4021 zero if one of gcc's own libgcc functions is called.
4023 Most ports don't need to define this macro.
4026 @cindex @code{EDOM}, implicit usage
4028 The value of @code{EDOM} on the target machine, as a C integer constant
4029 expression. If you don't define this macro, GCC does not attempt to
4030 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4031 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4034 If you do not define @code{TARGET_EDOM}, then compiled code reports
4035 domain errors by calling the library function and letting it report the
4036 error. If mathematical functions on your system use @code{matherr} when
4037 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4038 that @code{matherr} is used normally.
4040 @findex GEN_ERRNO_RTX
4041 @cindex @code{errno}, implicit usage
4043 Define this macro as a C expression to create an rtl expression that
4044 refers to the global ``variable'' @code{errno}. (On certain systems,
4045 @code{errno} may not actually be a variable.) If you don't define this
4046 macro, a reasonable default is used.
4048 @findex TARGET_MEM_FUNCTIONS
4049 @cindex @code{bcopy}, implicit usage
4050 @cindex @code{memcpy}, implicit usage
4051 @cindex @code{bzero}, implicit usage
4052 @cindex @code{memset}, implicit usage
4053 @item TARGET_MEM_FUNCTIONS
4054 Define this macro if GCC should generate calls to the System V
4055 (and ANSI C) library functions @code{memcpy} and @code{memset}
4056 rather than the BSD functions @code{bcopy} and @code{bzero}.
4058 @findex LIBGCC_NEEDS_DOUBLE
4059 @item LIBGCC_NEEDS_DOUBLE
4060 Define this macro if only @code{float} arguments cannot be passed to
4061 library routines (so they must be converted to @code{double}). This
4062 macro affects both how library calls are generated and how the library
4063 routines in @file{libgcc1.c} accept their arguments. It is useful on
4064 machines where floating and fixed point arguments are passed
4065 differently, such as the i860.
4067 @findex FLOAT_ARG_TYPE
4068 @item FLOAT_ARG_TYPE
4069 Define this macro to override the type used by the library routines to
4070 pick up arguments of type @code{float}. (By default, they use a union
4071 of @code{float} and @code{int}.)
4073 The obvious choice would be @code{float}---but that won't work with
4074 traditional C compilers that expect all arguments declared as @code{float}
4075 to arrive as @code{double}. To avoid this conversion, the library routines
4076 ask for the value as some other type and then treat it as a @code{float}.
4078 On some systems, no other type will work for this. For these systems,
4079 you must use @code{LIBGCC_NEEDS_DOUBLE} instead, to force conversion of
4080 the values @code{double} before they are passed.
4083 @item FLOATIFY (@var{passed-value})
4084 Define this macro to override the way library routines redesignate a
4085 @code{float} argument as a @code{float} instead of the type it was
4086 passed as. The default is an expression which takes the @code{float}
4089 @findex FLOAT_VALUE_TYPE
4090 @item FLOAT_VALUE_TYPE
4091 Define this macro to override the type used by the library routines to
4092 return values that ought to have type @code{float}. (By default, they
4095 The obvious choice would be @code{float}---but that won't work with
4096 traditional C compilers gratuitously convert values declared as
4097 @code{float} into @code{double}.
4100 @item INTIFY (@var{float-value})
4101 Define this macro to override the way the value of a
4102 @code{float}-returning library routine should be packaged in order to
4103 return it. These functions are actually declared to return type
4104 @code{FLOAT_VALUE_TYPE} (normally @code{int}).
4106 These values can't be returned as type @code{float} because traditional
4107 C compilers would gratuitously convert the value to a @code{double}.
4109 A local variable named @code{intify} is always available when the macro
4110 @code{INTIFY} is used. It is a union of a @code{float} field named
4111 @code{f} and a field named @code{i} whose type is
4112 @code{FLOAT_VALUE_TYPE} or @code{int}.
4114 If you don't define this macro, the default definition works by copying
4115 the value through that union.
4117 @findex nongcc_SI_type
4118 @item nongcc_SI_type
4119 Define this macro as the name of the data type corresponding to
4120 @code{SImode} in the system's own C compiler.
4122 You need not define this macro if that type is @code{long int}, as it usually
4125 @findex nongcc_word_type
4126 @item nongcc_word_type
4127 Define this macro as the name of the data type corresponding to the
4128 word_mode in the system's own C compiler.
4130 You need not define this macro if that type is @code{long int}, as it usually
4133 @findex perform_@dots{}
4134 @item perform_@dots{}
4135 Define these macros to supply explicit C statements to carry out various
4136 arithmetic operations on types @code{float} and @code{double} in the
4137 library routines in @file{libgcc1.c}. See that file for a full list
4138 of these macros and their arguments.
4140 On most machines, you don't need to define any of these macros, because
4141 the C compiler that comes with the system takes care of doing them.
4143 @findex NEXT_OBJC_RUNTIME
4144 @item NEXT_OBJC_RUNTIME
4145 Define this macro to generate code for Objective C message sending using
4146 the calling convention of the NeXT system. This calling convention
4147 involves passing the object, the selector and the method arguments all
4148 at once to the method-lookup library function.
4150 The default calling convention passes just the object and the selector
4151 to the lookup function, which returns a pointer to the method.
4154 @node Addressing Modes
4155 @section Addressing Modes
4156 @cindex addressing modes
4158 @c prevent bad page break with this line
4159 This is about addressing modes.
4162 @findex HAVE_POST_INCREMENT
4163 @item HAVE_POST_INCREMENT
4164 A C expression that is nonzero the machine supports post-increment addressing.
4166 @findex HAVE_PRE_INCREMENT
4167 @findex HAVE_POST_DECREMENT
4168 @findex HAVE_PRE_DECREMENT
4169 @item HAVE_PRE_INCREMENT
4170 @itemx HAVE_POST_DECREMENT
4171 @itemx HAVE_PRE_DECREMENT
4172 Similar for other kinds of addressing.
4174 @findex CONSTANT_ADDRESS_P
4175 @item CONSTANT_ADDRESS_P (@var{x})
4176 A C expression that is 1 if the RTX @var{x} is a constant which
4177 is a valid address. On most machines, this can be defined as
4178 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4179 in which constant addresses are supported.
4182 @code{CONSTANT_P} accepts integer-values expressions whose values are
4183 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4184 @code{high} expressions and @code{const} arithmetic expressions, in
4185 addition to @code{const_int} and @code{const_double} expressions.
4187 @findex MAX_REGS_PER_ADDRESS
4188 @item MAX_REGS_PER_ADDRESS
4189 A number, the maximum number of registers that can appear in a valid
4190 memory address. Note that it is up to you to specify a value equal to
4191 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4194 @findex GO_IF_LEGITIMATE_ADDRESS
4195 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4196 A C compound statement with a conditional @code{goto @var{label};}
4197 executed if @var{x} (an RTX) is a legitimate memory address on the
4198 target machine for a memory operand of mode @var{mode}.
4200 It usually pays to define several simpler macros to serve as
4201 subroutines for this one. Otherwise it may be too complicated to
4204 This macro must exist in two variants: a strict variant and a
4205 non-strict one. The strict variant is used in the reload pass. It
4206 must be defined so that any pseudo-register that has not been
4207 allocated a hard register is considered a memory reference. In
4208 contexts where some kind of register is required, a pseudo-register
4209 with no hard register must be rejected.
4211 The non-strict variant is used in other passes. It must be defined to
4212 accept all pseudo-registers in every context where some kind of
4213 register is required.
4215 @findex REG_OK_STRICT
4216 Compiler source files that want to use the strict variant of this
4217 macro define the macro @code{REG_OK_STRICT}. You should use an
4218 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4219 in that case and the non-strict variant otherwise.
4221 Subroutines to check for acceptable registers for various purposes (one
4222 for base registers, one for index registers, and so on) are typically
4223 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4224 Then only these subroutine macros need have two variants; the higher
4225 levels of macros may be the same whether strict or not.@refill
4227 Normally, constant addresses which are the sum of a @code{symbol_ref}
4228 and an integer are stored inside a @code{const} RTX to mark them as
4229 constant. Therefore, there is no need to recognize such sums
4230 specifically as legitimate addresses. Normally you would simply
4231 recognize any @code{const} as legitimate.
4233 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4234 sums that are not marked with @code{const}. It assumes that a naked
4235 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4236 naked constant sums as illegitimate addresses, so that none of them will
4237 be given to @code{PRINT_OPERAND_ADDRESS}.
4239 @cindex @code{ENCODE_SECTION_INFO} and address validation
4240 On some machines, whether a symbolic address is legitimate depends on
4241 the section that the address refers to. On these machines, define the
4242 macro @code{ENCODE_SECTION_INFO} to store the information into the
4243 @code{symbol_ref}, and then check for it here. When you see a
4244 @code{const}, you will have to look inside it to find the
4245 @code{symbol_ref} in order to determine the section. @xref{Assembler
4248 @findex saveable_obstack
4249 The best way to modify the name string is by adding text to the
4250 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4251 the new name in @code{saveable_obstack}. You will have to modify
4252 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4253 output the name accordingly, and define @code{STRIP_NAME_ENCODING} to
4254 access the original name string.
4256 You can check the information stored here into the @code{symbol_ref} in
4257 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4258 @code{PRINT_OPERAND_ADDRESS}.
4260 @findex REG_OK_FOR_BASE_P
4261 @item REG_OK_FOR_BASE_P (@var{x})
4262 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4263 RTX) is valid for use as a base register. For hard registers, it
4264 should always accept those which the hardware permits and reject the
4265 others. Whether the macro accepts or rejects pseudo registers must be
4266 controlled by @code{REG_OK_STRICT} as described above. This usually
4267 requires two variant definitions, of which @code{REG_OK_STRICT}
4268 controls the one actually used.
4270 @findex REG_MODE_OK_FOR_BASE_P
4271 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4272 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4273 that expression may examine the mode of the memory reference in
4274 @var{mode}. You should define this macro if the mode of the memory
4275 reference affects whether a register may be used as a base register. If
4276 you define this macro, the compiler will use it instead of
4277 @code{REG_OK_FOR_BASE_P}.
4279 @findex REG_OK_FOR_INDEX_P
4280 @item REG_OK_FOR_INDEX_P (@var{x})
4281 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4282 RTX) is valid for use as an index register.
4284 The difference between an index register and a base register is that
4285 the index register may be scaled. If an address involves the sum of
4286 two registers, neither one of them scaled, then either one may be
4287 labeled the ``base'' and the other the ``index''; but whichever
4288 labeling is used must fit the machine's constraints of which registers
4289 may serve in each capacity. The compiler will try both labelings,
4290 looking for one that is valid, and will reload one or both registers
4291 only if neither labeling works.
4293 @findex LEGITIMIZE_ADDRESS
4294 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4295 A C compound statement that attempts to replace @var{x} with a valid
4296 memory address for an operand of mode @var{mode}. @var{win} will be a
4297 C statement label elsewhere in the code; the macro definition may use
4300 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4304 to avoid further processing if the address has become legitimate.
4306 @findex break_out_memory_refs
4307 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4308 and @var{oldx} will be the operand that was given to that function to produce
4311 The code generated by this macro should not alter the substructure of
4312 @var{x}. If it transforms @var{x} into a more legitimate form, it
4313 should assign @var{x} (which will always be a C variable) a new value.
4315 It is not necessary for this macro to come up with a legitimate
4316 address. The compiler has standard ways of doing so in all cases. In
4317 fact, it is safe for this macro to do nothing. But often a
4318 machine-dependent strategy can generate better code.
4320 @findex LEGITIMIZE_RELOAD_ADDRESS
4321 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4322 A C compound statement that attempts to replace @var{x}, which is an address
4323 that needs reloading, with a valid memory address for an operand of mode
4324 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4325 It is not necessary to define this macro, but it might be useful for
4326 performance reasons.
4328 For example, on the i386, it is sometimes possible to use a single
4329 reload register instead of two by reloading a sum of two pseudo
4330 registers into a register. On the other hand, for number of RISC
4331 processors offsets are limited so that often an intermediate address
4332 needs to be generated in order to address a stack slot. By defining
4333 LEGITIMIZE_RELOAD_ADDRESS appropriately, the intermediate addresses
4334 generated for adjacent some stack slots can be made identical, and thus
4337 @emph{Note}: This macro should be used with caution. It is necessary
4338 to know something of how reload works in order to effectively use this,
4339 and it is quite easy to produce macros that build in too much knowledge
4340 of reload internals.
4342 @emph{Note}: This macro must be able to reload an address created by a
4343 previous invocation of this macro. If it fails to handle such addresses
4344 then the compiler may generate incorrect code or abort.
4347 The macro definition should use @code{push_reload} to indicate parts that
4348 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4349 suitable to be passed unaltered to @code{push_reload}.
4351 The code generated by this macro must not alter the substructure of
4352 @var{x}. If it transforms @var{x} into a more legitimate form, it
4353 should assign @var{x} (which will always be a C variable) a new value.
4354 This also applies to parts that you change indirectly by calling
4357 @findex strict_memory_address_p
4358 The macro definition may use @code{strict_memory_address_p} to test if
4359 the address has become legitimate.
4362 If you want to change only a part of @var{x}, one standard way of doing
4363 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4364 single level of rtl. Thus, if the part to be changed is not at the
4365 top level, you'll need to replace first the top leve
4366 It is not necessary for this macro to come up with a legitimate
4367 address; but often a machine-dependent strategy can generate better code.
4369 @findex GO_IF_MODE_DEPENDENT_ADDRESS
4370 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4371 A C statement or compound statement with a conditional @code{goto
4372 @var{label};} executed if memory address @var{x} (an RTX) can have
4373 different meanings depending on the machine mode of the memory
4374 reference it is used for or if the address is valid for some modes
4377 Autoincrement and autodecrement addresses typically have mode-dependent
4378 effects because the amount of the increment or decrement is the size
4379 of the operand being addressed. Some machines have other mode-dependent
4380 addresses. Many RISC machines have no mode-dependent addresses.
4382 You may assume that @var{addr} is a valid address for the machine.
4384 @findex LEGITIMATE_CONSTANT_P
4385 @item LEGITIMATE_CONSTANT_P (@var{x})
4386 A C expression that is nonzero if @var{x} is a legitimate constant for
4387 an immediate operand on the target machine. You can assume that
4388 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4389 @samp{1} is a suitable definition for this macro on machines where
4390 anything @code{CONSTANT_P} is valid.@refill
4393 @node Condition Code
4394 @section Condition Code Status
4395 @cindex condition code status
4397 @c prevent bad page break with this line
4398 This describes the condition code status.
4401 The file @file{conditions.h} defines a variable @code{cc_status} to
4402 describe how the condition code was computed (in case the interpretation of
4403 the condition code depends on the instruction that it was set by). This
4404 variable contains the RTL expressions on which the condition code is
4405 currently based, and several standard flags.
4407 Sometimes additional machine-specific flags must be defined in the machine
4408 description header file. It can also add additional machine-specific
4409 information by defining @code{CC_STATUS_MDEP}.
4412 @findex CC_STATUS_MDEP
4413 @item CC_STATUS_MDEP
4414 C code for a data type which is used for declaring the @code{mdep}
4415 component of @code{cc_status}. It defaults to @code{int}.
4417 This macro is not used on machines that do not use @code{cc0}.
4419 @findex CC_STATUS_MDEP_INIT
4420 @item CC_STATUS_MDEP_INIT
4421 A C expression to initialize the @code{mdep} field to ``empty''.
4422 The default definition does nothing, since most machines don't use
4423 the field anyway. If you want to use the field, you should probably
4424 define this macro to initialize it.
4426 This macro is not used on machines that do not use @code{cc0}.
4428 @findex NOTICE_UPDATE_CC
4429 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4430 A C compound statement to set the components of @code{cc_status}
4431 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4432 this macro's responsibility to recognize insns that set the condition
4433 code as a byproduct of other activity as well as those that explicitly
4436 This macro is not used on machines that do not use @code{cc0}.
4438 If there are insns that do not set the condition code but do alter
4439 other machine registers, this macro must check to see whether they
4440 invalidate the expressions that the condition code is recorded as
4441 reflecting. For example, on the 68000, insns that store in address
4442 registers do not set the condition code, which means that usually
4443 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4444 insns. But suppose that the previous insn set the condition code
4445 based on location @samp{a4@@(102)} and the current insn stores a new
4446 value in @samp{a4}. Although the condition code is not changed by
4447 this, it will no longer be true that it reflects the contents of
4448 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4449 @code{cc_status} in this case to say that nothing is known about the
4450 condition code value.
4452 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4453 with the results of peephole optimization: insns whose patterns are
4454 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4455 constants which are just the operands. The RTL structure of these
4456 insns is not sufficient to indicate what the insns actually do. What
4457 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4458 @code{CC_STATUS_INIT}.
4460 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4461 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4462 @samp{cc}. This avoids having detailed information about patterns in
4463 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4465 @findex EXTRA_CC_MODES
4466 @item EXTRA_CC_MODES
4467 A list of additional modes for condition code values in registers
4468 (@pxref{Jump Patterns}). This macro should expand to a sequence of
4469 calls of the macro @code{CC} separated by white space. @code{CC} takes
4470 two arguments. The first is the enumeration name of the mode, which
4471 should begin with @samp{CC} and end with @samp{mode}. The second is a C
4472 string giving the printable name of the mode; it should be the same as
4473 the first argument, but with the trailing @samp{mode} removed.
4475 You should only define this macro if additional modes are required.
4477 A sample definition of @code{EXTRA_CC_MODES} is:
4479 #define EXTRA_CC_MODES \
4480 CC(CC_NOOVmode, "CC_NOOV") \
4481 CC(CCFPmode, "CCFP") \
4482 CC(CCFPEmode, "CCFPE")
4485 @findex SELECT_CC_MODE
4486 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4487 Returns a mode from class @code{MODE_CC} to be used when comparison
4488 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
4489 example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
4490 @pxref{Jump Patterns} for a description of the reason for this
4494 #define SELECT_CC_MODE(OP,X,Y) \
4495 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4496 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
4497 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4498 || GET_CODE (X) == NEG) \
4499 ? CC_NOOVmode : CCmode))
4502 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
4504 @findex CANONICALIZE_COMPARISON
4505 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
4506 One some machines not all possible comparisons are defined, but you can
4507 convert an invalid comparison into a valid one. For example, the Alpha
4508 does not have a @code{GT} comparison, but you can use an @code{LT}
4509 comparison instead and swap the order of the operands.
