1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001
2 @c Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Target Description Macros
8 @cindex machine description macros
9 @cindex target description macros
10 @cindex macros, target description
11 @cindex @file{tm.h} macros
13 In addition to the file @file{@var{machine}.md}, a machine description
14 includes a C header file conventionally given the name
15 @file{@var{machine}.h}. This header file defines numerous macros
16 that convey the information about the target machine that does not fit
17 into the scheme of the @file{.md} file. The file @file{tm.h} should be
18 a link to @file{@var{machine}.h}. The header file @file{config.h}
19 includes @file{tm.h} and most compiler source files include
23 * Driver:: Controlling how the driver runs the compilation passes.
24 * Run-time Target:: Defining @samp{-m} options like @samp{-m68000} and @samp{-m68020}.
25 * Per-Function Data:: Defining data structures for per-function information.
26 * Storage Layout:: Defining sizes and alignments of data.
27 * Type Layout:: Defining sizes and properties of basic user data types.
28 * Registers:: Naming and describing the hardware registers.
29 * Register Classes:: Defining the classes of hardware registers.
30 * Stack and Calling:: Defining which way the stack grows and by how much.
31 * Varargs:: Defining the varargs macros.
32 * Trampolines:: Code set up at run time to enter a nested function.
33 * Library Calls:: Controlling how library routines are implicitly called.
34 * Addressing Modes:: Defining addressing modes valid for memory operands.
35 * Condition Code:: Defining how insns update the condition code.
36 * Costs:: Defining relative costs of different operations.
37 * Sections:: Dividing storage into text, data, and other sections.
38 * PIC:: Macros for position independent code.
39 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
40 * Debugging Info:: Defining the format of debugging output.
41 * Cross-compilation:: Handling floating point for cross-compilers.
42 * Mode Switching:: Insertion of mode-switching instructions.
43 * Misc:: Everything else.
47 @section Controlling the Compilation Driver, @file{gcc}
49 @cindex controlling the compilation driver
51 @c prevent bad page break with this line
52 You can control the compilation driver.
55 @findex SWITCH_TAKES_ARG
56 @item SWITCH_TAKES_ARG (@var{char})
57 A C expression which determines whether the option @samp{-@var{char}}
58 takes arguments. The value should be the number of arguments that
59 option takes--zero, for many options.
61 By default, this macro is defined as
62 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
63 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
64 wish to add additional options which take arguments. Any redefinition
65 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
68 @findex WORD_SWITCH_TAKES_ARG
69 @item WORD_SWITCH_TAKES_ARG (@var{name})
70 A C expression which determines whether the option @samp{-@var{name}}
71 takes arguments. The value should be the number of arguments that
72 option takes--zero, for many options. This macro rather than
73 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
75 By default, this macro is defined as
76 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
77 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
78 wish to add additional options which take arguments. Any redefinition
79 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
82 @findex SWITCH_CURTAILS_COMPILATION
83 @item SWITCH_CURTAILS_COMPILATION (@var{char})
84 A C expression which determines whether the option @samp{-@var{char}}
85 stops compilation before the generation of an executable. The value is
86 boolean, non-zero if the option does stop an executable from being
87 generated, zero otherwise.
89 By default, this macro is defined as
90 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
91 options properly. You need not define
92 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
93 options which affect the generation of an executable. Any redefinition
94 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
95 for additional options.
97 @findex SWITCHES_NEED_SPACES
98 @item SWITCHES_NEED_SPACES
99 A string-valued C expression which enumerates the options for which
100 the linker needs a space between the option and its argument.
102 If this macro is not defined, the default value is @code{""}.
106 A C string constant that tells the GCC driver program options to
107 pass to CPP. It can also specify how to translate options you
108 give to GCC into options for GCC to pass to the CPP.
110 Do not define this macro if it does not need to do anything.
112 @findex NO_BUILTIN_SIZE_TYPE
113 @item NO_BUILTIN_SIZE_TYPE
114 If this macro is defined, the preprocessor will not define the builtin macro
115 @code{__SIZE_TYPE__}. The macro @code{__SIZE_TYPE__} must then be defined
116 by @code{CPP_SPEC} instead.
118 This should be defined if @code{SIZE_TYPE} depends on target dependent flags
119 which are not accessible to the preprocessor. Otherwise, it should not
122 @findex NO_BUILTIN_PTRDIFF_TYPE
123 @item NO_BUILTIN_PTRDIFF_TYPE
124 If this macro is defined, the preprocessor will not define the builtin macro
125 @code{__PTRDIFF_TYPE__}. The macro @code{__PTRDIFF_TYPE__} must then be
126 defined by @code{CPP_SPEC} instead.
128 This should be defined if @code{PTRDIFF_TYPE} depends on target dependent flags
129 which are not accessible to the preprocessor. Otherwise, it should not
132 @findex NO_BUILTIN_WCHAR_TYPE
133 @item NO_BUILTIN_WCHAR_TYPE
134 If this macro is defined, the preprocessor will not define the builtin macro
135 @code{__WCHAR_TYPE__}. The macro @code{__WCHAR_TYPE__} must then be
136 defined by @code{CPP_SPEC} instead.
138 This should be defined if @code{WCHAR_TYPE} depends on target dependent flags
139 which are not accessible to the preprocessor. Otherwise, it should not
142 @findex NO_BUILTIN_WINT_TYPE
143 @item NO_BUILTIN_WINT_TYPE
144 If this macro is defined, the preprocessor will not define the builtin macro
145 @code{__WINT_TYPE__}. The macro @code{__WINT_TYPE__} must then be
146 defined by @code{CPP_SPEC} instead.
148 This should be defined if @code{WINT_TYPE} depends on target dependent flags
149 which are not accessible to the preprocessor. Otherwise, it should not
152 @findex SIGNED_CHAR_SPEC
153 @item SIGNED_CHAR_SPEC
154 A C string constant that tells the GCC driver program options to
155 pass to CPP. By default, this macro is defined to pass the option
156 @samp{-D__CHAR_UNSIGNED__} to CPP if @code{char} will be treated as
157 @code{unsigned char} by @code{cc1}.
159 Do not define this macro unless you need to override the default
164 A C string constant that tells the GCC driver program options to
165 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
167 It can also specify how to translate options you give to GCC into options
168 for GCC to pass to front ends..
170 Do not define this macro if it does not need to do anything.
174 A C string constant that tells the GCC driver program options to
175 pass to @code{cc1plus}. It can also specify how to translate options you
176 give to GCC into options for GCC to pass to the @code{cc1plus}.
178 Do not define this macro if it does not need to do anything.
179 Note that everything defined in CC1_SPEC is already passed to
180 @code{cc1plus} so there is no need to duplicate the contents of
181 CC1_SPEC in CC1PLUS_SPEC.
185 A C string constant that tells the GCC driver program options to
186 pass to the assembler. It can also specify how to translate options
187 you give to GCC into options for GCC to pass to the assembler.
188 See the file @file{sun3.h} for an example of this.
190 Do not define this macro if it does not need to do anything.
192 @findex ASM_FINAL_SPEC
194 A C string constant that tells the GCC driver program how to
195 run any programs which cleanup after the normal assembler.
196 Normally, this is not needed. See the file @file{mips.h} for
199 Do not define this macro if it does not need to do anything.
203 A C string constant that tells the GCC driver program options to
204 pass to the linker. It can also specify how to translate options you
205 give to GCC into options for GCC to pass to the linker.
207 Do not define this macro if it does not need to do anything.
211 Another C string constant used much like @code{LINK_SPEC}. The difference
212 between the two is that @code{LIB_SPEC} is used at the end of the
213 command given to the linker.
215 If this macro is not defined, a default is provided that
216 loads the standard C library from the usual place. See @file{gcc.c}.
220 Another C string constant that tells the GCC driver program
221 how and when to place a reference to @file{libgcc.a} into the
222 linker command line. This constant is placed both before and after
223 the value of @code{LIB_SPEC}.
225 If this macro is not defined, the GCC driver provides a default that
226 passes the string @samp{-lgcc} to the linker.
228 @findex STARTFILE_SPEC
230 Another C string constant used much like @code{LINK_SPEC}. The
231 difference between the two is that @code{STARTFILE_SPEC} is used at
232 the very beginning of the command given to the linker.
234 If this macro is not defined, a default is provided that loads the
235 standard C startup file from the usual place. See @file{gcc.c}.
239 Another C string constant used much like @code{LINK_SPEC}. The
240 difference between the two is that @code{ENDFILE_SPEC} is used at
241 the very end of the command given to the linker.
243 Do not define this macro if it does not need to do anything.
247 Define this macro to provide additional specifications to put in the
248 @file{specs} file that can be used in various specifications like
251 The definition should be an initializer for an array of structures,
252 containing a string constant, that defines the specification name, and a
253 string constant that provides the specification.
255 Do not define this macro if it does not need to do anything.
257 @code{EXTRA_SPECS} is useful when an architecture contains several
258 related targets, which have various @code{..._SPECS} which are similar
259 to each other, and the maintainer would like one central place to keep
262 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
263 define either @code{_CALL_SYSV} when the System V calling sequence is
264 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
267 The @file{config/rs6000/rs6000.h} target file defines:
270 #define EXTRA_SPECS \
271 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
273 #define CPP_SYS_DEFAULT ""
276 The @file{config/rs6000/sysv.h} target file defines:
280 "%@{posix: -D_POSIX_SOURCE @} \
281 %@{mcall-sysv: -D_CALL_SYSV @} %@{mcall-aix: -D_CALL_AIX @} \
282 %@{!mcall-sysv: %@{!mcall-aix: %(cpp_sysv_default) @}@} \
283 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
285 #undef CPP_SYSV_DEFAULT
286 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
289 while the @file{config/rs6000/eabiaix.h} target file defines
290 @code{CPP_SYSV_DEFAULT} as:
293 #undef CPP_SYSV_DEFAULT
294 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
297 @findex LINK_LIBGCC_SPECIAL
298 @item LINK_LIBGCC_SPECIAL
299 Define this macro if the driver program should find the library
300 @file{libgcc.a} itself and should not pass @samp{-L} options to the
301 linker. If you do not define this macro, the driver program will pass
302 the argument @samp{-lgcc} to tell the linker to do the search and will
303 pass @samp{-L} options to it.
305 @findex LINK_LIBGCC_SPECIAL_1
306 @item LINK_LIBGCC_SPECIAL_1
307 Define this macro if the driver program should find the library
308 @file{libgcc.a}. If you do not define this macro, the driver program will pass
309 the argument @samp{-lgcc} to tell the linker to do the search.
310 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
311 not affect @samp{-L} options.
313 @findex LINK_COMMAND_SPEC
314 @item LINK_COMMAND_SPEC
315 A C string constant giving the complete command line need to execute the
316 linker. When you do this, you will need to update your port each time a
317 change is made to the link command line within @file{gcc.c}. Therefore,
318 define this macro only if you need to completely redefine the command
319 line for invoking the linker and there is no other way to accomplish
322 @findex MULTILIB_DEFAULTS
323 @item MULTILIB_DEFAULTS
324 Define this macro as a C expression for the initializer of an array of
325 string to tell the driver program which options are defaults for this
326 target and thus do not need to be handled specially when using
327 @code{MULTILIB_OPTIONS}.
329 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
330 the target makefile fragment or if none of the options listed in
331 @code{MULTILIB_OPTIONS} are set by default.
332 @xref{Target Fragment}.
334 @findex RELATIVE_PREFIX_NOT_LINKDIR
335 @item RELATIVE_PREFIX_NOT_LINKDIR
336 Define this macro to tell @code{gcc} that it should only translate
337 a @samp{-B} prefix into a @samp{-L} linker option if the prefix
338 indicates an absolute file name.
340 @findex STANDARD_EXEC_PREFIX
341 @item STANDARD_EXEC_PREFIX
342 Define this macro as a C string constant if you wish to override the
343 standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
344 try when searching for the executable files of the compiler.
346 @findex MD_EXEC_PREFIX
348 If defined, this macro is an additional prefix to try after
349 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
350 when the @samp{-b} option is used, or the compiler is built as a cross
351 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
352 to the list of directories used to find the assembler in @file{configure.in}.
354 @findex STANDARD_STARTFILE_PREFIX
355 @item STANDARD_STARTFILE_PREFIX
356 Define this macro as a C string constant if you wish to override the
357 standard choice of @file{/usr/local/lib/} as the default prefix to
358 try when searching for startup files such as @file{crt0.o}.
360 @findex MD_STARTFILE_PREFIX
361 @item MD_STARTFILE_PREFIX
362 If defined, this macro supplies an additional prefix to try after the
363 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
364 @samp{-b} option is used, or when the compiler is built as a cross
367 @findex MD_STARTFILE_PREFIX_1
368 @item MD_STARTFILE_PREFIX_1
369 If defined, this macro supplies yet another prefix to try after the
370 standard prefixes. It is not searched when the @samp{-b} option is
371 used, or when the compiler is built as a cross compiler.
373 @findex INIT_ENVIRONMENT
374 @item INIT_ENVIRONMENT
375 Define this macro as a C string constant if you wish to set environment
376 variables for programs called by the driver, such as the assembler and
377 loader. The driver passes the value of this macro to @code{putenv} to
378 initialize the necessary environment variables.
380 @findex LOCAL_INCLUDE_DIR
381 @item LOCAL_INCLUDE_DIR
382 Define this macro as a C string constant if you wish to override the
383 standard choice of @file{/usr/local/include} as the default prefix to
384 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
385 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
387 Cross compilers do not use this macro and do not search either
388 @file{/usr/local/include} or its replacement.
390 @findex MODIFY_TARGET_NAME
391 @item MODIFY_TARGET_NAME
392 Define this macro if you with to define command-line switches that modify the
395 For each switch, you can include a string to be appended to the first
396 part of the configuration name or a string to be deleted from the
397 configuration name, if present. The definition should be an initializer
398 for an array of structures. Each array element should have three
399 elements: the switch name (a string constant, including the initial
400 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
401 indicate whether the string should be inserted or deleted, and the string
402 to be inserted or deleted (a string constant).
404 For example, on a machine where @samp{64} at the end of the
405 configuration name denotes a 64-bit target and you want the @samp{-32}
406 and @samp{-64} switches to select between 32- and 64-bit targets, you would
410 #define MODIFY_TARGET_NAME \
411 @{ @{ "-32", DELETE, "64"@}, \
412 @{"-64", ADD, "64"@}@}
416 @findex SYSTEM_INCLUDE_DIR
417 @item SYSTEM_INCLUDE_DIR
418 Define this macro as a C string constant if you wish to specify a
419 system-specific directory to search for header files before the standard
420 directory. @code{SYSTEM_INCLUDE_DIR} comes before
421 @code{STANDARD_INCLUDE_DIR} in the search order.
423 Cross compilers do not use this macro and do not search the directory
426 @findex STANDARD_INCLUDE_DIR
427 @item STANDARD_INCLUDE_DIR
428 Define this macro as a C string constant if you wish to override the
429 standard choice of @file{/usr/include} as the default prefix to
430 try when searching for header files.
432 Cross compilers do not use this macro and do not search either
433 @file{/usr/include} or its replacement.
435 @findex STANDARD_INCLUDE_COMPONENT
436 @item STANDARD_INCLUDE_COMPONENT
437 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
438 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
439 If you do not define this macro, no component is used.
441 @findex INCLUDE_DEFAULTS
442 @item INCLUDE_DEFAULTS
443 Define this macro if you wish to override the entire default search path
444 for include files. For a native compiler, the default search path
445 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
446 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
447 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
448 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
449 and specify private search areas for GCC. The directory
450 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
452 The definition should be an initializer for an array of structures.
453 Each array element should have four elements: the directory name (a
454 string constant), the component name (also a string constant), a flag
455 for C++-only directories,
456 and a flag showing that the includes in the directory don't need to be
457 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
458 the array with a null element.
460 The component name denotes what GNU package the include file is part of,
461 if any, in all upper-case letters. For example, it might be @samp{GCC}
462 or @samp{BINUTILS}. If the package is part of a vendor-supplied
463 operating system, code the component name as @samp{0}.
465 For example, here is the definition used for VAX/VMS:
468 #define INCLUDE_DEFAULTS \
470 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
471 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
472 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
479 Here is the order of prefixes tried for exec files:
483 Any prefixes specified by the user with @samp{-B}.
486 The environment variable @code{GCC_EXEC_PREFIX}, if any.
489 The directories specified by the environment variable @code{COMPILER_PATH}.
492 The macro @code{STANDARD_EXEC_PREFIX}.
495 @file{/usr/lib/gcc/}.
498 The macro @code{MD_EXEC_PREFIX}, if any.
501 Here is the order of prefixes tried for startfiles:
505 Any prefixes specified by the user with @samp{-B}.
508 The environment variable @code{GCC_EXEC_PREFIX}, if any.
511 The directories specified by the environment variable @code{LIBRARY_PATH}
512 (or port-specific name; native only, cross compilers do not use this).
515 The macro @code{STANDARD_EXEC_PREFIX}.
518 @file{/usr/lib/gcc/}.
521 The macro @code{MD_EXEC_PREFIX}, if any.
524 The macro @code{MD_STARTFILE_PREFIX}, if any.
527 The macro @code{STANDARD_STARTFILE_PREFIX}.
536 @node Run-time Target
537 @section Run-time Target Specification
538 @cindex run-time target specification
539 @cindex predefined macros
540 @cindex target specifications
542 @c prevent bad page break with this line
543 Here are run-time target specifications.
546 @findex CPP_PREDEFINES
548 Define this to be a string constant containing @samp{-D} options to
549 define the predefined macros that identify this machine and system.
550 These macros will be predefined unless the @option{-ansi} option (or a
551 @option{-std} option for strict ISO C conformance) is specified.
553 In addition, a parallel set of macros are predefined, whose names are
554 made by appending @samp{__} at the beginning and at the end. These
555 @samp{__} macros are permitted by the ISO standard, so they are
556 predefined regardless of whether @option{-ansi} or a @option{-std} option
559 For example, on the Sun, one can use the following value:
562 "-Dmc68000 -Dsun -Dunix"
565 The result is to define the macros @code{__mc68000__}, @code{__sun__}
566 and @code{__unix__} unconditionally, and the macros @code{mc68000},
567 @code{sun} and @code{unix} provided @samp{-ansi} is not specified.
569 @findex extern int target_flags
570 @item extern int target_flags;
571 This declaration should be present.
573 @cindex optional hardware or system features
574 @cindex features, optional, in system conventions
576 This series of macros is to allow compiler command arguments to
577 enable or disable the use of optional features of the target machine.
578 For example, one machine description serves both the 68000 and
579 the 68020; a command argument tells the compiler whether it should
580 use 68020-only instructions or not. This command argument works
581 by means of a macro @code{TARGET_68020} that tests a bit in
584 Define a macro @code{TARGET_@var{featurename}} for each such option.
585 Its definition should test a bit in @code{target_flags}. It is
586 recommended that a helper macro @code{TARGET_MASK_@var{featurename}}
587 is defined for each bit-value to test, and used in
588 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
592 #define TARGET_MASK_68020 1
593 #define TARGET_68020 (target_flags & TARGET_MASK_68020)
596 One place where these macros are used is in the condition-expressions
597 of instruction patterns. Note how @code{TARGET_68020} appears
598 frequently in the 68000 machine description file, @file{m68k.md}.
599 Another place they are used is in the definitions of the other
600 macros in the @file{@var{machine}.h} file.
602 @findex TARGET_SWITCHES
603 @item TARGET_SWITCHES
604 This macro defines names of command options to set and clear
605 bits in @code{target_flags}. Its definition is an initializer
606 with a subgrouping for each command option.
608 Each subgrouping contains a string constant, that defines the option
609 name, a number, which contains the bits to set in
610 @code{target_flags}, and a second string which is the description
611 displayed by --help. If the number is negative then the bits specified
612 by the number are cleared instead of being set. If the description
613 string is present but empty, then no help information will be displayed
614 for that option, but it will not count as an undocumented option. The
615 actual option name is made by appending @samp{-m} to the specified name.
617 One of the subgroupings should have a null string. The number in
618 this grouping is the default value for @code{target_flags}. Any
619 target options act starting with that value.
621 Here is an example which defines @samp{-m68000} and @samp{-m68020}
622 with opposite meanings, and picks the latter as the default:
625 #define TARGET_SWITCHES \
626 @{ @{ "68020", TARGET_MASK_68020, "" @}, \
627 @{ "68000", -TARGET_MASK_68020, "Compile for the 68000" @}, \
628 @{ "", TARGET_MASK_68020, "" @}@}
631 @findex TARGET_OPTIONS
633 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
634 options that have values. Its definition is an initializer with a
635 subgrouping for each command option.
637 Each subgrouping contains a string constant, that defines the fixed part
638 of the option name, the address of a variable, and a description string.
639 The variable, type @code{char *}, is set to the variable part of the
640 given option if the fixed part matches. The actual option name is made
641 by appending @samp{-m} to the specified name.
643 Here is an example which defines @samp{-mshort-data-@var{number}}. If the
644 given option is @samp{-mshort-data-512}, the variable @code{m88k_short_data}
645 will be set to the string @code{"512"}.
648 extern char *m88k_short_data;
649 #define TARGET_OPTIONS \
650 @{ @{ "short-data-", &m88k_short_data, "Specify the size of the short data section" @} @}
653 @findex TARGET_VERSION
655 This macro is a C statement to print on @code{stderr} a string
656 describing the particular machine description choice. Every machine
657 description should define @code{TARGET_VERSION}. For example:
661 #define TARGET_VERSION \
662 fprintf (stderr, " (68k, Motorola syntax)");
664 #define TARGET_VERSION \
665 fprintf (stderr, " (68k, MIT syntax)");
669 @findex OVERRIDE_OPTIONS
670 @item OVERRIDE_OPTIONS
671 Sometimes certain combinations of command options do not make sense on
672 a particular target machine. You can define a macro
673 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
674 defined, is executed once just after all the command options have been
677 Don't use this macro to turn on various extra optimizations for
678 @samp{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
680 @findex OPTIMIZATION_OPTIONS
681 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
682 Some machines may desire to change what optimizations are performed for
683 various optimization levels. This macro, if defined, is executed once
684 just after the optimization level is determined and before the remainder
685 of the command options have been parsed. Values set in this macro are
686 used as the default values for the other command line options.
688 @var{level} is the optimization level specified; 2 if @samp{-O2} is
689 specified, 1 if @samp{-O} is specified, and 0 if neither is specified.
691 @var{size} is non-zero if @samp{-Os} is specified and zero otherwise.
693 You should not use this macro to change options that are not
694 machine-specific. These should uniformly selected by the same
695 optimization level on all supported machines. Use this macro to enable
696 machine-specific optimizations.
698 @strong{Do not examine @code{write_symbols} in
699 this macro!} The debugging options are not supposed to alter the
702 @findex CAN_DEBUG_WITHOUT_FP
703 @item CAN_DEBUG_WITHOUT_FP
704 Define this macro if debugging can be performed even without a frame
705 pointer. If this macro is defined, GCC will turn on the
706 @samp{-fomit-frame-pointer} option whenever @samp{-O} is specified.
709 @node Per-Function Data
710 @section Defining data structures for per-function information.
711 @cindex per-function data
712 @cindex data structures
714 If the target needs to store information on a per-function basis, GCC
715 provides a macro and a couple of variables to allow this. Note, just
716 using statics to store the information is a bad idea, since GCC supports
717 nested functions, so you can be halfway through encoding one function
718 when another one comes along.
720 GCC defines a data structure called @code{struct function} which
721 contains all of the data specific to an individual function. This
722 structure contains a field called @code{machine} whose type is
723 @code{struct machine_function *}, which can be used by targets to point
724 to their own specific data.
726 If a target needs per-function specific data it should define the type
727 @code{struct machine_function} and also the macro
728 @code{INIT_EXPANDERS}. This macro should be used to initialise some or
729 all of the function pointers @code{init_machine_status},
730 @code{free_machine_status} and @code{mark_machine_status}. These
731 pointers are explained below.
733 One typical use of per-function, target specific data is to create an
734 RTX to hold the register containing the function's return address. This
735 RTX can then be used to implement the @code{__builtin_return_address}
736 function, for level 0.
738 Note - earlier implementations of GCC used a single data area to hold
739 all of the per-function information. Thus when processing of a nested
740 function began the old per-function data had to be pushed onto a
741 stack, and when the processing was finished, it had to be popped off the
742 stack. GCC used to provide function pointers called
743 @code{save_machine_status} and @code{restore_machine_status} to handle
744 the saving and restoring of the target specific information. Since the
745 single data area approach is no longer used, these pointers are no
748 The macro and function pointers are described below.
751 @findex INIT_EXPANDERS
753 Macro called to initialise any target specific information. This macro
754 is called once per function, before generation of any RTL has begun.
755 The intention of this macro is to allow the initialisation of the
756 function pointers below.
758 @findex init_machine_status
759 @item init_machine_status
760 This is a @code{void (*)(struct function *)} function pointer. If this
761 pointer is non-NULL it will be called once per function, before function
762 compilation starts, in order to allow the target to perform any target
763 specific initialisation of the @code{struct function} structure. It is
764 intended that this would be used to initialise the @code{machine} of
767 @findex free_machine_status
768 @item free_machine_status
769 This is a @code{void (*)(struct function *)} function pointer. If this
770 pointer is non-NULL it will be called once per function, after the
771 function has been compiled, in order to allow any memory allocated
772 during the @code{init_machine_status} function call to be freed.
774 @findex mark_machine_status
775 @item mark_machine_status
776 This is a @code{void (*)(struct function *)} function pointer. If this
777 pointer is non-NULL it will be called once per function in order to mark
778 any data items in the @code{struct machine_function} structure which
779 need garbage collection.
784 @section Storage Layout
785 @cindex storage layout
787 Note that the definitions of the macros in this table which are sizes or
788 alignments measured in bits do not need to be constant. They can be C
789 expressions that refer to static variables, such as the @code{target_flags}.
790 @xref{Run-time Target}.
793 @findex BITS_BIG_ENDIAN
794 @item BITS_BIG_ENDIAN
795 Define this macro to have the value 1 if the most significant bit in a
796 byte has the lowest number; otherwise define it to have the value zero.
797 This means that bit-field instructions count from the most significant
798 bit. If the machine has no bit-field instructions, then this must still
799 be defined, but it doesn't matter which value it is defined to. This
800 macro need not be a constant.
802 This macro does not affect the way structure fields are packed into
803 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
805 @findex BYTES_BIG_ENDIAN
806 @item BYTES_BIG_ENDIAN
807 Define this macro to have the value 1 if the most significant byte in a
808 word has the lowest number. This macro need not be a constant.
810 @findex WORDS_BIG_ENDIAN
811 @item WORDS_BIG_ENDIAN
812 Define this macro to have the value 1 if, in a multiword object, the
813 most significant word has the lowest number. This applies to both
814 memory locations and registers; GCC fundamentally assumes that the
815 order of words in memory is the same as the order in registers. This
816 macro need not be a constant.