4511 On such machines, define this macro to be a C statement to do any
4512 required conversions. @var{code} is the initial comparison code
4513 and @var{op0} and @var{op1} are the left and right operands of the
4514 comparison, respectively. You should modify @var{code}, @var{op0}, and
4515 @var{op1} as required.
4517 GCC will not assume that the comparison resulting from this macro is
4518 valid but will see if the resulting insn matches a pattern in the
4521 You need not define this macro if it would never change the comparison
4524 @findex REVERSIBLE_CC_MODE
4525 @item REVERSIBLE_CC_MODE (@var{mode})
4526 A C expression whose value is one if it is always safe to reverse a
4527 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4528 can ever return @var{mode} for a floating-point inequality comparison,
4529 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4531 You need not define this macro if it would always returns zero or if the
4532 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4533 For example, here is the definition used on the Sparc, where floating-point
4534 inequality comparisons are always given @code{CCFPEmode}:
4537 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
4543 @section Describing Relative Costs of Operations
4544 @cindex costs of instructions
4545 @cindex relative costs
4546 @cindex speed of instructions
4548 These macros let you describe the relative speed of various operations
4549 on the target machine.
4553 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
4554 A part of a C @code{switch} statement that describes the relative costs
4555 of constant RTL expressions. It must contain @code{case} labels for
4556 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
4557 @code{label_ref} and @code{const_double}. Each case must ultimately
4558 reach a @code{return} statement to return the relative cost of the use
4559 of that kind of constant value in an expression. The cost may depend on
4560 the precise value of the constant, which is available for examination in
4561 @var{x}, and the rtx code of the expression in which it is contained,
4562 found in @var{outer_code}.
4564 @var{code} is the expression code---redundant, since it can be
4565 obtained with @code{GET_CODE (@var{x})}.
4568 @findex COSTS_N_INSNS
4569 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4570 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
4571 This can be used, for example, to indicate how costly a multiply
4572 instruction is. In writing this macro, you can use the construct
4573 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
4574 instructions. @var{outer_code} is the code of the expression in which
4575 @var{x} is contained.
4577 This macro is optional; do not define it if the default cost assumptions
4578 are adequate for the target machine.
4580 @findex DEFAULT_RTX_COSTS
4581 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4582 This macro, if defined, is called for any case not handled by the
4583 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
4584 to put case labels into the macro, but the code, or any functions it
4585 calls, must assume that the RTL in @var{x} could be of any type that has
4586 not already been handled. The arguments are the same as for
4587 @code{RTX_COSTS}, and the macro should execute a return statement giving
4588 the cost of any RTL expressions that it can handle. The default cost
4589 calculation is used for any RTL for which this macro does not return a
4592 This macro is optional; do not define it if the default cost assumptions
4593 are adequate for the target machine.
4595 @findex ADDRESS_COST
4596 @item ADDRESS_COST (@var{address})
4597 An expression giving the cost of an addressing mode that contains
4598 @var{address}. If not defined, the cost is computed from
4599 the @var{address} expression and the @code{CONST_COSTS} values.
4601 For most CISC machines, the default cost is a good approximation of the
4602 true cost of the addressing mode. However, on RISC machines, all
4603 instructions normally have the same length and execution time. Hence
4604 all addresses will have equal costs.
4606 In cases where more than one form of an address is known, the form with
4607 the lowest cost will be used. If multiple forms have the same, lowest,
4608 cost, the one that is the most complex will be used.
4610 For example, suppose an address that is equal to the sum of a register
4611 and a constant is used twice in the same basic block. When this macro
4612 is not defined, the address will be computed in a register and memory
4613 references will be indirect through that register. On machines where
4614 the cost of the addressing mode containing the sum is no higher than
4615 that of a simple indirect reference, this will produce an additional
4616 instruction and possibly require an additional register. Proper
4617 specification of this macro eliminates this overhead for such machines.
4619 Similar use of this macro is made in strength reduction of loops.
4621 @var{address} need not be valid as an address. In such a case, the cost
4622 is not relevant and can be any value; invalid addresses need not be
4623 assigned a different cost.
4625 On machines where an address involving more than one register is as
4626 cheap as an address computation involving only one register, defining
4627 @code{ADDRESS_COST} to reflect this can cause two registers to be live
4628 over a region of code where only one would have been if
4629 @code{ADDRESS_COST} were not defined in that manner. This effect should
4630 be considered in the definition of this macro. Equivalent costs should
4631 probably only be given to addresses with different numbers of registers
4632 on machines with lots of registers.
4634 This macro will normally either not be defined or be defined as a
4637 @findex REGISTER_MOVE_COST
4638 @item REGISTER_MOVE_COST (@var{from}, @var{to})
4639 A C expression for the cost of moving data from a register in class
4640 @var{from} to one in class @var{to}. The classes are expressed using
4641 the enumeration values such as @code{GENERAL_REGS}. A value of 2 is the
4642 default; other values are interpreted relative to that.
4644 It is not required that the cost always equal 2 when @var{from} is the
4645 same as @var{to}; on some machines it is expensive to move between
4646 registers if they are not general registers.
4648 If reload sees an insn consisting of a single @code{set} between two
4649 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4650 classes returns a value of 2, reload does not check to ensure that the
4651 constraints of the insn are met. Setting a cost of other than 2 will
4652 allow reload to verify that the constraints are met. You should do this
4653 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4655 @findex MEMORY_MOVE_COST
4656 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4657 A C expression for the cost of moving data of mode @var{mode} between a
4658 register of class @var{class} and memory; @var{in} is zero if the value
4659 is to be written to memory, non-zero if it is to be read in. This cost
4660 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
4661 registers and memory is more expensive than between two registers, you
4662 should define this macro to express the relative cost.
4664 If you do not define this macro, GCC uses a default cost of 4 plus
4665 the cost of copying via a secondary reload register, if one is
4666 needed. If your machine requires a secondary reload register to copy
4667 between memory and a register of @var{class} but the reload mechanism is
4668 more complex than copying via an intermediate, define this macro to
4669 reflect the actual cost of the move.
4671 GCC defines the function @code{memory_move_secondary_cost} if
4672 secondary reloads are needed. It computes the costs due to copying via
4673 a secondary register. If your machine copies from memory using a
4674 secondary register in the conventional way but the default base value of
4675 4 is not correct for your machine, define this macro to add some other
4676 value to the result of that function. The arguments to that function
4677 are the same as to this macro.
4681 A C expression for the cost of a branch instruction. A value of 1 is
4682 the default; other values are interpreted relative to that.
4685 Here are additional macros which do not specify precise relative costs,
4686 but only that certain actions are more expensive than GCC would
4690 @findex SLOW_BYTE_ACCESS
4691 @item SLOW_BYTE_ACCESS
4692 Define this macro as a C expression which is nonzero if accessing less
4693 than a word of memory (i.e. a @code{char} or a @code{short}) is no
4694 faster than accessing a word of memory, i.e., if such access
4695 require more than one instruction or if there is no difference in cost
4696 between byte and (aligned) word loads.
4698 When this macro is not defined, the compiler will access a field by
4699 finding the smallest containing object; when it is defined, a fullword
4700 load will be used if alignment permits. Unless bytes accesses are
4701 faster than word accesses, using word accesses is preferable since it
4702 may eliminate subsequent memory access if subsequent accesses occur to
4703 other fields in the same word of the structure, but to different bytes.
4705 @findex SLOW_ZERO_EXTEND
4706 @item SLOW_ZERO_EXTEND
4707 Define this macro if zero-extension (of a @code{char} or @code{short}
4708 to an @code{int}) can be done faster if the destination is a register
4709 that is known to be zero.
4711 If you define this macro, you must have instruction patterns that
4712 recognize RTL structures like this:
4715 (set (strict_low_part (subreg:QI (reg:SI @dots{}) 0)) @dots{})
4719 and likewise for @code{HImode}.
4721 @findex SLOW_UNALIGNED_ACCESS
4722 @item SLOW_UNALIGNED_ACCESS
4723 Define this macro to be the value 1 if unaligned accesses have a cost
4724 many times greater than aligned accesses, for example if they are
4725 emulated in a trap handler.
4727 When this macro is non-zero, the compiler will act as if
4728 @code{STRICT_ALIGNMENT} were non-zero when generating code for block
4729 moves. This can cause significantly more instructions to be produced.
4730 Therefore, do not set this macro non-zero if unaligned accesses only add a
4731 cycle or two to the time for a memory access.
4733 If the value of this macro is always zero, it need not be defined.
4735 @findex DONT_REDUCE_ADDR
4736 @item DONT_REDUCE_ADDR
4737 Define this macro to inhibit strength reduction of memory addresses.
4738 (On some machines, such strength reduction seems to do harm rather
4743 The threshold of number of scalar memory-to-memory move insns, @emph{below}
4744 which a sequence of insns should be generated instead of a
4745 string move insn or a library call. Increasing the value will always
4746 make code faster, but eventually incurs high cost in increased code size.
4748 Note that on machines where the corresponding move insn is a
4749 @code{define_expand} that emits a sequence of insns, this macro counts
4750 the number of such sequences.
4752 If you don't define this, a reasonable default is used.
4754 @findex MOVE_BY_PIECES_P
4755 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
4756 A C expression used to determine whether @code{move_by_pieces} will be used to
4757 copy a chunk of memory, or whether some other block move mechanism
4758 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
4759 than @code{MOVE_RATIO}.
4761 @findex MOVE_MAX_PIECES
4762 @item MOVE_MAX_PIECES
4763 A C expression used by @code{move_by_pieces} to determine the largest unit
4764 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
4766 @findex USE_LOAD_POST_INCREMENT
4767 @item USE_LOAD_POST_INCREMENT (@var{mode})
4768 A C expression used to determine whether a load postincrement is a good
4769 thing to use for a given mode. Defaults to the value of
4770 @code{HAVE_POST_INCREMENT}.
4772 @findex USE_LOAD_POST_DECREMENT
4773 @item USE_LOAD_POST_DECREMENT (@var{mode})
4774 A C expression used to determine whether a load postdecrement is a good
4775 thing to use for a given mode. Defaults to the value of
4776 @code{HAVE_POST_DECREMENT}.
4778 @findex USE_LOAD_PRE_INCREMENT
4779 @item USE_LOAD_PRE_INCREMENT (@var{mode})
4780 A C expression used to determine whether a load preincrement is a good
4781 thing to use for a given mode. Defaults to the value of
4782 @code{HAVE_PRE_INCREMENT}.
4784 @findex USE_LOAD_PRE_DECREMENT
4785 @item USE_LOAD_PRE_DECREMENT (@var{mode})
4786 A C expression used to determine whether a load predecrement is a good
4787 thing to use for a given mode. Defaults to the value of
4788 @code{HAVE_PRE_DECREMENT}.
4790 @findex USE_STORE_POST_INCREMENT
4791 @item USE_STORE_POST_INCREMENT (@var{mode})
4792 A C expression used to determine whether a store postincrement is a good
4793 thing to use for a given mode. Defaults to the value of
4794 @code{HAVE_POST_INCREMENT}.
4796 @findex USE_STORE_POST_DECREMENT
4797 @item USE_STORE_POST_DECREMENT (@var{mode})
4798 A C expression used to determine whether a store postdeccrement is a good
4799 thing to use for a given mode. Defaults to the value of
4800 @code{HAVE_POST_DECREMENT}.
4802 @findex USE_STORE_PRE_INCREMENT
4803 @item USE_STORE_PRE_INCREMENT (@var{mode})
4804 This macro is used to determine whether a store preincrement is a good
4805 thing to use for a given mode. Defaults to the value of
4806 @code{HAVE_PRE_INCREMENT}.
4808 @findex USE_STORE_PRE_DECREMENT
4809 @item USE_STORE_PRE_DECREMENT (@var{mode})
4810 This macro is used to determine whether a store predecrement is a good
4811 thing to use for a given mode. Defaults to the value of
4812 @code{HAVE_PRE_DECREMENT}.
4814 @findex NO_FUNCTION_CSE
4815 @item NO_FUNCTION_CSE
4816 Define this macro if it is as good or better to call a constant
4817 function address than to call an address kept in a register.
4819 @findex NO_RECURSIVE_FUNCTION_CSE
4820 @item NO_RECURSIVE_FUNCTION_CSE
4821 Define this macro if it is as good or better for a function to call
4822 itself with an explicit address than to call an address kept in a
4826 @item ADJUST_COST (@var{insn}, @var{link}, @var{dep_insn}, @var{cost})
4827 A C statement (sans semicolon) to update the integer variable @var{cost}
4828 based on the relationship between @var{insn} that is dependent on
4829 @var{dep_insn} through the dependence @var{link}. The default is to
4830 make no adjustment to @var{cost}. This can be used for example to
4831 specify to the scheduler that an output- or anti-dependence does not
4832 incur the same cost as a data-dependence.
4834 @findex ADJUST_PRIORITY
4835 @item ADJUST_PRIORITY (@var{insn})
4836 A C statement (sans semicolon) to update the integer scheduling
4837 priority @code{INSN_PRIORITY(@var{insn})}. Reduce the priority
4838 to execute the @var{insn} earlier, increase the priority to execute
4839 @var{insn} later. Do not define this macro if you do not need to
4840 adjust the scheduling priorities of insns.
4844 @section Dividing the Output into Sections (Texts, Data, @dots{})
4845 @c the above section title is WAY too long. maybe cut the part between
4846 @c the (...)? --mew 10feb93
4848 An object file is divided into sections containing different types of
4849 data. In the most common case, there are three sections: the @dfn{text
4850 section}, which holds instructions and read-only data; the @dfn{data
4851 section}, which holds initialized writable data; and the @dfn{bss
4852 section}, which holds uninitialized data. Some systems have other kinds
4855 The compiler must tell the assembler when to switch sections. These
4856 macros control what commands to output to tell the assembler this. You
4857 can also define additional sections.
4860 @findex TEXT_SECTION_ASM_OP
4861 @item TEXT_SECTION_ASM_OP
4862 A C expression whose value is a string containing the assembler
4863 operation that should precede instructions and read-only data. Normally
4864 @code{".text"} is right.
4866 @findex DATA_SECTION_ASM_OP
4867 @item DATA_SECTION_ASM_OP
4868 A C expression whose value is a string containing the assembler
4869 operation to identify the following data as writable initialized data.
4870 Normally @code{".data"} is right.
4872 @findex SHARED_SECTION_ASM_OP
4873 @item SHARED_SECTION_ASM_OP
4874 If defined, a C expression whose value is a string containing the
4875 assembler operation to identify the following data as shared data. If
4876 not defined, @code{DATA_SECTION_ASM_OP} will be used.
4878 @findex BSS_SECTION_ASM_OP
4879 @item BSS_SECTION_ASM_OP
4880 If defined, a C expression whose value is a string containing the
4881 assembler operation to identify the following data as uninitialized global
4882 data. If not defined, and neither @code{ASM_OUTPUT_BSS} nor
4883 @code{ASM_OUTPUT_ALIGNED_BSS} are defined, uninitialized global data will be
4884 output in the data section if @samp{-fno-common} is passed, otherwise
4885 @code{ASM_OUTPUT_COMMON} will be used.
4887 @findex SHARED_BSS_SECTION_ASM_OP
4888 @item SHARED_BSS_SECTION_ASM_OP
4889 If defined, a C expression whose value is a string containing the
4890 assembler operation to identify the following data as uninitialized global
4891 shared data. If not defined, and @code{BSS_SECTION_ASM_OP} is, the latter
4894 @findex INIT_SECTION_ASM_OP
4895 @item INIT_SECTION_ASM_OP
4896 If defined, a C expression whose value is a string containing the
4897 assembler operation to identify the following data as initialization
4898 code. If not defined, GCC will assume such a section does not
4901 @findex EXTRA_SECTIONS
4904 @item EXTRA_SECTIONS
4905 A list of names for sections other than the standard two, which are
4906 @code{in_text} and @code{in_data}. You need not define this macro
4907 on a system with no other sections (that GCC needs to use).
4909 @findex EXTRA_SECTION_FUNCTIONS
4910 @findex text_section
4911 @findex data_section
4912 @item EXTRA_SECTION_FUNCTIONS
4913 One or more functions to be defined in @file{varasm.c}. These
4914 functions should do jobs analogous to those of @code{text_section} and
4915 @code{data_section}, for your additional sections. Do not define this
4916 macro if you do not define @code{EXTRA_SECTIONS}.
4918 @findex READONLY_DATA_SECTION
4919 @item READONLY_DATA_SECTION
4920 On most machines, read-only variables, constants, and jump tables are
4921 placed in the text section. If this is not the case on your machine,
4922 this macro should be defined to be the name of a function (either
4923 @code{data_section} or a function defined in @code{EXTRA_SECTIONS}) that
4924 switches to the section to be used for read-only items.
4926 If these items should be placed in the text section, this macro should
4929 @findex SELECT_SECTION
4930 @item SELECT_SECTION (@var{exp}, @var{reloc})
4931 A C statement or statements to switch to the appropriate section for
4932 output of @var{exp}. You can assume that @var{exp} is either a
4933 @code{VAR_DECL} node or a constant of some sort. @var{reloc}
4934 indicates whether the initial value of @var{exp} requires link-time
4935 relocations. Select the section by calling @code{text_section} or one
4936 of the alternatives for other sections.
4938 Do not define this macro if you put all read-only variables and
4939 constants in the read-only data section (usually the text section).
4941 @findex SELECT_RTX_SECTION
4942 @item SELECT_RTX_SECTION (@var{mode}, @var{rtx})
4943 A C statement or statements to switch to the appropriate section for
4944 output of @var{rtx} in mode @var{mode}. You can assume that @var{rtx}
4945 is some kind of constant in RTL. The argument @var{mode} is redundant
4946 except in the case of a @code{const_int} rtx. Select the section by
4947 calling @code{text_section} or one of the alternatives for other
4950 Do not define this macro if you put all constants in the read-only
4953 @findex JUMP_TABLES_IN_TEXT_SECTION
4954 @item JUMP_TABLES_IN_TEXT_SECTION
4955 Define this macro to be an expression with a non-zero value if jump
4956 tables (for @code{tablejump} insns) should be output in the text
4957 section, along with the assembler instructions. Otherwise, the
4958 readonly data section is used.
4960 This macro is irrelevant if there is no separate readonly data section.
4962 @findex ENCODE_SECTION_INFO
4963 @item ENCODE_SECTION_INFO (@var{decl})
4964 Define this macro if references to a symbol must be treated differently
4965 depending on something about the variable or function named by the
4966 symbol (such as what section it is in).
4968 The macro definition, if any, is executed immediately after the rtl for
4969 @var{decl} has been created and stored in @code{DECL_RTL (@var{decl})}.
4970 The value of the rtl will be a @code{mem} whose address is a
4973 @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO}
4974 The usual thing for this macro to do is to record a flag in the
4975 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
4976 modified name string in the @code{symbol_ref} (if one bit is not enough
4979 @findex STRIP_NAME_ENCODING
4980 @item STRIP_NAME_ENCODING (@var{var}, @var{sym_name})
4981 Decode @var{sym_name} and store the real name part in @var{var}, sans
4982 the characters that encode section info. Define this macro if
4983 @code{ENCODE_SECTION_INFO} alters the symbol's name string.
4985 @findex UNIQUE_SECTION_P
4986 @item UNIQUE_SECTION_P (@var{decl})
4987 A C expression which evaluates to true if @var{decl} should be placed
4988 into a unique section for some target-specific reason. If you do not
4989 define this macro, the default is @samp{0}. Note that the flag
4990 @samp{-ffunction-sections} will also cause functions to be placed into
4993 @findex UNIQUE_SECTION
4994 @item UNIQUE_SECTION (@var{decl}, @var{reloc})
4995 A C statement to build up a unique section name, expressed as a
4996 STRING_CST node, and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
4997 @var{reloc} indicates whether the initial value of @var{exp} requires
4998 link-time relocations. If you do not define this macro, GCC will use
4999 the symbol name prefixed by @samp{.} as the section name.
5003 @section Position Independent Code
5004 @cindex position independent code
5007 This section describes macros that help implement generation of position
5008 independent code. Simply defining these macros is not enough to
5009 generate valid PIC; you must also add support to the macros
5010 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5011 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5012 @samp{movsi} to do something appropriate when the source operand
5013 contains a symbolic address. You may also need to alter the handling of
5014 switch statements so that they use relative addresses.
5015 @c i rearranged the order of the macros above to try to force one of
5016 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5019 @findex PIC_OFFSET_TABLE_REGNUM
5020 @item PIC_OFFSET_TABLE_REGNUM
5021 The register number of the register used to address a table of static
5022 data addresses in memory. In some cases this register is defined by a
5023 processor's ``application binary interface'' (ABI). When this macro
5024 is defined, RTL is generated for this register once, as with the stack
5025 pointer and frame pointer registers. If this macro is not defined, it
5026 is up to the machine-dependent files to allocate such a register (if
5029 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5030 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5031 Define this macro if the register defined by
5032 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
5033 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5035 @findex FINALIZE_PIC
5037 By generating position-independent code, when two different programs (A
5038 and B) share a common library (libC.a), the text of the library can be
5039 shared whether or not the library is linked at the same address for both
5040 programs. In some of these environments, position-independent code
5041 requires not only the use of different addressing modes, but also
5042 special code to enable the use of these addressing modes.
5044 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
5045 codes once the function is being compiled into assembly code, but not
5046 before. (It is not done before, because in the case of compiling an
5047 inline function, it would lead to multiple PIC prologues being
5048 included in functions which used inline functions and were compiled to
5051 @findex LEGITIMATE_PIC_OPERAND_P
5052 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
5053 A C expression that is nonzero if @var{x} is a legitimate immediate
5054 operand on the target machine when generating position independent code.
5055 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5056 check this. You can also assume @var{flag_pic} is true, so you need not
5057 check it either. You need not define this macro if all constants
5058 (including @code{SYMBOL_REF}) can be immediate operands when generating
5059 position independent code.
5062 @node Assembler Format
5063 @section Defining the Output Assembler Language
5065 This section describes macros whose principal purpose is to describe how
5066 to write instructions in assembler language--rather than what the
5070 * File Framework:: Structural information for the assembler file.
5071 * Data Output:: Output of constants (numbers, strings, addresses).
5072 * Uninitialized Data:: Output of uninitialized variables.
5073 * Label Output:: Output and generation of labels.
5074 * Initialization:: General principles of initialization
5075 and termination routines.
5076 * Macros for Initialization::
5077 Specific macros that control the handling of
5078 initialization and termination routines.
5079 * Instruction Output:: Output of actual instructions.
5080 * Dispatch Tables:: Output of jump tables.
5081 * Exception Region Output:: Output of exception region code.
5082 * Alignment Output:: Pseudo ops for alignment and skipping data.
5085 @node File Framework
5086 @subsection The Overall Framework of an Assembler File
5087 @cindex assembler format
5088 @cindex output of assembler code
5090 @c prevent bad page break with this line
5091 This describes the overall framework of an assembler file.
5094 @findex ASM_FILE_START
5095 @item ASM_FILE_START (@var{stream})
5096 A C expression which outputs to the stdio stream @var{stream}
5097 some appropriate text to go at the start of an assembler file.
5099 Normally this macro is defined to output a line containing
5100 @samp{#NO_APP}, which is a comment that has no effect on most
5101 assemblers but tells the GNU assembler that it can save time by not
5102 checking for certain assembler constructs.
5104 On systems that use SDB, it is necessary to output certain commands;
5105 see @file{attasm.h}.
5107 @findex ASM_FILE_END
5108 @item ASM_FILE_END (@var{stream})
5109 A C expression which outputs to the stdio stream @var{stream}
5110 some appropriate text to go at the end of an assembler file.
5112 If this macro is not defined, the default is to output nothing
5113 special at the end of the file. Most systems don't require any
5116 On systems that use SDB, it is necessary to output certain commands;
5117 see @file{attasm.h}.
5119 @findex ASM_IDENTIFY_GCC
5120 @item ASM_IDENTIFY_GCC (@var{file})
5121 A C statement to output assembler commands which will identify
5122 the object file as having been compiled with GCC (or another
5125 If you don't define this macro, the string @samp{gcc_compiled.:}
5126 is output. This string is calculated to define a symbol which,
5127 on BSD systems, will never be defined for any other reason.
5128 GDB checks for the presence of this symbol when reading the
5129 symbol table of an executable.
5131 On non-BSD systems, you must arrange communication with GDB in
5132 some other fashion. If GDB is not used on your system, you can
5133 define this macro with an empty body.
5135 @findex ASM_COMMENT_START
5136 @item ASM_COMMENT_START
5137 A C string constant describing how to begin a comment in the target
5138 assembler language. The compiler assumes that the comment will end at
5139 the end of the line.
5143 A C string constant for text to be output before each @code{asm}
5144 statement or group of consecutive ones. Normally this is
5145 @code{"#APP"}, which is a comment that has no effect on most
5146 assemblers but tells the GNU assembler that it must check the lines
5147 that follow for all valid assembler constructs.
5151 A C string constant for text to be output after each @code{asm}
5152 statement or group of consecutive ones. Normally this is
5153 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5154 time-saving assumptions that are valid for ordinary compiler output.
5156 @findex ASM_OUTPUT_SOURCE_FILENAME
5157 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5158 A C statement to output COFF information or DWARF debugging information
5159 which indicates that filename @var{name} is the current source file to
5160 the stdio stream @var{stream}.
5162 This macro need not be defined if the standard form of output
5163 for the file format in use is appropriate.
5165 @findex OUTPUT_QUOTED_STRING
5166 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{name})
5167 A C statement to output the string @var{string} to the stdio stream
5168 @var{stream}. If you do not call the function @code{output_quoted_string}
5169 in your config files, GCC will only call it to output filenames to
5170 the assembler source. So you can use it to canonicalize the format
5171 of the filename using this macro.
5173 @findex ASM_OUTPUT_SOURCE_LINE
5174 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
5175 A C statement to output DBX or SDB debugging information before code
5176 for line number @var{line} of the current source file to the
5177 stdio stream @var{stream}.
5179 This macro need not be defined if the standard form of debugging
5180 information for the debugger in use is appropriate.
5182 @findex ASM_OUTPUT_IDENT
5183 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
5184 A C statement to output something to the assembler file to handle a
5185 @samp{#ident} directive containing the text @var{string}. If this
5186 macro is not defined, nothing is output for a @samp{#ident} directive.
5188 @findex ASM_OUTPUT_SECTION_NAME
5189 @item ASM_OUTPUT_SECTION_NAME (@var{stream}, @var{decl}, @var{name}, @var{reloc})
5190 A C statement to output something to the assembler file to switch to section
5191 @var{name} for object @var{decl} which is either a @code{FUNCTION_DECL}, a
5192 @code{VAR_DECL} or @code{NULL_TREE}. @var{reloc}
5193 indicates whether the initial value of @var{exp} requires link-time
5194 relocations. Some target formats do not support
5195 arbitrary sections. Do not define this macro in such cases.
5197 At present this macro is only used to support section attributes.
5198 When this macro is undefined, section attributes are disabled.
5200 @findex OBJC_PROLOGUE
5202 A C statement to output any assembler statements which are required to
5203 precede any Objective C object definitions or message sending. The
5204 statement is executed only when compiling an Objective C program.
5209 @subsection Output of Data
5211 @c prevent bad page break with this line
5212 This describes data output.
5215 @findex ASM_OUTPUT_LONG_DOUBLE
5216 @findex ASM_OUTPUT_DOUBLE
5217 @findex ASM_OUTPUT_FLOAT
5218 @item ASM_OUTPUT_LONG_DOUBLE (@var{stream}, @var{value})
5219 @itemx ASM_OUTPUT_DOUBLE (@var{stream}, @var{value})
5220 @itemx ASM_OUTPUT_FLOAT (@var{stream}, @var{value})
5221 @itemx ASM_OUTPUT_THREE_QUARTER_FLOAT (@var{stream}, @var{value})
5222 @itemx ASM_OUTPUT_SHORT_FLOAT (@var{stream}, @var{value})
5223 @itemx ASM_OUTPUT_BYTE_FLOAT (@var{stream}, @var{value})
5224 A C statement to output to the stdio stream @var{stream} an assembler
5225 instruction to assemble a floating-point constant of @code{TFmode},
5226 @code{DFmode}, @code{SFmode}, @code{TQFmode}, @code{HFmode}, or
5227 @code{QFmode}, respectively, whose value is @var{value}. @var{value}
5228 will be a C expression of type @code{REAL_VALUE_TYPE}. Macros such as
5229 @code{REAL_VALUE_TO_TARGET_DOUBLE} are useful for writing these
5232 @findex ASM_OUTPUT_QUADRUPLE_INT
5233 @findex ASM_OUTPUT_DOUBLE_INT
5234 @findex ASM_OUTPUT_INT
5235 @findex ASM_OUTPUT_SHORT
5236 @findex ASM_OUTPUT_CHAR
5237 @findex output_addr_const
5238 @item ASM_OUTPUT_QUADRUPLE_INT (@var{stream}, @var{exp})
5239 @itemx ASM_OUTPUT_DOUBLE_INT (@var{stream}, @var{exp})
5240 @itemx ASM_OUTPUT_INT (@var{stream}, @var{exp})
5241 @itemx ASM_OUTPUT_SHORT (@var{stream}, @var{exp})
5242 @itemx ASM_OUTPUT_CHAR (@var{stream}, @var{exp})
5243 A C statement to output to the stdio stream @var{stream} an assembler
5244 instruction to assemble an integer of 16, 8, 4, 2 or 1 bytes,
5245 respectively, whose value is @var{value}. The argument @var{exp} will
5246 be an RTL expression which represents a constant value. Use
5247 @samp{output_addr_const (@var{stream}, @var{exp})} to output this value
5248 as an assembler expression.@refill
5250 For sizes larger than @code{UNITS_PER_WORD}, if the action of a macro
5251 would be identical to repeatedly calling the macro corresponding to
5252 a size of @code{UNITS_PER_WORD}, once for each word, you need not define
5255 @findex ASM_OUTPUT_BYTE
5256 @item ASM_OUTPUT_BYTE (@var{stream}, @var{value})
5257 A C statement to output to the stdio stream @var{stream} an assembler
5258 instruction to assemble a single byte containing the number @var{value}.
5262 A C string constant giving the pseudo-op to use for a sequence of
5263 single-byte constants. If this macro is not defined, the default is
5266 @findex ASM_OUTPUT_ASCII
5267 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5268 A C statement to output to the stdio stream @var{stream} an assembler
5269 instruction to assemble a string constant containing the @var{len}
5270 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5271 @code{char *} and @var{len} a C expression of type @code{int}.
5273 If the assembler has a @code{.ascii} pseudo-op as found in the
5274 Berkeley Unix assembler, do not define the macro
5275 @code{ASM_OUTPUT_ASCII}.
5277 @findex CONSTANT_POOL_BEFORE_FUNCTION
5278 @item CONSTANT_POOL_BEFORE_FUNCTION
5279 You may define this macro as a C expression. You should define the
5280 expression to have a non-zero value if GCC should output the constant
5281 pool for a function before the code for the function, or a zero value if
5282 GCC should output the constant pool after the function. If you do
5283 not define this macro, the usual case, GCC will output the constant
5284 pool before the function.
5286 @findex ASM_OUTPUT_POOL_PROLOGUE
5287 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5288 A C statement to output assembler commands to define the start of the
5289 constant pool for a function. @var{funname} is a string giving
5290 the name of the function. Should the return type of the function
5291 be required, it can be obtained via @var{fundecl}. @var{size}
5292 is the size, in bytes, of the constant pool that will be written
5293 immediately after this call.
5295 If no constant-pool prefix is required, the usual case, this macro need
5298 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
5299 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5300 A C statement (with or without semicolon) to output a constant in the
5301 constant pool, if it needs special treatment. (This macro need not do
5302 anything for RTL expressions that can be output normally.)
5304 The argument @var{file} is the standard I/O stream to output the
5305 assembler code on. @var{x} is the RTL expression for the constant to
5306 output, and @var{mode} is the machine mode (in case @var{x} is a
5307 @samp{const_int}). @var{align} is the required alignment for the value
5308 @var{x}; you should output an assembler directive to force this much
5311 The argument @var{labelno} is a number to use in an internal label for
5312 the address of this pool entry. The definition of this macro is
5313 responsible for outputting the label definition at the proper place.
5314 Here is how to do this:
5317 ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
5320 When you output a pool entry specially, you should end with a
5321 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5322 entry from being output a second time in the usual manner.
5324 You need not define this macro if it would do nothing.
5326 @findex CONSTANT_AFTER_FUNCTION_P
5327 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
5328 Define this macro as a C expression which is nonzero if the constant
5329 @var{exp}, of type @code{tree}, should be output after the code for a
5330 function. The compiler will normally output all constants before the
5331 function; you need not define this macro if this is OK.
5333 @findex ASM_OUTPUT_POOL_EPILOGUE
5334 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5335 A C statement to output assembler commands to at the end of the constant
5336 pool for a function. @var{funname} is a string giving the name of the
5337 function. Should the return type of the function be required, you can
5338 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5339 constant pool that GCC wrote immediately before this call.
5341 If no constant-pool epilogue is required, the usual case, you need not
5344 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
5345 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
5346 Define this macro as a C expression which is nonzero if @var{C} is
5347 used as a logical line separator by the assembler.
5349 If you do not define this macro, the default is that only
5350 the character @samp{;} is treated as a logical line separator.
5353 @findex ASM_OPEN_PAREN
5354 @findex ASM_CLOSE_PAREN
5355 @item ASM_OPEN_PAREN
5356 @itemx ASM_CLOSE_PAREN
5357 These macros are defined as C string constant, describing the syntax
5358 in the assembler for grouping arithmetic expressions. The following
5359 definitions are correct for most assemblers:
5362 #define ASM_OPEN_PAREN "("
5363 #define ASM_CLOSE_PAREN ")"
5367 These macros are provided by @file{real.h} for writing the definitions
5368 of @code{ASM_OUTPUT_DOUBLE} and the like:
5371 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5372 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5373 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5374 @findex REAL_VALUE_TO_TARGET_SINGLE
5375 @findex REAL_VALUE_TO_TARGET_DOUBLE
5376 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
5377 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
5378 floating point representation, and store its bit pattern in the array of
5379 @code{long int} whose address is @var{l}. The number of elements in the
5380 output array is determined by the size of the desired target floating
5381 point data type: 32 bits of it go in each @code{long int} array
5382 element. Each array element holds 32 bits of the result, even if
5383 @code{long int} is wider than 32 bits on the host machine.
5385 The array element values are designed so that you can print them out
5386 using @code{fprintf} in the order they should appear in the target
5389 @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
5390 @findex REAL_VALUE_TO_DECIMAL
5391 This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
5392 decimal number and stores it as a string into @var{string}.
5393 You must pass, as @var{string}, the address of a long enough block
5394 of space to hold the result.
5396 The argument @var{format} is a @code{printf}-specification that serves
5397 as a suggestion for how to format the output string.
5400 @node Uninitialized Data
5401 @subsection Output of Uninitialized Variables
5403 Each of the macros in this section is used to do the whole job of
5404 outputting a single uninitialized variable.
5407 @findex ASM_OUTPUT_COMMON
5408 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5409 A C statement (sans semicolon) to output to the stdio stream
5410 @var{stream} the assembler definition of a common-label named
5411 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5412 is the size rounded up to whatever alignment the caller wants.