818 @findex LIBGCC2_WORDS_BIG_ENDIAN
819 @item LIBGCC2_WORDS_BIG_ENDIAN
820 Define this macro if WORDS_BIG_ENDIAN is not constant. This must be a
821 constant value with the same meaning as WORDS_BIG_ENDIAN, which will be
822 used only when compiling libgcc2.c. Typically the value will be set
823 based on preprocessor defines.
825 @findex FLOAT_WORDS_BIG_ENDIAN
826 @item FLOAT_WORDS_BIG_ENDIAN
827 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
828 @code{TFmode} floating point numbers are stored in memory with the word
829 containing the sign bit at the lowest address; otherwise define it to
830 have the value 0. This macro need not be a constant.
832 You need not define this macro if the ordering is the same as for
835 @findex BITS_PER_UNIT
837 Define this macro to be the number of bits in an addressable storage
838 unit (byte); normally 8.
840 @findex BITS_PER_WORD
842 Number of bits in a word; normally 32.
844 @findex MAX_BITS_PER_WORD
845 @item MAX_BITS_PER_WORD
846 Maximum number of bits in a word. If this is undefined, the default is
847 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
848 largest value that @code{BITS_PER_WORD} can have at run-time.
850 @findex UNITS_PER_WORD
852 Number of storage units in a word; normally 4.
854 @findex MIN_UNITS_PER_WORD
855 @item MIN_UNITS_PER_WORD
856 Minimum number of units in a word. If this is undefined, the default is
857 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
858 smallest value that @code{UNITS_PER_WORD} can have at run-time.
862 Width of a pointer, in bits. You must specify a value no wider than the
863 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
864 you must define @code{POINTERS_EXTEND_UNSIGNED}.
866 @findex POINTERS_EXTEND_UNSIGNED
867 @item POINTERS_EXTEND_UNSIGNED
868 A C expression whose value is nonzero if pointers that need to be
869 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
870 be zero-extended and zero if they are to be sign-extended.
872 You need not define this macro if the @code{POINTER_SIZE} is equal
873 to the width of @code{Pmode}.
876 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
877 A macro to update @var{m} and @var{unsignedp} when an object whose type
878 is @var{type} and which has the specified mode and signedness is to be
879 stored in a register. This macro is only called when @var{type} is a
882 On most RISC machines, which only have operations that operate on a full
883 register, define this macro to set @var{m} to @code{word_mode} if
884 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
885 cases, only integer modes should be widened because wider-precision
886 floating-point operations are usually more expensive than their narrower
889 For most machines, the macro definition does not change @var{unsignedp}.
890 However, some machines, have instructions that preferentially handle
891 either signed or unsigned quantities of certain modes. For example, on
892 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
893 sign-extend the result to 64 bits. On such machines, set
894 @var{unsignedp} according to which kind of extension is more efficient.
896 Do not define this macro if it would never modify @var{m}.
898 @findex PROMOTE_FUNCTION_ARGS
899 @item PROMOTE_FUNCTION_ARGS
900 Define this macro if the promotion described by @code{PROMOTE_MODE}
901 should also be done for outgoing function arguments.
903 @findex PROMOTE_FUNCTION_RETURN
904 @item PROMOTE_FUNCTION_RETURN
905 Define this macro if the promotion described by @code{PROMOTE_MODE}
906 should also be done for the return value of functions.
908 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
909 promotions done by @code{PROMOTE_MODE}.
911 @findex PROMOTE_FOR_CALL_ONLY
912 @item PROMOTE_FOR_CALL_ONLY
913 Define this macro if the promotion described by @code{PROMOTE_MODE}
914 should @emph{only} be performed for outgoing function arguments or
915 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
916 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
918 @findex PARM_BOUNDARY
920 Normal alignment required for function parameters on the stack, in
921 bits. All stack parameters receive at least this much alignment
922 regardless of data type. On most machines, this is the same as the
925 @findex STACK_BOUNDARY
927 Define this macro if there is a guaranteed alignment for the stack
928 pointer on this machine. The definition is a C expression
929 for the desired alignment (measured in bits). This value is used as a
930 default if PREFERRED_STACK_BOUNDARY is not defined.
932 @findex PREFERRED_STACK_BOUNDARY
933 @item PREFERRED_STACK_BOUNDARY
934 Define this macro if you wish to preserve a certain alignment for
935 the stack pointer. The definition is a C expression
936 for the desired alignment (measured in bits). If STACK_BOUNDARY is
937 also defined, this macro must evaluate to a value equal to or larger
940 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
941 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
942 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
943 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
944 be momentarily unaligned while pushing arguments.
946 @findex FUNCTION_BOUNDARY
947 @item FUNCTION_BOUNDARY
948 Alignment required for a function entry point, in bits.
950 @findex BIGGEST_ALIGNMENT
951 @item BIGGEST_ALIGNMENT
952 Biggest alignment that any data type can require on this machine, in bits.
954 @findex MINIMUM_ATOMIC_ALIGNMENT
955 @item MINIMUM_ATOMIC_ALIGNMENT
956 If defined, the smallest alignment, in bits, that can be given to an
957 object that can be referenced in one operation, without disturbing any
958 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
959 on machines that don't have byte or half-word store operations.
961 @findex BIGGEST_FIELD_ALIGNMENT
962 @item BIGGEST_FIELD_ALIGNMENT
963 Biggest alignment that any structure or union field can require on this
964 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
965 structure and union fields only, unless the field alignment has been set
966 by the @code{__attribute__ ((aligned (@var{n})))} construct.
968 @findex ADJUST_FIELD_ALIGN
969 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
970 An expression for the alignment of a structure field @var{field} if the
971 alignment computed in the usual way is @var{computed}. GCC uses
972 this value instead of the value in @code{BIGGEST_ALIGNMENT} or
973 @code{BIGGEST_FIELD_ALIGNMENT}, if defined, for structure fields only.
975 @findex MAX_OFILE_ALIGNMENT
976 @item MAX_OFILE_ALIGNMENT
977 Biggest alignment supported by the object file format of this machine.
978 Use this macro to limit the alignment which can be specified using the
979 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
980 the default value is @code{BIGGEST_ALIGNMENT}.
982 @findex DATA_ALIGNMENT
983 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
984 If defined, a C expression to compute the alignment for a variable in
985 the static store. @var{type} is the data type, and @var{basic-align} is
986 the alignment that the object would ordinarily have. The value of this
987 macro is used instead of that alignment to align the object.
989 If this macro is not defined, then @var{basic-align} is used.
992 One use of this macro is to increase alignment of medium-size data to
993 make it all fit in fewer cache lines. Another is to cause character
994 arrays to be word-aligned so that @code{strcpy} calls that copy
995 constants to character arrays can be done inline.
997 @findex CONSTANT_ALIGNMENT
998 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
999 If defined, a C expression to compute the alignment given to a constant
1000 that is being placed in memory. @var{constant} is the constant and
1001 @var{basic-align} is the alignment that the object would ordinarily
1002 have. The value of this macro is used instead of that alignment to
1005 If this macro is not defined, then @var{basic-align} is used.
1007 The typical use of this macro is to increase alignment for string
1008 constants to be word aligned so that @code{strcpy} calls that copy
1009 constants can be done inline.
1011 @findex LOCAL_ALIGNMENT
1012 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1013 If defined, a C expression to compute the alignment for a variable in
1014 the local store. @var{type} is the data type, and @var{basic-align} is
1015 the alignment that the object would ordinarily have. The value of this
1016 macro is used instead of that alignment to align the object.
1018 If this macro is not defined, then @var{basic-align} is used.
1020 One use of this macro is to increase alignment of medium-size data to
1021 make it all fit in fewer cache lines.
1023 @findex EMPTY_FIELD_BOUNDARY
1024 @item EMPTY_FIELD_BOUNDARY
1025 Alignment in bits to be given to a structure bit field that follows an
1026 empty field such as @code{int : 0;}.
1028 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
1029 that results from an empty field.
1031 @findex STRUCTURE_SIZE_BOUNDARY
1032 @item STRUCTURE_SIZE_BOUNDARY
1033 Number of bits which any structure or union's size must be a multiple of.
1034 Each structure or union's size is rounded up to a multiple of this.
1036 If you do not define this macro, the default is the same as
1037 @code{BITS_PER_UNIT}.
1039 @findex STRICT_ALIGNMENT
1040 @item STRICT_ALIGNMENT
1041 Define this macro to be the value 1 if instructions will fail to work
1042 if given data not on the nominal alignment. If instructions will merely
1043 go slower in that case, define this macro as 0.
1045 @findex PCC_BITFIELD_TYPE_MATTERS
1046 @item PCC_BITFIELD_TYPE_MATTERS
1047 Define this if you wish to imitate the way many other C compilers handle
1048 alignment of bitfields and the structures that contain them.
1050 The behavior is that the type written for a bitfield (@code{int},
1051 @code{short}, or other integer type) imposes an alignment for the
1052 entire structure, as if the structure really did contain an ordinary
1053 field of that type. In addition, the bitfield is placed within the
1054 structure so that it would fit within such a field, not crossing a
1057 Thus, on most machines, a bitfield whose type is written as @code{int}
1058 would not cross a four-byte boundary, and would force four-byte
1059 alignment for the whole structure. (The alignment used may not be four
1060 bytes; it is controlled by the other alignment parameters.)
1062 If the macro is defined, its definition should be a C expression;
1063 a nonzero value for the expression enables this behavior.
1065 Note that if this macro is not defined, or its value is zero, some
1066 bitfields may cross more than one alignment boundary. The compiler can
1067 support such references if there are @samp{insv}, @samp{extv}, and
1068 @samp{extzv} insns that can directly reference memory.
1070 The other known way of making bitfields work is to define
1071 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1072 Then every structure can be accessed with fullwords.
1074 Unless the machine has bitfield instructions or you define
1075 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1076 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1078 If your aim is to make GCC use the same conventions for laying out
1079 bitfields as are used by another compiler, here is how to investigate
1080 what the other compiler does. Compile and run this program:
1099 printf ("Size of foo1 is %d\n",
1100 sizeof (struct foo1));
1101 printf ("Size of foo2 is %d\n",
1102 sizeof (struct foo2));
1107 If this prints 2 and 5, then the compiler's behavior is what you would
1108 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1110 @findex BITFIELD_NBYTES_LIMITED
1111 @item BITFIELD_NBYTES_LIMITED
1112 Like PCC_BITFIELD_TYPE_MATTERS except that its effect is limited to
1113 aligning a bitfield within the structure.
1115 @findex STRUCT_FORCE_BLK
1116 @item STRUCT_FORCE_BLK (@var{field})
1117 Return 1 if a structure containing @var{field} should be accessed using
1120 Normally, this is not needed. See the file @file{c4x.h} for an example
1121 of how to use this macro to prevent a structure having a floating point
1122 field from being accessed in an integer mode.
1124 @findex ROUND_TYPE_SIZE
1125 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
1126 Define this macro as an expression for the overall size of a type
1127 (given by @var{type} as a tree node) when the size computed in the
1128 usual way is @var{computed} and the alignment is @var{specified}.
1130 The default is to round @var{computed} up to a multiple of @var{specified}.
1132 @findex ROUND_TYPE_SIZE_UNIT
1133 @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified})
1134 Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are
1135 specified in units (bytes). If you define @code{ROUND_TYPE_SIZE},
1136 you must also define this macro and they must be defined consistently
1139 @findex ROUND_TYPE_ALIGN
1140 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1141 Define this macro as an expression for the alignment of a type (given
1142 by @var{type} as a tree node) if the alignment computed in the usual
1143 way is @var{computed} and the alignment explicitly specified was
1146 The default is to use @var{specified} if it is larger; otherwise, use
1147 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1149 @findex MAX_FIXED_MODE_SIZE
1150 @item MAX_FIXED_MODE_SIZE
1151 An integer expression for the size in bits of the largest integer
1152 machine mode that should actually be used. All integer machine modes of
1153 this size or smaller can be used for structures and unions with the
1154 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1155 (DImode)} is assumed.
1157 @findex VECTOR_MODE_SUPPORTED_P
1158 @item VECTOR_MODE_SUPPORTED_P(@var{mode})
1159 Define this macro to be nonzero if the port is prepared to handle insns
1160 involving vector mode @var{mode}. At the very least, it must have move
1161 patterns for this mode.
1163 @findex STACK_SAVEAREA_MODE
1164 @item STACK_SAVEAREA_MODE (@var{save_level})
1165 If defined, an expression of type @code{enum machine_mode} that
1166 specifies the mode of the save area operand of a
1167 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1168 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1169 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1170 having its mode specified.
1172 You need not define this macro if it always returns @code{Pmode}. You
1173 would most commonly define this macro if the
1174 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1177 @findex STACK_SIZE_MODE
1178 @item STACK_SIZE_MODE
1179 If defined, an expression of type @code{enum machine_mode} that
1180 specifies the mode of the size increment operand of an
1181 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1183 You need not define this macro if it always returns @code{word_mode}.
1184 You would most commonly define this macro if the @code{allocate_stack}
1185 pattern needs to support both a 32- and a 64-bit mode.
1187 @findex CHECK_FLOAT_VALUE
1188 @item CHECK_FLOAT_VALUE (@var{mode}, @var{value}, @var{overflow})
1189 A C statement to validate the value @var{value} (of type
1190 @code{double}) for mode @var{mode}. This means that you check whether
1191 @var{value} fits within the possible range of values for mode
1192 @var{mode} on this target machine. The mode @var{mode} is always
1193 a mode of class @code{MODE_FLOAT}. @var{overflow} is nonzero if
1194 the value is already known to be out of range.
1196 If @var{value} is not valid or if @var{overflow} is nonzero, you should
1197 set @var{overflow} to 1 and then assign some valid value to @var{value}.
1198 Allowing an invalid value to go through the compiler can produce
1199 incorrect assembler code which may even cause Unix assemblers to crash.
1201 This macro need not be defined if there is no work for it to do.
1203 @findex TARGET_FLOAT_FORMAT
1204 @item TARGET_FLOAT_FORMAT
1205 A code distinguishing the floating point format of the target machine.
1206 There are three defined values:
1209 @findex IEEE_FLOAT_FORMAT
1210 @item IEEE_FLOAT_FORMAT
1211 This code indicates IEEE floating point. It is the default; there is no
1212 need to define this macro when the format is IEEE.
1214 @findex VAX_FLOAT_FORMAT
1215 @item VAX_FLOAT_FORMAT
1216 This code indicates the peculiar format used on the Vax.
1218 @findex UNKNOWN_FLOAT_FORMAT
1219 @item UNKNOWN_FLOAT_FORMAT
1220 This code indicates any other format.
1223 The value of this macro is compared with @code{HOST_FLOAT_FORMAT}
1224 (@pxref{Config}) to determine whether the target machine has the same
1225 format as the host machine. If any other formats are actually in use on
1226 supported machines, new codes should be defined for them.
1228 The ordering of the component words of floating point values stored in
1229 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN} for the target
1230 machine and @code{HOST_FLOAT_WORDS_BIG_ENDIAN} for the host.
1232 @findex DEFAULT_VTABLE_THUNKS
1233 @item DEFAULT_VTABLE_THUNKS
1234 GCC supports two ways of implementing C++ vtables: traditional or with
1235 so-called ``thunks''. The flag @samp{-fvtable-thunk} chooses between them.
1236 Define this macro to be a C expression for the default value of that flag.
1237 If @code{DEFAULT_VTABLE_THUNKS} is 0, GCC uses the traditional
1238 implementation by default. The ``thunk'' implementation is more efficient
1239 (especially if you have provided an implementation of
1240 @code{ASM_OUTPUT_MI_THUNK}, see @ref{Function Entry}), but is not binary
1241 compatible with code compiled using the traditional implementation.
1242 If you are writing a new port, define @code{DEFAULT_VTABLE_THUNKS} to 1.
1244 If you do not define this macro, the default for @samp{-fvtable-thunk} is 0.
1248 @section Layout of Source Language Data Types
1250 These macros define the sizes and other characteristics of the standard
1251 basic data types used in programs being compiled. Unlike the macros in
1252 the previous section, these apply to specific features of C and related
1253 languages, rather than to fundamental aspects of storage layout.
1256 @findex INT_TYPE_SIZE
1258 A C expression for the size in bits of the type @code{int} on the
1259 target machine. If you don't define this, the default is one word.
1261 @findex MAX_INT_TYPE_SIZE
1262 @item MAX_INT_TYPE_SIZE
1263 Maximum number for the size in bits of the type @code{int} on the target
1264 machine. If this is undefined, the default is @code{INT_TYPE_SIZE}.
1265 Otherwise, it is the constant value that is the largest value that
1266 @code{INT_TYPE_SIZE} can have at run-time. This is used in @code{cpp}.
1268 @findex SHORT_TYPE_SIZE
1269 @item SHORT_TYPE_SIZE
1270 A C expression for the size in bits of the type @code{short} on the
1271 target machine. If you don't define this, the default is half a word.
1272 (If this would be less than one storage unit, it is rounded up to one
1275 @findex LONG_TYPE_SIZE
1276 @item LONG_TYPE_SIZE
1277 A C expression for the size in bits of the type @code{long} on the
1278 target machine. If you don't define this, the default is one word.
1280 @findex MAX_LONG_TYPE_SIZE
1281 @item MAX_LONG_TYPE_SIZE
1282 Maximum number for the size in bits of the type @code{long} on the
1283 target machine. If this is undefined, the default is
1284 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1285 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1288 @findex LONG_LONG_TYPE_SIZE
1289 @item LONG_LONG_TYPE_SIZE
1290 A C expression for the size in bits of the type @code{long long} on the
1291 target machine. If you don't define this, the default is two
1292 words. If you want to support GNU Ada on your machine, the value of this
1293 macro must be at least 64.
1295 @findex CHAR_TYPE_SIZE
1296 @item CHAR_TYPE_SIZE
1297 A C expression for the size in bits of the type @code{char} on the
1298 target machine. If you don't define this, the default is
1299 @code{BITS_PER_UNIT}.
1301 @findex MAX_CHAR_TYPE_SIZE
1302 @item MAX_CHAR_TYPE_SIZE
1303 Maximum number for the size in bits of the type @code{char} on the
1304 target machine. If this is undefined, the default is
1305 @code{CHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1306 largest value that @code{CHAR_TYPE_SIZE} can have at run-time. This is
1309 @findex FLOAT_TYPE_SIZE
1310 @item FLOAT_TYPE_SIZE
1311 A C expression for the size in bits of the type @code{float} on the
1312 target machine. If you don't define this, the default is one word.
1314 @findex DOUBLE_TYPE_SIZE
1315 @item DOUBLE_TYPE_SIZE
1316 A C expression for the size in bits of the type @code{double} on the
1317 target machine. If you don't define this, the default is two
1320 @findex LONG_DOUBLE_TYPE_SIZE
1321 @item LONG_DOUBLE_TYPE_SIZE
1322 A C expression for the size in bits of the type @code{long double} on
1323 the target machine. If you don't define this, the default is two
1326 @findex WIDEST_HARDWARE_FP_SIZE
1327 @item WIDEST_HARDWARE_FP_SIZE
1328 A C expression for the size in bits of the widest floating-point format
1329 supported by the hardware. If you define this macro, you must specify a
1330 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1331 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1334 @findex DEFAULT_SIGNED_CHAR
1335 @item DEFAULT_SIGNED_CHAR
1336 An expression whose value is 1 or 0, according to whether the type
1337 @code{char} should be signed or unsigned by default. The user can
1338 always override this default with the options @samp{-fsigned-char}
1339 and @samp{-funsigned-char}.
1341 @findex DEFAULT_SHORT_ENUMS
1342 @item DEFAULT_SHORT_ENUMS
1343 A C expression to determine whether to give an @code{enum} type
1344 only as many bytes as it takes to represent the range of possible values
1345 of that type. A nonzero value means to do that; a zero value means all
1346 @code{enum} types should be allocated like @code{int}.
1348 If you don't define the macro, the default is 0.
1352 A C expression for a string describing the name of the data type to use
1353 for size values. The typedef name @code{size_t} is defined using the
1354 contents of the string.
1356 The string can contain more than one keyword. If so, separate them with
1357 spaces, and write first any length keyword, then @code{unsigned} if
1358 appropriate, and finally @code{int}. The string must exactly match one
1359 of the data type names defined in the function
1360 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1361 omit @code{int} or change the order---that would cause the compiler to
1364 If you don't define this macro, the default is @code{"long unsigned
1367 @findex PTRDIFF_TYPE
1369 A C expression for a string describing the name of the data type to use
1370 for the result of subtracting two pointers. The typedef name
1371 @code{ptrdiff_t} is defined using the contents of the string. See
1372 @code{SIZE_TYPE} above for more information.
1374 If you don't define this macro, the default is @code{"long int"}.
1378 A C expression for a string describing the name of the data type to use
1379 for wide characters. The typedef name @code{wchar_t} is defined using
1380 the contents of the string. See @code{SIZE_TYPE} above for more
1383 If you don't define this macro, the default is @code{"int"}.
1385 @findex WCHAR_TYPE_SIZE
1386 @item WCHAR_TYPE_SIZE
1387 A C expression for the size in bits of the data type for wide
1388 characters. This is used in @code{cpp}, which cannot make use of
1391 @findex MAX_WCHAR_TYPE_SIZE
1392 @item MAX_WCHAR_TYPE_SIZE
1393 Maximum number for the size in bits of the data type for wide
1394 characters. If this is undefined, the default is
1395 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1396 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1401 A C expression for a string describing the name of the data type to
1402 use for wide characters passed to @code{printf} and returned from
1403 @code{getwc}. The typedef name @code{wint_t} is defined using the
1404 contents of the string. See @code{SIZE_TYPE} above for more
1407 If you don't define this macro, the default is @code{"unsigned int"}.
1411 A C expression for a string describing the name of the data type that
1412 can represent any value of any standard or extended signed integer type.
1413 The typedef name @code{intmax_t} is defined using the contents of the
1414 string. See @code{SIZE_TYPE} above for more information.
1416 If you don't define this macro, the default is the first of
1417 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1418 much precision as @code{long long int}.
1420 @findex UINTMAX_TYPE
1422 A C expression for a string describing the name of the data type that
1423 can represent any value of any standard or extended unsigned integer
1424 type. The typedef name @code{uintmax_t} is defined using the contents
1425 of the string. See @code{SIZE_TYPE} above for more information.
1427 If you don't define this macro, the default is the first of
1428 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1429 unsigned int"} that has as much precision as @code{long long unsigned
1432 @findex OBJC_INT_SELECTORS
1433 @item OBJC_INT_SELECTORS
1434 Define this macro if the type of Objective C selectors should be
1437 If this macro is not defined, then selectors should have the type
1438 @code{struct objc_selector *}.
1440 @findex OBJC_SELECTORS_WITHOUT_LABELS
1441 @item OBJC_SELECTORS_WITHOUT_LABELS
1442 Define this macro if the compiler can group all the selectors together
1443 into a vector and use just one label at the beginning of the vector.
1444 Otherwise, the compiler must give each selector its own assembler
1447 On certain machines, it is important to have a separate label for each
1448 selector because this enables the linker to eliminate duplicate selectors.
1452 A C constant expression for the integer value for escape sequence
1457 @findex TARGET_NEWLINE
1460 @itemx TARGET_NEWLINE
1461 C constant expressions for the integer values for escape sequences
1462 @samp{\b}, @samp{\t} and @samp{\n}.
1470 C constant expressions for the integer values for escape sequences
1471 @samp{\v}, @samp{\f} and @samp{\r}.
1475 @section Register Usage
1476 @cindex register usage
1478 This section explains how to describe what registers the target machine
1479 has, and how (in general) they can be used.
1481 The description of which registers a specific instruction can use is
1482 done with register classes; see @ref{Register Classes}. For information
1483 on using registers to access a stack frame, see @ref{Frame Registers}.
1484 For passing values in registers, see @ref{Register Arguments}.
1485 For returning values in registers, see @ref{Scalar Return}.
1488 * Register Basics:: Number and kinds of registers.
1489 * Allocation Order:: Order in which registers are allocated.
1490 * Values in Registers:: What kinds of values each reg can hold.
1491 * Leaf Functions:: Renumbering registers for leaf functions.
1492 * Stack Registers:: Handling a register stack such as 80387.
1495 @node Register Basics
1496 @subsection Basic Characteristics of Registers
1498 @c prevent bad page break with this line
1499 Registers have various characteristics.
1502 @findex FIRST_PSEUDO_REGISTER
1503 @item FIRST_PSEUDO_REGISTER
1504 Number of hardware registers known to the compiler. They receive
1505 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1506 pseudo register's number really is assigned the number
1507 @code{FIRST_PSEUDO_REGISTER}.
1509 @item FIXED_REGISTERS
1510 @findex FIXED_REGISTERS
1511 @cindex fixed register
1512 An initializer that says which registers are used for fixed purposes
1513 all throughout the compiled code and are therefore not available for
1514 general allocation. These would include the stack pointer, the frame
1515 pointer (except on machines where that can be used as a general
1516 register when no frame pointer is needed), the program counter on
1517 machines where that is considered one of the addressable registers,
1518 and any other numbered register with a standard use.
1520 This information is expressed as a sequence of numbers, separated by
1521 commas and surrounded by braces. The @var{n}th number is 1 if
1522 register @var{n} is fixed, 0 otherwise.
1524 The table initialized from this macro, and the table initialized by
1525 the following one, may be overridden at run time either automatically,
1526 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1527 the user with the command options @samp{-ffixed-@var{reg}},
1528 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}}.
1530 @findex CALL_USED_REGISTERS
1531 @item CALL_USED_REGISTERS
1532 @cindex call-used register
1533 @cindex call-clobbered register
1534 @cindex call-saved register
1535 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1536 clobbered (in general) by function calls as well as for fixed
1537 registers. This macro therefore identifies the registers that are not
1538 available for general allocation of values that must live across
1541 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1542 automatically saves it on function entry and restores it on function
1543 exit, if the register is used within the function.
1545 @findex HARD_REGNO_CALL_PART_CLOBBERED
1546 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1547 @cindex call-used register
1548 @cindex call-clobbered register
1549 @cindex call-saved register
1550 A C expression that is non-zero if it is not permissible to store a
1551 value of mode @var{mode} in hard register number @var{regno} across a
1552 call without some part of it being clobbered. For most machines this
1553 macro need not be defined. It is only required for machines that do not
1554 preserve the entire contents of a register across a call.
1556 @findex CONDITIONAL_REGISTER_USAGE
1558 @findex call_used_regs
1559 @item CONDITIONAL_REGISTER_USAGE
1560 Zero or more C statements that may conditionally modify five variables
1561 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1562 (these three are of type @code{char []}), @code{reg_names} (of type
1563 @code{const char * []}) and @code{reg_class_contents} (of type
1564 @code{HARD_REG_SET}).
1565 Before the macro is called @code{fixed_regs}, @code{call_used_regs}
1566 @code{reg_class_contents} and @code{reg_names} have been initialized
1567 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1568 @code{REG_CLASS_CONTENTS} and @code{REGISTER_NAMES}, respectively,
1569 @code{global_regs} has been cleared, and any @samp{-ffixed-@var{reg}},
1570 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}} command
1571 options have been applied.