5414 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5415 output the name itself; before and after that, output the additional
5416 assembler syntax for defining the name, and a newline.
5418 This macro controls how the assembler definitions of uninitialized
5419 common global variables are output.
5421 @findex ASM_OUTPUT_ALIGNED_COMMON
5422 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5423 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5424 separate, explicit argument. If you define this macro, it is used in
5425 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5426 handling the required alignment of the variable. The alignment is specified
5427 as the number of bits.
5429 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
5430 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5431 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5432 variable to be output, if there is one, or @code{NULL_TREE} if there
5433 is not corresponding variable. If you define this macro, GCC wil use it
5434 in place of both @code{ASM_OUTPUT_COMMON} and
5435 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5436 the variable's decl in order to chose what to output.
5438 @findex ASM_OUTPUT_SHARED_COMMON
5439 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5440 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
5441 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
5444 @findex ASM_OUTPUT_BSS
5445 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5446 A C statement (sans semicolon) to output to the stdio stream
5447 @var{stream} the assembler definition of uninitialized global @var{decl} named
5448 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5449 is the size rounded up to whatever alignment the caller wants.
5451 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
5452 defining this macro. If unable, use the expression
5453 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5454 before and after that, output the additional assembler syntax for defining
5455 the name, and a newline.
5457 This macro controls how the assembler definitions of uninitialized global
5458 variables are output. This macro exists to properly support languages like
5459 @code{c++} which do not have @code{common} data. However, this macro currently
5460 is not defined for all targets. If this macro and
5461 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
5462 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
5463 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
5465 @findex ASM_OUTPUT_ALIGNED_BSS
5466 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5467 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
5468 separate, explicit argument. If you define this macro, it is used in
5469 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
5470 handling the required alignment of the variable. The alignment is specified
5471 as the number of bits.
5473 Try to use function @code{asm_output_aligned_bss} defined in file
5474 @file{varasm.c} when defining this macro.
5476 @findex ASM_OUTPUT_SHARED_BSS
5477 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5478 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
5479 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
5482 @findex ASM_OUTPUT_LOCAL
5483 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5484 A C statement (sans semicolon) to output to the stdio stream
5485 @var{stream} the assembler definition of a local-common-label named
5486 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5487 is the size rounded up to whatever alignment the caller wants.
5489 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5490 output the name itself; before and after that, output the additional
5491 assembler syntax for defining the name, and a newline.
5493 This macro controls how the assembler definitions of uninitialized
5494 static variables are output.
5496 @findex ASM_OUTPUT_ALIGNED_LOCAL
5497 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5498 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5499 separate, explicit argument. If you define this macro, it is used in
5500 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5501 handling the required alignment of the variable. The alignment is specified
5502 as the number of bits.
5504 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
5505 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5506 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5507 variable to be output, if there is one, or @code{NULL_TREE} if there
5508 is not corresponding variable. If you define this macro, GCC wil use it
5509 in place of both @code{ASM_OUTPUT_DECL} and
5510 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5511 the variable's decl in order to chose what to output.
5514 @findex ASM_OUTPUT_SHARED_LOCAL
5515 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5516 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
5517 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
5522 @subsection Output and Generation of Labels
5524 @c prevent bad page break with this line
5525 This is about outputting labels.
5528 @findex ASM_OUTPUT_LABEL
5529 @findex assemble_name
5530 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5531 A C statement (sans semicolon) to output to the stdio stream
5532 @var{stream} the assembler definition of a label named @var{name}.
5533 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5534 output the name itself; before and after that, output the additional
5535 assembler syntax for defining the name, and a newline.
5537 @findex ASM_DECLARE_FUNCTION_NAME
5538 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5539 A C statement (sans semicolon) to output to the stdio stream
5540 @var{stream} any text necessary for declaring the name @var{name} of a
5541 function which is being defined. This macro is responsible for
5542 outputting the label definition (perhaps using
5543 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
5544 @code{FUNCTION_DECL} tree node representing the function.
5546 If this macro is not defined, then the function name is defined in the
5547 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5549 @findex ASM_DECLARE_FUNCTION_SIZE
5550 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5551 A C statement (sans semicolon) to output to the stdio stream
5552 @var{stream} any text necessary for declaring the size of a function
5553 which is being defined. The argument @var{name} is the name of the
5554 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5555 representing the function.
5557 If this macro is not defined, then the function size is not defined.
5559 @findex ASM_DECLARE_OBJECT_NAME
5560 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5561 A C statement (sans semicolon) to output to the stdio stream
5562 @var{stream} any text necessary for declaring the name @var{name} of an
5563 initialized variable which is being defined. This macro must output the
5564 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
5565 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
5567 If this macro is not defined, then the variable name is defined in the
5568 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5570 @findex ASM_DECLARE_REGISTER_GLOBAL
5571 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
5572 A C statement (sans semicolon) to output to the stdio stream
5573 @var{stream} any text necessary for claiming a register @var{regno}
5574 for a global variable @var{decl} with name @var{name}.
5576 If you don't define this macro, that is equivalent to defining it to do
5579 @findex ASM_FINISH_DECLARE_OBJECT
5580 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5581 A C statement (sans semicolon) to finish up declaring a variable name
5582 once the compiler has processed its initializer fully and thus has had a
5583 chance to determine the size of an array when controlled by an
5584 initializer. This is used on systems where it's necessary to declare
5585 something about the size of the object.
5587 If you don't define this macro, that is equivalent to defining it to do
5590 @findex ASM_GLOBALIZE_LABEL
5591 @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
5592 A C statement (sans semicolon) to output to the stdio stream
5593 @var{stream} some commands that will make the label @var{name} global;
5594 that is, available for reference from other files. Use the expression
5595 @code{assemble_name (@var{stream}, @var{name})} to output the name
5596 itself; before and after that, output the additional assembler syntax
5597 for making that name global, and a newline.
5599 @findex ASM_WEAKEN_LABEL
5600 @item ASM_WEAKEN_LABEL
5601 A C statement (sans semicolon) to output to the stdio stream
5602 @var{stream} some commands that will make the label @var{name} weak;
5603 that is, available for reference from other files but only used if
5604 no other definition is available. Use the expression
5605 @code{assemble_name (@var{stream}, @var{name})} to output the name
5606 itself; before and after that, output the additional assembler syntax
5607 for making that name weak, and a newline.
5609 If you don't define this macro, GCC will not support weak
5610 symbols and you should not define the @code{SUPPORTS_WEAK} macro.
5612 @findex SUPPORTS_WEAK
5614 A C expression which evaluates to true if the target supports weak symbols.
5616 If you don't define this macro, @file{defaults.h} provides a default
5617 definition. If @code{ASM_WEAKEN_LABEL} is defined, the default
5618 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5619 you want to control weak symbol support with a compiler flag such as
5622 @findex MAKE_DECL_ONE_ONLY (@var{decl})
5623 @item MAKE_DECL_ONE_ONLY
5624 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5625 public symbol such that extra copies in multiple translation units will
5626 be discarded by the linker. Define this macro if your object file
5627 format provides support for this concept, such as the @samp{COMDAT}
5628 section flags in the Microsoft Windows PE/COFF format, and this support
5629 requires changes to @var{decl}, such as putting it in a separate section.
5631 @findex SUPPORTS_ONE_ONLY
5632 @item SUPPORTS_ONE_ONLY
5633 A C expression which evaluates to true if the target supports one-only
5636 If you don't define this macro, @file{varasm.c} provides a default
5637 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
5638 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5639 you want to control one-only symbol support with a compiler flag, or if
5640 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
5641 be emitted as one-only.
5643 @findex ASM_OUTPUT_EXTERNAL
5644 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
5645 A C statement (sans semicolon) to output to the stdio stream
5646 @var{stream} any text necessary for declaring the name of an external
5647 symbol named @var{name} which is referenced in this compilation but
5648 not defined. The value of @var{decl} is the tree node for the
5651 This macro need not be defined if it does not need to output anything.
5652 The GNU assembler and most Unix assemblers don't require anything.
5654 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
5655 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
5656 A C statement (sans semicolon) to output on @var{stream} an assembler
5657 pseudo-op to declare a library function name external. The name of the
5658 library function is given by @var{symref}, which has type @code{rtx} and
5659 is a @code{symbol_ref}.
5661 This macro need not be defined if it does not need to output anything.
5662 The GNU assembler and most Unix assemblers don't require anything.
5664 @findex ASM_OUTPUT_LABELREF
5665 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
5666 A C statement (sans semicolon) to output to the stdio stream
5667 @var{stream} a reference in assembler syntax to a label named
5668 @var{name}. This should add @samp{_} to the front of the name, if that
5669 is customary on your operating system, as it is in most Berkeley Unix
5670 systems. This macro is used in @code{assemble_name}.
5672 @ignore @c Seems not to exist anymore.
5673 @findex ASM_OUTPUT_LABELREF_AS_INT
5674 @item ASM_OUTPUT_LABELREF_AS_INT (@var{file}, @var{label})
5675 Define this macro for systems that use the program @code{collect2}.
5676 The definition should be a C statement to output a word containing
5677 a reference to the label @var{label}.
5680 @findex ASM_OUTPUT_INTERNAL_LABEL
5681 @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
5682 A C statement to output to the stdio stream @var{stream} a label whose
5683 name is made from the string @var{prefix} and the number @var{num}.
5685 It is absolutely essential that these labels be distinct from the labels
5686 used for user-level functions and variables. Otherwise, certain programs
5687 will have name conflicts with internal labels.
5689 It is desirable to exclude internal labels from the symbol table of the
5690 object file. Most assemblers have a naming convention for labels that
5691 should be excluded; on many systems, the letter @samp{L} at the
5692 beginning of a label has this effect. You should find out what
5693 convention your system uses, and follow it.
5695 The usual definition of this macro is as follows:
5698 fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
5701 @findex ASM_OUTPUT_ALTERNATE_LABEL_NAME
5702 @item ASM_OUTPUT_ALTERNATE_LABEL_NAME (@var{stream}, @var{string})
5703 A C statement to output to the stdio stream @var{stream} the string
5706 The default definition of this macro is as follows:
5709 fprintf (@var{stream}, "%s:\n", LABEL_ALTERNATE_NAME (INSN))
5712 @findex ASM_GENERATE_INTERNAL_LABEL
5713 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
5714 A C statement to store into the string @var{string} a label whose name
5715 is made from the string @var{prefix} and the number @var{num}.
5717 This string, when output subsequently by @code{assemble_name}, should
5718 produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
5719 with the same @var{prefix} and @var{num}.
5721 If the string begins with @samp{*}, then @code{assemble_name} will
5722 output the rest of the string unchanged. It is often convenient for
5723 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
5724 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
5725 to output the string, and may change it. (Of course,
5726 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
5727 you should know what it does on your machine.)
5729 @findex ASM_FORMAT_PRIVATE_NAME
5730 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
5731 A C expression to assign to @var{outvar} (which is a variable of type
5732 @code{char *}) a newly allocated string made from the string
5733 @var{name} and the number @var{number}, with some suitable punctuation
5734 added. Use @code{alloca} to get space for the string.
5736 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
5737 produce an assembler label for an internal static variable whose name is
5738 @var{name}. Therefore, the string must be such as to result in valid
5739 assembler code. The argument @var{number} is different each time this
5740 macro is executed; it prevents conflicts between similarly-named
5741 internal static variables in different scopes.
5743 Ideally this string should not be a valid C identifier, to prevent any
5744 conflict with the user's own symbols. Most assemblers allow periods
5745 or percent signs in assembler symbols; putting at least one of these
5746 between the name and the number will suffice.
5748 @findex ASM_OUTPUT_DEF
5749 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
5750 A C statement to output to the stdio stream @var{stream} assembler code
5751 which defines (equates) the symbol @var{name} to have the value @var{value}.
5754 If SET_ASM_OP is defined, a default definition is provided which is
5755 correct for most systems.
5757 @findex ASM_OUTPUT_DEF_FROM_DECLS
5758 @item ASM_OUTPUT_DEF (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
5759 A C statement to output to the stdio stream @var{stream} assembler code
5760 which defines (equates) the symbol whoes tree node is @var{decl_of_name}
5761 to have the value of the tree node @var{decl_of_value}. This macro will
5762 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
5763 the tree nodes are available.
5765 @findex ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL
5766 @item ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL (@var{stream}, @var{symbol}, @var{high}, @var{low})
5767 A C statement to output to the stdio stream @var{stream} assembler code
5768 which defines (equates) the symbol @var{symbol} to have a value equal to
5769 the difference of the two symbols @var{high} and @var{low}, i.e.
5770 @var{high} minus @var{low}. GCC guarantees that the symbols @var{high}
5771 and @var{low} are already known by the assembler so that the difference
5772 resolves into a constant.
5775 If SET_ASM_OP is defined, a default definition is provided which is
5776 correct for most systems.
5778 @findex ASM_OUTPUT_WEAK_ALIAS
5779 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
5780 A C statement to output to the stdio stream @var{stream} assembler code
5781 which defines (equates) the weak symbol @var{name} to have the value
5784 Define this macro if the target only supports weak aliases; define
5785 ASM_OUTPUT_DEF instead if possible.
5787 @findex OBJC_GEN_METHOD_LABEL
5788 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
5789 Define this macro to override the default assembler names used for
5790 Objective C methods.
5792 The default name is a unique method number followed by the name of the
5793 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
5794 the category is also included in the assembler name (e.g.@:
5797 These names are safe on most systems, but make debugging difficult since
5798 the method's selector is not present in the name. Therefore, particular
5799 systems define other ways of computing names.
5801 @var{buf} is an expression of type @code{char *} which gives you a
5802 buffer in which to store the name; its length is as long as
5803 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
5804 50 characters extra.
5806 The argument @var{is_inst} specifies whether the method is an instance
5807 method or a class method; @var{class_name} is the name of the class;
5808 @var{cat_name} is the name of the category (or NULL if the method is not
5809 in a category); and @var{sel_name} is the name of the selector.
5811 On systems where the assembler can handle quoted names, you can use this
5812 macro to provide more human-readable names.
5815 @node Initialization
5816 @subsection How Initialization Functions Are Handled
5817 @cindex initialization routines
5818 @cindex termination routines
5819 @cindex constructors, output of
5820 @cindex destructors, output of
5822 The compiled code for certain languages includes @dfn{constructors}
5823 (also called @dfn{initialization routines})---functions to initialize
5824 data in the program when the program is started. These functions need
5825 to be called before the program is ``started''---that is to say, before
5826 @code{main} is called.
5828 Compiling some languages generates @dfn{destructors} (also called
5829 @dfn{termination routines}) that should be called when the program
5832 To make the initialization and termination functions work, the compiler
5833 must output something in the assembler code to cause those functions to
5834 be called at the appropriate time. When you port the compiler to a new
5835 system, you need to specify how to do this.
5837 There are two major ways that GCC currently supports the execution of
5838 initialization and termination functions. Each way has two variants.
5839 Much of the structure is common to all four variations.
5841 @findex __CTOR_LIST__
5842 @findex __DTOR_LIST__
5843 The linker must build two lists of these functions---a list of
5844 initialization functions, called @code{__CTOR_LIST__}, and a list of
5845 termination functions, called @code{__DTOR_LIST__}.
5847 Each list always begins with an ignored function pointer (which may hold
5848 0, @minus{}1, or a count of the function pointers after it, depending on
5849 the environment). This is followed by a series of zero or more function
5850 pointers to constructors (or destructors), followed by a function
5851 pointer containing zero.
5853 Depending on the operating system and its executable file format, either
5854 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
5855 time and exit time. Constructors are called in reverse order of the
5856 list; destructors in forward order.
5858 The best way to handle static constructors works only for object file
5859 formats which provide arbitrarily-named sections. A section is set
5860 aside for a list of constructors, and another for a list of destructors.
5861 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
5862 object file that defines an initialization function also puts a word in
5863 the constructor section to point to that function. The linker
5864 accumulates all these words into one contiguous @samp{.ctors} section.
5865 Termination functions are handled similarly.
5867 To use this method, you need appropriate definitions of the macros
5868 @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}. Usually
5869 you can get them by including @file{svr4.h}.
5871 When arbitrary sections are available, there are two variants, depending
5872 upon how the code in @file{crtstuff.c} is called. On systems that
5873 support an @dfn{init} section which is executed at program startup,
5874 parts of @file{crtstuff.c} are compiled into that section. The
5875 program is linked by the @code{gcc} driver like this:
5878 ld -o @var{output_file} crtbegin.o @dots{} crtend.o -lgcc
5881 The head of a function (@code{__do_global_ctors}) appears in the init
5882 section of @file{crtbegin.o}; the remainder of the function appears in
5883 the init section of @file{crtend.o}. The linker will pull these two
5884 parts of the section together, making a whole function. If any of the
5885 user's object files linked into the middle of it contribute code, then that
5886 code will be executed as part of the body of @code{__do_global_ctors}.
5888 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
5891 If no init section is available, do not define
5892 @code{INIT_SECTION_ASM_OP}. Then @code{__do_global_ctors} is built into
5893 the text section like all other functions, and resides in
5894 @file{libgcc.a}. When GCC compiles any function called @code{main}, it
5895 inserts a procedure call to @code{__main} as the first executable code
5896 after the function prologue. The @code{__main} function, also defined
5897 in @file{libgcc2.c}, simply calls @file{__do_global_ctors}.
5899 In file formats that don't support arbitrary sections, there are again
5900 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
5901 and an `a.out' format must be used. In this case,
5902 @code{ASM_OUTPUT_CONSTRUCTOR} is defined to produce a @code{.stabs}
5903 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
5904 and with the address of the void function containing the initialization
5905 code as its value. The GNU linker recognizes this as a request to add
5906 the value to a ``set''; the values are accumulated, and are eventually
5907 placed in the executable as a vector in the format described above, with
5908 a leading (ignored) count and a trailing zero element.
5909 @code{ASM_OUTPUT_DESTRUCTOR} is handled similarly. Since no init
5910 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
5911 the compilation of @code{main} to call @code{__main} as above, starting
5912 the initialization process.