1573 This is necessary in case the fixed or call-clobbered registers depend
1576 You need not define this macro if it has no work to do.
1578 @cindex disabling certain registers
1579 @cindex controlling register usage
1580 If the usage of an entire class of registers depends on the target
1581 flags, you may indicate this to GCC by using this macro to modify
1582 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1583 registers in the classes which should not be used by GCC. Also define
1584 the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1585 is called with a letter for a class that shouldn't be used.
1587 (However, if this class is not included in @code{GENERAL_REGS} and all
1588 of the insn patterns whose constraints permit this class are
1589 controlled by target switches, then GCC will automatically avoid using
1590 these registers when the target switches are opposed to them.)
1592 @findex NON_SAVING_SETJMP
1593 @item NON_SAVING_SETJMP
1594 If this macro is defined and has a nonzero value, it means that
1595 @code{setjmp} and related functions fail to save the registers, or that
1596 @code{longjmp} fails to restore them. To compensate, the compiler
1597 avoids putting variables in registers in functions that use
1600 @findex INCOMING_REGNO
1601 @item INCOMING_REGNO (@var{out})
1602 Define this macro if the target machine has register windows. This C
1603 expression returns the register number as seen by the called function
1604 corresponding to the register number @var{out} as seen by the calling
1605 function. Return @var{out} if register number @var{out} is not an
1608 @findex OUTGOING_REGNO
1609 @item OUTGOING_REGNO (@var{in})
1610 Define this macro if the target machine has register windows. This C
1611 expression returns the register number as seen by the calling function
1612 corresponding to the register number @var{in} as seen by the called
1613 function. Return @var{in} if register number @var{in} is not an inbound
1617 @item LOCAL_REGNO (@var{regno})
1618 Define this macro if the target machine has register windows. This C
1619 expression returns true if the register is call-saved but is in the
1620 register window. Unlike most call-saved registers, such registers
1621 need not be explicitly restored on function exit or during non-local
1627 If the program counter has a register number, define this as that
1628 register number. Otherwise, do not define it.
1632 @node Allocation Order
1633 @subsection Order of Allocation of Registers
1634 @cindex order of register allocation
1635 @cindex register allocation order
1637 @c prevent bad page break with this line
1638 Registers are allocated in order.
1641 @findex REG_ALLOC_ORDER
1642 @item REG_ALLOC_ORDER
1643 If defined, an initializer for a vector of integers, containing the
1644 numbers of hard registers in the order in which GCC should prefer
1645 to use them (from most preferred to least).
1647 If this macro is not defined, registers are used lowest numbered first
1648 (all else being equal).
1650 One use of this macro is on machines where the highest numbered
1651 registers must always be saved and the save-multiple-registers
1652 instruction supports only sequences of consecutive registers. On such
1653 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1654 the highest numbered allocable register first.
1656 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1657 @item ORDER_REGS_FOR_LOCAL_ALLOC
1658 A C statement (sans semicolon) to choose the order in which to allocate
1659 hard registers for pseudo-registers local to a basic block.
1661 Store the desired register order in the array @code{reg_alloc_order}.
1662 Element 0 should be the register to allocate first; element 1, the next
1663 register; and so on.
1665 The macro body should not assume anything about the contents of
1666 @code{reg_alloc_order} before execution of the macro.
1668 On most machines, it is not necessary to define this macro.
1671 @node Values in Registers
1672 @subsection How Values Fit in Registers
1674 This section discusses the macros that describe which kinds of values
1675 (specifically, which machine modes) each register can hold, and how many
1676 consecutive registers are needed for a given mode.
1679 @findex HARD_REGNO_NREGS
1680 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1681 A C expression for the number of consecutive hard registers, starting
1682 at register number @var{regno}, required to hold a value of mode
1685 On a machine where all registers are exactly one word, a suitable
1686 definition of this macro is
1689 #define HARD_REGNO_NREGS(REGNO, MODE) \
1690 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1694 @findex ALTER_HARD_SUBREG
1695 @item ALTER_HARD_SUBREG (@var{tgt_mode}, @var{word}, @var{src_mode}, @var{regno})
1696 A C expression that returns an adjusted hard register number for
1699 (subreg:@var{tgt_mode} (reg:@var{src_mode} @var{regno}) @var{word})
1702 This may be needed if the target machine has mixed sized big-endian
1703 registers, like Sparc v9.
1705 @findex HARD_REGNO_MODE_OK
1706 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1707 A C expression that is nonzero if it is permissible to store a value
1708 of mode @var{mode} in hard register number @var{regno} (or in several
1709 registers starting with that one). For a machine where all registers
1710 are equivalent, a suitable definition is
1713 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1716 You need not include code to check for the numbers of fixed registers,
1717 because the allocation mechanism considers them to be always occupied.
1719 @cindex register pairs
1720 On some machines, double-precision values must be kept in even/odd
1721 register pairs. You can implement that by defining this macro to reject
1722 odd register numbers for such modes.
1724 The minimum requirement for a mode to be OK in a register is that the
1725 @samp{mov@var{mode}} instruction pattern support moves between the
1726 register and other hard register in the same class and that moving a
1727 value into the register and back out not alter it.
1729 Since the same instruction used to move @code{word_mode} will work for
1730 all narrower integer modes, it is not necessary on any machine for
1731 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1732 you define patterns @samp{movhi}, etc., to take advantage of this. This
1733 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1734 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1737 Many machines have special registers for floating point arithmetic.
1738 Often people assume that floating point machine modes are allowed only
1739 in floating point registers. This is not true. Any registers that
1740 can hold integers can safely @emph{hold} a floating point machine
1741 mode, whether or not floating arithmetic can be done on it in those
1742 registers. Integer move instructions can be used to move the values.
1744 On some machines, though, the converse is true: fixed-point machine
1745 modes may not go in floating registers. This is true if the floating
1746 registers normalize any value stored in them, because storing a
1747 non-floating value there would garble it. In this case,
1748 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1749 floating registers. But if the floating registers do not automatically
1750 normalize, if you can store any bit pattern in one and retrieve it
1751 unchanged without a trap, then any machine mode may go in a floating
1752 register, so you can define this macro to say so.
1754 The primary significance of special floating registers is rather that
1755 they are the registers acceptable in floating point arithmetic
1756 instructions. However, this is of no concern to
1757 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1758 constraints for those instructions.
1760 On some machines, the floating registers are especially slow to access,
1761 so that it is better to store a value in a stack frame than in such a
1762 register if floating point arithmetic is not being done. As long as the
1763 floating registers are not in class @code{GENERAL_REGS}, they will not
1764 be used unless some pattern's constraint asks for one.
1766 @findex MODES_TIEABLE_P
1767 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
1768 A C expression that is nonzero if a value of mode
1769 @var{mode1} is accessible in mode @var{mode2} without copying.
1771 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
1772 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
1773 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
1774 should be nonzero. If they differ for any @var{r}, you should define
1775 this macro to return zero unless some other mechanism ensures the
1776 accessibility of the value in a narrower mode.
1778 You should define this macro to return nonzero in as many cases as
1779 possible since doing so will allow GCC to perform better register
1782 @findex AVOID_CCMODE_COPIES
1783 @item AVOID_CCMODE_COPIES
1784 Define this macro if the compiler should avoid copies to/from @code{CCmode}
1785 registers. You should only define this macro if support for copying to/from
1786 @code{CCmode} is incomplete.
1789 @node Leaf Functions
1790 @subsection Handling Leaf Functions
1792 @cindex leaf functions
1793 @cindex functions, leaf
1794 On some machines, a leaf function (i.e., one which makes no calls) can run
1795 more efficiently if it does not make its own register window. Often this
1796 means it is required to receive its arguments in the registers where they
1797 are passed by the caller, instead of the registers where they would
1800 The special treatment for leaf functions generally applies only when
1801 other conditions are met; for example, often they may use only those
1802 registers for its own variables and temporaries. We use the term ``leaf
1803 function'' to mean a function that is suitable for this special
1804 handling, so that functions with no calls are not necessarily ``leaf
1807 GCC assigns register numbers before it knows whether the function is
1808 suitable for leaf function treatment. So it needs to renumber the
1809 registers in order to output a leaf function. The following macros
1813 @findex LEAF_REGISTERS
1814 @item LEAF_REGISTERS
1815 Name of a char vector, indexed by hard register number, which
1816 contains 1 for a register that is allowable in a candidate for leaf
1819 If leaf function treatment involves renumbering the registers, then the
1820 registers marked here should be the ones before renumbering---those that
1821 GCC would ordinarily allocate. The registers which will actually be
1822 used in the assembler code, after renumbering, should not be marked with 1
1825 Define this macro only if the target machine offers a way to optimize
1826 the treatment of leaf functions.
1828 @findex LEAF_REG_REMAP
1829 @item LEAF_REG_REMAP (@var{regno})
1830 A C expression whose value is the register number to which @var{regno}
1831 should be renumbered, when a function is treated as a leaf function.
1833 If @var{regno} is a register number which should not appear in a leaf
1834 function before renumbering, then the expression should yield -1, which
1835 will cause the compiler to abort.
1837 Define this macro only if the target machine offers a way to optimize the
1838 treatment of leaf functions, and registers need to be renumbered to do
1842 @findex current_function_is_leaf
1843 @findex current_function_uses_only_leaf_regs
1844 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
1845 treat leaf functions specially. They can test the C variable
1846 @code{current_function_is_leaf} which is nonzero for leaf functions.
1847 @code{current_function_is_leaf} is set prior to local register allocation
1848 and is valid for the remaining compiler passes. They can also test the C
1849 variable @code{current_function_uses_only_leaf_regs} which is nonzero for
1850 leaf functions which only use leaf registers.
1851 @code{current_function_uses_only_leaf_regs} is valid after reload and is
1852 only useful if @code{LEAF_REGISTERS} is defined.
1853 @c changed this to fix overfull. ALSO: why the "it" at the beginning
1854 @c of the next paragraph?! --mew 2feb93
1856 @node Stack Registers
1857 @subsection Registers That Form a Stack
1859 There are special features to handle computers where some of the
1860 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
1861 Stack registers are normally written by pushing onto the stack, and are
1862 numbered relative to the top of the stack.
1864 Currently, GCC can only handle one group of stack-like registers, and
1865 they must be consecutively numbered.
1870 Define this if the machine has any stack-like registers.
1872 @findex FIRST_STACK_REG
1873 @item FIRST_STACK_REG
1874 The number of the first stack-like register. This one is the top
1877 @findex LAST_STACK_REG
1878 @item LAST_STACK_REG
1879 The number of the last stack-like register. This one is the bottom of
1883 @node Register Classes
1884 @section Register Classes
1885 @cindex register class definitions
1886 @cindex class definitions, register
1888 On many machines, the numbered registers are not all equivalent.
1889 For example, certain registers may not be allowed for indexed addressing;
1890 certain registers may not be allowed in some instructions. These machine
1891 restrictions are described to the compiler using @dfn{register classes}.
1893 You define a number of register classes, giving each one a name and saying
1894 which of the registers belong to it. Then you can specify register classes
1895 that are allowed as operands to particular instruction patterns.
1899 In general, each register will belong to several classes. In fact, one
1900 class must be named @code{ALL_REGS} and contain all the registers. Another
1901 class must be named @code{NO_REGS} and contain no registers. Often the
1902 union of two classes will be another class; however, this is not required.
1904 @findex GENERAL_REGS
1905 One of the classes must be named @code{GENERAL_REGS}. There is nothing
1906 terribly special about the name, but the operand constraint letters
1907 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
1908 the same as @code{ALL_REGS}, just define it as a macro which expands
1911 Order the classes so that if class @var{x} is contained in class @var{y}
1912 then @var{x} has a lower class number than @var{y}.
1914 The way classes other than @code{GENERAL_REGS} are specified in operand
1915 constraints is through machine-dependent operand constraint letters.
1916 You can define such letters to correspond to various classes, then use
1917 them in operand constraints.
1919 You should define a class for the union of two classes whenever some
1920 instruction allows both classes. For example, if an instruction allows
1921 either a floating point (coprocessor) register or a general register for a
1922 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
1923 which includes both of them. Otherwise you will get suboptimal code.
1925 You must also specify certain redundant information about the register
1926 classes: for each class, which classes contain it and which ones are
1927 contained in it; for each pair of classes, the largest class contained
1930 When a value occupying several consecutive registers is expected in a
1931 certain class, all the registers used must belong to that class.
1932 Therefore, register classes cannot be used to enforce a requirement for
1933 a register pair to start with an even-numbered register. The way to
1934 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
1936 Register classes used for input-operands of bitwise-and or shift
1937 instructions have a special requirement: each such class must have, for
1938 each fixed-point machine mode, a subclass whose registers can transfer that
1939 mode to or from memory. For example, on some machines, the operations for
1940 single-byte values (@code{QImode}) are limited to certain registers. When
1941 this is so, each register class that is used in a bitwise-and or shift
1942 instruction must have a subclass consisting of registers from which
1943 single-byte values can be loaded or stored. This is so that
1944 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
1947 @findex enum reg_class
1948 @item enum reg_class
1949 An enumeral type that must be defined with all the register class names
1950 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
1951 must be the last register class, followed by one more enumeral value,
1952 @code{LIM_REG_CLASSES}, which is not a register class but rather
1953 tells how many classes there are.
1955 Each register class has a number, which is the value of casting
1956 the class name to type @code{int}. The number serves as an index
1957 in many of the tables described below.
1959 @findex N_REG_CLASSES
1961 The number of distinct register classes, defined as follows:
1964 #define N_REG_CLASSES (int) LIM_REG_CLASSES
1967 @findex REG_CLASS_NAMES
1968 @item REG_CLASS_NAMES
1969 An initializer containing the names of the register classes as C string
1970 constants. These names are used in writing some of the debugging dumps.
1972 @findex REG_CLASS_CONTENTS
1973 @item REG_CLASS_CONTENTS
1974 An initializer containing the contents of the register classes, as integers
1975 which are bit masks. The @var{n}th integer specifies the contents of class
1976 @var{n}. The way the integer @var{mask} is interpreted is that
1977 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
1979 When the machine has more than 32 registers, an integer does not suffice.
1980 Then the integers are replaced by sub-initializers, braced groupings containing
1981 several integers. Each sub-initializer must be suitable as an initializer
1982 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
1983 In this situation, the first integer in each sub-initializer corresponds to
1984 registers 0 through 31, the second integer to registers 32 through 63, and
1987 @findex REGNO_REG_CLASS
1988 @item REGNO_REG_CLASS (@var{regno})
1989 A C expression whose value is a register class containing hard register
1990 @var{regno}. In general there is more than one such class; choose a class
1991 which is @dfn{minimal}, meaning that no smaller class also contains the
1994 @findex BASE_REG_CLASS
1995 @item BASE_REG_CLASS
1996 A macro whose definition is the name of the class to which a valid
1997 base register must belong. A base register is one used in an address
1998 which is the register value plus a displacement.
2000 @findex INDEX_REG_CLASS
2001 @item INDEX_REG_CLASS
2002 A macro whose definition is the name of the class to which a valid
2003 index register must belong. An index register is one used in an
2004 address where its value is either multiplied by a scale factor or
2005 added to another register (as well as added to a displacement).
2007 @findex REG_CLASS_FROM_LETTER
2008 @item REG_CLASS_FROM_LETTER (@var{char})
2009 A C expression which defines the machine-dependent operand constraint
2010 letters for register classes. If @var{char} is such a letter, the
2011 value should be the register class corresponding to it. Otherwise,
2012 the value should be @code{NO_REGS}. The register letter @samp{r},
2013 corresponding to class @code{GENERAL_REGS}, will not be passed
2014 to this macro; you do not need to handle it.
2016 @findex REGNO_OK_FOR_BASE_P
2017 @item REGNO_OK_FOR_BASE_P (@var{num})
2018 A C expression which is nonzero if register number @var{num} is
2019 suitable for use as a base register in operand addresses. It may be
2020 either a suitable hard register or a pseudo register that has been
2021 allocated such a hard register.
2023 @findex REGNO_MODE_OK_FOR_BASE_P
2024 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2025 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2026 that expression may examine the mode of the memory reference in
2027 @var{mode}. You should define this macro if the mode of the memory
2028 reference affects whether a register may be used as a base register. If
2029 you define this macro, the compiler will use it instead of
2030 @code{REGNO_OK_FOR_BASE_P}.
2032 @findex REGNO_OK_FOR_INDEX_P
2033 @item REGNO_OK_FOR_INDEX_P (@var{num})
2034 A C expression which is nonzero if register number @var{num} is
2035 suitable for use as an index register in operand addresses. It may be
2036 either a suitable hard register or a pseudo register that has been
2037 allocated such a hard register.
2039 The difference between an index register and a base register is that
2040 the index register may be scaled. If an address involves the sum of
2041 two registers, neither one of them scaled, then either one may be
2042 labeled the ``base'' and the other the ``index''; but whichever
2043 labeling is used must fit the machine's constraints of which registers
2044 may serve in each capacity. The compiler will try both labelings,
2045 looking for one that is valid, and will reload one or both registers
2046 only if neither labeling works.
2048 @findex PREFERRED_RELOAD_CLASS
2049 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2050 A C expression that places additional restrictions on the register class
2051 to use when it is necessary to copy value @var{x} into a register in class
2052 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2053 another, smaller class. On many machines, the following definition is
2057 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2060 Sometimes returning a more restrictive class makes better code. For
2061 example, on the 68000, when @var{x} is an integer constant that is in range
2062 for a @samp{moveq} instruction, the value of this macro is always
2063 @code{DATA_REGS} as long as @var{class} includes the data registers.
2064 Requiring a data register guarantees that a @samp{moveq} will be used.
2066 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2067 you can force @var{x} into a memory constant. This is useful on
2068 certain machines where immediate floating values cannot be loaded into
2069 certain kinds of registers.
2071 @findex PREFERRED_OUTPUT_RELOAD_CLASS
2072 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2073 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2074 input reloads. If you don't define this macro, the default is to use
2075 @var{class}, unchanged.
2077 @findex LIMIT_RELOAD_CLASS
2078 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2079 A C expression that places additional restrictions on the register class
2080 to use when it is necessary to be able to hold a value of mode
2081 @var{mode} in a reload register for which class @var{class} would
2084 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2085 there are certain modes that simply can't go in certain reload classes.
2087 The value is a register class; perhaps @var{class}, or perhaps another,
2090 Don't define this macro unless the target machine has limitations which
2091 require the macro to do something nontrivial.
2093 @findex SECONDARY_RELOAD_CLASS
2094 @findex SECONDARY_INPUT_RELOAD_CLASS
2095 @findex SECONDARY_OUTPUT_RELOAD_CLASS
2096 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2097 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2098 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2099 Many machines have some registers that cannot be copied directly to or
2100 from memory or even from other types of registers. An example is the
2101 @samp{MQ} register, which on most machines, can only be copied to or
2102 from general registers, but not memory. Some machines allow copying all
2103 registers to and from memory, but require a scratch register for stores
2104 to some memory locations (e.g., those with symbolic address on the RT,
2105 and those with certain symbolic address on the Sparc when compiling
2106 PIC). In some cases, both an intermediate and a scratch register are
2109 You should define these macros to indicate to the reload phase that it may
2110 need to allocate at least one register for a reload in addition to the
2111 register to contain the data. Specifically, if copying @var{x} to a
2112 register @var{class} in @var{mode} requires an intermediate register,
2113 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2114 largest register class all of whose registers can be used as
2115 intermediate registers or scratch registers.
2117 If copying a register @var{class} in @var{mode} to @var{x} requires an
2118 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2119 should be defined to return the largest register class required. If the
2120 requirements for input and output reloads are the same, the macro
2121 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2124 The values returned by these macros are often @code{GENERAL_REGS}.
2125 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2126 can be directly copied to or from a register of @var{class} in
2127 @var{mode} without requiring a scratch register. Do not define this
2128 macro if it would always return @code{NO_REGS}.
2130 If a scratch register is required (either with or without an
2131 intermediate register), you should define patterns for
2132 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2133 (@pxref{Standard Names}. These patterns, which will normally be
2134 implemented with a @code{define_expand}, should be similar to the
2135 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2138 Define constraints for the reload register and scratch register that
2139 contain a single register class. If the original reload register (whose
2140 class is @var{class}) can meet the constraint given in the pattern, the
2141 value returned by these macros is used for the class of the scratch
2142 register. Otherwise, two additional reload registers are required.
2143 Their classes are obtained from the constraints in the insn pattern.
2145 @var{x} might be a pseudo-register or a @code{subreg} of a
2146 pseudo-register, which could either be in a hard register or in memory.
2147 Use @code{true_regnum} to find out; it will return -1 if the pseudo is
2148 in memory and the hard register number if it is in a register.
2150 These macros should not be used in the case where a particular class of
2151 registers can only be copied to memory and not to another class of
2152 registers. In that case, secondary reload registers are not needed and
2153 would not be helpful. Instead, a stack location must be used to perform
2154 the copy and the @code{mov@var{m}} pattern should use memory as a
2155 intermediate storage. This case often occurs between floating-point and
2158 @findex SECONDARY_MEMORY_NEEDED
2159 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2160 Certain machines have the property that some registers cannot be copied
2161 to some other registers without using memory. Define this macro on
2162 those machines to be a C expression that is non-zero if objects of mode
2163 @var{m} in registers of @var{class1} can only be copied to registers of
2164 class @var{class2} by storing a register of @var{class1} into memory
2165 and loading that memory location into a register of @var{class2}.
2167 Do not define this macro if its value would always be zero.
2169 @findex SECONDARY_MEMORY_NEEDED_RTX
2170 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2171 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2172 allocates a stack slot for a memory location needed for register copies.
2173 If this macro is defined, the compiler instead uses the memory location
2174 defined by this macro.
2176 Do not define this macro if you do not define
2177 @code{SECONDARY_MEMORY_NEEDED}.
2179 @findex SECONDARY_MEMORY_NEEDED_MODE
2180 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2181 When the compiler needs a secondary memory location to copy between two
2182 registers of mode @var{mode}, it normally allocates sufficient memory to
2183 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2184 load operations in a mode that many bits wide and whose class is the
2185 same as that of @var{mode}.
2187 This is right thing to do on most machines because it ensures that all
2188 bits of the register are copied and prevents accesses to the registers
2189 in a narrower mode, which some machines prohibit for floating-point
2192 However, this default behavior is not correct on some machines, such as
2193 the DEC Alpha, that store short integers in floating-point registers
2194 differently than in integer registers. On those machines, the default
2195 widening will not work correctly and you must define this macro to
2196 suppress that widening in some cases. See the file @file{alpha.h} for
2199 Do not define this macro if you do not define
2200 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2201 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2203 @findex SMALL_REGISTER_CLASSES
2204 @item SMALL_REGISTER_CLASSES
2205 On some machines, it is risky to let hard registers live across arbitrary
2206 insns. Typically, these machines have instructions that require values
2207 to be in specific registers (like an accumulator), and reload will fail
2208 if the required hard register is used for another purpose across such an
2211 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a non-zero
2212 value on these machines. When this macro has a non-zero value, the
2213 compiler will try to minimize the lifetime of hard registers.
2215 It is always safe to define this macro with a non-zero value, but if you
2216 unnecessarily define it, you will reduce the amount of optimizations
2217 that can be performed in some cases. If you do not define this macro
2218 with a non-zero value when it is required, the compiler will run out of
2219 spill registers and print a fatal error message. For most machines, you
2220 should not define this macro at all.
2222 @findex CLASS_LIKELY_SPILLED_P
2223 @item CLASS_LIKELY_SPILLED_P (@var{class})
2224 A C expression whose value is nonzero if pseudos that have been assigned
2225 to registers of class @var{class} would likely be spilled because
2226 registers of @var{class} are needed for spill registers.
2228 The default value of this macro returns 1 if @var{class} has exactly one
2229 register and zero otherwise. On most machines, this default should be
2230 used. Only define this macro to some other expression if pseudos
2231 allocated by @file{local-alloc.c} end up in memory because their hard
2232 registers were needed for spill registers. If this macro returns nonzero
2233 for those classes, those pseudos will only be allocated by
2234 @file{global.c}, which knows how to reallocate the pseudo to another
2235 register. If there would not be another register available for
2236 reallocation, you should not change the definition of this macro since
2237 the only effect of such a definition would be to slow down register
2240 @findex CLASS_MAX_NREGS
2241 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2242 A C expression for the maximum number of consecutive registers
2243 of class @var{class} needed to hold a value of mode @var{mode}.
2245 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2246 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2247 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2248 @var{mode})} for all @var{regno} values in the class @var{class}.
2250 This macro helps control the handling of multiple-word values
2253 @item CLASS_CANNOT_CHANGE_MODE
2254 If defined, a C expression for a class that contains registers for
2255 which the compiler may not change modes arbitrarily.
2257 @item CLASS_CANNOT_CHANGE_MODE_P(@var{from}, @var{to})
2258 A C expression that is true if, for a register in
2259 @code{CLASS_CANNOT_CHANGE_MODE}, the requested mode punning is illegal.
2261 For the example, loading 32-bit integer or floating-point objects into
2262 floating-point registers on the Alpha extends them to 64-bits.
2263 Therefore loading a 64-bit object and then storing it as a 32-bit object
2264 does not store the low-order 32-bits, as would be the case for a normal
2265 register. Therefore, @file{alpha.h} defines @code{CLASS_CANNOT_CHANGE_MODE}
2266 as @code{FLOAT_REGS} and @code{CLASS_CANNOT_CHANGE_MODE_P} restricts
2267 mode changes to same-size modes.
2269 Compare this to IA-64, which extends floating-point values to 82-bits,
2270 and stores 64-bit integers in a different format than 64-bit doubles.
2271 Therefore @code{CLASS_CANNOT_CHANGE_MODE_P} is always true.
2274 Three other special macros describe which operands fit which constraint
2278 @findex CONST_OK_FOR_LETTER_P
2279 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2280 A C expression that defines the machine-dependent operand constraint
2281 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2282 particular ranges of integer values. If @var{c} is one of those
2283 letters, the expression should check that @var{value}, an integer, is in
2284 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2285 not one of those letters, the value should be 0 regardless of
2288 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2289 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2290 A C expression that defines the machine-dependent operand constraint
2291 letters that specify particular ranges of @code{const_double} values
2292 (@samp{G} or @samp{H}).
2294 If @var{c} is one of those letters, the expression should check that
2295 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2296 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2297 letters, the value should be 0 regardless of @var{value}.
2299 @code{const_double} is used for all floating-point constants and for
2300 @code{DImode} fixed-point constants. A given letter can accept either
2301 or both kinds of values. It can use @code{GET_MODE} to distinguish
2302 between these kinds.
2304 @findex EXTRA_CONSTRAINT
2305 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2306 A C expression that defines the optional machine-dependent constraint
2307 letters that can be used to segregate specific types of operands, usually
2308 memory references, for the target machine. Any letter that is not
2309 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER}
2310 may be used. Normally this macro will not be defined.
2312 If it is required for a particular target machine, it should return 1
2313 if @var{value} corresponds to the operand type represented by the
2314 constraint letter @var{c}. If @var{c} is not defined as an extra
2315 constraint, the value returned should be 0 regardless of @var{value}.
2317 For example, on the ROMP, load instructions cannot have their output
2318 in r0 if the memory reference contains a symbolic address. Constraint
2319 letter @samp{Q} is defined as representing a memory address that does
2320 @emph{not} contain a symbolic address. An alternative is specified with
2321 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2322 alternative specifies @samp{m} on the input and a register class that
2323 does not include r0 on the output.
2326 @node Stack and Calling
2327 @section Stack Layout and Calling Conventions
2328 @cindex calling conventions
2330 @c prevent bad page break with this line
2331 This describes the stack layout and calling conventions.
2339 * Register Arguments::
2341 * Aggregate Return::
2350 @subsection Basic Stack Layout
2351 @cindex stack frame layout
2352 @cindex frame layout
2354 @c prevent bad page break with this line
2355 Here is the basic stack layout.
2358 @findex STACK_GROWS_DOWNWARD
2359 @item STACK_GROWS_DOWNWARD
2360 Define this macro if pushing a word onto the stack moves the stack
2361 pointer to a smaller address.
2363 When we say, ``define this macro if @dots{},'' it means that the
2364 compiler checks this macro only with @code{#ifdef} so the precise
2365 definition used does not matter.
2367 @findex FRAME_GROWS_DOWNWARD
2368 @item FRAME_GROWS_DOWNWARD
2369 Define this macro if the addresses of local variable slots are at negative
2370 offsets from the frame pointer.
2372 @findex ARGS_GROW_DOWNWARD
2373 @item ARGS_GROW_DOWNWARD
2374 Define this macro if successive arguments to a function occupy decreasing
2375 addresses on the stack.
2377 @findex STARTING_FRAME_OFFSET
2378 @item STARTING_FRAME_OFFSET
2379 Offset from the frame pointer to the first local variable slot to be allocated.
2381 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2382 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2383 Otherwise, it is found by adding the length of the first slot to the
2384 value @code{STARTING_FRAME_OFFSET}.
2385 @c i'm not sure if the above is still correct.. had to change it to get
2386 @c rid of an overfull. --mew 2feb93
2388 @findex STACK_POINTER_OFFSET
2389 @item STACK_POINTER_OFFSET
2390 Offset from the stack pointer register to the first location at which
2391 outgoing arguments are placed. If not specified, the default value of
2392 zero is used. This is the proper value for most machines.
2394 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2395 the first location at which outgoing arguments are placed.
2397 @findex FIRST_PARM_OFFSET
2398 @item FIRST_PARM_OFFSET (@var{fundecl})
2399 Offset from the argument pointer register to the first argument's
2400 address. On some machines it may depend on the data type of the
2403 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2404 the first argument's address.
2406 @findex STACK_DYNAMIC_OFFSET
2407 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2408 Offset from the stack pointer register to an item dynamically allocated
2409 on the stack, e.g., by @code{alloca}.
2411 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2412 length of the outgoing arguments. The default is correct for most
2413 machines. See @file{function.c} for details.
2415 @findex DYNAMIC_CHAIN_ADDRESS
2416 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2417 A C expression whose value is RTL representing the address in a stack
2418 frame where the pointer to the caller's frame is stored. Assume that
2419 @var{frameaddr} is an RTL expression for the address of the stack frame
2422 If you don't define this macro, the default is to return the value
2423 of @var{frameaddr}---that is, the stack frame address is also the
2424 address of the stack word that points to the previous frame.
2426 @findex SETUP_FRAME_ADDRESSES
2427 @item SETUP_FRAME_ADDRESSES
2428 If defined, a C expression that produces the machine-specific code to
2429 setup the stack so that arbitrary frames can be accessed. For example,
2430 on the Sparc, we must flush all of the register windows to the stack
2431 before we can access arbitrary stack frames. You will seldom need to
2434 @findex BUILTIN_SETJMP_FRAME_VALUE
2435 @item BUILTIN_SETJMP_FRAME_VALUE
2436 If defined, a C expression that contains an rtx that is used to store
2437 the address of the current frame into the built in @code{setjmp} buffer.
2438 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2439 machines. One reason you may need to define this macro is if
2440 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2442 @findex RETURN_ADDR_RTX
2443 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2444 A C expression whose value is RTL representing the value of the return
2445 address for the frame @var{count} steps up from the current frame, after
2446 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2447 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2448 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2450 The value of the expression must always be the correct address when
2451 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2452 determine the return address of other frames.
2454 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2455 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2456 Define this if the return address of a particular stack frame is accessed
2457 from the frame pointer of the previous stack frame.
2459 @findex INCOMING_RETURN_ADDR_RTX
2460 @item INCOMING_RETURN_ADDR_RTX
2461 A C expression whose value is RTL representing the location of the
2462 incoming return address at the beginning of any function, before the
2463 prologue. This RTL is either a @code{REG}, indicating that the return
2464 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2467 You only need to define this macro if you want to support call frame
2468 debugging information like that provided by DWARF 2.
2470 If this RTL is a @code{REG}, you should also define
2471 DWARF_FRAME_RETURN_COLUMN to @code{DWARF_FRAME_REGNUM (REGNO)}.
2473 @findex INCOMING_FRAME_SP_OFFSET
2474 @item INCOMING_FRAME_SP_OFFSET
2475 A C expression whose value is an integer giving the offset, in bytes,
2476 from the value of the stack pointer register to the top of the stack
2477 frame at the beginning of any function, before the prologue. The top of
2478 the frame is defined to be the value of the stack pointer in the
2479 previous frame, just before the call instruction.
2481 You only need to define this macro if you want to support call frame
2482 debugging information like that provided by DWARF 2.
2484 @findex ARG_POINTER_CFA_OFFSET
2485 @item ARG_POINTER_CFA_OFFSET (@var{fundecl})
2486 A C expression whose value is an integer giving the offset, in bytes,
2487 from the argument pointer to the canonical frame address (cfa). The
2488 final value should coincide with that calculated by
2489 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2490 during virtual register instantiation.
2492 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2493 which is correct for most machines; in general, the arguments are found
2494 immediately before the stack frame. Note that this is not the case on
2495 some targets that save registers into the caller's frame, such as SPARC
2496 and rs6000, and so such targets need to define this macro.
2498 You only need to define this macro if the default is incorrect, and you
2499 want to support call frame debugging information like that provided by
2504 Define this macro if the stack size for the target is very small. This
2505 has the effect of disabling gcc's builtin @samp{alloca}, though
2506 @samp{__builtin_alloca} is not affected.
2509 @node Stack Checking
2510 @subsection Specifying How Stack Checking is Done
2512 GCC will check that stack references are within the boundaries of
2513 the stack, if the @samp{-fstack-check} is specified, in one of three ways:
2517 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2518 will assume that you have arranged for stack checking to be done at
2519 appropriate places in the configuration files, e.g., in
2520 @code{FUNCTION_PROLOGUE}. GCC will do not other special processing.
2523 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
2524 called @code{check_stack} in your @file{md} file, GCC will call that
2525 pattern with one argument which is the address to compare the stack
2526 value against. You must arrange for this pattern to report an error if
2527 the stack pointer is out of range.
2530 If neither of the above are true, GCC will generate code to periodically
2531 ``probe'' the stack pointer using the values of the macros defined below.
2534 Normally, you will use the default values of these macros, so GCC
2535 will use the third approach.
2538 @findex STACK_CHECK_BUILTIN
2539 @item STACK_CHECK_BUILTIN
2540 A nonzero value if stack checking is done by the configuration files in a
2541 machine-dependent manner. You should define this macro if stack checking
2542 is require by the ABI of your machine or if you would like to have to stack
2543 checking in some more efficient way than GCC's portable approach.
2544 The default value of this macro is zero.
2546 @findex STACK_CHECK_PROBE_INTERVAL
2547 @item STACK_CHECK_PROBE_INTERVAL
2548 An integer representing the interval at which GCC must generate stack
2549 probe instructions. You will normally define this macro to be no larger
2550 than the size of the ``guard pages'' at the end of a stack area. The
2551 default value of 4096 is suitable for most systems.
2553 @findex STACK_CHECK_PROBE_LOAD
2554 @item STACK_CHECK_PROBE_LOAD
2555 A integer which is nonzero if GCC should perform the stack probe
2556 as a load instruction and zero if GCC should use a store instruction.
2557 The default is zero, which is the most efficient choice on most systems.
2559 @findex STACK_CHECK_PROTECT
2560 @item STACK_CHECK_PROTECT
2561 The number of bytes of stack needed to recover from a stack overflow,
2562 for languages where such a recovery is supported. The default value of
2563 75 words should be adequate for most machines.
2565 @findex STACK_CHECK_MAX_FRAME_SIZE
2566 @item STACK_CHECK_MAX_FRAME_SIZE
2567 The maximum size of a stack frame, in bytes. GCC will generate probe
2568 instructions in non-leaf functions to ensure at least this many bytes of
2569 stack are available. If a stack frame is larger than this size, stack
2570 checking will not be reliable and GCC will issue a warning. The
2571 default is chosen so that GCC only generates one instruction on most
2572 systems. You should normally not change the default value of this macro.
2574 @findex STACK_CHECK_FIXED_FRAME_SIZE
2575 @item STACK_CHECK_FIXED_FRAME_SIZE
2576 GCC uses this value to generate the above warning message. It
2577 represents the amount of fixed frame used by a function, not including
2578 space for any callee-saved registers, temporaries and user variables.
2579 You need only specify an upper bound for this amount and will normally
2580 use the default of four words.
2582 @findex STACK_CHECK_MAX_VAR_SIZE
2583 @item STACK_CHECK_MAX_VAR_SIZE
2584 The maximum size, in bytes, of an object that GCC will place in the
2585 fixed area of the stack frame when the user specifies
2586 @samp{-fstack-check}.
2587 GCC computed the default from the values of the above macros and you will
2588 normally not need to override that default.
2592 @node Frame Registers
2593 @subsection Registers That Address the Stack Frame
2595 @c prevent bad page break with this line
2596 This discusses registers that address the stack frame.
2599 @findex STACK_POINTER_REGNUM
2600 @item STACK_POINTER_REGNUM
2601 The register number of the stack pointer register, which must also be a
2602 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2603 the hardware determines which register this is.
2605 @findex FRAME_POINTER_REGNUM
2606 @item FRAME_POINTER_REGNUM
2607 The register number of the frame pointer register, which is used to
2608 access automatic variables in the stack frame. On some machines, the
2609 hardware determines which register this is. On other machines, you can
2610 choose any register you wish for this purpose.
2612 @findex HARD_FRAME_POINTER_REGNUM
2613 @item HARD_FRAME_POINTER_REGNUM
2614 On some machines the offset between the frame pointer and starting
2615 offset of the automatic variables is not known until after register
2616 allocation has been done (for example, because the saved registers are
2617 between these two locations). On those machines, define
2618 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
2619 be used internally until the offset is known, and define
2620 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
2621 used for the frame pointer.
2623 You should define this macro only in the very rare circumstances when it
2624 is not possible to calculate the offset between the frame pointer and
2625 the automatic variables until after register allocation has been
2626 completed. When this macro is defined, you must also indicate in your
2627 definition of @code{ELIMINABLE_REGS} how to eliminate
2628 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
2629 or @code{STACK_POINTER_REGNUM}.
2631 Do not define this macro if it would be the same as
2632 @code{FRAME_POINTER_REGNUM}.
2634 @findex ARG_POINTER_REGNUM
2635 @item ARG_POINTER_REGNUM
2636 The register number of the arg pointer register, which is used to access
2637 the function's argument list. On some machines, this is the same as the
2638 frame pointer register. On some machines, the hardware determines which
2639 register this is. On other machines, you can choose any register you
2640 wish for this purpose. If this is not the same register as the frame
2641 pointer register, then you must mark it as a fixed register according to
2642 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2643 (@pxref{Elimination}).
2645 @findex RETURN_ADDRESS_POINTER_REGNUM
2646 @item RETURN_ADDRESS_POINTER_REGNUM
2647 The register number of the return address pointer register, which is used to
2648 access the current function's return address from the stack. On some
2649 machines, the return address is not at a fixed offset from the frame
2650 pointer or stack pointer or argument pointer. This register can be defined
2651 to point to the return address on the stack, and then be converted by
2652 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
2654 Do not define this macro unless there is no other way to get the return
2655 address from the stack.
2657 @findex STATIC_CHAIN_REGNUM
2658 @findex STATIC_CHAIN_INCOMING_REGNUM
2659 @item STATIC_CHAIN_REGNUM
2660 @itemx STATIC_CHAIN_INCOMING_REGNUM
2661 Register numbers used for passing a function's static chain pointer. If
2662 register windows are used, the register number as seen by the called
2663 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
2664 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
2665 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
2666 not be defined.@refill
2668 The static chain register need not be a fixed register.
2670 If the static chain is passed in memory, these macros should not be
2671 defined; instead, the next two macros should be defined.
2673 @findex STATIC_CHAIN
2674 @findex STATIC_CHAIN_INCOMING
2676 @itemx STATIC_CHAIN_INCOMING
2677 If the static chain is passed in memory, these macros provide rtx giving
2678 @code{mem} expressions that denote where they are stored.
2679 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
2680 as seen by the calling and called functions, respectively. Often the former
2681 will be at an offset from the stack pointer and the latter at an offset from
2682 the frame pointer.@refill
2684 @findex stack_pointer_rtx
2685 @findex frame_pointer_rtx
2686 @findex arg_pointer_rtx
2687 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
2688 @code{arg_pointer_rtx} will have been initialized prior to the use of these
2689 macros and should be used to refer to those items.
2691 If the static chain is passed in a register, the two previous macros should
2696 @subsection Eliminating Frame Pointer and Arg Pointer
2698 @c prevent bad page break with this line
2699 This is about eliminating the frame pointer and arg pointer.
2702 @findex FRAME_POINTER_REQUIRED
2703 @item FRAME_POINTER_REQUIRED
2704 A C expression which is nonzero if a function must have and use a frame
2705 pointer. This expression is evaluated in the reload pass. If its value is
2706 nonzero the function will have a frame pointer.
2708 The expression can in principle examine the current function and decide
2709 according to the facts, but on most machines the constant 0 or the
2710 constant 1 suffices. Use 0 when the machine allows code to be generated
2711 with no frame pointer, and doing so saves some time or space. Use 1
2712 when there is no possible advantage to avoiding a frame pointer.
2714 In certain cases, the compiler does not know how to produce valid code
2715 without a frame pointer. The compiler recognizes those cases and
2716 automatically gives the function a frame pointer regardless of what
2717 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
2720 In a function that does not require a frame pointer, the frame pointer
2721 register can be allocated for ordinary usage, unless you mark it as a
2722 fixed register. See @code{FIXED_REGISTERS} for more information.
2724 @findex INITIAL_FRAME_POINTER_OFFSET
2725 @findex get_frame_size
2726 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
2727 A C statement to store in the variable @var{depth-var} the difference
2728 between the frame pointer and the stack pointer values immediately after
2729 the function prologue. The value would be computed from information
2730 such as the result of @code{get_frame_size ()} and the tables of
2731 registers @code{regs_ever_live} and @code{call_used_regs}.
2733 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
2734 need not be defined. Otherwise, it must be defined even if
2735 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
2736 case, you may set @var{depth-var} to anything.
2738 @findex ELIMINABLE_REGS
2739 @item ELIMINABLE_REGS
2740 If defined, this macro specifies a table of register pairs used to
2741 eliminate unneeded registers that point into the stack frame. If it is not
2742 defined, the only elimination attempted by the compiler is to replace
2743 references to the frame pointer with references to the stack pointer.
2745 The definition of this macro is a list of structure initializations, each
2746 of which specifies an original and replacement register.
2748 On some machines, the position of the argument pointer is not known until
2749 the compilation is completed. In such a case, a separate hard register
2750 must be used for the argument pointer. This register can be eliminated by
2751 replacing it with either the frame pointer or the argument pointer,
2752 depending on whether or not the frame pointer has been eliminated.
2754 In this case, you might specify:
2756 #define ELIMINABLE_REGS \
2757 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
2758 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
2759 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
2762 Note that the elimination of the argument pointer with the stack pointer is
2763 specified first since that is the preferred elimination.
2765 @findex CAN_ELIMINATE
2766 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
2767 A C expression that returns non-zero if the compiler is allowed to try
2768 to replace register number @var{from-reg} with register number
2769 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
2770 is defined, and will usually be the constant 1, since most of the cases
2771 preventing register elimination are things that the compiler already
2774 @findex INITIAL_ELIMINATION_OFFSET
2775 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
2776 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
2777 specifies the initial difference between the specified pair of
2778 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
2781 @findex LONGJMP_RESTORE_FROM_STACK
2782 @item LONGJMP_RESTORE_FROM_STACK
2783 Define this macro if the @code{longjmp} function restores registers from
2784 the stack frames, rather than from those saved specifically by
2785 @code{setjmp}. Certain quantities must not be kept in registers across
2786 a call to @code{setjmp} on such machines.
2789 @node Stack Arguments
2790 @subsection Passing Function Arguments on the Stack
2791 @cindex arguments on stack
2792 @cindex stack arguments
2794 The macros in this section control how arguments are passed
2795 on the stack. See the following section for other macros that
2796 control passing certain arguments in registers.
2799 @findex PROMOTE_PROTOTYPES
2800 @item PROMOTE_PROTOTYPES
2801 A C expression whose value is nonzero if an argument declared in
2802 a prototype as an integral type smaller than @code{int} should
2803 actually be passed as an @code{int}. In addition to avoiding
2804 errors in certain cases of mismatch, it also makes for better
2805 code on certain machines. If the macro is not defined in target
2806 header files, it defaults to 0.
2810 A C expression. If nonzero, push insns will be used to pass
2812 If the target machine does not have a push instruction, set it to zero.
2813 That directs GCC to use an alternate strategy: to
2814 allocate the entire argument block and then store the arguments into
2815 it. When PUSH_ARGS is nonzero, PUSH_ROUNDING must be defined too.
2816 On some machines, the definition
2818 @findex PUSH_ROUNDING
2819 @item PUSH_ROUNDING (@var{npushed})
2820 A C expression that is the number of bytes actually pushed onto the
2821 stack when an instruction attempts to push @var{npushed} bytes.
2823 On some machines, the definition
2826 #define PUSH_ROUNDING(BYTES) (BYTES)
2830 will suffice. But on other machines, instructions that appear
2831 to push one byte actually push two bytes in an attempt to maintain
2832 alignment. Then the definition should be
2835 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
2838 @findex ACCUMULATE_OUTGOING_ARGS
2839 @findex current_function_outgoing_args_size
2840 @item ACCUMULATE_OUTGOING_ARGS
2841 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
2842 will be computed and placed into the variable
2843 @code{current_function_outgoing_args_size}. No space will be pushed
2844 onto the stack for each call; instead, the function prologue should
2845 increase the stack frame size by this amount.
2847 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
2850 @findex REG_PARM_STACK_SPACE
2851 @item REG_PARM_STACK_SPACE (@var{fndecl})
2852 Define this macro if functions should assume that stack space has been
2853 allocated for arguments even when their values are passed in
2856 The value of this macro is the size, in bytes, of the area reserved for
2857 arguments passed in registers for the function represented by @var{fndecl},
2858 which can be zero if GCC is calling a library function.
2860 This space can be allocated by the caller, or be a part of the
2861 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
2863 @c above is overfull. not sure what to do. --mew 5feb93 did
2864 @c something, not sure if it looks good. --mew 10feb93
2866 @findex MAYBE_REG_PARM_STACK_SPACE
2867 @findex FINAL_REG_PARM_STACK_SPACE
2868 @item MAYBE_REG_PARM_STACK_SPACE
2869 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
2870 Define these macros in addition to the one above if functions might
2871 allocate stack space for arguments even when their values are passed
2872 in registers. These should be used when the stack space allocated
2873 for arguments in registers is not a simple constant independent of the
2874 function declaration.
2876 The value of the first macro is the size, in bytes, of the area that
2877 we should initially assume would be reserved for arguments passed in registers.
2879 The value of the second macro is the actual size, in bytes, of the area
2880 that will be reserved for arguments passed in registers. This takes two
2881 arguments: an integer representing the number of bytes of fixed sized
2882 arguments on the stack, and a tree representing the number of bytes of
2883 variable sized arguments on the stack.
2885 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
2886 called for libcall functions, the current function, or for a function
2887 being called when it is known that such stack space must be allocated.
2888 In each case this value can be easily computed.
2890 When deciding whether a called function needs such stack space, and how
2891 much space to reserve, GCC uses these two macros instead of
2892 @code{REG_PARM_STACK_SPACE}.
2894 @findex OUTGOING_REG_PARM_STACK_SPACE
2895 @item OUTGOING_REG_PARM_STACK_SPACE
2896 Define this if it is the responsibility of the caller to allocate the area
2897 reserved for arguments passed in registers.
2899 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
2900 whether the space for these arguments counts in the value of
2901 @code{current_function_outgoing_args_size}.
2903 @findex STACK_PARMS_IN_REG_PARM_AREA
2904 @item STACK_PARMS_IN_REG_PARM_AREA
2905 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
2906 stack parameters don't skip the area specified by it.
2907 @c i changed this, makes more sens and it should have taken care of the
2908 @c overfull.. not as specific, tho. --mew 5feb93
2910 Normally, when a parameter is not passed in registers, it is placed on the
2911 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
2912 suppresses this behavior and causes the parameter to be passed on the
2913 stack in its natural location.
2915 @findex RETURN_POPS_ARGS
2916 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
2917 A C expression that should indicate the number of bytes of its own
2918 arguments that a function pops on returning, or 0 if the
2919 function pops no arguments and the caller must therefore pop them all
2920 after the function returns.
2922 @var{fundecl} is a C variable whose value is a tree node that describes
2923 the function in question. Normally it is a node of type
2924 @code{FUNCTION_DECL} that describes the declaration of the function.
2925 From this you can obtain the DECL_MACHINE_ATTRIBUTES of the function.
2927 @var{funtype} is a C variable whose value is a tree node that
2928 describes the function in question. Normally it is a node of type
2929 @code{FUNCTION_TYPE} that describes the data type of the function.
2930 From this it is possible to obtain the data types of the value and
2931 arguments (if known).
2933 When a call to a library function is being considered, @var{fundecl}
2934 will contain an identifier node for the library function. Thus, if
2935 you need to distinguish among various library functions, you can do so
2936 by their names. Note that ``library function'' in this context means
2937 a function used to perform arithmetic, whose name is known specially
2938 in the compiler and was not mentioned in the C code being compiled.
2940 @var{stack-size} is the number of bytes of arguments passed on the
2941 stack. If a variable number of bytes is passed, it is zero, and
2942 argument popping will always be the responsibility of the calling function.
2944 On the Vax, all functions always pop their arguments, so the definition
2945 of this macro is @var{stack-size}. On the 68000, using the standard
2946 calling convention, no functions pop their arguments, so the value of
2947 the macro is always 0 in this case. But an alternative calling
2948 convention is available in which functions that take a fixed number of
2949 arguments pop them but other functions (such as @code{printf}) pop
2950 nothing (the caller pops all). When this convention is in use,
2951 @var{funtype} is examined to determine whether a function takes a fixed
2952 number of arguments.
2955 @node Register Arguments
2956 @subsection Passing Arguments in Registers
2957 @cindex arguments in registers
2958 @cindex registers arguments
2960 This section describes the macros which let you control how various
2961 types of arguments are passed in registers or how they are arranged in
2965 @findex FUNCTION_ARG
2966 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2967 A C expression that controls whether a function argument is passed
2968 in a register, and which register.
2970 The arguments are @var{cum}, which summarizes all the previous
2971 arguments; @var{mode}, the machine mode of the argument; @var{type},
2972 the data type of the argument as a tree node or 0 if that is not known
2973 (which happens for C support library functions); and @var{named},
2974 which is 1 for an ordinary argument and 0 for nameless arguments that
2975 correspond to @samp{@dots{}} in the called function's prototype.
2976 @var{type} can be an incomplete type if a syntax error has previously
2979 The value of the expression is usually either a @code{reg} RTX for the
2980 hard register in which to pass the argument, or zero to pass the
2981 argument on the stack.
2983 For machines like the Vax and 68000, where normally all arguments are
2984 pushed, zero suffices as a definition.
2986 The value of the expression can also be a @code{parallel} RTX. This is
2987 used when an argument is passed in multiple locations. The mode of the
2988 of the @code{parallel} should be the mode of the entire argument. The
2989 @code{parallel} holds any number of @code{expr_list} pairs; each one
2990 describes where part of the argument is passed. In each
2991 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
2992 register in which to pass this part of the argument, and the mode of the
2993 register RTX indicates how large this part of the argument is. The
2994 second operand of the @code{expr_list} is a @code{const_int} which gives
2995 the offset in bytes into the entire argument of where this part starts.
2996 As a special exception the first @code{expr_list} in the @code{parallel}
2997 RTX may have a first operand of zero. This indicates that the entire
2998 argument is also stored on the stack.
3000 @cindex @file{stdarg.h} and register arguments
3001 The usual way to make the ISO library @file{stdarg.h} work on a machine
3002 where some arguments are usually passed in registers, is to cause
3003 nameless arguments to be passed on the stack instead. This is done
3004 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3006 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3007 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3008 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3009 in the definition of this macro to determine if this argument is of a
3010 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3011 is not defined and @code{FUNCTION_ARG} returns non-zero for such an
3012 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3013 defined, the argument will be computed in the stack and then loaded into
3016 @findex MUST_PASS_IN_STACK
3017 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
3018 Define as a C expression that evaluates to nonzero if we do not know how
3019 to pass TYPE solely in registers. The file @file{expr.h} defines a
3020 definition that is usually appropriate, refer to @file{expr.h} for additional
3023 @findex FUNCTION_INCOMING_ARG
3024 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3025 Define this macro if the target machine has ``register windows'', so
3026 that the register in which a function sees an arguments is not
3027 necessarily the same as the one in which the caller passed the
3030 For such machines, @code{FUNCTION_ARG} computes the register in which
3031 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3032 be defined in a similar fashion to tell the function being called
3033 where the arguments will arrive.
3035 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3036 serves both purposes.@refill
3038 @findex FUNCTION_ARG_PARTIAL_NREGS
3039 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3040 A C expression for the number of words, at the beginning of an
3041 argument, that must be put in registers. The value must be zero for
3042 arguments that are passed entirely in registers or that are entirely
3043 pushed on the stack.
3045 On some machines, certain arguments must be passed partially in
3046 registers and partially in memory. On these machines, typically the
3047 first @var{n} words of arguments are passed in registers, and the rest
3048 on the stack. If a multi-word argument (a @code{double} or a
3049 structure) crosses that boundary, its first few words must be passed
3050 in registers and the rest must be pushed. This macro tells the
3051 compiler when this occurs, and how many of the words should go in
3054 @code{FUNCTION_ARG} for these arguments should return the first
3055 register to be used by the caller for this argument; likewise
3056 @code{FUNCTION_INCOMING_ARG}, for the called function.