5914 The last variant uses neither arbitrary sections nor the GNU linker.
5915 This is preferable when you want to do dynamic linking and when using
5916 file formats which the GNU linker does not support, such as `ECOFF'. In
5917 this case, @code{ASM_OUTPUT_CONSTRUCTOR} does not produce an
5918 @code{N_SETT} symbol; initialization and termination functions are
5919 recognized simply by their names. This requires an extra program in the
5920 linkage step, called @code{collect2}. This program pretends to be the
5921 linker, for use with GCC; it does its job by running the ordinary
5922 linker, but also arranges to include the vectors of initialization and
5923 termination functions. These functions are called via @code{__main} as
5926 Choosing among these configuration options has been simplified by a set
5927 of operating-system-dependent files in the @file{config} subdirectory.
5928 These files define all of the relevant parameters. Usually it is
5929 sufficient to include one into your specific machine-dependent
5930 configuration file. These files are:
5934 For operating systems using the `a.out' format.
5937 For operating systems using the `MachO' format.
5940 For System V Release 3 and similar systems using `COFF' format.
5943 For System V Release 4 and similar systems using `ELF' format.
5946 For the VMS operating system.
5950 The following section describes the specific macros that control and
5951 customize the handling of initialization and termination functions.
5954 @node Macros for Initialization
5955 @subsection Macros Controlling Initialization Routines
5957 Here are the macros that control how the compiler handles initialization
5958 and termination functions:
5961 @findex INIT_SECTION_ASM_OP
5962 @item INIT_SECTION_ASM_OP
5963 If defined, a C string constant for the assembler operation to identify
5964 the following data as initialization code. If not defined, GCC will
5965 assume such a section does not exist. When you are using special
5966 sections for initialization and termination functions, this macro also
5967 controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to run the
5968 initialization functions.
5970 @item HAS_INIT_SECTION
5971 @findex HAS_INIT_SECTION
5972 If defined, @code{main} will not call @code{__main} as described above.
5973 This macro should be defined for systems that control the contents of the
5974 init section on a symbol-by-symbol basis, such as OSF/1, and should not
5975 be defined explicitly for systems that support
5976 @code{INIT_SECTION_ASM_OP}.
5978 @item LD_INIT_SWITCH
5979 @findex LD_INIT_SWITCH
5980 If defined, a C string constant for a switch that tells the linker that
5981 the following symbol is an initialization routine.
5983 @item LD_FINI_SWITCH
5984 @findex LD_FINI_SWITCH
5985 If defined, a C string constant for a switch that tells the linker that
5986 the following symbol is a finalization routine.
5989 @findex INVOKE__main
5990 If defined, @code{main} will call @code{__main} despite the presence of
5991 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
5992 where the init section is not actually run automatically, but is still
5993 useful for collecting the lists of constructors and destructors.
5995 @item ASM_OUTPUT_CONSTRUCTOR (@var{stream}, @var{name})
5996 @findex ASM_OUTPUT_CONSTRUCTOR
5997 Define this macro as a C statement to output on the stream @var{stream}
5998 the assembler code to arrange to call the function named @var{name} at
5999 initialization time.
6001 Assume that @var{name} is the name of a C function generated
6002 automatically by the compiler. This function takes no arguments. Use
6003 the function @code{assemble_name} to output the name @var{name}; this
6004 performs any system-specific syntactic transformations such as adding an
6007 If you don't define this macro, nothing special is output to arrange to
6008 call the function. This is correct when the function will be called in
6009 some other manner---for example, by means of the @code{collect2} program,
6010 which looks through the symbol table to find these functions by their
6013 @item ASM_OUTPUT_DESTRUCTOR (@var{stream}, @var{name})
6014 @findex ASM_OUTPUT_DESTRUCTOR
6015 This is like @code{ASM_OUTPUT_CONSTRUCTOR} but used for termination
6016 functions rather than initialization functions.
6018 When @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR} are
6019 defined, the initializaiton routine generated for the generated object
6020 file will have static linkage.
6023 If your system uses @code{collect2} as the means of processing
6024 constructors, then that program normally uses @code{nm} to scan an
6025 object file for constructor functions to be called. On such systems you
6026 must not define @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}
6027 as the object file's initialization routine must have global scope.
6029 On certain kinds of systems, you can define these macros to make
6030 @code{collect2} work faster (and, in some cases, make it work at all):
6033 @findex OBJECT_FORMAT_COFF
6034 @item OBJECT_FORMAT_COFF
6035 Define this macro if the system uses COFF (Common Object File Format)
6036 object files, so that @code{collect2} can assume this format and scan
6037 object files directly for dynamic constructor/destructor functions.
6039 @findex OBJECT_FORMAT_ROSE
6040 @item OBJECT_FORMAT_ROSE
6041 Define this macro if the system uses ROSE format object files, so that
6042 @code{collect2} can assume this format and scan object files directly
6043 for dynamic constructor/destructor functions.
6045 These macros are effective only in a native compiler; @code{collect2} as
6046 part of a cross compiler always uses @code{nm} for the target machine.
6048 @findex REAL_NM_FILE_NAME
6049 @item REAL_NM_FILE_NAME
6050 Define this macro as a C string constant containing the file name to use
6051 to execute @code{nm}. The default is to search the path normally for
6054 If your system supports shared libraries and has a program to list the
6055 dynamic dependencies of a given library or executable, you can define
6056 these macros to enable support for running initialization and
6057 termination functions in shared libraries:
6061 Define this macro to a C string constant containing the name of the
6062 program which lists dynamic dependencies, like @code{"ldd"} under SunOS 4.
6064 @findex PARSE_LDD_OUTPUT
6065 @item PARSE_LDD_OUTPUT (@var{PTR})
6066 Define this macro to be C code that extracts filenames from the output
6067 of the program denoted by @code{LDD_SUFFIX}. @var{PTR} is a variable
6068 of type @code{char *} that points to the beginning of a line of output
6069 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6070 code must advance @var{PTR} to the beginning of the filename on that
6071 line. Otherwise, it must set @var{PTR} to @code{NULL}.
6075 @node Instruction Output
6076 @subsection Output of Assembler Instructions
6078 @c prevent bad page break with this line
6079 This describes assembler instruction output.
6082 @findex REGISTER_NAMES
6083 @item REGISTER_NAMES
6084 A C initializer containing the assembler's names for the machine
6085 registers, each one as a C string constant. This is what translates
6086 register numbers in the compiler into assembler language.
6088 @findex ADDITIONAL_REGISTER_NAMES
6089 @item ADDITIONAL_REGISTER_NAMES
6090 If defined, a C initializer for an array of structures containing a name
6091 and a register number. This macro defines additional names for hard
6092 registers, thus allowing the @code{asm} option in declarations to refer
6093 to registers using alternate names.
6095 @findex ASM_OUTPUT_OPCODE
6096 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6097 Define this macro if you are using an unusual assembler that
6098 requires different names for the machine instructions.
6100 The definition is a C statement or statements which output an
6101 assembler instruction opcode to the stdio stream @var{stream}. The
6102 macro-operand @var{ptr} is a variable of type @code{char *} which
6103 points to the opcode name in its ``internal'' form---the form that is
6104 written in the machine description. The definition should output the
6105 opcode name to @var{stream}, performing any translation you desire, and
6106 increment the variable @var{ptr} to point at the end of the opcode
6107 so that it will not be output twice.
6109 In fact, your macro definition may process less than the entire opcode
6110 name, or more than the opcode name; but if you want to process text
6111 that includes @samp{%}-sequences to substitute operands, you must take
6112 care of the substitution yourself. Just be sure to increment
6113 @var{ptr} over whatever text should not be output normally.
6115 @findex recog_operand
6116 If you need to look at the operand values, they can be found as the
6117 elements of @code{recog_operand}.
6119 If the macro definition does nothing, the instruction is output
6122 @findex FINAL_PRESCAN_INSN
6123 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6124 If defined, a C statement to be executed just prior to the output of
6125 assembler code for @var{insn}, to modify the extracted operands so
6126 they will be output differently.
6128 Here the argument @var{opvec} is the vector containing the operands
6129 extracted from @var{insn}, and @var{noperands} is the number of
6130 elements of the vector which contain meaningful data for this insn.
6131 The contents of this vector are what will be used to convert the insn
6132 template into assembler code, so you can change the assembler output
6133 by changing the contents of the vector.
6135 This macro is useful when various assembler syntaxes share a single
6136 file of instruction patterns; by defining this macro differently, you
6137 can cause a large class of instructions to be output differently (such
6138 as with rearranged operands). Naturally, variations in assembler
6139 syntax affecting individual insn patterns ought to be handled by
6140 writing conditional output routines in those patterns.
6142 If this macro is not defined, it is equivalent to a null statement.
6144 @findex FINAL_PRESCAN_LABEL
6145 @item FINAL_PRESCAN_LABEL
6146 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
6147 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
6148 @var{noperands} will be zero.
6150 @findex PRINT_OPERAND
6151 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6152 A C compound statement to output to stdio stream @var{stream} the
6153 assembler syntax for an instruction operand @var{x}. @var{x} is an
6156 @var{code} is a value that can be used to specify one of several ways
6157 of printing the operand. It is used when identical operands must be
6158 printed differently depending on the context. @var{code} comes from
6159 the @samp{%} specification that was used to request printing of the
6160 operand. If the specification was just @samp{%@var{digit}} then
6161 @var{code} is 0; if the specification was @samp{%@var{ltr}
6162 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6165 If @var{x} is a register, this macro should print the register's name.
6166 The names can be found in an array @code{reg_names} whose type is
6167 @code{char *[]}. @code{reg_names} is initialized from
6168 @code{REGISTER_NAMES}.
6170 When the machine description has a specification @samp{%@var{punct}}
6171 (a @samp{%} followed by a punctuation character), this macro is called
6172 with a null pointer for @var{x} and the punctuation character for
6175 @findex PRINT_OPERAND_PUNCT_VALID_P
6176 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6177 A C expression which evaluates to true if @var{code} is a valid
6178 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6179 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6180 punctuation characters (except for the standard one, @samp{%}) are used
6183 @findex PRINT_OPERAND_ADDRESS
6184 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6185 A C compound statement to output to stdio stream @var{stream} the
6186 assembler syntax for an instruction operand that is a memory reference
6187 whose address is @var{x}. @var{x} is an RTL expression.
6189 @cindex @code{ENCODE_SECTION_INFO} usage
6190 On some machines, the syntax for a symbolic address depends on the
6191 section that the address refers to. On these machines, define the macro
6192 @code{ENCODE_SECTION_INFO} to store the information into the
6193 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
6195 @findex DBR_OUTPUT_SEQEND
6196 @findex dbr_sequence_length
6197 @item DBR_OUTPUT_SEQEND(@var{file})
6198 A C statement, to be executed after all slot-filler instructions have
6199 been output. If necessary, call @code{dbr_sequence_length} to
6200 determine the number of slots filled in a sequence (zero if not
6201 currently outputting a sequence), to decide how many no-ops to output,
6204 Don't define this macro if it has nothing to do, but it is helpful in
6205 reading assembly output if the extent of the delay sequence is made
6206 explicit (e.g. with white space).
6208 @findex final_sequence
6209 Note that output routines for instructions with delay slots must be
6210 prepared to deal with not being output as part of a sequence (i.e.
6211 when the scheduling pass is not run, or when no slot fillers could be
6212 found.) The variable @code{final_sequence} is null when not
6213 processing a sequence, otherwise it contains the @code{sequence} rtx
6216 @findex REGISTER_PREFIX
6217 @findex LOCAL_LABEL_PREFIX
6218 @findex USER_LABEL_PREFIX
6219 @findex IMMEDIATE_PREFIX
6221 @item REGISTER_PREFIX
6222 @itemx LOCAL_LABEL_PREFIX
6223 @itemx USER_LABEL_PREFIX
6224 @itemx IMMEDIATE_PREFIX
6225 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6226 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6227 @file{final.c}). These are useful when a single @file{md} file must
6228 support multiple assembler formats. In that case, the various @file{tm.h}
6229 files can define these macros differently.
6231 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
6232 @findex ASM_FPRINTF_EXTENSIONS
6233 If defiend this macro should expand to a series of @code{case}
6234 statements which will be parsed inside the @code{switch} statement of
6235 the @code{asm_fprintf} function. This allows targets to define extra
6236 printf formats which may useful when generating their assembler
6237 statements. Noet that upper case letters are reserved for future
6238 generic extensions to asm_fprintf, and so are not available to target
6239 specific code. The output file is given by the parameter @var{file}.
6240 The varargs input pointer is @var{argptr} and the rest of the format
6241 string, starting the character after the one that is being switched
6242 upon, is pointed to by @var{format}.
6244 @findex ASSEMBLER_DIALECT
6245 @item ASSEMBLER_DIALECT
6246 If your target supports multiple dialects of assembler language (such as
6247 different opcodes), define this macro as a C expression that gives the
6248 numeric index of the assembler language dialect to use, with zero as the
6251 If this macro is defined, you may use constructs of the form
6252 @samp{@{option0|option1|option2@dots{}@}} in the output
6253 templates of patterns (@pxref{Output Template}) or in the first argument
6254 of @code{asm_fprintf}. This construct outputs @samp{option0},
6255 @samp{option1} or @samp{option2}, etc., if the value of
6256 @code{ASSEMBLER_DIALECT} is zero, one or two, etc. Any special
6257 characters within these strings retain their usual meaning.
6259 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6260 @samp{@}} do not have any special meaning when used in templates or
6261 operands to @code{asm_fprintf}.
6263 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6264 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6265 the variations in assembler language syntax with that mechanism. Define
6266 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6267 if the syntax variant are larger and involve such things as different
6268 opcodes or operand order.
6270 @findex ASM_OUTPUT_REG_PUSH
6271 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6272 A C expression to output to @var{stream} some assembler code
6273 which will push hard register number @var{regno} onto the stack.
6274 The code need not be optimal, since this macro is used only when
6277 @findex ASM_OUTPUT_REG_POP
6278 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6279 A C expression to output to @var{stream} some assembler code
6280 which will pop hard register number @var{regno} off of the stack.
6281 The code need not be optimal, since this macro is used only when
6285 @node Dispatch Tables
6286 @subsection Output of Dispatch Tables
6288 @c prevent bad page break with this line
6289 This concerns dispatch tables.
6292 @cindex dispatch table
6293 @findex ASM_OUTPUT_ADDR_DIFF_ELT
6294 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6295 A C statement to output to the stdio stream @var{stream} an assembler
6296 pseudo-instruction to generate a difference between two labels.
6297 @var{value} and @var{rel} are the numbers of two internal labels. The
6298 definitions of these labels are output using
6299 @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
6300 way here. For example,
6303 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6304 @var{value}, @var{rel})
6307 You must provide this macro on machines where the addresses in a
6308 dispatch table are relative to the table's own address. If defined, GNU
6309 CC will also use this macro on all machines when producing PIC.
6310 @var{body} is the body of the ADDR_DIFF_VEC; it is provided so that the
6311 mode and flags can be read.
6313 @findex ASM_OUTPUT_ADDR_VEC_ELT
6314 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6315 This macro should be provided on machines where the addresses
6316 in a dispatch table are absolute.
6318 The definition should be a C statement to output to the stdio stream
6319 @var{stream} an assembler pseudo-instruction to generate a reference to
6320 a label. @var{value} is the number of an internal label whose
6321 definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
6325 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6328 @findex ASM_OUTPUT_CASE_LABEL
6329 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6330 Define this if the label before a jump-table needs to be output
6331 specially. The first three arguments are the same as for
6332 @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
6333 jump-table which follows (a @code{jump_insn} containing an
6334 @code{addr_vec} or @code{addr_diff_vec}).
6336 This feature is used on system V to output a @code{swbeg} statement
6339 If this macro is not defined, these labels are output with
6340 @code{ASM_OUTPUT_INTERNAL_LABEL}.
6342 @findex ASM_OUTPUT_CASE_END
6343 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6344 Define this if something special must be output at the end of a
6345 jump-table. The definition should be a C statement to be executed
6346 after the assembler code for the table is written. It should write
6347 the appropriate code to stdio stream @var{stream}. The argument
6348 @var{table} is the jump-table insn, and @var{num} is the label-number
6349 of the preceding label.
6351 If this macro is not defined, nothing special is output at the end of
6355 @node Exception Region Output
6356 @subsection Assembler Commands for Exception Regions
6358 @c prevent bad page break with this line
6360 This describes commands marking the start and the end of an exception
6364 @findex ASM_OUTPUT_EH_REGION_BEG
6365 @item ASM_OUTPUT_EH_REGION_BEG ()
6366 A C expression to output text to mark the start of an exception region.
6368 This macro need not be defined on most platforms.
6370 @findex ASM_OUTPUT_EH_REGION_END
6371 @item ASM_OUTPUT_EH_REGION_END ()
6372 A C expression to output text to mark the end of an exception region.
6374 This macro need not be defined on most platforms.
6376 @findex EXCEPTION_SECTION
6377 @item EXCEPTION_SECTION ()
6378 A C expression to switch to the section in which the main
6379 exception table is to be placed (@pxref{Sections}). The default is a
6380 section named @code{.gcc_except_table} on machines that support named
6381 sections via @code{ASM_OUTPUT_SECTION_NAME}, otherwise if @samp{-fpic}
6382 or @samp{-fPIC} is in effect, the @code{data_section}, otherwise the
6383 @code{readonly_data_section}.
6385 @findex EH_FRAME_SECTION_ASM_OP
6386 @item EH_FRAME_SECTION_ASM_OP
6387 If defined, a C string constant for the assembler operation to switch to
6388 the section for exception handling frame unwind information. If not
6389 defined, GCC will provide a default definition if the target supports
6390 named sections. @file{crtstuff.c} uses this macro to switch to the
6391 appropriate section.
6393 You should define this symbol if your target supports DWARF 2 frame
6394 unwind information and the default definition does not work.