3058 @findex FUNCTION_ARG_PASS_BY_REFERENCE
3059 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3060 A C expression that indicates when an argument must be passed by reference.
3061 If nonzero for an argument, a copy of that argument is made in memory and a
3062 pointer to the argument is passed instead of the argument itself.
3063 The pointer is passed in whatever way is appropriate for passing a pointer
3066 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3067 definition of this macro might be
3069 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3070 (CUM, MODE, TYPE, NAMED) \
3071 MUST_PASS_IN_STACK (MODE, TYPE)
3073 @c this is *still* too long. --mew 5feb93
3075 @findex FUNCTION_ARG_CALLEE_COPIES
3076 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3077 If defined, a C expression that indicates when it is the called function's
3078 responsibility to make a copy of arguments passed by invisible reference.
3079 Normally, the caller makes a copy and passes the address of the copy to the
3080 routine being called. When FUNCTION_ARG_CALLEE_COPIES is defined and is
3081 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3082 ``live'' value. The called function must not modify this value. If it can be
3083 determined that the value won't be modified, it need not make a copy;
3084 otherwise a copy must be made.
3086 @findex CUMULATIVE_ARGS
3087 @item CUMULATIVE_ARGS
3088 A C type for declaring a variable that is used as the first argument of
3089 @code{FUNCTION_ARG} and other related values. For some target machines,
3090 the type @code{int} suffices and can hold the number of bytes of
3093 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3094 arguments that have been passed on the stack. The compiler has other
3095 variables to keep track of that. For target machines on which all
3096 arguments are passed on the stack, there is no need to store anything in
3097 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3098 should not be empty, so use @code{int}.
3100 @findex INIT_CUMULATIVE_ARGS
3101 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
3102 A C statement (sans semicolon) for initializing the variable @var{cum}
3103 for the state at the beginning of the argument list. The variable has
3104 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
3105 for the data type of the function which will receive the args, or 0
3106 if the args are to a compiler support library function. The value of
3107 @var{indirect} is nonzero when processing an indirect call, for example
3108 a call through a function pointer. The value of @var{indirect} is zero
3109 for a call to an explicitly named function, a library function call, or when
3110 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3113 When processing a call to a compiler support library function,
3114 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3115 contains the name of the function, as a string. @var{libname} is 0 when
3116 an ordinary C function call is being processed. Thus, each time this
3117 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3118 never both of them at once.
3120 @findex INIT_CUMULATIVE_LIBCALL_ARGS
3121 @item INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3122 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3123 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3124 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3125 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3126 0)} is used instead.
3128 @findex INIT_CUMULATIVE_INCOMING_ARGS
3129 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3130 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3131 finding the arguments for the function being compiled. If this macro is
3132 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3134 The value passed for @var{libname} is always 0, since library routines
3135 with special calling conventions are never compiled with GCC. The
3136 argument @var{libname} exists for symmetry with
3137 @code{INIT_CUMULATIVE_ARGS}.
3138 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3139 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3141 @findex FUNCTION_ARG_ADVANCE
3142 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3143 A C statement (sans semicolon) to update the summarizer variable
3144 @var{cum} to advance past an argument in the argument list. The
3145 values @var{mode}, @var{type} and @var{named} describe that argument.
3146 Once this is done, the variable @var{cum} is suitable for analyzing
3147 the @emph{following} argument with @code{FUNCTION_ARG}, etc.@refill
3149 This macro need not do anything if the argument in question was passed
3150 on the stack. The compiler knows how to track the amount of stack space
3151 used for arguments without any special help.
3153 @findex FUNCTION_ARG_PADDING
3154 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3155 If defined, a C expression which determines whether, and in which direction,
3156 to pad out an argument with extra space. The value should be of type
3157 @code{enum direction}: either @code{upward} to pad above the argument,
3158 @code{downward} to pad below, or @code{none} to inhibit padding.
3160 The @emph{amount} of padding is always just enough to reach the next
3161 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3164 This macro has a default definition which is right for most systems.
3165 For little-endian machines, the default is to pad upward. For
3166 big-endian machines, the default is to pad downward for an argument of
3167 constant size shorter than an @code{int}, and upward otherwise.
3169 @findex PAD_VARARGS_DOWN
3170 @item PAD_VARARGS_DOWN
3171 If defined, a C expression which determines whether the default
3172 implementation of va_arg will attempt to pad down before reading the
3173 next argument, if that argument is smaller than its aligned space as
3174 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3175 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3177 @findex FUNCTION_ARG_BOUNDARY
3178 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3179 If defined, a C expression that gives the alignment boundary, in bits,
3180 of an argument with the specified mode and type. If it is not defined,
3181 @code{PARM_BOUNDARY} is used for all arguments.
3183 @findex FUNCTION_ARG_REGNO_P
3184 @item FUNCTION_ARG_REGNO_P (@var{regno})
3185 A C expression that is nonzero if @var{regno} is the number of a hard
3186 register in which function arguments are sometimes passed. This does
3187 @emph{not} include implicit arguments such as the static chain and
3188 the structure-value address. On many machines, no registers can be
3189 used for this purpose since all function arguments are pushed on the
3192 @findex LOAD_ARGS_REVERSED
3193 @item LOAD_ARGS_REVERSED
3194 If defined, the order in which arguments are loaded into their
3195 respective argument registers is reversed so that the last
3196 argument is loaded first. This macro only affects arguments
3197 passed in registers.
3202 @subsection How Scalar Function Values Are Returned
3203 @cindex return values in registers
3204 @cindex values, returned by functions
3205 @cindex scalars, returned as values
3207 This section discusses the macros that control returning scalars as
3208 values---values that can fit in registers.
3211 @findex TRADITIONAL_RETURN_FLOAT
3212 @item TRADITIONAL_RETURN_FLOAT
3213 Define this macro if @samp{-traditional} should not cause functions
3214 declared to return @code{float} to convert the value to @code{double}.
3216 @findex FUNCTION_VALUE
3217 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3218 A C expression to create an RTX representing the place where a
3219 function returns a value of data type @var{valtype}. @var{valtype} is
3220 a tree node representing a data type. Write @code{TYPE_MODE
3221 (@var{valtype})} to get the machine mode used to represent that type.
3222 On many machines, only the mode is relevant. (Actually, on most
3223 machines, scalar values are returned in the same place regardless of
3226 The value of the expression is usually a @code{reg} RTX for the hard
3227 register where the return value is stored. The value can also be a
3228 @code{parallel} RTX, if the return value is in multiple places. See
3229 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3231 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3232 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3235 If the precise function being called is known, @var{func} is a tree
3236 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3237 pointer. This makes it possible to use a different value-returning
3238 convention for specific functions when all their calls are
3241 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3242 types, because these are returned in another way. See
3243 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3245 @findex FUNCTION_OUTGOING_VALUE
3246 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3247 Define this macro if the target machine has ``register windows''
3248 so that the register in which a function returns its value is not
3249 the same as the one in which the caller sees the value.
3251 For such machines, @code{FUNCTION_VALUE} computes the register in which
3252 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3253 defined in a similar fashion to tell the function where to put the
3256 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3257 @code{FUNCTION_VALUE} serves both purposes.@refill
3259 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3260 aggregate data types, because these are returned in another way. See
3261 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3263 @findex LIBCALL_VALUE
3264 @item LIBCALL_VALUE (@var{mode})
3265 A C expression to create an RTX representing the place where a library
3266 function returns a value of mode @var{mode}. If the precise function
3267 being called is known, @var{func} is a tree node
3268 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3269 pointer. This makes it possible to use a different value-returning
3270 convention for specific functions when all their calls are
3273 Note that ``library function'' in this context means a compiler
3274 support routine, used to perform arithmetic, whose name is known
3275 specially by the compiler and was not mentioned in the C code being
3278 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3279 data types, because none of the library functions returns such types.
3281 @findex FUNCTION_VALUE_REGNO_P
3282 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3283 A C expression that is nonzero if @var{regno} is the number of a hard
3284 register in which the values of called function may come back.
3286 A register whose use for returning values is limited to serving as the
3287 second of a pair (for a value of type @code{double}, say) need not be
3288 recognized by this macro. So for most machines, this definition
3292 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3295 If the machine has register windows, so that the caller and the called
3296 function use different registers for the return value, this macro
3297 should recognize only the caller's register numbers.
3299 @findex APPLY_RESULT_SIZE
3300 @item APPLY_RESULT_SIZE
3301 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3302 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3303 saving and restoring an arbitrary return value.
3306 @node Aggregate Return
3307 @subsection How Large Values Are Returned
3308 @cindex aggregates as return values
3309 @cindex large return values
3310 @cindex returning aggregate values
3311 @cindex structure value address
3313 When a function value's mode is @code{BLKmode} (and in some other
3314 cases), the value is not returned according to @code{FUNCTION_VALUE}
3315 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3316 block of memory in which the value should be stored. This address
3317 is called the @dfn{structure value address}.
3319 This section describes how to control returning structure values in
3323 @findex RETURN_IN_MEMORY
3324 @item RETURN_IN_MEMORY (@var{type})
3325 A C expression which can inhibit the returning of certain function
3326 values in registers, based on the type of value. A nonzero value says
3327 to return the function value in memory, just as large structures are
3328 always returned. Here @var{type} will be a C expression of type
3329 @code{tree}, representing the data type of the value.
3331 Note that values of mode @code{BLKmode} must be explicitly handled
3332 by this macro. Also, the option @samp{-fpcc-struct-return}
3333 takes effect regardless of this macro. On most systems, it is
3334 possible to leave the macro undefined; this causes a default
3335 definition to be used, whose value is the constant 1 for @code{BLKmode}
3336 values, and 0 otherwise.
3338 Do not use this macro to indicate that structures and unions should always
3339 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3342 @findex DEFAULT_PCC_STRUCT_RETURN
3343 @item DEFAULT_PCC_STRUCT_RETURN
3344 Define this macro to be 1 if all structure and union return values must be
3345 in memory. Since this results in slower code, this should be defined
3346 only if needed for compatibility with other compilers or with an ABI.
3347 If you define this macro to be 0, then the conventions used for structure
3348 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3350 If not defined, this defaults to the value 1.
3352 @findex STRUCT_VALUE_REGNUM
3353 @item STRUCT_VALUE_REGNUM
3354 If the structure value address is passed in a register, then
3355 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3357 @findex STRUCT_VALUE
3359 If the structure value address is not passed in a register, define
3360 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3361 where the address is passed. If it returns 0, the address is passed as
3362 an ``invisible'' first argument.
3364 @findex STRUCT_VALUE_INCOMING_REGNUM
3365 @item STRUCT_VALUE_INCOMING_REGNUM
3366 On some architectures the place where the structure value address
3367 is found by the called function is not the same place that the
3368 caller put it. This can be due to register windows, or it could
3369 be because the function prologue moves it to a different place.
3371 If the incoming location of the structure value address is in a
3372 register, define this macro as the register number.
3374 @findex STRUCT_VALUE_INCOMING
3375 @item STRUCT_VALUE_INCOMING
3376 If the incoming location is not a register, then you should define
3377 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3378 called function should find the value. If it should find the value on
3379 the stack, define this to create a @code{mem} which refers to the frame
3380 pointer. A definition of 0 means that the address is passed as an
3381 ``invisible'' first argument.
3383 @findex PCC_STATIC_STRUCT_RETURN
3384 @item PCC_STATIC_STRUCT_RETURN
3385 Define this macro if the usual system convention on the target machine
3386 for returning structures and unions is for the called function to return
3387 the address of a static variable containing the value.
3389 Do not define this if the usual system convention is for the caller to
3390 pass an address to the subroutine.
3392 This macro has effect in @samp{-fpcc-struct-return} mode, but it does
3393 nothing when you use @samp{-freg-struct-return} mode.
3397 @subsection Caller-Saves Register Allocation
3399 If you enable it, GCC can save registers around function calls. This
3400 makes it possible to use call-clobbered registers to hold variables that
3401 must live across calls.
3404 @findex DEFAULT_CALLER_SAVES
3405 @item DEFAULT_CALLER_SAVES
3406 Define this macro if function calls on the target machine do not preserve
3407 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3408 for all registers. When defined, this macro enables @samp{-fcaller-saves}
3409 by default for all optimization levels. It has no effect for optimization
3410 levels 2 and higher, where @samp{-fcaller-saves} is the default.
3412 @findex CALLER_SAVE_PROFITABLE
3413 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3414 A C expression to determine whether it is worthwhile to consider placing
3415 a pseudo-register in a call-clobbered hard register and saving and
3416 restoring it around each function call. The expression should be 1 when
3417 this is worth doing, and 0 otherwise.
3419 If you don't define this macro, a default is used which is good on most
3420 machines: @code{4 * @var{calls} < @var{refs}}.
3422 @findex HARD_REGNO_CALLER_SAVE_MODE
3423 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3424 A C expression specifying which mode is required for saving @var{nregs}
3425 of a pseudo-register in call-clobbered hard register @var{regno}. If
3426 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3427 returned. For most machines this macro need not be defined since GCC
3428 will select the smallest suitable mode.
3431 @node Function Entry
3432 @subsection Function Entry and Exit
3433 @cindex function entry and exit
3437 This section describes the macros that output function entry
3438 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3441 @findex FUNCTION_PROLOGUE
3442 @item FUNCTION_PROLOGUE (@var{file}, @var{size})
3443 A C compound statement that outputs the assembler code for entry to a
3444 function. The prologue is responsible for setting up the stack frame,
3445 initializing the frame pointer register, saving registers that must be
3446 saved, and allocating @var{size} additional bytes of storage for the
3447 local variables. @var{size} is an integer. @var{file} is a stdio
3448 stream to which the assembler code should be output.
3450 The label for the beginning of the function need not be output by this
3451 macro. That has already been done when the macro is run.
3453 @findex regs_ever_live
3454 To determine which registers to save, the macro can refer to the array
3455 @code{regs_ever_live}: element @var{r} is nonzero if hard register
3456 @var{r} is used anywhere within the function. This implies the function
3457 prologue should save register @var{r}, provided it is not one of the
3458 call-used registers. (@code{FUNCTION_EPILOGUE} must likewise use
3459 @code{regs_ever_live}.)
3461 On machines that have ``register windows'', the function entry code does
3462 not save on the stack the registers that are in the windows, even if
3463 they are supposed to be preserved by function calls; instead it takes
3464 appropriate steps to ``push'' the register stack, if any non-call-used
3465 registers are used in the function.
3467 @findex frame_pointer_needed
3468 On machines where functions may or may not have frame-pointers, the
3469 function entry code must vary accordingly; it must set up the frame
3470 pointer if one is wanted, and not otherwise. To determine whether a
3471 frame pointer is in wanted, the macro can refer to the variable
3472 @code{frame_pointer_needed}. The variable's value will be 1 at run
3473 time in a function that needs a frame pointer. @xref{Elimination}.
3475 The function entry code is responsible for allocating any stack space
3476 required for the function. This stack space consists of the regions
3477 listed below. In most cases, these regions are allocated in the
3478 order listed, with the last listed region closest to the top of the
3479 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
3480 the highest address if it is not defined). You can use a different order
3481 for a machine if doing so is more convenient or required for
3482 compatibility reasons. Except in cases where required by standard
3483 or by a debugger, there is no reason why the stack layout used by GCC
3484 need agree with that used by other compilers for a machine.
3488 @findex current_function_pretend_args_size
3489 A region of @code{current_function_pretend_args_size} bytes of
3490 uninitialized space just underneath the first argument arriving on the
3491 stack. (This may not be at the very start of the allocated stack region
3492 if the calling sequence has pushed anything else since pushing the stack
3493 arguments. But usually, on such machines, nothing else has been pushed
3494 yet, because the function prologue itself does all the pushing.) This
3495 region is used on machines where an argument may be passed partly in
3496 registers and partly in memory, and, in some cases to support the
3497 features in @file{varargs.h} and @file{stdargs.h}.
3500 An area of memory used to save certain registers used by the function.
3501 The size of this area, which may also include space for such things as
3502 the return address and pointers to previous stack frames, is
3503 machine-specific and usually depends on which registers have been used
3504 in the function. Machines with register windows often do not require
3508 A region of at least @var{size} bytes, possibly rounded up to an allocation
3509 boundary, to contain the local variables of the function. On some machines,
3510 this region and the save area may occur in the opposite order, with the
3511 save area closer to the top of the stack.
3514 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3515 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3516 @code{current_function_outgoing_args_size} bytes to be used for outgoing
3517 argument lists of the function. @xref{Stack Arguments}.
3520 Normally, it is necessary for the macros @code{FUNCTION_PROLOGUE} and
3521 @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C
3522 variable @code{current_function_is_leaf} is nonzero for such a function.
3524 @findex EXIT_IGNORE_STACK
3525 @item EXIT_IGNORE_STACK
3526 Define this macro as a C expression that is nonzero if the return
3527 instruction or the function epilogue ignores the value of the stack
3528 pointer; in other words, if it is safe to delete an instruction to
3529 adjust the stack pointer before a return from the function.
3531 Note that this macro's value is relevant only for functions for which
3532 frame pointers are maintained. It is never safe to delete a final
3533 stack adjustment in a function that has no frame pointer, and the
3534 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3536 @findex EPILOGUE_USES
3537 @item EPILOGUE_USES (@var{regno})
3538 Define this macro as a C expression that is nonzero for registers that are
3539 used by the epilogue or the @samp{return} pattern. The stack and frame
3540 pointer registers are already be assumed to be used as needed.
3542 @findex FUNCTION_EPILOGUE
3543 @item FUNCTION_EPILOGUE (@var{file}, @var{size})
3544 A C compound statement that outputs the assembler code for exit from a
3545 function. The epilogue is responsible for restoring the saved
3546 registers and stack pointer to their values when the function was
3547 called, and returning control to the caller. This macro takes the
3548 same arguments as the macro @code{FUNCTION_PROLOGUE}, and the
3549 registers to restore are determined from @code{regs_ever_live} and
3550 @code{CALL_USED_REGISTERS} in the same way.
3552 On some machines, there is a single instruction that does all the work
3553 of returning from the function. On these machines, give that
3554 instruction the name @samp{return} and do not define the macro
3555 @code{FUNCTION_EPILOGUE} at all.
3557 Do not define a pattern named @samp{return} if you want the
3558 @code{FUNCTION_EPILOGUE} to be used. If you want the target switches
3559 to control whether return instructions or epilogues are used, define a
3560 @samp{return} pattern with a validity condition that tests the target
3561 switches appropriately. If the @samp{return} pattern's validity
3562 condition is false, epilogues will be used.
3564 On machines where functions may or may not have frame-pointers, the
3565 function exit code must vary accordingly. Sometimes the code for these
3566 two cases is completely different. To determine whether a frame pointer
3567 is wanted, the macro can refer to the variable
3568 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
3569 a function that needs a frame pointer.
3571 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
3572 treat leaf functions specially. The C variable @code{current_function_is_leaf}
3573 is nonzero for such a function. @xref{Leaf Functions}.
3575 On some machines, some functions pop their arguments on exit while
3576 others leave that for the caller to do. For example, the 68020 when
3577 given @samp{-mrtd} pops arguments in functions that take a fixed
3578 number of arguments.
3580 @findex current_function_pops_args
3581 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
3582 functions pop their own arguments. @code{FUNCTION_EPILOGUE} needs to
3583 know what was decided. The variable that is called
3584 @code{current_function_pops_args} is the number of bytes of its
3585 arguments that a function should pop. @xref{Scalar Return}.
3586 @c what is the "its arguments" in the above sentence referring to, pray
3587 @c tell? --mew 5feb93
3589 @findex DELAY_SLOTS_FOR_EPILOGUE
3590 @item DELAY_SLOTS_FOR_EPILOGUE
3591 Define this macro if the function epilogue contains delay slots to which
3592 instructions from the rest of the function can be ``moved''. The
3593 definition should be a C expression whose value is an integer
3594 representing the number of delay slots there.
3596 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
3597 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
3598 A C expression that returns 1 if @var{insn} can be placed in delay
3599 slot number @var{n} of the epilogue.
3601 The argument @var{n} is an integer which identifies the delay slot now
3602 being considered (since different slots may have different rules of
3603 eligibility). It is never negative and is always less than the number
3604 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
3605 If you reject a particular insn for a given delay slot, in principle, it
3606 may be reconsidered for a subsequent delay slot. Also, other insns may
3607 (at least in principle) be considered for the so far unfilled delay
3610 @findex current_function_epilogue_delay_list
3611 @findex final_scan_insn
3612 The insns accepted to fill the epilogue delay slots are put in an RTL
3613 list made with @code{insn_list} objects, stored in the variable
3614 @code{current_function_epilogue_delay_list}. The insn for the first
3615 delay slot comes first in the list. Your definition of the macro
3616 @code{FUNCTION_EPILOGUE} should fill the delay slots by outputting the
3617 insns in this list, usually by calling @code{final_scan_insn}.
3619 You need not define this macro if you did not define
3620 @code{DELAY_SLOTS_FOR_EPILOGUE}.
3622 @findex ASM_OUTPUT_MI_THUNK
3623 @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function})
3624 A C compound statement that outputs the assembler code for a thunk
3625 function, used to implement C++ virtual function calls with multiple
3626 inheritance. The thunk acts as a wrapper around a virtual function,
3627 adjusting the implicit object parameter before handing control off to
3630 First, emit code to add the integer @var{delta} to the location that
3631 contains the incoming first argument. Assume that this argument
3632 contains a pointer, and is the one used to pass the @code{this} pointer
3633 in C++. This is the incoming argument @emph{before} the function prologue,
3634 e.g. @samp{%o0} on a sparc. The addition must preserve the values of
3635 all other incoming arguments.
3637 After the addition, emit code to jump to @var{function}, which is a
3638 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
3639 not touch the return address. Hence returning from @var{FUNCTION} will
3640 return to whoever called the current @samp{thunk}.
3642 The effect must be as if @var{function} had been called directly with
3643 the adjusted first argument. This macro is responsible for emitting all
3644 of the code for a thunk function; @code{FUNCTION_PROLOGUE} and
3645 @code{FUNCTION_EPILOGUE} are not invoked.
3647 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
3648 have already been extracted from it.) It might possibly be useful on
3649 some targets, but probably not.
3651 If you do not define this macro, the target-independent code in the C++
3652 frontend will generate a less efficient heavyweight thunk that calls
3653 @var{function} instead of jumping to it. The generic approach does
3654 not support varargs.
3658 @subsection Generating Code for Profiling
3659 @cindex profiling, code generation
3661 These macros will help you generate code for profiling.
3664 @findex FUNCTION_PROFILER
3665 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
3666 A C statement or compound statement to output to @var{file} some
3667 assembler code to call the profiling subroutine @code{mcount}.
3670 The details of how @code{mcount} expects to be called are determined by
3671 your operating system environment, not by GCC. To figure them out,
3672 compile a small program for profiling using the system's installed C
3673 compiler and look at the assembler code that results.
3675 Older implementations of @code{mcount} expect the address of a counter
3676 variable to be loaded into some register. The name of this variable is
3677 @samp{LP} followed by the number @var{labelno}, so you would generate
3678 the name using @samp{LP%d} in a @code{fprintf}.
3680 @findex NO_PROFILE_COUNTERS
3681 @item NO_PROFILE_COUNTERS
3682 Define this macro if the @code{mcount} subroutine on your system does
3683 not need a counter variable allocated for each function. This is true
3684 for almost all modern implementations. If you define this macro, you
3685 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
3687 @findex PROFILE_BEFORE_PROLOGUE
3688 @item PROFILE_BEFORE_PROLOGUE
3689 Define this macro if the code for function profiling should come before
3690 the function prologue. Normally, the profiling code comes after.
3692 @findex FUNCTION_BLOCK_PROFILER
3693 @vindex profile_block_flag
3694 @item FUNCTION_BLOCK_PROFILER (@var{file}, @var{labelno})
3695 A C statement or compound statement to output to @var{file} some
3696 assembler code to initialize basic-block profiling for the current
3697 object module. The global compile flag @code{profile_block_flag}
3698 distinguishes two profile modes.
3701 @findex __bb_init_func
3702 @item profile_block_flag != 2
3703 Output code to call the subroutine @code{__bb_init_func} once per
3704 object module, passing it as its sole argument the address of a block
3705 allocated in the object module.
3707 The name of the block is a local symbol made with this statement:
3710 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3713 Of course, since you are writing the definition of
3714 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3715 can take a short cut in the definition of this macro and use the name
3716 that you know will result.
3718 The first word of this block is a flag which will be nonzero if the
3719 object module has already been initialized. So test this word first,
3720 and do not call @code{__bb_init_func} if the flag is
3721 nonzero. BLOCK_OR_LABEL contains a unique number which may be used to
3722 generate a label as a branch destination when @code{__bb_init_func}
3725 Described in assembler language, the code to be output looks like:
3735 @findex __bb_init_trace_func
3736 @item profile_block_flag == 2
3737 Output code to call the subroutine @code{__bb_init_trace_func}
3738 and pass two parameters to it. The first parameter is the same as
3739 for @code{__bb_init_func}. The second parameter is the number of the
3740 first basic block of the function as given by BLOCK_OR_LABEL. Note
3741 that @code{__bb_init_trace_func} has to be called, even if the object
3742 module has been initialized already.
3744 Described in assembler language, the code to be output looks like:
3747 parameter2 <- BLOCK_OR_LABEL
3748 call __bb_init_trace_func
3752 @findex BLOCK_PROFILER
3753 @vindex profile_block_flag
3754 @item BLOCK_PROFILER (@var{file}, @var{blockno})
3755 A C statement or compound statement to output to @var{file} some
3756 assembler code to increment the count associated with the basic
3757 block number @var{blockno}. The global compile flag
3758 @code{profile_block_flag} distinguishes two profile modes.
3761 @item profile_block_flag != 2
3762 Output code to increment the counter directly. Basic blocks are
3763 numbered separately from zero within each compilation. The count
3764 associated with block number @var{blockno} is at index
3765 @var{blockno} in a vector of words; the name of this array is a local
3766 symbol made with this statement:
3769 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 2);
3772 @c This paragraph is the same as one a few paragraphs up.
3773 @c That is not an error.
3774 Of course, since you are writing the definition of
3775 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3776 can take a short cut in the definition of this macro and use the name
3777 that you know will result.
3779 Described in assembler language, the code to be output looks like:
3782 inc (LPBX2+4*BLOCKNO)
3786 @findex __bb_trace_func
3787 @item profile_block_flag == 2
3788 Output code to initialize the global structure @code{__bb} and
3789 call the function @code{__bb_trace_func}, which will increment the
3792 @code{__bb} consists of two words. In the first word, the current
3793 basic block number, as given by BLOCKNO, has to be stored. In
3794 the second word, the address of a block allocated in the object
3795 module has to be stored. The address is given by the label created
3796 with this statement:
3799 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3802 Described in assembler language, the code to be output looks like:
3804 move BLOCKNO -> (__bb)
3805 move LPBX0 -> (__bb+4)
3806 call __bb_trace_func
3810 @findex FUNCTION_BLOCK_PROFILER_EXIT
3811 @findex __bb_trace_ret
3812 @vindex profile_block_flag
3813 @item FUNCTION_BLOCK_PROFILER_EXIT (@var{file})
3814 A C statement or compound statement to output to @var{file}
3815 assembler code to call function @code{__bb_trace_ret}. The
3816 assembler code should only be output
3817 if the global compile flag @code{profile_block_flag} == 2. This
3818 macro has to be used at every place where code for returning from
3819 a function is generated (e.g. @code{FUNCTION_EPILOGUE}). Although
3820 you have to write the definition of @code{FUNCTION_EPILOGUE}
3821 as well, you have to define this macro to tell the compiler, that
3822 the proper call to @code{__bb_trace_ret} is produced.
3824 @findex MACHINE_STATE_SAVE
3825 @findex __bb_init_trace_func
3826 @findex __bb_trace_func
3827 @findex __bb_trace_ret
3828 @item MACHINE_STATE_SAVE (@var{id})
3829 A C statement or compound statement to save all registers, which may
3830 be clobbered by a function call, including condition codes. The
3831 @code{asm} statement will be mostly likely needed to handle this
3832 task. Local labels in the assembler code can be concatenated with the
3833 string @var{id}, to obtain a unique label name.
3835 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3836 @code{FUNCTION_EPILOGUE} must be saved in the macros
3837 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3838 @code{BLOCK_PROFILER} prior calling @code{__bb_init_trace_func},
3839 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3841 @findex MACHINE_STATE_RESTORE
3842 @findex __bb_init_trace_func
3843 @findex __bb_trace_func
3844 @findex __bb_trace_ret
3845 @item MACHINE_STATE_RESTORE (@var{id})
3846 A C statement or compound statement to restore all registers, including
3847 condition codes, saved by @code{MACHINE_STATE_SAVE}.
3849 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3850 @code{FUNCTION_EPILOGUE} must be restored in the macros
3851 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3852 @code{BLOCK_PROFILER} after calling @code{__bb_init_trace_func},
3853 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3855 @findex BLOCK_PROFILER_CODE
3856 @item BLOCK_PROFILER_CODE
3857 A C function or functions which are needed in the library to
3858 support block profiling.
3862 @subsection Permitting inlining of functions with attributes
3865 By default if a function has a target specific attribute attached to it,
3866 it will not be inlined. This behaviour can be overridden if the target
3867 defines the @samp{FUNCTION_ATTRIBUTE_INLINABLE_P} macro. This macro
3868 takes one argument, a @samp{DECL} describing the function. It should
3869 return non-zero if the function can be inlined, otherwise it should
3873 @subsection Permitting tail calls to functions
3875 @cindex sibling calls
3878 @findex FUNCTION_OK_FOR_SIBCALL
3879 @item FUNCTION_OK_FOR_SIBCALL (@var{decl})
3880 A C expression that evaluates to true if it is ok to perform a sibling
3883 It is not uncommon for limitations of calling conventions to prevent
3884 tail calls to functions outside the current unit of translation, or
3885 during PIC compilation. Use this macro to enforce these restrictions,
3886 as the @code{sibcall} md pattern can not fail, or fall over to a
3891 @section Implementing the Varargs Macros
3892 @cindex varargs implementation
3894 GCC comes with an implementation of @file{varargs.h} and
3895 @file{stdarg.h} that work without change on machines that pass arguments
3896 on the stack. Other machines require their own implementations of
3897 varargs, and the two machine independent header files must have
3898 conditionals to include it.
3900 ISO @file{stdarg.h} differs from traditional @file{varargs.h} mainly in
3901 the calling convention for @code{va_start}. The traditional
3902 implementation takes just one argument, which is the variable in which
3903 to store the argument pointer. The ISO implementation of
3904 @code{va_start} takes an additional second argument. The user is
3905 supposed to write the last named argument of the function here.
3907 However, @code{va_start} should not use this argument. The way to find
3908 the end of the named arguments is with the built-in functions described
3912 @findex __builtin_saveregs
3913 @item __builtin_saveregs ()
3914 Use this built-in function to save the argument registers in memory so
3915 that the varargs mechanism can access them. Both ISO and traditional
3916 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3917 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
3919 On some machines, @code{__builtin_saveregs} is open-coded under the
3920 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
3921 it calls a routine written in assembler language, found in
3924 Code generated for the call to @code{__builtin_saveregs} appears at the
3925 beginning of the function, as opposed to where the call to
3926 @code{__builtin_saveregs} is written, regardless of what the code is.
3927 This is because the registers must be saved before the function starts
3928 to use them for its own purposes.
3929 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3932 @findex __builtin_args_info
3933 @item __builtin_args_info (@var{category})
3934 Use this built-in function to find the first anonymous arguments in
3937 In general, a machine may have several categories of registers used for
3938 arguments, each for a particular category of data types. (For example,
3939 on some machines, floating-point registers are used for floating-point
3940 arguments while other arguments are passed in the general registers.)
3941 To make non-varargs functions use the proper calling convention, you
3942 have defined the @code{CUMULATIVE_ARGS} data type to record how many
3943 registers in each category have been used so far
3945 @code{__builtin_args_info} accesses the same data structure of type
3946 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
3947 with it, with @var{category} specifying which word to access. Thus, the
3948 value indicates the first unused register in a given category.
3950 Normally, you would use @code{__builtin_args_info} in the implementation
3951 of @code{va_start}, accessing each category just once and storing the
3952 value in the @code{va_list} object. This is because @code{va_list} will
3953 have to update the values, and there is no way to alter the
3954 values accessed by @code{__builtin_args_info}.
3956 @findex __builtin_next_arg
3957 @item __builtin_next_arg (@var{lastarg})
3958 This is the equivalent of @code{__builtin_args_info}, for stack
3959 arguments. It returns the address of the first anonymous stack
3960 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
3961 returns the address of the location above the first anonymous stack
3962 argument. Use it in @code{va_start} to initialize the pointer for
3963 fetching arguments from the stack. Also use it in @code{va_start} to
3964 verify that the second parameter @var{lastarg} is the last named argument
3965 of the current function.
3967 @findex __builtin_classify_type
3968 @item __builtin_classify_type (@var{object})
3969 Since each machine has its own conventions for which data types are
3970 passed in which kind of register, your implementation of @code{va_arg}
3971 has to embody these conventions. The easiest way to categorize the
3972 specified data type is to use @code{__builtin_classify_type} together
3973 with @code{sizeof} and @code{__alignof__}.
3975 @code{__builtin_classify_type} ignores the value of @var{object},
3976 considering only its data type. It returns an integer describing what
3977 kind of type that is---integer, floating, pointer, structure, and so on.
3979 The file @file{typeclass.h} defines an enumeration that you can use to
3980 interpret the values of @code{__builtin_classify_type}.
3983 These machine description macros help implement varargs:
3986 @findex EXPAND_BUILTIN_SAVEREGS
3987 @item EXPAND_BUILTIN_SAVEREGS ()
3988 If defined, is a C expression that produces the machine-specific code
3989 for a call to @code{__builtin_saveregs}. This code will be moved to the
3990 very beginning of the function, before any parameter access are made.
3991 The return value of this function should be an RTX that contains the
3992 value to use as the return of @code{__builtin_saveregs}.
3994 @findex SETUP_INCOMING_VARARGS
3995 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
3996 This macro offers an alternative to using @code{__builtin_saveregs} and
3997 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
3998 anonymous register arguments into the stack so that all the arguments
3999 appear to have been passed consecutively on the stack. Once this is
4000 done, you can use the standard implementation of varargs that works for
4001 machines that pass all their arguments on the stack.
4003 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
4004 structure, containing the values that are obtained after processing the
4005 named arguments. The arguments @var{mode} and @var{type} describe the
4006 last named argument---its machine mode and its data type as a tree node.
4008 The macro implementation should do two things: first, push onto the
4009 stack all the argument registers @emph{not} used for the named
4010 arguments, and second, store the size of the data thus pushed into the
4011 @code{int}-valued variable whose name is supplied as the argument
4012 @var{pretend_args_size}. The value that you store here will serve as
4013 additional offset for setting up the stack frame.
4015 Because you must generate code to push the anonymous arguments at
4016 compile time without knowing their data types,
4017 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
4018 a single category of argument register and use it uniformly for all data
4021 If the argument @var{second_time} is nonzero, it means that the
4022 arguments of the function are being analyzed for the second time. This
4023 happens for an inline function, which is not actually compiled until the
4024 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
4025 not generate any instructions in this case.
4027 @findex STRICT_ARGUMENT_NAMING
4028 @item STRICT_ARGUMENT_NAMING
4029 Define this macro to be a nonzero value if the location where a function
4030 argument is passed depends on whether or not it is a named argument.
4032 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
4033 is set for varargs and stdarg functions. If this macro returns a
4034 nonzero value, the @var{named} argument is always true for named
4035 arguments, and false for unnamed arguments. If it returns a value of
4036 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
4037 are treated as named. Otherwise, all named arguments except the last
4038 are treated as named.
4040 You need not define this macro if it always returns zero.
4042 @findex PRETEND_OUTGOING_VARARGS_NAMED
4043 @item PRETEND_OUTGOING_VARARGS_NAMED
4044 If you need to conditionally change ABIs so that one works with
4045 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
4046 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
4047 defined, then define this macro to return nonzero if
4048 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
4049 Otherwise, you should not define this macro.
4053 @section Trampolines for Nested Functions
4054 @cindex trampolines for nested functions
4055 @cindex nested functions, trampolines for
4057 A @dfn{trampoline} is a small piece of code that is created at run time
4058 when the address of a nested function is taken. It normally resides on
4059 the stack, in the stack frame of the containing function. These macros
4060 tell GCC how to generate code to allocate and initialize a
4063 The instructions in the trampoline must do two things: load a constant
4064 address into the static chain register, and jump to the real address of
4065 the nested function. On CISC machines such as the m68k, this requires
4066 two instructions, a move immediate and a jump. Then the two addresses
4067 exist in the trampoline as word-long immediate operands. On RISC
4068 machines, it is often necessary to load each address into a register in
4069 two parts. Then pieces of each address form separate immediate
4072 The code generated to initialize the trampoline must store the variable
4073 parts---the static chain value and the function address---into the
4074 immediate operands of the instructions. On a CISC machine, this is
4075 simply a matter of copying each address to a memory reference at the
4076 proper offset from the start of the trampoline. On a RISC machine, it
4077 may be necessary to take out pieces of the address and store them
4081 @findex TRAMPOLINE_TEMPLATE
4082 @item TRAMPOLINE_TEMPLATE (@var{file})
4083 A C statement to output, on the stream @var{file}, assembler code for a
4084 block of data that contains the constant parts of a trampoline. This
4085 code should not include a label---the label is taken care of
4088 If you do not define this macro, it means no template is needed
4089 for the target. Do not define this macro on systems where the block move
4090 code to copy the trampoline into place would be larger than the code
4091 to generate it on the spot.
4093 @findex TRAMPOLINE_SECTION
4094 @item TRAMPOLINE_SECTION
4095 The name of a subroutine to switch to the section in which the
4096 trampoline template is to be placed (@pxref{Sections}). The default is
4097 a value of @samp{readonly_data_section}, which places the trampoline in
4098 the section containing read-only data.
4100 @findex TRAMPOLINE_SIZE
4101 @item TRAMPOLINE_SIZE
4102 A C expression for the size in bytes of the trampoline, as an integer.
4104 @findex TRAMPOLINE_ALIGNMENT
4105 @item TRAMPOLINE_ALIGNMENT
4106 Alignment required for trampolines, in bits.
4108 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4109 is used for aligning trampolines.
4111 @findex INITIALIZE_TRAMPOLINE
4112 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4113 A C statement to initialize the variable parts of a trampoline.
4114 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4115 an RTX for the address of the nested function; @var{static_chain} is an
4116 RTX for the static chain value that should be passed to the function
4119 @findex TRAMPOLINE_ADJUST_ADDRESS
4120 @item TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4121 A C statement that should perform any machine-specific adjustment in
4122 the address of the trampoline. Its argument contains the address that
4123 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4124 used for a function call should be different from the address in which
4125 the template was stored, the different address should be assigned to
4126 @var{addr}. If this macro is not defined, @var{addr} will be used for
4129 @findex ALLOCATE_TRAMPOLINE
4130 @item ALLOCATE_TRAMPOLINE (@var{fp})
4131 A C expression to allocate run-time space for a trampoline. The
4132 expression value should be an RTX representing a memory reference to the
4133 space for the trampoline.
4135 @cindex @code{FUNCTION_EPILOGUE} and trampolines
4136 @cindex @code{FUNCTION_PROLOGUE} and trampolines
4137 If this macro is not defined, by default the trampoline is allocated as
4138 a stack slot. This default is right for most machines. The exceptions
4139 are machines where it is impossible to execute instructions in the stack
4140 area. On such machines, you may have to implement a separate stack,
4141 using this macro in conjunction with @code{FUNCTION_PROLOGUE} and
4142 @code{FUNCTION_EPILOGUE}.
4144 @var{fp} points to a data structure, a @code{struct function}, which
4145 describes the compilation status of the immediate containing function of
4146 the function which the trampoline is for. Normally (when
4147 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
4148 trampoline is in the stack frame of this containing function. Other
4149 allocation strategies probably must do something analogous with this
4153 Implementing trampolines is difficult on many machines because they have
4154 separate instruction and data caches. Writing into a stack location
4155 fails to clear the memory in the instruction cache, so when the program
4156 jumps to that location, it executes the old contents.
4158 Here are two possible solutions. One is to clear the relevant parts of
4159 the instruction cache whenever a trampoline is set up. The other is to
4160 make all trampolines identical, by having them jump to a standard
4161 subroutine. The former technique makes trampoline execution faster; the
4162 latter makes initialization faster.
4164 To clear the instruction cache when a trampoline is initialized, define
4165 the following macros which describe the shape of the cache.
4168 @findex INSN_CACHE_SIZE
4169 @item INSN_CACHE_SIZE
4170 The total size in bytes of the cache.
4172 @findex INSN_CACHE_LINE_WIDTH
4173 @item INSN_CACHE_LINE_WIDTH
4174 The length in bytes of each cache line. The cache is divided into cache
4175 lines which are disjoint slots, each holding a contiguous chunk of data
4176 fetched from memory. Each time data is brought into the cache, an
4177 entire line is read at once. The data loaded into a cache line is
4178 always aligned on a boundary equal to the line size.
4180 @findex INSN_CACHE_DEPTH
4181 @item INSN_CACHE_DEPTH
4182 The number of alternative cache lines that can hold any particular memory
4186 Alternatively, if the machine has system calls or instructions to clear
4187 the instruction cache directly, you can define the following macro.
4190 @findex CLEAR_INSN_CACHE
4191 @item CLEAR_INSN_CACHE (@var{BEG}, @var{END})
4192 If defined, expands to a C expression clearing the @emph{instruction
4193 cache} in the specified interval. If it is not defined, and the macro
4194 INSN_CACHE_SIZE is defined, some generic code is generated to clear the
4195 cache. The definition of this macro would typically be a series of
4196 @code{asm} statements. Both @var{BEG} and @var{END} are both pointer
4200 To use a standard subroutine, define the following macro. In addition,
4201 you must make sure that the instructions in a trampoline fill an entire
4202 cache line with identical instructions, or else ensure that the
4203 beginning of the trampoline code is always aligned at the same point in
4204 its cache line. Look in @file{m68k.h} as a guide.
4207 @findex TRANSFER_FROM_TRAMPOLINE
4208 @item TRANSFER_FROM_TRAMPOLINE
4209 Define this macro if trampolines need a special subroutine to do their
4210 work. The macro should expand to a series of @code{asm} statements
4211 which will be compiled with GCC. They go in a library function named
4212 @code{__transfer_from_trampoline}.
4214 If you need to avoid executing the ordinary prologue code of a compiled
4215 C function when you jump to the subroutine, you can do so by placing a
4216 special label of your own in the assembler code. Use one @code{asm}
4217 statement to generate an assembler label, and another to make the label
4218 global. Then trampolines can use that label to jump directly to your
4219 special assembler code.
4223 @section Implicit Calls to Library Routines
4224 @cindex library subroutine names
4225 @cindex @file{libgcc.a}
4227 @c prevent bad page break with this line
4228 Here is an explanation of implicit calls to library routines.
4231 @findex MULSI3_LIBCALL
4232 @item MULSI3_LIBCALL
4233 A C string constant giving the name of the function to call for
4234 multiplication of one signed full-word by another. If you do not
4235 define this macro, the default name is used, which is @code{__mulsi3},
4236 a function defined in @file{libgcc.a}.
4238 @findex DIVSI3_LIBCALL
4239 @item DIVSI3_LIBCALL
4240 A C string constant giving the name of the function to call for
4241 division of one signed full-word by another. If you do not define
4242 this macro, the default name is used, which is @code{__divsi3}, a
4243 function defined in @file{libgcc.a}.
4245 @findex UDIVSI3_LIBCALL
4246 @item UDIVSI3_LIBCALL
4247 A C string constant giving the name of the function to call for
4248 division of one unsigned full-word by another. If you do not define
4249 this macro, the default name is used, which is @code{__udivsi3}, a
4250 function defined in @file{libgcc.a}.
4252 @findex MODSI3_LIBCALL
4253 @item MODSI3_LIBCALL
4254 A C string constant giving the name of the function to call for the
4255 remainder in division of one signed full-word by another. If you do
4256 not define this macro, the default name is used, which is
4257 @code{__modsi3}, a function defined in @file{libgcc.a}.
4259 @findex UMODSI3_LIBCALL
4260 @item UMODSI3_LIBCALL
4261 A C string constant giving the name of the function to call for the
4262 remainder in division of one unsigned full-word by another. If you do
4263 not define this macro, the default name is used, which is
4264 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4266 @findex MULDI3_LIBCALL
4267 @item MULDI3_LIBCALL
4268 A C string constant giving the name of the function to call for
4269 multiplication of one signed double-word by another. If you do not
4270 define this macro, the default name is used, which is @code{__muldi3},
4271 a function defined in @file{libgcc.a}.
4273 @findex DIVDI3_LIBCALL
4274 @item DIVDI3_LIBCALL
4275 A C string constant giving the name of the function to call for
4276 division of one signed double-word by another. If you do not define
4277 this macro, the default name is used, which is @code{__divdi3}, a
4278 function defined in @file{libgcc.a}.
4280 @findex UDIVDI3_LIBCALL
4281 @item UDIVDI3_LIBCALL
4282 A C string constant giving the name of the function to call for
4283 division of one unsigned full-word by another. If you do not define
4284 this macro, the default name is used, which is @code{__udivdi3}, a
4285 function defined in @file{libgcc.a}.
4287 @findex MODDI3_LIBCALL
4288 @item MODDI3_LIBCALL
4289 A C string constant giving the name of the function to call for the
4290 remainder in division of one signed double-word by another. If you do
4291 not define this macro, the default name is used, which is
4292 @code{__moddi3}, a function defined in @file{libgcc.a}.
4294 @findex UMODDI3_LIBCALL
4295 @item UMODDI3_LIBCALL
4296 A C string constant giving the name of the function to call for the
4297 remainder in division of one unsigned full-word by another. If you do
4298 not define this macro, the default name is used, which is
4299 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4301 @findex INIT_TARGET_OPTABS
4302 @item INIT_TARGET_OPTABS
4303 Define this macro as a C statement that declares additional library
4304 routines renames existing ones. @code{init_optabs} calls this macro after
4305 initializing all the normal library routines.
4307 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4308 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4309 Define this macro as a C statement that returns nonzero if a call to
4310 the floating point comparison library function will return a boolean
4311 value that indicates the result of the comparison. It should return
4312 zero if one of gcc's own libgcc functions is called.
4314 Most ports don't need to define this macro.
4317 @cindex @code{EDOM}, implicit usage
4319 The value of @code{EDOM} on the target machine, as a C integer constant
4320 expression. If you don't define this macro, GCC does not attempt to
4321 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4322 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4325 If you do not define @code{TARGET_EDOM}, then compiled code reports
4326 domain errors by calling the library function and letting it report the
4327 error. If mathematical functions on your system use @code{matherr} when
4328 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4329 that @code{matherr} is used normally.
4331 @findex GEN_ERRNO_RTX
4332 @cindex @code{errno}, implicit usage
4334 Define this macro as a C expression to create an rtl expression that
4335 refers to the global ``variable'' @code{errno}. (On certain systems,
4336 @code{errno} may not actually be a variable.) If you don't define this
4337 macro, a reasonable default is used.
4339 @findex TARGET_MEM_FUNCTIONS
4340 @cindex @code{bcopy}, implicit usage
4341 @cindex @code{memcpy}, implicit usage
4342 @cindex @code{bzero}, implicit usage
4343 @cindex @code{memset}, implicit usage
4344 @item TARGET_MEM_FUNCTIONS
4345 Define this macro if GCC should generate calls to the ISO C
4346 (and System V) library functions @code{memcpy} and @code{memset}
4347 rather than the BSD functions @code{bcopy} and @code{bzero}.
4349 @findex LIBGCC_NEEDS_DOUBLE
4350 @item LIBGCC_NEEDS_DOUBLE
4351 Define this macro if only @code{float} arguments cannot be passed to
4352 library routines (so they must be converted to @code{double}). This
4353 macro affects both how library calls are generated and how the library
4354 routines in @file{libgcc1.c} accept their arguments. It is useful on
4355 machines where floating and fixed point arguments are passed
4356 differently, such as the i860.
4358 @findex FLOAT_ARG_TYPE
4359 @item FLOAT_ARG_TYPE
4360 Define this macro to override the type used by the library routines to
4361 pick up arguments of type @code{float}. (By default, they use a union
4362 of @code{float} and @code{int}.)
4364 The obvious choice would be @code{float}---but that won't work with
4365 traditional C compilers that expect all arguments declared as @code{float}
4366 to arrive as @code{double}. To avoid this conversion, the library routines
4367 ask for the value as some other type and then treat it as a @code{float}.
4369 On some systems, no other type will work for this. For these systems,
4370 you must use @code{LIBGCC_NEEDS_DOUBLE} instead, to force conversion of
4371 the values @code{double} before they are passed.
4374 @item FLOATIFY (@var{passed-value})
4375 Define this macro to override the way library routines redesignate a
4376 @code{float} argument as a @code{float} instead of the type it was
4377 passed as. The default is an expression which takes the @code{float}
4380 @findex FLOAT_VALUE_TYPE
4381 @item FLOAT_VALUE_TYPE
4382 Define this macro to override the type used by the library routines to
4383 return values that ought to have type @code{float}. (By default, they
4386 The obvious choice would be @code{float}---but that won't work with
4387 traditional C compilers gratuitously convert values declared as
4388 @code{float} into @code{double}.
4391 @item INTIFY (@var{float-value})
4392 Define this macro to override the way the value of a
4393 @code{float}-returning library routine should be packaged in order to
4394 return it. These functions are actually declared to return type
4395 @code{FLOAT_VALUE_TYPE} (normally @code{int}).
4397 These values can't be returned as type @code{float} because traditional
4398 C compilers would gratuitously convert the value to a @code{double}.
4400 A local variable named @code{intify} is always available when the macro
4401 @code{INTIFY} is used. It is a union of a @code{float} field named
4402 @code{f} and a field named @code{i} whose type is
4403 @code{FLOAT_VALUE_TYPE} or @code{int}.
4405 If you don't define this macro, the default definition works by copying
4406 the value through that union.
4408 @findex nongcc_SI_type
4409 @item nongcc_SI_type
4410 Define this macro as the name of the data type corresponding to
4411 @code{SImode} in the system's own C compiler.
4413 You need not define this macro if that type is @code{long int}, as it usually
4416 @findex nongcc_word_type
4417 @item nongcc_word_type
4418 Define this macro as the name of the data type corresponding to the
4419 word_mode in the system's own C compiler.
4421 You need not define this macro if that type is @code{long int}, as it usually
4424 @findex perform_@dots{}
4425 @item perform_@dots{}
4426 Define these macros to supply explicit C statements to carry out various
4427 arithmetic operations on types @code{float} and @code{double} in the
4428 library routines in @file{libgcc1.c}. See that file for a full list
4429 of these macros and their arguments.
4431 On most machines, you don't need to define any of these macros, because
4432 the C compiler that comes with the system takes care of doing them.
4434 @findex NEXT_OBJC_RUNTIME
4435 @item NEXT_OBJC_RUNTIME
4436 Define this macro to generate code for Objective C message sending using
4437 the calling convention of the NeXT system. This calling convention
4438 involves passing the object, the selector and the method arguments all
4439 at once to the method-lookup library function.
4441 The default calling convention passes just the object and the selector
4442 to the lookup function, which returns a pointer to the method.
4445 @node Addressing Modes
4446 @section Addressing Modes
4447 @cindex addressing modes
4449 @c prevent bad page break with this line
4450 This is about addressing modes.
4453 @findex HAVE_PRE_INCREMENT
4454 @findex HAVE_PRE_DECREMENT
4455 @findex HAVE_POST_INCREMENT
4456 @findex HAVE_POST_DECREMENT
4457 @item HAVE_PRE_INCREMENT
4458 @itemx HAVE_PRE_DECREMENT
4459 @itemx HAVE_POST_INCREMENT
4460 @itemx HAVE_POST_DECREMENT
4461 A C expression that is non-zero if the machine supports pre-increment,
4462 pre-decrement, post-increment, or post-decrement addressing respectively.
4464 @findex HAVE_POST_MODIFY_DISP
4465 @findex HAVE_PRE_MODIFY_DISP
4466 @item HAVE_PRE_MODIFY_DISP
4467 @itemx HAVE_POST_MODIFY_DISP
4468 A C expression that is non-zero if the machine supports pre- or
4469 post-address side-effect generation involving constants other than
4470 the size of the memory operand.
4472 @findex HAVE_POST_MODIFY_REG
4473 @findex HAVE_PRE_MODIFY_REG
4474 @item HAVE_PRE_MODIFY_REG
4475 @itemx HAVE_POST_MODIFY_REG
4476 A C expression that is non-zero if the machine supports pre- or
4477 post-address side-effect generation involving a register displacement.
4479 @findex CONSTANT_ADDRESS_P
4480 @item CONSTANT_ADDRESS_P (@var{x})
4481 A C expression that is 1 if the RTX @var{x} is a constant which
4482 is a valid address. On most machines, this can be defined as
4483 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4484 in which constant addresses are supported.
4487 @code{CONSTANT_P} accepts integer-values expressions whose values are
4488 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4489 @code{high} expressions and @code{const} arithmetic expressions, in
4490 addition to @code{const_int} and @code{const_double} expressions.