6396 @findex OMIT_EH_TABLE
6397 @item OMIT_EH_TABLE ()
6398 A C expression that is nonzero if the normal exception table output
6401 This macro need not be defined on most platforms.
6403 @findex EH_TABLE_LOOKUP
6404 @item EH_TABLE_LOOKUP ()
6405 Alternate runtime support for looking up an exception at runtime and
6406 finding the associated handler, if the default method won't work.
6408 This macro need not be defined on most platforms.
6410 @findex DOESNT_NEED_UNWINDER
6411 @item DOESNT_NEED_UNWINDER
6412 A C expression that decides whether or not the current function needs to
6413 have a function unwinder generated for it. See the file @code{except.c}
6414 for details on when to define this, and how.
6416 @findex MASK_RETURN_ADDR
6417 @item MASK_RETURN_ADDR
6418 An rtx used to mask the return address found via RETURN_ADDR_RTX, so
6419 that it does not contain any extraneous set bits in it.
6421 @findex DWARF2_UNWIND_INFO
6422 @item DWARF2_UNWIND_INFO
6423 Define this macro to 0 if your target supports DWARF 2 frame unwind
6424 information, but it does not yet work with exception handling.
6425 Otherwise, if your target supports this information (if it defines
6426 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
6427 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
6430 If this macro is defined to 1, the DWARF 2 unwinder will be the default
6431 exception handling mechanism; otherwise, setjmp/longjmp will be used by
6434 If this macro is defined to anything, the DWARF 2 unwinder will be used
6435 instead of inline unwinders and __unwind_function in the non-setjmp case.
6439 @node Alignment Output
6440 @subsection Assembler Commands for Alignment
6442 @c prevent bad page break with this line
6443 This describes commands for alignment.
6446 @findex LABEL_ALIGN_AFTER_BARRIER
6447 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
6448 The alignment (log base 2) to put in front of @var{label}, which follows
6451 This macro need not be defined if you don't want any special alignment
6452 to be done at such a time. Most machine descriptions do not currently
6455 Unless it's necessary to inspect the @var{label} parameter, it is better
6456 to set the variable @var{align_jumps} in the target's
6457 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6458 selection in @var{align_jumps} in a @code{LABEL_ALIGN_AFTER_BARRIER}
6461 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6462 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6463 The maximum number of bytes to skip when applying
6464 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
6465 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6468 @item LOOP_ALIGN (@var{label})
6469 The alignment (log base 2) to put in front of @var{label}, which follows
6470 a NOTE_INSN_LOOP_BEG note.
6472 This macro need not be defined if you don't want any special alignment
6473 to be done at such a time. Most machine descriptions do not currently
6476 Unless it's necessary to inspect the @var{label} parameter, it is better
6477 to set the variable @var{align_loops} in the target's
6478 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6479 selection in @var{align_loops} in a @code{LOOP_ALIGN} implementation.
6481 @findex LOOP_ALIGN_MAX_SKIP
6482 @item LOOP_ALIGN_MAX_SKIP
6483 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
6484 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6487 @item LABEL_ALIGN (@var{label})
6488 The alignment (log base 2) to put in front of @var{label}.
6489 If LABEL_ALIGN_AFTER_BARRIER / LOOP_ALIGN specify a different alignment,
6490 the maximum of the specified values is used.
6492 Unless it's necessary to inspect the @var{label} parameter, it is better
6493 to set the variable @var{align_labels} in the target's
6494 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6495 selection in @var{align_labels} in a @code{LABEL_ALIGN} implementation.
6497 @findex LABEL_ALIGN_MAX_SKIP
6498 @item LABEL_ALIGN_MAX_SKIP
6499 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
6500 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6502 @findex ASM_OUTPUT_SKIP
6503 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6504 A C statement to output to the stdio stream @var{stream} an assembler
6505 instruction to advance the location counter by @var{nbytes} bytes.
6506 Those bytes should be zero when loaded. @var{nbytes} will be a C
6507 expression of type @code{int}.
6509 @findex ASM_NO_SKIP_IN_TEXT
6510 @item ASM_NO_SKIP_IN_TEXT
6511 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6512 text section because it fails to put zeros in the bytes that are skipped.
6513 This is true on many Unix systems, where the pseudo--op to skip bytes
6514 produces no-op instructions rather than zeros when used in the text
6517 @findex ASM_OUTPUT_ALIGN
6518 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6519 A C statement to output to the stdio stream @var{stream} an assembler
6520 command to advance the location counter to a multiple of 2 to the
6521 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
6523 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
6524 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
6525 A C statement to output to the stdio stream @var{stream} an assembler
6526 command to advance the location counter to a multiple of 2 to the
6527 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
6528 satisfy the alignment request. @var{power} and @var{max_skip} will be
6529 a C expression of type @code{int}.
6533 @node Debugging Info
6534 @section Controlling Debugging Information Format
6536 @c prevent bad page break with this line
6537 This describes how to specify debugging information.
6540 * All Debuggers:: Macros that affect all debugging formats uniformly.
6541 * DBX Options:: Macros enabling specific options in DBX format.
6542 * DBX Hooks:: Hook macros for varying DBX format.
6543 * File Names and DBX:: Macros controlling output of file names in DBX format.
6544 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
6548 @subsection Macros Affecting All Debugging Formats
6550 @c prevent bad page break with this line
6551 These macros affect all debugging formats.
6554 @findex DBX_REGISTER_NUMBER
6555 @item DBX_REGISTER_NUMBER (@var{regno})
6556 A C expression that returns the DBX register number for the compiler
6557 register number @var{regno}. In simple cases, the value of this
6558 expression may be @var{regno} itself. But sometimes there are some
6559 registers that the compiler knows about and DBX does not, or vice
6560 versa. In such cases, some register may need to have one number in
6561 the compiler and another for DBX.
6563 If two registers have consecutive numbers inside GCC, and they can be
6564 used as a pair to hold a multiword value, then they @emph{must} have
6565 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
6566 Otherwise, debuggers will be unable to access such a pair, because they
6567 expect register pairs to be consecutive in their own numbering scheme.
6569 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
6570 does not preserve register pairs, then what you must do instead is
6571 redefine the actual register numbering scheme.
6573 @findex DEBUGGER_AUTO_OFFSET
6574 @item DEBUGGER_AUTO_OFFSET (@var{x})
6575 A C expression that returns the integer offset value for an automatic
6576 variable having address @var{x} (an RTL expression). The default
6577 computation assumes that @var{x} is based on the frame-pointer and
6578 gives the offset from the frame-pointer. This is required for targets
6579 that produce debugging output for DBX or COFF-style debugging output
6580 for SDB and allow the frame-pointer to be eliminated when the
6581 @samp{-g} options is used.
6583 @findex DEBUGGER_ARG_OFFSET
6584 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6585 A C expression that returns the integer offset value for an argument
6586 having address @var{x} (an RTL expression). The nominal offset is
6589 @findex PREFERRED_DEBUGGING_TYPE
6590 @item PREFERRED_DEBUGGING_TYPE
6591 A C expression that returns the type of debugging output GCC should
6592 produce when the user specifies just @samp{-g}. Define
6593 this if you have arranged for GCC to support more than one format of
6594 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
6595 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG}, and
6598 When the user specifies @samp{-ggdb}, GCC normally also uses the
6599 value of this macro to select the debugging output format, but with two
6600 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
6601 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
6602 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
6603 defined, GCC uses @code{DBX_DEBUG}.
6605 The value of this macro only affects the default debugging output; the
6606 user can always get a specific type of output by using @samp{-gstabs},
6607 @samp{-gcoff}, @samp{-gdwarf-1}, @samp{-gdwarf-2}, or @samp{-gxcoff}.
6611 @subsection Specific Options for DBX Output
6613 @c prevent bad page break with this line
6614 These are specific options for DBX output.
6617 @findex DBX_DEBUGGING_INFO
6618 @item DBX_DEBUGGING_INFO
6619 Define this macro if GCC should produce debugging output for DBX
6620 in response to the @samp{-g} option.
6622 @findex XCOFF_DEBUGGING_INFO
6623 @item XCOFF_DEBUGGING_INFO
6624 Define this macro if GCC should produce XCOFF format debugging output
6625 in response to the @samp{-g} option. This is a variant of DBX format.
6627 @findex DEFAULT_GDB_EXTENSIONS
6628 @item DEFAULT_GDB_EXTENSIONS
6629 Define this macro to control whether GCC should by default generate
6630 GDB's extended version of DBX debugging information (assuming DBX-format
6631 debugging information is enabled at all). If you don't define the
6632 macro, the default is 1: always generate the extended information
6633 if there is any occasion to.
6635 @findex DEBUG_SYMS_TEXT
6636 @item DEBUG_SYMS_TEXT
6637 Define this macro if all @code{.stabs} commands should be output while
6638 in the text section.
6640 @findex ASM_STABS_OP
6642 A C string constant naming the assembler pseudo op to use instead of
6643 @code{.stabs} to define an ordinary debugging symbol. If you don't
6644 define this macro, @code{.stabs} is used. This macro applies only to
6645 DBX debugging information format.
6647 @findex ASM_STABD_OP
6649 A C string constant naming the assembler pseudo op to use instead of
6650 @code{.stabd} to define a debugging symbol whose value is the current
6651 location. If you don't define this macro, @code{.stabd} is used.
6652 This macro applies only to DBX debugging information format.
6654 @findex ASM_STABN_OP
6656 A C string constant naming the assembler pseudo op to use instead of
6657 @code{.stabn} to define a debugging symbol with no name. If you don't
6658 define this macro, @code{.stabn} is used. This macro applies only to
6659 DBX debugging information format.
6661 @findex DBX_NO_XREFS
6663 Define this macro if DBX on your system does not support the construct
6664 @samp{xs@var{tagname}}. On some systems, this construct is used to
6665 describe a forward reference to a structure named @var{tagname}.
6666 On other systems, this construct is not supported at all.
6668 @findex DBX_CONTIN_LENGTH
6669 @item DBX_CONTIN_LENGTH
6670 A symbol name in DBX-format debugging information is normally
6671 continued (split into two separate @code{.stabs} directives) when it
6672 exceeds a certain length (by default, 80 characters). On some
6673 operating systems, DBX requires this splitting; on others, splitting
6674 must not be done. You can inhibit splitting by defining this macro
6675 with the value zero. You can override the default splitting-length by
6676 defining this macro as an expression for the length you desire.
6678 @findex DBX_CONTIN_CHAR
6679 @item DBX_CONTIN_CHAR
6680 Normally continuation is indicated by adding a @samp{\} character to
6681 the end of a @code{.stabs} string when a continuation follows. To use
6682 a different character instead, define this macro as a character
6683 constant for the character you want to use. Do not define this macro
6684 if backslash is correct for your system.
6686 @findex DBX_STATIC_STAB_DATA_SECTION
6687 @item DBX_STATIC_STAB_DATA_SECTION
6688 Define this macro if it is necessary to go to the data section before
6689 outputting the @samp{.stabs} pseudo-op for a non-global static
6692 @findex DBX_TYPE_DECL_STABS_CODE
6693 @item DBX_TYPE_DECL_STABS_CODE
6694 The value to use in the ``code'' field of the @code{.stabs} directive
6695 for a typedef. The default is @code{N_LSYM}.
6697 @findex DBX_STATIC_CONST_VAR_CODE
6698 @item DBX_STATIC_CONST_VAR_CODE
6699 The value to use in the ``code'' field of the @code{.stabs} directive
6700 for a static variable located in the text section. DBX format does not
6701 provide any ``right'' way to do this. The default is @code{N_FUN}.
6703 @findex DBX_REGPARM_STABS_CODE
6704 @item DBX_REGPARM_STABS_CODE
6705 The value to use in the ``code'' field of the @code{.stabs} directive
6706 for a parameter passed in registers. DBX format does not provide any
6707 ``right'' way to do this. The default is @code{N_RSYM}.
6709 @findex DBX_REGPARM_STABS_LETTER
6710 @item DBX_REGPARM_STABS_LETTER
6711 The letter to use in DBX symbol data to identify a symbol as a parameter
6712 passed in registers. DBX format does not customarily provide any way to
6713 do this. The default is @code{'P'}.
6715 @findex DBX_MEMPARM_STABS_LETTER
6716 @item DBX_MEMPARM_STABS_LETTER
6717 The letter to use in DBX symbol data to identify a symbol as a stack
6718 parameter. The default is @code{'p'}.
6720 @findex DBX_FUNCTION_FIRST
6721 @item DBX_FUNCTION_FIRST
6722 Define this macro if the DBX information for a function and its
6723 arguments should precede the assembler code for the function. Normally,
6724 in DBX format, the debugging information entirely follows the assembler
6727 @findex DBX_LBRAC_FIRST
6728 @item DBX_LBRAC_FIRST
6729 Define this macro if the @code{N_LBRAC} symbol for a block should
6730 precede the debugging information for variables and functions defined in
6731 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
6734 @findex DBX_BLOCKS_FUNCTION_RELATIVE
6735 @item DBX_BLOCKS_FUNCTION_RELATIVE
6736 Define this macro if the value of a symbol describing the scope of a
6737 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
6738 of the enclosing function. Normally, GNU C uses an absolute address.
6740 @findex DBX_USE_BINCL
6742 Define this macro if GNU C should generate @code{N_BINCL} and
6743 @code{N_EINCL} stabs for included header files, as on Sun systems. This
6744 macro also directs GNU C to output a type number as a pair of a file
6745 number and a type number within the file. Normally, GNU C does not
6746 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
6747 number for a type number.
6751 @subsection Open-Ended Hooks for DBX Format
6753 @c prevent bad page break with this line
6754 These are hooks for DBX format.
6757 @findex DBX_OUTPUT_LBRAC
6758 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
6759 Define this macro to say how to output to @var{stream} the debugging
6760 information for the start of a scope level for variable names. The
6761 argument @var{name} is the name of an assembler symbol (for use with
6762 @code{assemble_name}) whose value is the address where the scope begins.
6764 @findex DBX_OUTPUT_RBRAC
6765 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
6766 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
6768 @findex DBX_OUTPUT_ENUM
6769 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
6770 Define this macro if the target machine requires special handling to
6771 output an enumeration type. The definition should be a C statement
6772 (sans semicolon) to output the appropriate information to @var{stream}
6773 for the type @var{type}.
6775 @findex DBX_OUTPUT_FUNCTION_END
6776 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
6777 Define this macro if the target machine requires special output at the
6778 end of the debugging information for a function. The definition should
6779 be a C statement (sans semicolon) to output the appropriate information
6780 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
6783 @findex DBX_OUTPUT_STANDARD_TYPES
6784 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
6785 Define this macro if you need to control the order of output of the
6786 standard data types at the beginning of compilation. The argument
6787 @var{syms} is a @code{tree} which is a chain of all the predefined
6788 global symbols, including names of data types.
6790 Normally, DBX output starts with definitions of the types for integers
6791 and characters, followed by all the other predefined types of the
6792 particular language in no particular order.
6794 On some machines, it is necessary to output different particular types
6795 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
6796 those symbols in the necessary order. Any predefined types that you
6797 don't explicitly output will be output afterward in no particular order.
6799 Be careful not to define this macro so that it works only for C. There
6800 are no global variables to access most of the built-in types, because
6801 another language may have another set of types. The way to output a
6802 particular type is to look through @var{syms} to see if you can find it.
6808 for (decl = syms; decl; decl = TREE_CHAIN (decl))
6809 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
6811 dbxout_symbol (decl);
6817 This does nothing if the expected type does not exist.
6819 See the function @code{init_decl_processing} in @file{c-decl.c} to find
6820 the names to use for all the built-in C types.
6822 Here is another way of finding a particular type:
6824 @c this is still overfull. --mew 10feb93
6828 for (decl = syms; decl; decl = TREE_CHAIN (decl))
6829 if (TREE_CODE (decl) == TYPE_DECL
6830 && (TREE_CODE (TREE_TYPE (decl))
6832 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
6833 && TYPE_UNSIGNED (TREE_TYPE (decl)))
6835 /* @r{This must be @code{unsigned short}.} */
6836 dbxout_symbol (decl);
6842 @findex NO_DBX_FUNCTION_END
6843 @item NO_DBX_FUNCTION_END
6844 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
6845 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extention construct.
6846 On those machines, define this macro to turn this feature off without
6847 disturbing the rest of the gdb extensions.
6851 @node File Names and DBX
6852 @subsection File Names in DBX Format
6854 @c prevent bad page break with this line
6855 This describes file names in DBX format.
6858 @findex DBX_WORKING_DIRECTORY
6859 @item DBX_WORKING_DIRECTORY
6860 Define this if DBX wants to have the current directory recorded in each
6863 Note that the working directory is always recorded if GDB extensions are
6866 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
6867 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
6868 A C statement to output DBX debugging information to the stdio stream
6869 @var{stream} which indicates that file @var{name} is the main source
6870 file---the file specified as the input file for compilation.
6871 This macro is called only once, at the beginning of compilation.
6873 This macro need not be defined if the standard form of output
6874 for DBX debugging information is appropriate.
6876 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
6877 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
6878 A C statement to output DBX debugging information to the stdio stream
6879 @var{stream} which indicates that the current directory during
6880 compilation is named @var{name}.
6882 This macro need not be defined if the standard form of output
6883 for DBX debugging information is appropriate.
6885 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
6886 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
6887 A C statement to output DBX debugging information at the end of
6888 compilation of the main source file @var{name}.
6890 If you don't define this macro, nothing special is output at the end
6891 of compilation, which is correct for most machines.
6893 @findex DBX_OUTPUT_SOURCE_FILENAME
6894 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6895 A C statement to output DBX debugging information to the stdio stream
6896 @var{stream} which indicates that file @var{name} is the current source
6897 file. This output is generated each time input shifts to a different
6898 source file as a result of @samp{#include}, the end of an included file,
6899 or a @samp{#line} command.
6901 This macro need not be defined if the standard form of output
6902 for DBX debugging information is appropriate.