4492 @findex MAX_REGS_PER_ADDRESS
4493 @item MAX_REGS_PER_ADDRESS
4494 A number, the maximum number of registers that can appear in a valid
4495 memory address. Note that it is up to you to specify a value equal to
4496 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4499 @findex GO_IF_LEGITIMATE_ADDRESS
4500 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4501 A C compound statement with a conditional @code{goto @var{label};}
4502 executed if @var{x} (an RTX) is a legitimate memory address on the
4503 target machine for a memory operand of mode @var{mode}.
4505 It usually pays to define several simpler macros to serve as
4506 subroutines for this one. Otherwise it may be too complicated to
4509 This macro must exist in two variants: a strict variant and a
4510 non-strict one. The strict variant is used in the reload pass. It
4511 must be defined so that any pseudo-register that has not been
4512 allocated a hard register is considered a memory reference. In
4513 contexts where some kind of register is required, a pseudo-register
4514 with no hard register must be rejected.
4516 The non-strict variant is used in other passes. It must be defined to
4517 accept all pseudo-registers in every context where some kind of
4518 register is required.
4520 @findex REG_OK_STRICT
4521 Compiler source files that want to use the strict variant of this
4522 macro define the macro @code{REG_OK_STRICT}. You should use an
4523 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4524 in that case and the non-strict variant otherwise.
4526 Subroutines to check for acceptable registers for various purposes (one
4527 for base registers, one for index registers, and so on) are typically
4528 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4529 Then only these subroutine macros need have two variants; the higher
4530 levels of macros may be the same whether strict or not.@refill
4532 Normally, constant addresses which are the sum of a @code{symbol_ref}
4533 and an integer are stored inside a @code{const} RTX to mark them as
4534 constant. Therefore, there is no need to recognize such sums
4535 specifically as legitimate addresses. Normally you would simply
4536 recognize any @code{const} as legitimate.
4538 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4539 sums that are not marked with @code{const}. It assumes that a naked
4540 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4541 naked constant sums as illegitimate addresses, so that none of them will
4542 be given to @code{PRINT_OPERAND_ADDRESS}.
4544 @cindex @code{ENCODE_SECTION_INFO} and address validation
4545 On some machines, whether a symbolic address is legitimate depends on
4546 the section that the address refers to. On these machines, define the
4547 macro @code{ENCODE_SECTION_INFO} to store the information into the
4548 @code{symbol_ref}, and then check for it here. When you see a
4549 @code{const}, you will have to look inside it to find the
4550 @code{symbol_ref} in order to determine the section. @xref{Assembler
4553 @findex saveable_obstack
4554 The best way to modify the name string is by adding text to the
4555 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4556 the new name in @code{saveable_obstack}. You will have to modify
4557 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4558 output the name accordingly, and define @code{STRIP_NAME_ENCODING} to
4559 access the original name string.
4561 You can check the information stored here into the @code{symbol_ref} in
4562 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4563 @code{PRINT_OPERAND_ADDRESS}.
4565 @findex REG_OK_FOR_BASE_P
4566 @item REG_OK_FOR_BASE_P (@var{x})
4567 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4568 RTX) is valid for use as a base register. For hard registers, it
4569 should always accept those which the hardware permits and reject the
4570 others. Whether the macro accepts or rejects pseudo registers must be
4571 controlled by @code{REG_OK_STRICT} as described above. This usually
4572 requires two variant definitions, of which @code{REG_OK_STRICT}
4573 controls the one actually used.
4575 @findex REG_MODE_OK_FOR_BASE_P
4576 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4577 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4578 that expression may examine the mode of the memory reference in
4579 @var{mode}. You should define this macro if the mode of the memory
4580 reference affects whether a register may be used as a base register. If
4581 you define this macro, the compiler will use it instead of
4582 @code{REG_OK_FOR_BASE_P}.
4584 @findex REG_OK_FOR_INDEX_P
4585 @item REG_OK_FOR_INDEX_P (@var{x})
4586 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4587 RTX) is valid for use as an index register.
4589 The difference between an index register and a base register is that
4590 the index register may be scaled. If an address involves the sum of
4591 two registers, neither one of them scaled, then either one may be
4592 labeled the ``base'' and the other the ``index''; but whichever
4593 labeling is used must fit the machine's constraints of which registers
4594 may serve in each capacity. The compiler will try both labelings,
4595 looking for one that is valid, and will reload one or both registers
4596 only if neither labeling works.
4598 @findex FIND_BASE_TERM
4599 @item FIND_BASE_TERM (@var{x})
4600 A C expression to determine the base term of address @var{x}.
4601 This macro is used in only one place: `find_base_term' in alias.c.
4603 It is always safe for this macro to not be defined. It exists so
4604 that alias analysis can understand machine-dependent addresses.
4606 The typical use of this macro is to handle addresses containing
4607 a label_ref or symbol_ref within an UNSPEC.
4609 @findex LEGITIMIZE_ADDRESS
4610 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4611 A C compound statement that attempts to replace @var{x} with a valid
4612 memory address for an operand of mode @var{mode}. @var{win} will be a
4613 C statement label elsewhere in the code; the macro definition may use
4616 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4620 to avoid further processing if the address has become legitimate.
4622 @findex break_out_memory_refs
4623 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4624 and @var{oldx} will be the operand that was given to that function to produce
4627 The code generated by this macro should not alter the substructure of
4628 @var{x}. If it transforms @var{x} into a more legitimate form, it
4629 should assign @var{x} (which will always be a C variable) a new value.
4631 It is not necessary for this macro to come up with a legitimate
4632 address. The compiler has standard ways of doing so in all cases. In
4633 fact, it is safe for this macro to do nothing. But often a
4634 machine-dependent strategy can generate better code.
4636 @findex LEGITIMIZE_RELOAD_ADDRESS
4637 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4638 A C compound statement that attempts to replace @var{x}, which is an address
4639 that needs reloading, with a valid memory address for an operand of mode
4640 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4641 It is not necessary to define this macro, but it might be useful for
4642 performance reasons.
4644 For example, on the i386, it is sometimes possible to use a single
4645 reload register instead of two by reloading a sum of two pseudo
4646 registers into a register. On the other hand, for number of RISC
4647 processors offsets are limited so that often an intermediate address
4648 needs to be generated in order to address a stack slot. By defining
4649 LEGITIMIZE_RELOAD_ADDRESS appropriately, the intermediate addresses
4650 generated for adjacent some stack slots can be made identical, and thus
4653 @emph{Note}: This macro should be used with caution. It is necessary
4654 to know something of how reload works in order to effectively use this,
4655 and it is quite easy to produce macros that build in too much knowledge
4656 of reload internals.
4658 @emph{Note}: This macro must be able to reload an address created by a
4659 previous invocation of this macro. If it fails to handle such addresses
4660 then the compiler may generate incorrect code or abort.
4663 The macro definition should use @code{push_reload} to indicate parts that
4664 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4665 suitable to be passed unaltered to @code{push_reload}.
4667 The code generated by this macro must not alter the substructure of
4668 @var{x}. If it transforms @var{x} into a more legitimate form, it
4669 should assign @var{x} (which will always be a C variable) a new value.
4670 This also applies to parts that you change indirectly by calling
4673 @findex strict_memory_address_p
4674 The macro definition may use @code{strict_memory_address_p} to test if
4675 the address has become legitimate.
4678 If you want to change only a part of @var{x}, one standard way of doing
4679 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4680 single level of rtl. Thus, if the part to be changed is not at the
4681 top level, you'll need to replace first the top leve
4682 It is not necessary for this macro to come up with a legitimate
4683 address; but often a machine-dependent strategy can generate better code.
4685 @findex GO_IF_MODE_DEPENDENT_ADDRESS
4686 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4687 A C statement or compound statement with a conditional @code{goto
4688 @var{label};} executed if memory address @var{x} (an RTX) can have
4689 different meanings depending on the machine mode of the memory
4690 reference it is used for or if the address is valid for some modes
4693 Autoincrement and autodecrement addresses typically have mode-dependent
4694 effects because the amount of the increment or decrement is the size
4695 of the operand being addressed. Some machines have other mode-dependent
4696 addresses. Many RISC machines have no mode-dependent addresses.
4698 You may assume that @var{addr} is a valid address for the machine.
4700 @findex LEGITIMATE_CONSTANT_P
4701 @item LEGITIMATE_CONSTANT_P (@var{x})
4702 A C expression that is nonzero if @var{x} is a legitimate constant for
4703 an immediate operand on the target machine. You can assume that
4704 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4705 @samp{1} is a suitable definition for this macro on machines where
4706 anything @code{CONSTANT_P} is valid.@refill
4709 @node Condition Code
4710 @section Condition Code Status
4711 @cindex condition code status
4713 @c prevent bad page break with this line
4714 This describes the condition code status.
4717 The file @file{conditions.h} defines a variable @code{cc_status} to
4718 describe how the condition code was computed (in case the interpretation of
4719 the condition code depends on the instruction that it was set by). This
4720 variable contains the RTL expressions on which the condition code is
4721 currently based, and several standard flags.
4723 Sometimes additional machine-specific flags must be defined in the machine
4724 description header file. It can also add additional machine-specific
4725 information by defining @code{CC_STATUS_MDEP}.
4728 @findex CC_STATUS_MDEP
4729 @item CC_STATUS_MDEP
4730 C code for a data type which is used for declaring the @code{mdep}
4731 component of @code{cc_status}. It defaults to @code{int}.
4733 This macro is not used on machines that do not use @code{cc0}.
4735 @findex CC_STATUS_MDEP_INIT
4736 @item CC_STATUS_MDEP_INIT
4737 A C expression to initialize the @code{mdep} field to ``empty''.
4738 The default definition does nothing, since most machines don't use
4739 the field anyway. If you want to use the field, you should probably
4740 define this macro to initialize it.
4742 This macro is not used on machines that do not use @code{cc0}.
4744 @findex NOTICE_UPDATE_CC
4745 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4746 A C compound statement to set the components of @code{cc_status}
4747 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4748 this macro's responsibility to recognize insns that set the condition
4749 code as a byproduct of other activity as well as those that explicitly
4752 This macro is not used on machines that do not use @code{cc0}.
4754 If there are insns that do not set the condition code but do alter
4755 other machine registers, this macro must check to see whether they
4756 invalidate the expressions that the condition code is recorded as
4757 reflecting. For example, on the 68000, insns that store in address
4758 registers do not set the condition code, which means that usually
4759 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4760 insns. But suppose that the previous insn set the condition code
4761 based on location @samp{a4@@(102)} and the current insn stores a new
4762 value in @samp{a4}. Although the condition code is not changed by
4763 this, it will no longer be true that it reflects the contents of
4764 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4765 @code{cc_status} in this case to say that nothing is known about the
4766 condition code value.
4768 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4769 with the results of peephole optimization: insns whose patterns are
4770 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4771 constants which are just the operands. The RTL structure of these
4772 insns is not sufficient to indicate what the insns actually do. What
4773 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4774 @code{CC_STATUS_INIT}.
4776 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4777 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4778 @samp{cc}. This avoids having detailed information about patterns in
4779 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4781 @findex EXTRA_CC_MODES
4782 @item EXTRA_CC_MODES
4783 A list of additional modes for condition code values in registers
4784 (@pxref{Jump Patterns}). This macro should expand to a sequence of
4785 calls of the macro @code{CC} separated by white space. @code{CC} takes
4786 two arguments. The first is the enumeration name of the mode, which
4787 should begin with @samp{CC} and end with @samp{mode}. The second is a C
4788 string giving the printable name of the mode; it should be the same as
4789 the first argument, but with the trailing @samp{mode} removed.
4791 You should only define this macro if additional modes are required.
4793 A sample definition of @code{EXTRA_CC_MODES} is:
4795 #define EXTRA_CC_MODES \
4796 CC(CC_NOOVmode, "CC_NOOV") \
4797 CC(CCFPmode, "CCFP") \
4798 CC(CCFPEmode, "CCFPE")
4801 @findex SELECT_CC_MODE
4802 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4803 Returns a mode from class @code{MODE_CC} to be used when comparison
4804 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
4805 example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
4806 @pxref{Jump Patterns} for a description of the reason for this
4810 #define SELECT_CC_MODE(OP,X,Y) \
4811 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4812 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
4813 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4814 || GET_CODE (X) == NEG) \
4815 ? CC_NOOVmode : CCmode))
4818 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
4820 @findex CANONICALIZE_COMPARISON
4821 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
4822 On some machines not all possible comparisons are defined, but you can
4823 convert an invalid comparison into a valid one. For example, the Alpha
4824 does not have a @code{GT} comparison, but you can use an @code{LT}
4825 comparison instead and swap the order of the operands.
4827 On such machines, define this macro to be a C statement to do any
4828 required conversions. @var{code} is the initial comparison code
4829 and @var{op0} and @var{op1} are the left and right operands of the
4830 comparison, respectively. You should modify @var{code}, @var{op0}, and
4831 @var{op1} as required.
4833 GCC will not assume that the comparison resulting from this macro is
4834 valid but will see if the resulting insn matches a pattern in the
4837 You need not define this macro if it would never change the comparison
4840 @findex REVERSIBLE_CC_MODE
4841 @item REVERSIBLE_CC_MODE (@var{mode})
4842 A C expression whose value is one if it is always safe to reverse a
4843 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4844 can ever return @var{mode} for a floating-point inequality comparison,
4845 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4847 You need not define this macro if it would always returns zero or if the
4848 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4849 For example, here is the definition used on the Sparc, where floating-point
4850 inequality comparisons are always given @code{CCFPEmode}:
4853 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
4856 @findex REVERSE_CONDITION (@var{code}, @var{mode})
4857 A C expression whose value is reversed condition code of the @var{code} for
4858 comparison done in CC_MODE @var{mode}. The macro is used only in case
4859 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
4860 machine has some non-standard way how to reverse certain conditionals. For
4861 instance in case all floating point conditions are non-trapping, compiler may
4862 freely convert unordered compares to ordered one. Then definition may look
4866 #define REVERSE_CONDITION(CODE, MODE) \
4867 ((MODE) != CCFPmode ? reverse_condtion (CODE) \
4868 : reverse_condition_maybe_unordered (CODE))
4874 @section Describing Relative Costs of Operations
4875 @cindex costs of instructions
4876 @cindex relative costs
4877 @cindex speed of instructions
4879 These macros let you describe the relative speed of various operations
4880 on the target machine.
4884 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
4885 A part of a C @code{switch} statement that describes the relative costs
4886 of constant RTL expressions. It must contain @code{case} labels for
4887 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
4888 @code{label_ref} and @code{const_double}. Each case must ultimately
4889 reach a @code{return} statement to return the relative cost of the use
4890 of that kind of constant value in an expression. The cost may depend on
4891 the precise value of the constant, which is available for examination in
4892 @var{x}, and the rtx code of the expression in which it is contained,
4893 found in @var{outer_code}.
4895 @var{code} is the expression code---redundant, since it can be
4896 obtained with @code{GET_CODE (@var{x})}.
4899 @findex COSTS_N_INSNS
4900 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4901 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
4902 This can be used, for example, to indicate how costly a multiply
4903 instruction is. In writing this macro, you can use the construct
4904 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
4905 instructions. @var{outer_code} is the code of the expression in which
4906 @var{x} is contained.
4908 This macro is optional; do not define it if the default cost assumptions
4909 are adequate for the target machine.
4911 @findex DEFAULT_RTX_COSTS
4912 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4913 This macro, if defined, is called for any case not handled by the
4914 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
4915 to put case labels into the macro, but the code, or any functions it
4916 calls, must assume that the RTL in @var{x} could be of any type that has
4917 not already been handled. The arguments are the same as for
4918 @code{RTX_COSTS}, and the macro should execute a return statement giving
4919 the cost of any RTL expressions that it can handle. The default cost
4920 calculation is used for any RTL for which this macro does not return a
4923 This macro is optional; do not define it if the default cost assumptions
4924 are adequate for the target machine.
4926 @findex ADDRESS_COST
4927 @item ADDRESS_COST (@var{address})
4928 An expression giving the cost of an addressing mode that contains
4929 @var{address}. If not defined, the cost is computed from
4930 the @var{address} expression and the @code{CONST_COSTS} values.
4932 For most CISC machines, the default cost is a good approximation of the
4933 true cost of the addressing mode. However, on RISC machines, all
4934 instructions normally have the same length and execution time. Hence
4935 all addresses will have equal costs.
4937 In cases where more than one form of an address is known, the form with
4938 the lowest cost will be used. If multiple forms have the same, lowest,
4939 cost, the one that is the most complex will be used.
4941 For example, suppose an address that is equal to the sum of a register
4942 and a constant is used twice in the same basic block. When this macro
4943 is not defined, the address will be computed in a register and memory
4944 references will be indirect through that register. On machines where
4945 the cost of the addressing mode containing the sum is no higher than
4946 that of a simple indirect reference, this will produce an additional
4947 instruction and possibly require an additional register. Proper
4948 specification of this macro eliminates this overhead for such machines.
4950 Similar use of this macro is made in strength reduction of loops.
4952 @var{address} need not be valid as an address. In such a case, the cost
4953 is not relevant and can be any value; invalid addresses need not be
4954 assigned a different cost.
4956 On machines where an address involving more than one register is as
4957 cheap as an address computation involving only one register, defining
4958 @code{ADDRESS_COST} to reflect this can cause two registers to be live
4959 over a region of code where only one would have been if
4960 @code{ADDRESS_COST} were not defined in that manner. This effect should
4961 be considered in the definition of this macro. Equivalent costs should
4962 probably only be given to addresses with different numbers of registers
4963 on machines with lots of registers.
4965 This macro will normally either not be defined or be defined as a
4968 @findex REGISTER_MOVE_COST
4969 @item REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
4970 A C expression for the cost of moving data of mode @var{mode} from a
4971 register in class @var{from} to one in class @var{to}. The classes are
4972 expressed using the enumeration values such as @code{GENERAL_REGS}. A
4973 value of 2 is the default; other values are interpreted relative to
4976 It is not required that the cost always equal 2 when @var{from} is the
4977 same as @var{to}; on some machines it is expensive to move between
4978 registers if they are not general registers.
4980 If reload sees an insn consisting of a single @code{set} between two
4981 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4982 classes returns a value of 2, reload does not check to ensure that the
4983 constraints of the insn are met. Setting a cost of other than 2 will
4984 allow reload to verify that the constraints are met. You should do this
4985 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4987 @findex MEMORY_MOVE_COST
4988 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4989 A C expression for the cost of moving data of mode @var{mode} between a
4990 register of class @var{class} and memory; @var{in} is zero if the value
4991 is to be written to memory, non-zero if it is to be read in. This cost
4992 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
4993 registers and memory is more expensive than between two registers, you
4994 should define this macro to express the relative cost.
4996 If you do not define this macro, GCC uses a default cost of 4 plus
4997 the cost of copying via a secondary reload register, if one is
4998 needed. If your machine requires a secondary reload register to copy
4999 between memory and a register of @var{class} but the reload mechanism is
5000 more complex than copying via an intermediate, define this macro to
5001 reflect the actual cost of the move.
5003 GCC defines the function @code{memory_move_secondary_cost} if
5004 secondary reloads are needed. It computes the costs due to copying via
5005 a secondary register. If your machine copies from memory using a
5006 secondary register in the conventional way but the default base value of
5007 4 is not correct for your machine, define this macro to add some other
5008 value to the result of that function. The arguments to that function
5009 are the same as to this macro.
5013 A C expression for the cost of a branch instruction. A value of 1 is
5014 the default; other values are interpreted relative to that.
5017 Here are additional macros which do not specify precise relative costs,
5018 but only that certain actions are more expensive than GCC would
5022 @findex SLOW_BYTE_ACCESS
5023 @item SLOW_BYTE_ACCESS
5024 Define this macro as a C expression which is nonzero if accessing less
5025 than a word of memory (i.e. a @code{char} or a @code{short}) is no
5026 faster than accessing a word of memory, i.e., if such access
5027 require more than one instruction or if there is no difference in cost
5028 between byte and (aligned) word loads.
5030 When this macro is not defined, the compiler will access a field by
5031 finding the smallest containing object; when it is defined, a fullword
5032 load will be used if alignment permits. Unless bytes accesses are
5033 faster than word accesses, using word accesses is preferable since it
5034 may eliminate subsequent memory access if subsequent accesses occur to
5035 other fields in the same word of the structure, but to different bytes.
5037 @findex SLOW_ZERO_EXTEND
5038 @item SLOW_ZERO_EXTEND
5039 Define this macro if zero-extension (of a @code{char} or @code{short}
5040 to an @code{int}) can be done faster if the destination is a register
5041 that is known to be zero.
5043 If you define this macro, you must have instruction patterns that
5044 recognize RTL structures like this:
5047 (set (strict_low_part (subreg:QI (reg:SI @dots{}) 0)) @dots{})
5051 and likewise for @code{HImode}.
5053 @findex SLOW_UNALIGNED_ACCESS
5054 @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5055 Define this macro to be the value 1 if memory accesses described by the
5056 @var{mode} and @var{alignment} parameters have a cost many times greater
5057 than aligned accesses, for example if they are emulated in a trap
5060 When this macro is non-zero, the compiler will act as if
5061 @code{STRICT_ALIGNMENT} were non-zero when generating code for block
5062 moves. This can cause significantly more instructions to be produced.
5063 Therefore, do not set this macro non-zero if unaligned accesses only add a
5064 cycle or two to the time for a memory access.
5066 If the value of this macro is always zero, it need not be defined. If
5067 this macro is defined, it should produce a non-zero value when
5068 @code{STRICT_ALIGNMENT} is non-zero.
5070 @findex DONT_REDUCE_ADDR
5071 @item DONT_REDUCE_ADDR
5072 Define this macro to inhibit strength reduction of memory addresses.
5073 (On some machines, such strength reduction seems to do harm rather
5078 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5079 which a sequence of insns should be generated instead of a
5080 string move insn or a library call. Increasing the value will always
5081 make code faster, but eventually incurs high cost in increased code size.
5083 Note that on machines where the corresponding move insn is a
5084 @code{define_expand} that emits a sequence of insns, this macro counts
5085 the number of such sequences.
5087 If you don't define this, a reasonable default is used.
5089 @findex MOVE_BY_PIECES_P
5090 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5091 A C expression used to determine whether @code{move_by_pieces} will be used to
5092 copy a chunk of memory, or whether some other block move mechanism
5093 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5094 than @code{MOVE_RATIO}.
5096 @findex MOVE_MAX_PIECES
5097 @item MOVE_MAX_PIECES
5098 A C expression used by @code{move_by_pieces} to determine the largest unit
5099 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5101 @findex USE_LOAD_POST_INCREMENT
5102 @item USE_LOAD_POST_INCREMENT (@var{mode})
5103 A C expression used to determine whether a load postincrement is a good
5104 thing to use for a given mode. Defaults to the value of
5105 @code{HAVE_POST_INCREMENT}.
5107 @findex USE_LOAD_POST_DECREMENT
5108 @item USE_LOAD_POST_DECREMENT (@var{mode})
5109 A C expression used to determine whether a load postdecrement is a good
5110 thing to use for a given mode. Defaults to the value of
5111 @code{HAVE_POST_DECREMENT}.
5113 @findex USE_LOAD_PRE_INCREMENT
5114 @item USE_LOAD_PRE_INCREMENT (@var{mode})
5115 A C expression used to determine whether a load preincrement is a good
5116 thing to use for a given mode. Defaults to the value of
5117 @code{HAVE_PRE_INCREMENT}.
5119 @findex USE_LOAD_PRE_DECREMENT
5120 @item USE_LOAD_PRE_DECREMENT (@var{mode})
5121 A C expression used to determine whether a load predecrement is a good
5122 thing to use for a given mode. Defaults to the value of
5123 @code{HAVE_PRE_DECREMENT}.
5125 @findex USE_STORE_POST_INCREMENT
5126 @item USE_STORE_POST_INCREMENT (@var{mode})
5127 A C expression used to determine whether a store postincrement is a good
5128 thing to use for a given mode. Defaults to the value of
5129 @code{HAVE_POST_INCREMENT}.
5131 @findex USE_STORE_POST_DECREMENT
5132 @item USE_STORE_POST_DECREMENT (@var{mode})
5133 A C expression used to determine whether a store postdeccrement is a good
5134 thing to use for a given mode. Defaults to the value of
5135 @code{HAVE_POST_DECREMENT}.
5137 @findex USE_STORE_PRE_INCREMENT
5138 @item USE_STORE_PRE_INCREMENT (@var{mode})
5139 This macro is used to determine whether a store preincrement is a good
5140 thing to use for a given mode. Defaults to the value of
5141 @code{HAVE_PRE_INCREMENT}.
5143 @findex USE_STORE_PRE_DECREMENT
5144 @item USE_STORE_PRE_DECREMENT (@var{mode})
5145 This macro is used to determine whether a store predecrement is a good
5146 thing to use for a given mode. Defaults to the value of
5147 @code{HAVE_PRE_DECREMENT}.
5149 @findex NO_FUNCTION_CSE
5150 @item NO_FUNCTION_CSE
5151 Define this macro if it is as good or better to call a constant
5152 function address than to call an address kept in a register.
5154 @findex NO_RECURSIVE_FUNCTION_CSE
5155 @item NO_RECURSIVE_FUNCTION_CSE
5156 Define this macro if it is as good or better for a function to call
5157 itself with an explicit address than to call an address kept in a
5161 @item ADJUST_COST (@var{insn}, @var{link}, @var{dep_insn}, @var{cost})
5162 A C statement (sans semicolon) to update the integer variable @var{cost}
5163 based on the relationship between @var{insn} that is dependent on
5164 @var{dep_insn} through the dependence @var{link}. The default is to
5165 make no adjustment to @var{cost}. This can be used for example to
5166 specify to the scheduler that an output- or anti-dependence does not
5167 incur the same cost as a data-dependence.
5169 @findex ADJUST_PRIORITY
5170 @item ADJUST_PRIORITY (@var{insn})
5171 A C statement (sans semicolon) to update the integer scheduling
5172 priority @code{INSN_PRIORITY(@var{insn})}. Reduce the priority
5173 to execute the @var{insn} earlier, increase the priority to execute
5174 @var{insn} later. Do not define this macro if you do not need to
5175 adjust the scheduling priorities of insns.
5179 @section Dividing the Output into Sections (Texts, Data, @dots{})
5180 @c the above section title is WAY too long. maybe cut the part between
5181 @c the (...)? --mew 10feb93
5183 An object file is divided into sections containing different types of
5184 data. In the most common case, there are three sections: the @dfn{text
5185 section}, which holds instructions and read-only data; the @dfn{data
5186 section}, which holds initialized writable data; and the @dfn{bss
5187 section}, which holds uninitialized data. Some systems have other kinds
5190 The compiler must tell the assembler when to switch sections. These
5191 macros control what commands to output to tell the assembler this. You
5192 can also define additional sections.
5195 @findex TEXT_SECTION_ASM_OP
5196 @item TEXT_SECTION_ASM_OP
5197 A C expression whose value is a string, including spacing, containing the
5198 assembler operation that should precede instructions and read-only data.
5199 Normally @code{"\t.text"} is right.
5201 @findex DATA_SECTION_ASM_OP
5202 @item DATA_SECTION_ASM_OP
5203 A C expression whose value is a string, including spacing, containing the
5204 assembler operation to identify the following data as writable initialized
5205 data. Normally @code{"\t.data"} is right.