6907 @subsection Macros for SDB and DWARF Output
6909 @c prevent bad page break with this line
6910 Here are macros for SDB and DWARF output.
6913 @findex SDB_DEBUGGING_INFO
6914 @item SDB_DEBUGGING_INFO
6915 Define this macro if GCC should produce COFF-style debugging output
6916 for SDB in response to the @samp{-g} option.
6918 @findex DWARF_DEBUGGING_INFO
6919 @item DWARF_DEBUGGING_INFO
6920 Define this macro if GCC should produce dwarf format debugging output
6921 in response to the @samp{-g} option.
6923 @findex DWARF2_DEBUGGING_INFO
6924 @item DWARF2_DEBUGGING_INFO
6925 Define this macro if GCC should produce dwarf version 2 format
6926 debugging output in response to the @samp{-g} option.
6928 To support optional call frame debugging information, you must also
6929 define @code{INCOMING_RETURN_ADDR_RTX} and either set
6930 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
6931 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
6932 as appropriate from @code{FUNCTION_PROLOGUE} if you don't.
6934 @findex DWARF2_FRAME_INFO
6935 @item DWARF2_FRAME_INFO
6936 Define this macro to a nonzero value if GCC should always output
6937 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
6938 (@pxref{Exception Region Output} is nonzero, GCC will output this
6939 information not matter how you define @code{DWARF2_FRAME_INFO}.
6941 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
6942 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
6943 Define this macro if the linker does not work with Dwarf version 2.
6944 Normally, if the user specifies only @samp{-ggdb} GCC will use Dwarf
6945 version 2 if available; this macro disables this. See the description
6946 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
6948 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
6949 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
6950 By default, the Dwarf 2 debugging information generator will generate a
6951 label to mark the beginning of the text section. If it is better simply
6952 to use the name of the text section itself, rather than an explicit label,
6953 to indicate the beginning of the text section, define this macro to zero.
6955 @findex DWARF2_ASM_LINE_DEBUG_INFO
6956 @item DWARF2_ASM_LINE_DEBUG_INFO
6957 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
6958 line debug info sections. This will result in much more compact line number
6959 tables, and hence is desirable if it works.
6961 @findex PUT_SDB_@dots{}
6962 @item PUT_SDB_@dots{}
6963 Define these macros to override the assembler syntax for the special
6964 SDB assembler directives. See @file{sdbout.c} for a list of these
6965 macros and their arguments. If the standard syntax is used, you need
6966 not define them yourself.
6970 Some assemblers do not support a semicolon as a delimiter, even between
6971 SDB assembler directives. In that case, define this macro to be the
6972 delimiter to use (usually @samp{\n}). It is not necessary to define
6973 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
6976 @findex SDB_GENERATE_FAKE
6977 @item SDB_GENERATE_FAKE
6978 Define this macro to override the usual method of constructing a dummy
6979 name for anonymous structure and union types. See @file{sdbout.c} for
6982 @findex SDB_ALLOW_UNKNOWN_REFERENCES
6983 @item SDB_ALLOW_UNKNOWN_REFERENCES
6984 Define this macro to allow references to unknown structure,
6985 union, or enumeration tags to be emitted. Standard COFF does not
6986 allow handling of unknown references, MIPS ECOFF has support for
6989 @findex SDB_ALLOW_FORWARD_REFERENCES
6990 @item SDB_ALLOW_FORWARD_REFERENCES
6991 Define this macro to allow references to structure, union, or
6992 enumeration tags that have not yet been seen to be handled. Some
6993 assemblers choke if forward tags are used, while some require it.
6996 @node Cross-compilation
6997 @section Cross Compilation and Floating Point
6998 @cindex cross compilation and floating point
6999 @cindex floating point and cross compilation
7001 While all modern machines use 2's complement representation for integers,
7002 there are a variety of representations for floating point numbers. This
7003 means that in a cross-compiler the representation of floating point numbers
7004 in the compiled program may be different from that used in the machine
7005 doing the compilation.
7008 Because different representation systems may offer different amounts of
7009 range and precision, the cross compiler cannot safely use the host
7010 machine's floating point arithmetic. Therefore, floating point constants
7011 must be represented in the target machine's format. This means that the
7012 cross compiler cannot use @code{atof} to parse a floating point constant;
7013 it must have its own special routine to use instead. Also, constant
7014 folding must emulate the target machine's arithmetic (or must not be done
7017 The macros in the following table should be defined only if you are cross
7018 compiling between different floating point formats.
7020 Otherwise, don't define them. Then default definitions will be set up which
7021 use @code{double} as the data type, @code{==} to test for equality, etc.
7023 You don't need to worry about how many times you use an operand of any
7024 of these macros. The compiler never uses operands which have side effects.
7027 @findex REAL_VALUE_TYPE
7028 @item REAL_VALUE_TYPE
7029 A macro for the C data type to be used to hold a floating point value
7030 in the target machine's format. Typically this would be a
7031 @code{struct} containing an array of @code{int}.
7033 @findex REAL_VALUES_EQUAL
7034 @item REAL_VALUES_EQUAL (@var{x}, @var{y})
7035 A macro for a C expression which compares for equality the two values,
7036 @var{x} and @var{y}, both of type @code{REAL_VALUE_TYPE}.
7038 @findex REAL_VALUES_LESS
7039 @item REAL_VALUES_LESS (@var{x}, @var{y})
7040 A macro for a C expression which tests whether @var{x} is less than
7041 @var{y}, both values being of type @code{REAL_VALUE_TYPE} and
7042 interpreted as floating point numbers in the target machine's
7045 @findex REAL_VALUE_LDEXP
7047 @item REAL_VALUE_LDEXP (@var{x}, @var{scale})
7048 A macro for a C expression which performs the standard library
7049 function @code{ldexp}, but using the target machine's floating point
7050 representation. Both @var{x} and the value of the expression have
7051 type @code{REAL_VALUE_TYPE}. The second argument, @var{scale}, is an
7054 @findex REAL_VALUE_FIX
7055 @item REAL_VALUE_FIX (@var{x})
7056 A macro whose definition is a C expression to convert the target-machine
7057 floating point value @var{x} to a signed integer. @var{x} has type
7058 @code{REAL_VALUE_TYPE}.
7060 @findex REAL_VALUE_UNSIGNED_FIX
7061 @item REAL_VALUE_UNSIGNED_FIX (@var{x})
7062 A macro whose definition is a C expression to convert the target-machine
7063 floating point value @var{x} to an unsigned integer. @var{x} has type
7064 @code{REAL_VALUE_TYPE}.
7066 @findex REAL_VALUE_RNDZINT
7067 @item REAL_VALUE_RNDZINT (@var{x})
7068 A macro whose definition is a C expression to round the target-machine
7069 floating point value @var{x} towards zero to an integer value (but still
7070 as a floating point number). @var{x} has type @code{REAL_VALUE_TYPE},
7071 and so does the value.
7073 @findex REAL_VALUE_UNSIGNED_RNDZINT
7074 @item REAL_VALUE_UNSIGNED_RNDZINT (@var{x})
7075 A macro whose definition is a C expression to round the target-machine
7076 floating point value @var{x} towards zero to an unsigned integer value
7077 (but still represented as a floating point number). @var{x} has type
7078 @code{REAL_VALUE_TYPE}, and so does the value.
7080 @findex REAL_VALUE_ATOF
7081 @item REAL_VALUE_ATOF (@var{string}, @var{mode})
7082 A macro for a C expression which converts @var{string}, an expression of
7083 type @code{char *}, into a floating point number in the target machine's
7084 representation for mode @var{mode}. The value has type
7085 @code{REAL_VALUE_TYPE}.
7087 @findex REAL_INFINITY
7089 Define this macro if infinity is a possible floating point value, and
7090 therefore division by 0 is legitimate.
7092 @findex REAL_VALUE_ISINF
7094 @item REAL_VALUE_ISINF (@var{x})
7095 A macro for a C expression which determines whether @var{x}, a floating
7096 point value, is infinity. The value has type @code{int}.
7097 By default, this is defined to call @code{isinf}.
7099 @findex REAL_VALUE_ISNAN
7101 @item REAL_VALUE_ISNAN (@var{x})
7102 A macro for a C expression which determines whether @var{x}, a floating
7103 point value, is a ``nan'' (not-a-number). The value has type
7104 @code{int}. By default, this is defined to call @code{isnan}.
7107 @cindex constant folding and floating point
7108 Define the following additional macros if you want to make floating
7109 point constant folding work while cross compiling. If you don't
7110 define them, cross compilation is still possible, but constant folding
7111 will not happen for floating point values.
7114 @findex REAL_ARITHMETIC
7115 @item REAL_ARITHMETIC (@var{output}, @var{code}, @var{x}, @var{y})
7116 A macro for a C statement which calculates an arithmetic operation of
7117 the two floating point values @var{x} and @var{y}, both of type
7118 @code{REAL_VALUE_TYPE} in the target machine's representation, to
7119 produce a result of the same type and representation which is stored
7120 in @var{output} (which will be a variable).
7122 The operation to be performed is specified by @var{code}, a tree code
7123 which will always be one of the following: @code{PLUS_EXPR},
7124 @code{MINUS_EXPR}, @code{MULT_EXPR}, @code{RDIV_EXPR},
7125 @code{MAX_EXPR}, @code{MIN_EXPR}.@refill
7127 @cindex overflow while constant folding
7128 The expansion of this macro is responsible for checking for overflow.
7129 If overflow happens, the macro expansion should execute the statement
7130 @code{return 0;}, which indicates the inability to perform the
7131 arithmetic operation requested.
7133 @findex REAL_VALUE_NEGATE
7134 @item REAL_VALUE_NEGATE (@var{x})
7135 A macro for a C expression which returns the negative of the floating
7136 point value @var{x}. Both @var{x} and the value of the expression
7137 have type @code{REAL_VALUE_TYPE} and are in the target machine's
7138 floating point representation.
7140 There is no way for this macro to report overflow, since overflow
7141 can't happen in the negation operation.
7143 @findex REAL_VALUE_TRUNCATE
7144 @item REAL_VALUE_TRUNCATE (@var{mode}, @var{x})
7145 A macro for a C expression which converts the floating point value
7146 @var{x} to mode @var{mode}.
7148 Both @var{x} and the value of the expression are in the target machine's
7149 floating point representation and have type @code{REAL_VALUE_TYPE}.
7150 However, the value should have an appropriate bit pattern to be output
7151 properly as a floating constant whose precision accords with mode
7154 There is no way for this macro to report overflow.
7156 @findex REAL_VALUE_TO_INT
7157 @item REAL_VALUE_TO_INT (@var{low}, @var{high}, @var{x})
7158 A macro for a C expression which converts a floating point value
7159 @var{x} into a double-precision integer which is then stored into
7160 @var{low} and @var{high}, two variables of type @var{int}.
7162 @item REAL_VALUE_FROM_INT (@var{x}, @var{low}, @var{high}, @var{mode})
7163 @findex REAL_VALUE_FROM_INT
7164 A macro for a C expression which converts a double-precision integer
7165 found in @var{low} and @var{high}, two variables of type @var{int},
7166 into a floating point value which is then stored into @var{x}.
7167 The value is in the target machine's representation for mode @var{mode}
7168 and has the type @code{REAL_VALUE_TYPE}.
7172 @section Miscellaneous Parameters
7173 @cindex parameters, miscellaneous
7175 @c prevent bad page break with this line
7176 Here are several miscellaneous parameters.
7179 @item PREDICATE_CODES
7180 @findex PREDICATE_CODES
7181 Define this if you have defined special-purpose predicates in the file
7182 @file{@var{machine}.c}. This macro is called within an initializer of an
7183 array of structures. The first field in the structure is the name of a
7184 predicate and the second field is an array of rtl codes. For each
7185 predicate, list all rtl codes that can be in expressions matched by the
7186 predicate. The list should have a trailing comma. Here is an example
7187 of two entries in the list for a typical RISC machine:
7190 #define PREDICATE_CODES \
7191 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
7192 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
7195 Defining this macro does not affect the generated code (however,
7196 incorrect definitions that omit an rtl code that may be matched by the
7197 predicate can cause the compiler to malfunction). Instead, it allows
7198 the table built by @file{genrecog} to be more compact and efficient,
7199 thus speeding up the compiler. The most important predicates to include
7200 in the list specified by this macro are those used in the most insn
7203 @item SPECIAL_MODE_PREDICATES
7204 @findex SPECIAL_MODE_PREDICATES
7205 Define this if you have special predicates that know special things
7206 about modes. Genrecog will warn about certain forms of
7207 @code{match_operand} without a mode; if the operand predicate is
7208 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
7211 Here is an example from the IA-32 port (@code{ext_register_operand}
7212 specially checks for @code{HImode} or @code{SImode} in preparation
7213 for a byte extraction from @code{%ah} etc.).
7216 #define SPECIAL_MODE_PREDICATES \
7217 "ext_register_operand",
7220 @findex CASE_VECTOR_MODE
7221 @item CASE_VECTOR_MODE
7222 An alias for a machine mode name. This is the machine mode that
7223 elements of a jump-table should have.
7225 @findex CASE_VECTOR_SHORTEN_MODE
7226 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7227 Optional: return the preferred mode for an @code{addr_diff_vec}
7228 when the minimum and maximum offset are known. If you define this,
7229 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7230 To make this work, you also have to define INSN_ALIGN and
7231 make the alignment for @code{addr_diff_vec} explicit.
7232 The @var{body} argument is provided so that the offset_unsigned and scale
7233 flags can be updated.
7235 @findex CASE_VECTOR_PC_RELATIVE
7236 @item CASE_VECTOR_PC_RELATIVE
7237 Define this macro to be a C expression to indicate when jump-tables
7238 should contain relative addresses. If jump-tables never contain
7239 relative addresses, then you need not define this macro.
7241 @findex CASE_DROPS_THROUGH
7242 @item CASE_DROPS_THROUGH
7243 Define this if control falls through a @code{case} insn when the index
7244 value is out of range. This means the specified default-label is
7245 actually ignored by the @code{case} insn proper.
7247 @findex CASE_VALUES_THRESHOLD
7248 @item CASE_VALUES_THRESHOLD
7249 Define this to be the smallest number of different values for which it
7250 is best to use a jump-table instead of a tree of conditional branches.
7251 The default is four for machines with a @code{casesi} instruction and
7252 five otherwise. This is best for most machines.
7254 @findex WORD_REGISTER_OPERATIONS
7255 @item WORD_REGISTER_OPERATIONS
7256 Define this macro if operations between registers with integral mode
7257 smaller than a word are always performed on the entire register.
7258 Most RISC machines have this property and most CISC machines do not.
7260 @findex LOAD_EXTEND_OP
7261 @item LOAD_EXTEND_OP (@var{mode})
7262 Define this macro to be a C expression indicating when insns that read
7263 memory in @var{mode}, an integral mode narrower than a word, set the
7264 bits outside of @var{mode} to be either the sign-extension or the
7265 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7266 of @var{mode} for which the
7267 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7268 @code{NIL} for other modes.
7270 This macro is not called with @var{mode} non-integral or with a width
7271 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7272 value in this case. Do not define this macro if it would always return
7273 @code{NIL}. On machines where this macro is defined, you will normally
7274 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7276 @findex SHORT_IMMEDIATES_SIGN_EXTEND
7277 @item SHORT_IMMEDIATES_SIGN_EXTEND
7278 Define this macro if loading short immediate values into registers sign
7281 @findex IMPLICIT_FIX_EXPR
7282 @item IMPLICIT_FIX_EXPR
7283 An alias for a tree code that should be used by default for conversion
7284 of floating point values to fixed point. Normally,
7285 @code{FIX_ROUND_EXPR} is used.@refill
7287 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
7288 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
7289 Define this macro if the same instructions that convert a floating
7290 point number to a signed fixed point number also convert validly to an
7293 @findex EASY_DIV_EXPR
7295 An alias for a tree code that is the easiest kind of division to
7296 compile code for in the general case. It may be
7297 @code{TRUNC_DIV_EXPR}, @code{FLOOR_DIV_EXPR}, @code{CEIL_DIV_EXPR} or
7298 @code{ROUND_DIV_EXPR}. These four division operators differ in how
7299 they round the result to an integer. @code{EASY_DIV_EXPR} is used
7300 when it is permissible to use any of those kinds of division and the
7301 choice should be made on the basis of efficiency.@refill
7305 The maximum number of bytes that a single instruction can move quickly
7306 between memory and registers or between two memory locations.
7308 @findex MAX_MOVE_MAX
7310 The maximum number of bytes that a single instruction can move quickly
7311 between memory and registers or between two memory locations. If this
7312 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7313 constant value that is the largest value that @code{MOVE_MAX} can have
7316 @findex SHIFT_COUNT_TRUNCATED
7317 @item SHIFT_COUNT_TRUNCATED
7318 A C expression that is nonzero if on this machine the number of bits
7319 actually used for the count of a shift operation is equal to the number
7320 of bits needed to represent the size of the object being shifted. When
7321 this macro is non-zero, the compiler will assume that it is safe to omit
7322 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7323 truncates the count of a shift operation. On machines that have
7324 instructions that act on bitfields at variable positions, which may
7325 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7326 also enables deletion of truncations of the values that serve as
7327 arguments to bitfield instructions.
7329 If both types of instructions truncate the count (for shifts) and
7330 position (for bitfield operations), or if no variable-position bitfield
7331 instructions exist, you should define this macro.
7333 However, on some machines, such as the 80386 and the 680x0, truncation
7334 only applies to shift operations and not the (real or pretended)
7335 bitfield operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7336 such machines. Instead, add patterns to the @file{md} file that include
7337 the implied truncation of the shift instructions.
7339 You need not define this macro if it would always have the value of zero.
7341 @findex TRULY_NOOP_TRUNCATION
7342 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
7343 A C expression which is nonzero if on this machine it is safe to
7344 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
7345 bits (where @var{outprec} is smaller than @var{inprec}) by merely
7346 operating on it as if it had only @var{outprec} bits.
7348 On many machines, this expression can be 1.
7350 @c rearranged this, removed the phrase "it is reported that". this was
7351 @c to fix an overfull hbox. --mew 10feb93
7352 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
7353 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
7354 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
7355 such cases may improve things.
7357 @findex STORE_FLAG_VALUE
7358 @item STORE_FLAG_VALUE
7359 A C expression describing the value returned by a comparison operator
7360 with an integral mode and stored by a store-flag instruction
7361 (@samp{s@var{cond}}) when the condition is true. This description must
7362 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
7363 comparison operators whose results have a @code{MODE_INT} mode.
7365 A value of 1 or -1 means that the instruction implementing the
7366 comparison operator returns exactly 1 or -1 when the comparison is true
7367 and 0 when the comparison is false. Otherwise, the value indicates
7368 which bits of the result are guaranteed to be 1 when the comparison is
7369 true. This value is interpreted in the mode of the comparison
7370 operation, which is given by the mode of the first operand in the
7371 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
7372 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7375 If @code{STORE_FLAG_VALUE} is neither 1 or -1, the compiler will
7376 generate code that depends only on the specified bits. It can also
7377 replace comparison operators with equivalent operations if they cause
7378 the required bits to be set, even if the remaining bits are undefined.
7379 For example, on a machine whose comparison operators return an
7380 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7381 @samp{0x80000000}, saying that just the sign bit is relevant, the
7385 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7392 (ashift:SI @var{x} (const_int @var{n}))
7396 where @var{n} is the appropriate shift count to move the bit being
7397 tested into the sign bit.
7399 There is no way to describe a machine that always sets the low-order bit
7400 for a true value, but does not guarantee the value of any other bits,
7401 but we do not know of any machine that has such an instruction. If you
7402 are trying to port GCC to such a machine, include an instruction to
7403 perform a logical-and of the result with 1 in the pattern for the
7404 comparison operators and let us know
7406 (@pxref{Bug Reporting,,How to Report Bugs}).
7409 (@pxref{Bug Reporting,,How to Report Bugs,gcc.info,Using GCC}).
7412 Often, a machine will have multiple instructions that obtain a value
7413 from a comparison (or the condition codes). Here are rules to guide the
7414 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7419 Use the shortest sequence that yields a valid definition for
7420 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7421 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7422 comparison operators to do so because there may be opportunities to
7423 combine the normalization with other operations.
7426 For equal-length sequences, use a value of 1 or -1, with -1 being
7427 slightly preferred on machines with expensive jumps and 1 preferred on
7431 As a second choice, choose a value of @samp{0x80000001} if instructions
7432 exist that set both the sign and low-order bits but do not define the
7436 Otherwise, use a value of @samp{0x80000000}.
7439 Many machines can produce both the value chosen for
7440 @code{STORE_FLAG_VALUE} and its negation in the same number of
7441 instructions. On those machines, you should also define a pattern for
7442 those cases, e.g., one matching
7445 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7448 Some machines can also perform @code{and} or @code{plus} operations on
7449 condition code values with less instructions than the corresponding
7450 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
7451 machines, define the appropriate patterns. Use the names @code{incscc}
7452 and @code{decscc}, respectively, for the patterns which perform
7453 @code{plus} or @code{minus} operations on condition code values. See
7454 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
7455 find such instruction sequences on other machines.
7457 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7460 @findex FLOAT_STORE_FLAG_VALUE
7461 @item FLOAT_STORE_FLAG_VALUE
7462 A C expression that gives a non-zero floating point value that is
7463 returned when comparison operators with floating-point results are true.
7464 Define this macro on machine that have comparison operations that return
7465 floating-point values. If there are no such operations, do not define
7470 An alias for the machine mode for pointers. On most machines, define
7471 this to be the integer mode corresponding to the width of a hardware
7472 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7473 On some machines you must define this to be one of the partial integer
7474 modes, such as @code{PSImode}.
7476 The width of @code{Pmode} must be at least as large as the value of
7477 @code{POINTER_SIZE}. If it is not equal, you must define the macro
7478 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7481 @findex FUNCTION_MODE
7483 An alias for the machine mode used for memory references to functions
7484 being called, in @code{call} RTL expressions. On most machines this
7485 should be @code{QImode}.
7487 @findex INTEGRATE_THRESHOLD
7488 @item INTEGRATE_THRESHOLD (@var{decl})
7489 A C expression for the maximum number of instructions above which the
7490 function @var{decl} should not be inlined. @var{decl} is a
7491 @code{FUNCTION_DECL} node.
7493 The default definition of this macro is 64 plus 8 times the number of
7494 arguments that the function accepts. Some people think a larger
7495 threshold should be used on RISC machines.
7497 @findex SCCS_DIRECTIVE
7498 @item SCCS_DIRECTIVE
7499 Define this if the preprocessor should ignore @code{#sccs} directives
7500 and print no error message.
7502 @findex NO_IMPLICIT_EXTERN_C
7503 @item NO_IMPLICIT_EXTERN_C
7504 Define this macro if the system header files support C++ as well as C.
7505 This macro inhibits the usual method of using system header files in
7506 C++, which is to pretend that the file's contents are enclosed in
7507 @samp{extern "C" @{@dots{}@}}.
7509 @findex HANDLE_PRAGMA
7512 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
7513 Define this macro if you want to implement any pragmas. If defined, it
7514 is a C expression whose value is 1 if the pragma was handled by the
7515 macro, zero otherwise. The argument @var{getc} is a function of type
7516 @samp{int (*)(void)} which will return the next character in the input
7517 stream, or EOF if no characters are left. The argument @var{ungetc} is
7518 a function of type @samp{void (*)(int)} which will push a character back
7519 into the input stream. The argument @var{name} is the word following
7520 #pragma in the input stream. The input stream pointer will be pointing
7521 just beyond the end of this word. The input stream should be left
7522 undistrubed if the expression returns zero, otherwise it should be
7523 pointing at the next character after the end of the pragma. Any
7524 characters remaining on the line will be ignored.
7526 It is generally a bad idea to implement new uses of @code{#pragma}. The
7527 only reason to define this macro is for compatibility with other
7528 compilers that do support @code{#pragma} for the sake of any user
7529 programs which already use it.
7531 If the pragma can be implemented by atttributes then the macro
7532 @samp{INSERT_ATTRIBUTES} might be a useful one to define as well.
7534 Note: older versions of this macro only had two arguments: @var{stream}
7535 and @var{token}. The macro was changed in order to allow it to work
7536 when gcc is built both with and without a cpp library.
7538 @findex HANDLE_SYSV_PRAGMA
7541 @item HANDLE_SYSV_PRAGMA
7542 Define this macro (to a value of 1) if you want the System V style
7543 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
7544 [=<value>]} to be supported by gcc.
7546 The pack pragma specifies the maximum alignment (in bytes) of fields
7547 within a structure, in much the same way as the @samp{__aligned__} and
7548 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
7549 the behaviour to the default.
7551 The weak pragma only works if @code{SUPPORTS_WEAK} and
7552 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
7553 of specifically named weak labels, optionally with a value.
7555 @findex HANDLE_PRAGMA_PACK_PUSH_POP
7558 @item HANDLE_PRAGMA_PACK_PUSH_POP
7559 Define this macro (to a value of 1) if you want to support the Win32
7560 style pragmas @samp{#pragma pack(push,<n>)} and @samp{#pragma
7561 pack(pop)}. The pack(push,<n>) pragma specifies the maximum alignment
7562 (in bytes) of fields within a structure, in much the same way as the
7563 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
7564 pack value of zero resets the behaviour to the default. Successive
7565 invocations of this pragma cause the previous values to be stacked, so
7566 that invocations of @samp{#pragma pack(pop)} will return to the previous
7569 @findex VALID_MACHINE_DECL_ATTRIBUTE
7570 @item VALID_MACHINE_DECL_ATTRIBUTE (@var{decl}, @var{attributes}, @var{identifier}, @var{args})
7571 If defined, a C expression whose value is nonzero if @var{identifier} with
7572 arguments @var{args} is a valid machine specific attribute for @var{decl}.
7573 The attributes in @var{attributes} have previously been assigned to @var{decl}.
7575 @findex VALID_MACHINE_TYPE_ATTRIBUTE
7576 @item VALID_MACHINE_TYPE_ATTRIBUTE (@var{type}, @var{attributes}, @var{identifier}, @var{args})
7577 If defined, a C expression whose value is nonzero if @var{identifier} with
7578 arguments @var{args} is a valid machine specific attribute for @var{type}.
7579 The attributes in @var{attributes} have previously been assigned to @var{type}.
7581 @findex COMP_TYPE_ATTRIBUTES
7582 @item COMP_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
7583 If defined, a C expression whose value is zero if the attributes on
7584 @var{type1} and @var{type2} are incompatible, one if they are compatible,
7585 and two if they are nearly compatible (which causes a warning to be
7588 @findex SET_DEFAULT_TYPE_ATTRIBUTES
7589 @item SET_DEFAULT_TYPE_ATTRIBUTES (@var{type})
7590 If defined, a C statement that assigns default attributes to
7591 newly defined @var{type}.
7593 @findex MERGE_MACHINE_TYPE_ATTRIBUTES
7594 @item MERGE_MACHINE_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
7595 Define this macro if the merging of type attributes needs special handling.
7596 If defined, the result is a list of the combined TYPE_ATTRIBUTES of
7597 @var{type1} and @var{type2}. It is assumed that comptypes has already been
7598 called and returned 1.
7600 @findex MERGE_MACHINE_DECL_ATTRIBUTES
7601 @item MERGE_MACHINE_DECL_ATTRIBUTES (@var{olddecl}, @var{newdecl})
7602 Define this macro if the merging of decl attributes needs special handling.
7603 If defined, the result is a list of the combined DECL_MACHINE_ATTRIBUTES of
7604 @var{olddecl} and @var{newdecl}. @var{newdecl} is a duplicate declaration
7605 of @var{olddecl}. Examples of when this is needed are when one attribute
7606 overrides another, or when an attribute is nullified by a subsequent
7609 @findex INSERT_ATTRIBUTES
7610 @item INSERT_ATTRIBUTES (@var{node}, @var{attr_ptr}, @var{prefix_ptr})
7611 Define this macro if you want to be able to add attributes to a decl
7612 when it is being created. This is normally useful for backends which
7613 wish to implement a pragma by using the attributes which correspond to
7614 the pragma's effect. The @var{node} argument is the decl which is being
7615 created. The @var{attr_ptr} argument is a pointer to the attribute list
7616 for this decl. The @var{prefix_ptr} is a pointer to the list of
7617 attributes that have appeared after the specifiers and modifiers of the
7618 declaration, but before the declaration proper.
7620 @findex SET_DEFAULT_DECL_ATTRIBUTES
7621 @item SET_DEFAULT_DECL_ATTRIBUTES (@var{decl}, @var{attributes})
7622 If defined, a C statement that assigns default attributes to
7623 newly defined @var{decl}.
7625 @findex DOLLARS_IN_IDENTIFIERS
7626 @item DOLLARS_IN_IDENTIFIERS
7627 Define this macro to control use of the character @samp{$} in identifier
7628 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
7629 1 is the default; there is no need to define this macro in that case.
7630 This macro controls the compiler proper; it does not affect the preprocessor.
7632 @findex NO_DOLLAR_IN_LABEL
7633 @item NO_DOLLAR_IN_LABEL
7634 Define this macro if the assembler does not accept the character
7635 @samp{$} in label names. By default constructors and destructors in
7636 G++ have @samp{$} in the identifiers. If this macro is defined,
7637 @samp{.} is used instead.
7639 @findex NO_DOT_IN_LABEL
7640 @item NO_DOT_IN_LABEL
7641 Define this macro if the assembler does not accept the character
7642 @samp{.} in label names. By default constructors and destructors in G++
7643 have names that use @samp{.}. If this macro is defined, these names
7644 are rewritten to avoid @samp{.}.
7646 @findex DEFAULT_MAIN_RETURN
7647 @item DEFAULT_MAIN_RETURN
7648 Define this macro if the target system expects every program's @code{main}
7649 function to return a standard ``success'' value by default (if no other
7650 value is explicitly returned).
7652 The definition should be a C statement (sans semicolon) to generate the
7653 appropriate rtl instructions. It is used only when compiling the end of
7658 Define this if the target system lacks the function @code{atexit}
7659 from the ANSI C standard. If this macro is defined, a default definition
7660 will be provided to support C++. If @code{ON_EXIT} is not defined,
7661 a default @code{exit} function will also be provided.
7665 Define this macro if the target has another way to implement atexit
7666 functionality without replacing @code{exit}. For instance, SunOS 4 has
7667 a similar @code{on_exit} library function.
7669 The definition should be a functional macro which can be used just like
7670 the @code{atexit} function.
7674 Define this if your @code{exit} function needs to do something
7675 besides calling an external function @code{_cleanup} before
7676 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
7677 only needed if neither @code{HAVE_ATEXIT} nor
7678 @code{INIT_SECTION_ASM_OP} are defined.
7680 @findex INSN_SETS_ARE_DELAYED
7681 @item INSN_SETS_ARE_DELAYED (@var{insn})
7682 Define this macro as a C expression that is nonzero if it is safe for the
7683 delay slot scheduler to place instructions in the delay slot of @var{insn},
7684 even if they appear to use a resource set or clobbered in @var{insn}.
7685 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
7686 every @code{call_insn} has this behavior. On machines where some @code{insn}
7687 or @code{jump_insn} is really a function call and hence has this behavior,
7688 you should define this macro.
7690 You need not define this macro if it would always return zero.
7692 @findex INSN_REFERENCES_ARE_DELAYED
7693 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
7694 Define this macro as a C expression that is nonzero if it is safe for the
7695 delay slot scheduler to place instructions in the delay slot of @var{insn},
7696 even if they appear to set or clobber a resource referenced in @var{insn}.
7697 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
7698 some @code{insn} or @code{jump_insn} is really a function call and its operands
7699 are registers whose use is actually in the subroutine it calls, you should
7700 define this macro. Doing so allows the delay slot scheduler to move
7701 instructions which copy arguments into the argument registers into the delay
7704 You need not define this macro if it would always return zero.
7706 @findex MACHINE_DEPENDENT_REORG
7707 @item MACHINE_DEPENDENT_REORG (@var{insn})
7708 In rare cases, correct code generation requires extra machine
7709 dependent processing between the second jump optimization pass and
7710 delayed branch scheduling. On those machines, define this macro as a C
7711 statement to act on the code starting at @var{insn}.
7713 @findex MULTIPLE_SYMBOL_SPACES
7714 @item MULTIPLE_SYMBOL_SPACES
7715 Define this macro if in some cases global symbols from one translation
7716 unit may not be bound to undefined symbols in another translation unit
7717 without user intervention. For instance, under Microsoft Windows
7718 symbols must be explicitly imported from shared libraries (DLLs).
7720 @findex MD_ASM_CLOBBERS
7721 @item MD_ASM_CLOBBERS
7722 A C statement that adds to @var{CLOBBERS} @code{STRING_CST} trees for
7723 any hard regs the port wishes to automatically clobber for all asms.
7727 A C expression that returns how many instructions can be issued at the
7728 same time if the machine is a superscalar machine. This is only used by
7729 the @samp{Haifa} scheduler, and not the traditional scheduler.
7731 @findex MD_SCHED_INIT
7732 @item MD_SCHED_INIT (@var{file}, @var{verbose})
7733 A C statement which is executed by the @samp{Haifa} scheduler at the
7734 beginning of each block of instructions that are to be scheduled.
7735 @var{file} is either a null pointer, or a stdio stream to write any
7736 debug output to. @var{verbose} is the verbose level provided by
7737 @samp{-fsched-verbose-}@var{n}.
7739 @findex MD_SCHED_REORDER
7740 @item MD_SCHED_REORDER (@var{file}, @var{verbose}, @var{ready}, @var{n_ready},
7741 @var{clock}, @var{can_issue_more})
7742 A C statement which is executed by the @samp{Haifa} scheduler after it
7743 has scheduled the ready list to allow the machine description to reorder
7744 it (for example to combine two small instructions together on
7745 @samp{VLIW} machines). @var{file} is either a null pointer, or a stdio
7746 stream to write any debug output to. @var{verbose} is the verbose level
7747 provided by @samp{-fsched-verbose-}@var{n}. @var{ready} is a pointer to
7748 the ready list of instructions that are ready to be scheduled.
7749 @var{n_ready} is the number of elements in the ready list. The
7750 scheduler reads the ready list in reverse order, starting with
7751 @var{ready}[@var{n_ready}-1] and going to @var{ready}[0]. @var{clock}
7752 is the timer tick of the scheduler. @var{can_issue_more} is an output
7753 parameter that is set to the number of insns that can issue this clock;
7754 normally this is just @code{issue_rate}.
7756 @findex MD_SCHED_VARIABLE_ISSUE
7757 @item MD_SCHED_VARIABLE_ISSUE (@var{file}, @var{verbose}, @var{insn}, @var{more})
7758 A C statement which is executed by the @samp{Haifa} scheduler after it
7759 has scheduled an insn from the ready list. @var{file} is either a null
7760 pointer, or a stdio stream to write any debug output to. @var{verbose}
7761 is the verbose level provided by @samp{-fsched-verbose-}@var{n}.
7762 @var{insn} is the instruction that was scheduled. @var{more} is the
7763 number of instructions that can be issued in the current cycle. The
7764 @samp{MD_SCHED_VARIABLE_ISSUE} macro is responsible for updating the
7765 value of @var{more} (typically by @var{more}--).
7767 @findex MAX_INTEGER_COMPUTATION_MODE
7768 @item MAX_INTEGER_COMPUTATION_MODE
7769 Define this to the largest integer machine mode which can be used for
7770 operations other than load, store and copy operations.
7772 You need only define this macro if the target holds values larger than
7773 @code{word_mode} in general purpose registers. Most targets should not define
7776 @findex MATH_LIBRARY
7778 Define this macro as a C string constant for the linker argument to link
7779 in the system math library, or @samp{""} if the target does not have a
7780 separate math library.
7782 You need only define this macro if the default of @samp{"-lm"} is wrong.