5207 @findex SHARED_SECTION_ASM_OP
5208 @item SHARED_SECTION_ASM_OP
5209 If defined, a C expression whose value is a string, including spacing,
5210 containing the assembler operation to identify the following data as
5211 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5213 @findex BSS_SECTION_ASM_OP
5214 @item BSS_SECTION_ASM_OP
5215 If defined, a C expression whose value is a string, including spacing,
5216 containing the assembler operation to identify the following data as
5217 uninitialized global data. If not defined, and neither
5218 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5219 uninitialized global data will be output in the data section if
5220 @samp{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5223 @findex SHARED_BSS_SECTION_ASM_OP
5224 @item SHARED_BSS_SECTION_ASM_OP
5225 If defined, a C expression whose value is a string, including spacing,
5226 containing the assembler operation to identify the following data as
5227 uninitialized global shared data. If not defined, and
5228 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5230 @findex INIT_SECTION_ASM_OP
5231 @item INIT_SECTION_ASM_OP
5232 If defined, a C expression whose value is a string, including spacing,
5233 containing the assembler operation to identify the following data as
5234 initialization code. If not defined, GCC will assume such a section does
5237 @findex FINI_SECTION_ASM_OP
5238 @item FINI_SECTION_ASM_OP
5239 If defined, a C expression whose value is a string, including spacing,
5240 containing the assembler operation to identify the following data as
5241 finalization code. If not defined, GCC will assume such a section does
5244 @findex CRT_CALL_STATIC_FUNCTION
5245 @item CRT_CALL_STATIC_FUNCTION
5246 If defined, a C statement that calls the function named as the sole
5247 argument of this macro. This is used in @file{crtstuff.c} if
5248 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls to
5249 initialization and finalization functions from the init and fini
5250 sections. By default, this macro is a simple function call. Some
5251 ports need hand-crafted assembly code to avoid dependencies on
5252 registers initialized in the function prologue or to ensure that
5253 constant pools don't end up too far way in the text section.
5255 @findex EXTRA_SECTIONS
5258 @item EXTRA_SECTIONS
5259 A list of names for sections other than the standard two, which are
5260 @code{in_text} and @code{in_data}. You need not define this macro
5261 on a system with no other sections (that GCC needs to use).
5263 @findex EXTRA_SECTION_FUNCTIONS
5264 @findex text_section
5265 @findex data_section
5266 @item EXTRA_SECTION_FUNCTIONS
5267 One or more functions to be defined in @file{varasm.c}. These
5268 functions should do jobs analogous to those of @code{text_section} and
5269 @code{data_section}, for your additional sections. Do not define this
5270 macro if you do not define @code{EXTRA_SECTIONS}.
5272 @findex READONLY_DATA_SECTION
5273 @item READONLY_DATA_SECTION
5274 On most machines, read-only variables, constants, and jump tables are
5275 placed in the text section. If this is not the case on your machine,
5276 this macro should be defined to be the name of a function (either
5277 @code{data_section} or a function defined in @code{EXTRA_SECTIONS}) that
5278 switches to the section to be used for read-only items.
5280 If these items should be placed in the text section, this macro should
5283 @findex SELECT_SECTION
5284 @item SELECT_SECTION (@var{exp}, @var{reloc})
5285 A C statement or statements to switch to the appropriate section for
5286 output of @var{exp}. You can assume that @var{exp} is either a
5287 @code{VAR_DECL} node or a constant of some sort. @var{reloc}
5288 indicates whether the initial value of @var{exp} requires link-time
5289 relocations. Select the section by calling @code{text_section} or one
5290 of the alternatives for other sections.
5292 Do not define this macro if you put all read-only variables and
5293 constants in the read-only data section (usually the text section).
5295 @findex SELECT_RTX_SECTION
5296 @item SELECT_RTX_SECTION (@var{mode}, @var{rtx})
5297 A C statement or statements to switch to the appropriate section for
5298 output of @var{rtx} in mode @var{mode}. You can assume that @var{rtx}
5299 is some kind of constant in RTL. The argument @var{mode} is redundant
5300 except in the case of a @code{const_int} rtx. Select the section by
5301 calling @code{text_section} or one of the alternatives for other
5304 Do not define this macro if you put all constants in the read-only
5307 @findex JUMP_TABLES_IN_TEXT_SECTION
5308 @item JUMP_TABLES_IN_TEXT_SECTION
5309 Define this macro to be an expression with a non-zero value if jump
5310 tables (for @code{tablejump} insns) should be output in the text
5311 section, along with the assembler instructions. Otherwise, the
5312 readonly data section is used.
5314 This macro is irrelevant if there is no separate readonly data section.
5316 @findex ENCODE_SECTION_INFO
5317 @item ENCODE_SECTION_INFO (@var{decl})
5318 Define this macro if references to a symbol must be treated differently
5319 depending on something about the variable or function named by the
5320 symbol (such as what section it is in).
5322 The macro definition, if any, is executed immediately after the rtl for
5323 @var{decl} has been created and stored in @code{DECL_RTL (@var{decl})}.
5324 The value of the rtl will be a @code{mem} whose address is a
5327 @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO}
5328 The usual thing for this macro to do is to record a flag in the
5329 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
5330 modified name string in the @code{symbol_ref} (if one bit is not enough
5333 @findex STRIP_NAME_ENCODING
5334 @item STRIP_NAME_ENCODING (@var{var}, @var{sym_name})
5335 Decode @var{sym_name} and store the real name part in @var{var}, sans
5336 the characters that encode section info. Define this macro if
5337 @code{ENCODE_SECTION_INFO} alters the symbol's name string.
5339 @findex UNIQUE_SECTION_P
5340 @item UNIQUE_SECTION_P (@var{decl})
5341 A C expression which evaluates to true if @var{decl} should be placed
5342 into a unique section for some target-specific reason. If you do not
5343 define this macro, the default is @samp{0}. Note that the flag
5344 @samp{-ffunction-sections} will also cause functions to be placed into
5347 @findex UNIQUE_SECTION
5348 @item UNIQUE_SECTION (@var{decl}, @var{reloc})
5349 A C statement to build up a unique section name, expressed as a
5350 STRING_CST node, and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5351 @var{reloc} indicates whether the initial value of @var{exp} requires
5352 link-time relocations. If you do not define this macro, GCC will use
5353 the symbol name prefixed by @samp{.} as the section name. Note - this
5354 macro can now be called for unitialised data items as well as
5355 initialised data and functions.
5359 @section Position Independent Code
5360 @cindex position independent code
5363 This section describes macros that help implement generation of position
5364 independent code. Simply defining these macros is not enough to
5365 generate valid PIC; you must also add support to the macros
5366 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5367 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5368 @samp{movsi} to do something appropriate when the source operand
5369 contains a symbolic address. You may also need to alter the handling of
5370 switch statements so that they use relative addresses.
5371 @c i rearranged the order of the macros above to try to force one of
5372 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5375 @findex PIC_OFFSET_TABLE_REGNUM
5376 @item PIC_OFFSET_TABLE_REGNUM
5377 The register number of the register used to address a table of static
5378 data addresses in memory. In some cases this register is defined by a
5379 processor's ``application binary interface'' (ABI). When this macro
5380 is defined, RTL is generated for this register once, as with the stack
5381 pointer and frame pointer registers. If this macro is not defined, it
5382 is up to the machine-dependent files to allocate such a register (if
5385 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5386 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5387 Define this macro if the register defined by
5388 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
5389 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5391 @findex FINALIZE_PIC
5393 By generating position-independent code, when two different programs (A
5394 and B) share a common library (libC.a), the text of the library can be
5395 shared whether or not the library is linked at the same address for both
5396 programs. In some of these environments, position-independent code
5397 requires not only the use of different addressing modes, but also
5398 special code to enable the use of these addressing modes.
5400 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
5401 codes once the function is being compiled into assembly code, but not
5402 before. (It is not done before, because in the case of compiling an
5403 inline function, it would lead to multiple PIC prologues being
5404 included in functions which used inline functions and were compiled to
5407 @findex LEGITIMATE_PIC_OPERAND_P
5408 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
5409 A C expression that is nonzero if @var{x} is a legitimate immediate
5410 operand on the target machine when generating position independent code.
5411 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5412 check this. You can also assume @var{flag_pic} is true, so you need not
5413 check it either. You need not define this macro if all constants
5414 (including @code{SYMBOL_REF}) can be immediate operands when generating
5415 position independent code.
5418 @node Assembler Format
5419 @section Defining the Output Assembler Language
5421 This section describes macros whose principal purpose is to describe how
5422 to write instructions in assembler language--rather than what the
5426 * File Framework:: Structural information for the assembler file.
5427 * Data Output:: Output of constants (numbers, strings, addresses).
5428 * Uninitialized Data:: Output of uninitialized variables.
5429 * Label Output:: Output and generation of labels.
5430 * Initialization:: General principles of initialization
5431 and termination routines.
5432 * Macros for Initialization::
5433 Specific macros that control the handling of
5434 initialization and termination routines.
5435 * Instruction Output:: Output of actual instructions.
5436 * Dispatch Tables:: Output of jump tables.
5437 * Exception Region Output:: Output of exception region code.
5438 * Alignment Output:: Pseudo ops for alignment and skipping data.
5441 @node File Framework
5442 @subsection The Overall Framework of an Assembler File
5443 @cindex assembler format
5444 @cindex output of assembler code
5446 @c prevent bad page break with this line
5447 This describes the overall framework of an assembler file.
5450 @findex ASM_FILE_START
5451 @item ASM_FILE_START (@var{stream})
5452 A C expression which outputs to the stdio stream @var{stream}
5453 some appropriate text to go at the start of an assembler file.
5455 Normally this macro is defined to output a line containing
5456 @samp{#NO_APP}, which is a comment that has no effect on most
5457 assemblers but tells the GNU assembler that it can save time by not
5458 checking for certain assembler constructs.
5460 On systems that use SDB, it is necessary to output certain commands;
5461 see @file{attasm.h}.
5463 @findex ASM_FILE_END
5464 @item ASM_FILE_END (@var{stream})
5465 A C expression which outputs to the stdio stream @var{stream}
5466 some appropriate text to go at the end of an assembler file.
5468 If this macro is not defined, the default is to output nothing
5469 special at the end of the file. Most systems don't require any
5472 On systems that use SDB, it is necessary to output certain commands;
5473 see @file{attasm.h}.
5475 @findex ASM_IDENTIFY_GCC
5476 @item ASM_IDENTIFY_GCC (@var{file})
5477 A C statement to output assembler commands which will identify
5478 the object file as having been compiled with GCC (or another
5481 If you don't define this macro, the string @samp{gcc_compiled.:}
5482 is output. This string is calculated to define a symbol which,
5483 on BSD systems, will never be defined for any other reason.
5484 GDB checks for the presence of this symbol when reading the
5485 symbol table of an executable.
5487 On non-BSD systems, you must arrange communication with GDB in
5488 some other fashion. If GDB is not used on your system, you can
5489 define this macro with an empty body.
5491 @findex ASM_COMMENT_START
5492 @item ASM_COMMENT_START
5493 A C string constant describing how to begin a comment in the target
5494 assembler language. The compiler assumes that the comment will end at
5495 the end of the line.
5499 A C string constant for text to be output before each @code{asm}
5500 statement or group of consecutive ones. Normally this is
5501 @code{"#APP"}, which is a comment that has no effect on most
5502 assemblers but tells the GNU assembler that it must check the lines
5503 that follow for all valid assembler constructs.
5507 A C string constant for text to be output after each @code{asm}
5508 statement or group of consecutive ones. Normally this is
5509 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5510 time-saving assumptions that are valid for ordinary compiler output.
5512 @findex ASM_OUTPUT_SOURCE_FILENAME
5513 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5514 A C statement to output COFF information or DWARF debugging information
5515 which indicates that filename @var{name} is the current source file to
5516 the stdio stream @var{stream}.
5518 This macro need not be defined if the standard form of output
5519 for the file format in use is appropriate.
5521 @findex OUTPUT_QUOTED_STRING
5522 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5523 A C statement to output the string @var{string} to the stdio stream
5524 @var{stream}. If you do not call the function @code{output_quoted_string}
5525 in your config files, GCC will only call it to output filenames to
5526 the assembler source. So you can use it to canonicalize the format
5527 of the filename using this macro.
5529 @findex ASM_OUTPUT_SOURCE_LINE
5530 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
5531 A C statement to output DBX or SDB debugging information before code
5532 for line number @var{line} of the current source file to the
5533 stdio stream @var{stream}.
5535 This macro need not be defined if the standard form of debugging
5536 information for the debugger in use is appropriate.
5538 @findex ASM_OUTPUT_IDENT
5539 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
5540 A C statement to output something to the assembler file to handle a
5541 @samp{#ident} directive containing the text @var{string}. If this
5542 macro is not defined, nothing is output for a @samp{#ident} directive.
5544 @findex ASM_OUTPUT_SECTION_NAME
5545 @item ASM_OUTPUT_SECTION_NAME (@var{stream}, @var{decl}, @var{name}, @var{reloc})
5546 A C statement to output something to the assembler file to switch to section
5547 @var{name} for object @var{decl} which is either a @code{FUNCTION_DECL}, a
5548 @code{VAR_DECL} or @code{NULL_TREE}. @var{reloc}
5549 indicates whether the initial value of @var{exp} requires link-time
5550 relocations. The string given by @var{name} will always be the
5551 canonical version stored in the global stringpool.
5553 Some target formats do not support arbitrary sections. Do not define
5554 this macro in such cases.
5556 At present this macro is only used to support section attributes.
5557 When this macro is undefined, section attributes are disabled.
5559 @findex OBJC_PROLOGUE
5561 A C statement to output any assembler statements which are required to
5562 precede any Objective C object definitions or message sending. The
5563 statement is executed only when compiling an Objective C program.
5568 @subsection Output of Data
5570 @c prevent bad page break with this line
5571 This describes data output.
5574 @findex ASM_OUTPUT_LONG_DOUBLE
5575 @findex ASM_OUTPUT_DOUBLE
5576 @findex ASM_OUTPUT_FLOAT
5577 @item ASM_OUTPUT_LONG_DOUBLE (@var{stream}, @var{value})
5578 @itemx ASM_OUTPUT_DOUBLE (@var{stream}, @var{value})
5579 @itemx ASM_OUTPUT_FLOAT (@var{stream}, @var{value})
5580 @itemx ASM_OUTPUT_THREE_QUARTER_FLOAT (@var{stream}, @var{value})
5581 @itemx ASM_OUTPUT_SHORT_FLOAT (@var{stream}, @var{value})
5582 @itemx ASM_OUTPUT_BYTE_FLOAT (@var{stream}, @var{value})
5583 A C statement to output to the stdio stream @var{stream} an assembler
5584 instruction to assemble a floating-point constant of @code{TFmode},
5585 @code{DFmode}, @code{SFmode}, @code{TQFmode}, @code{HFmode}, or
5586 @code{QFmode}, respectively, whose value is @var{value}. @var{value}
5587 will be a C expression of type @code{REAL_VALUE_TYPE}. Macros such as
5588 @code{REAL_VALUE_TO_TARGET_DOUBLE} are useful for writing these
5591 @findex ASM_OUTPUT_QUADRUPLE_INT
5592 @findex ASM_OUTPUT_DOUBLE_INT
5593 @findex ASM_OUTPUT_INT
5594 @findex ASM_OUTPUT_SHORT
5595 @findex ASM_OUTPUT_CHAR
5596 @findex output_addr_const
5597 @item ASM_OUTPUT_QUADRUPLE_INT (@var{stream}, @var{exp})
5598 @itemx ASM_OUTPUT_DOUBLE_INT (@var{stream}, @var{exp})
5599 @itemx ASM_OUTPUT_INT (@var{stream}, @var{exp})
5600 @itemx ASM_OUTPUT_SHORT (@var{stream}, @var{exp})
5601 @itemx ASM_OUTPUT_CHAR (@var{stream}, @var{exp})
5602 A C statement to output to the stdio stream @var{stream} an assembler
5603 instruction to assemble an integer of 16, 8, 4, 2 or 1 bytes,
5604 respectively, whose value is @var{value}. The argument @var{exp} will
5605 be an RTL expression which represents a constant value. Use
5606 @samp{output_addr_const (@var{stream}, @var{exp})} to output this value
5607 as an assembler expression.@refill
5609 For sizes larger than @code{UNITS_PER_WORD}, if the action of a macro
5610 would be identical to repeatedly calling the macro corresponding to
5611 a size of @code{UNITS_PER_WORD}, once for each word, you need not define
5614 @findex OUTPUT_ADDR_CONST_EXTRA
5615 @item OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
5616 A C statement to recognize @var{rtx} patterns that
5617 @code{output_addr_const} can't deal with, and output assembly code to
5618 @var{stream} corresponding to the pattern @var{x}. This may be used to
5619 allow machine-dependent @code{UNSPEC}s to appear within constants.
5621 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
5622 @code{goto fail}, so that a standard error message is printed. If it
5623 prints an error message itself, by calling, for example,
5624 @code{output_operand_lossage}, it may just complete normally.
5626 @findex ASM_OUTPUT_BYTE
5627 @item ASM_OUTPUT_BYTE (@var{stream}, @var{value})
5628 A C statement to output to the stdio stream @var{stream} an assembler
5629 instruction to assemble a single byte containing the number @var{value}.
5633 A C string constant, including spacing, giving the pseudo-op to use for a
5634 sequence of single-byte constants. If this macro is not defined, the
5635 default is @code{"\t.byte\t"}.
5637 @findex ASM_OUTPUT_ASCII
5638 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5639 A C statement to output to the stdio stream @var{stream} an assembler
5640 instruction to assemble a string constant containing the @var{len}
5641 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5642 @code{char *} and @var{len} a C expression of type @code{int}.
5644 If the assembler has a @code{.ascii} pseudo-op as found in the
5645 Berkeley Unix assembler, do not define the macro
5646 @code{ASM_OUTPUT_ASCII}.
5648 @findex CONSTANT_POOL_BEFORE_FUNCTION
5649 @item CONSTANT_POOL_BEFORE_FUNCTION
5650 You may define this macro as a C expression. You should define the
5651 expression to have a non-zero value if GCC should output the constant
5652 pool for a function before the code for the function, or a zero value if
5653 GCC should output the constant pool after the function. If you do
5654 not define this macro, the usual case, GCC will output the constant
5655 pool before the function.
5657 @findex ASM_OUTPUT_POOL_PROLOGUE
5658 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
5659 A C statement to output assembler commands to define the start of the
5660 constant pool for a function. @var{funname} is a string giving
5661 the name of the function. Should the return type of the function
5662 be required, it can be obtained via @var{fundecl}. @var{size}
5663 is the size, in bytes, of the constant pool that will be written
5664 immediately after this call.
5666 If no constant-pool prefix is required, the usual case, this macro need
5669 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
5670 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5671 A C statement (with or without semicolon) to output a constant in the
5672 constant pool, if it needs special treatment. (This macro need not do
5673 anything for RTL expressions that can be output normally.)
5675 The argument @var{file} is the standard I/O stream to output the
5676 assembler code on. @var{x} is the RTL expression for the constant to
5677 output, and @var{mode} is the machine mode (in case @var{x} is a
5678 @samp{const_int}). @var{align} is the required alignment for the value
5679 @var{x}; you should output an assembler directive to force this much
5682 The argument @var{labelno} is a number to use in an internal label for
5683 the address of this pool entry. The definition of this macro is
5684 responsible for outputting the label definition at the proper place.
5685 Here is how to do this:
5688 ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
5691 When you output a pool entry specially, you should end with a
5692 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5693 entry from being output a second time in the usual manner.
5695 You need not define this macro if it would do nothing.
5697 @findex CONSTANT_AFTER_FUNCTION_P
5698 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
5699 Define this macro as a C expression which is nonzero if the constant
5700 @var{exp}, of type @code{tree}, should be output after the code for a
5701 function. The compiler will normally output all constants before the
5702 function; you need not define this macro if this is OK.
5704 @findex ASM_OUTPUT_POOL_EPILOGUE
5705 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5706 A C statement to output assembler commands to at the end of the constant
5707 pool for a function. @var{funname} is a string giving the name of the
5708 function. Should the return type of the function be required, you can
5709 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5710 constant pool that GCC wrote immediately before this call.
5712 If no constant-pool epilogue is required, the usual case, you need not
5715 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
5716 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
5717 Define this macro as a C expression which is nonzero if @var{C} is
5718 used as a logical line separator by the assembler.
5720 If you do not define this macro, the default is that only
5721 the character @samp{;} is treated as a logical line separator.
5724 @findex ASM_OPEN_PAREN
5725 @findex ASM_CLOSE_PAREN
5726 @item ASM_OPEN_PAREN
5727 @itemx ASM_CLOSE_PAREN
5728 These macros are defined as C string constants, describing the syntax
5729 in the assembler for grouping arithmetic expressions. The following
5730 definitions are correct for most assemblers:
5733 #define ASM_OPEN_PAREN "("
5734 #define ASM_CLOSE_PAREN ")"
5738 These macros are provided by @file{real.h} for writing the definitions
5739 of @code{ASM_OUTPUT_DOUBLE} and the like:
5742 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5743 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5744 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5745 @findex REAL_VALUE_TO_TARGET_SINGLE
5746 @findex REAL_VALUE_TO_TARGET_DOUBLE
5747 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
5748 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
5749 floating point representation, and store its bit pattern in the array of
5750 @code{long int} whose address is @var{l}. The number of elements in the
5751 output array is determined by the size of the desired target floating
5752 point data type: 32 bits of it go in each @code{long int} array
5753 element. Each array element holds 32 bits of the result, even if
5754 @code{long int} is wider than 32 bits on the host machine.
5756 The array element values are designed so that you can print them out
5757 using @code{fprintf} in the order they should appear in the target
5760 @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
5761 @findex REAL_VALUE_TO_DECIMAL
5762 This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
5763 decimal number and stores it as a string into @var{string}.
5764 You must pass, as @var{string}, the address of a long enough block
5765 of space to hold the result.
5767 The argument @var{format} is a @code{printf}-specification that serves
5768 as a suggestion for how to format the output string.
5771 @node Uninitialized Data
5772 @subsection Output of Uninitialized Variables
5774 Each of the macros in this section is used to do the whole job of
5775 outputting a single uninitialized variable.
5778 @findex ASM_OUTPUT_COMMON
5779 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5780 A C statement (sans semicolon) to output to the stdio stream
5781 @var{stream} the assembler definition of a common-label named
5782 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5783 is the size rounded up to whatever alignment the caller wants.
5785 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5786 output the name itself; before and after that, output the additional
5787 assembler syntax for defining the name, and a newline.
5789 This macro controls how the assembler definitions of uninitialized
5790 common global variables are output.
5792 @findex ASM_OUTPUT_ALIGNED_COMMON
5793 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5794 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5795 separate, explicit argument. If you define this macro, it is used in
5796 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5797 handling the required alignment of the variable. The alignment is specified
5798 as the number of bits.
5800 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
5801 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5802 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5803 variable to be output, if there is one, or @code{NULL_TREE} if there
5804 is no corresponding variable. If you define this macro, GCC will use it
5805 in place of both @code{ASM_OUTPUT_COMMON} and
5806 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5807 the variable's decl in order to chose what to output.
5809 @findex ASM_OUTPUT_SHARED_COMMON
5810 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5811 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
5812 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
5815 @findex ASM_OUTPUT_BSS
5816 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5817 A C statement (sans semicolon) to output to the stdio stream
5818 @var{stream} the assembler definition of uninitialized global @var{decl} named
5819 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5820 is the size rounded up to whatever alignment the caller wants.
5822 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
5823 defining this macro. If unable, use the expression
5824 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5825 before and after that, output the additional assembler syntax for defining
5826 the name, and a newline.
5828 This macro controls how the assembler definitions of uninitialized global
5829 variables are output. This macro exists to properly support languages like
5830 @code{c++} which do not have @code{common} data. However, this macro currently
5831 is not defined for all targets. If this macro and
5832 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
5833 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
5834 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
5836 @findex ASM_OUTPUT_ALIGNED_BSS
5837 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5838 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
5839 separate, explicit argument. If you define this macro, it is used in
5840 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
5841 handling the required alignment of the variable. The alignment is specified
5842 as the number of bits.
5844 Try to use function @code{asm_output_aligned_bss} defined in file
5845 @file{varasm.c} when defining this macro.
5847 @findex ASM_OUTPUT_SHARED_BSS
5848 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5849 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
5850 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
5853 @findex ASM_OUTPUT_LOCAL
5854 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5855 A C statement (sans semicolon) to output to the stdio stream
5856 @var{stream} the assembler definition of a local-common-label named
5857 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5858 is the size rounded up to whatever alignment the caller wants.
5860 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5861 output the name itself; before and after that, output the additional
5862 assembler syntax for defining the name, and a newline.
5864 This macro controls how the assembler definitions of uninitialized
5865 static variables are output.
5867 @findex ASM_OUTPUT_ALIGNED_LOCAL
5868 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5869 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5870 separate, explicit argument. If you define this macro, it is used in
5871 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5872 handling the required alignment of the variable. The alignment is specified
5873 as the number of bits.
5875 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
5876 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5877 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5878 variable to be output, if there is one, or @code{NULL_TREE} if there
5879 is no corresponding variable. If you define this macro, GCC will use it
5880 in place of both @code{ASM_OUTPUT_DECL} and
5881 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5882 the variable's decl in order to chose what to output.
5884 @findex ASM_OUTPUT_SHARED_LOCAL
5885 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5886 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
5887 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
5892 @subsection Output and Generation of Labels
5894 @c prevent bad page break with this line
5895 This is about outputting labels.
5898 @findex ASM_OUTPUT_LABEL
5899 @findex assemble_name
5900 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5901 A C statement (sans semicolon) to output to the stdio stream
5902 @var{stream} the assembler definition of a label named @var{name}.
5903 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5904 output the name itself; before and after that, output the additional
5905 assembler syntax for defining the name, and a newline.
5907 @findex ASM_DECLARE_FUNCTION_NAME
5908 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5909 A C statement (sans semicolon) to output to the stdio stream
5910 @var{stream} any text necessary for declaring the name @var{name} of a
5911 function which is being defined. This macro is responsible for
5912 outputting the label definition (perhaps using
5913 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
5914 @code{FUNCTION_DECL} tree node representing the function.
5916 If this macro is not defined, then the function name is defined in the
5917 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5919 @findex ASM_DECLARE_FUNCTION_SIZE
5920 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5921 A C statement (sans semicolon) to output to the stdio stream
5922 @var{stream} any text necessary for declaring the size of a function
5923 which is being defined. The argument @var{name} is the name of the
5924 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5925 representing the function.
5927 If this macro is not defined, then the function size is not defined.
5929 @findex ASM_DECLARE_OBJECT_NAME
5930 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5931 A C statement (sans semicolon) to output to the stdio stream
5932 @var{stream} any text necessary for declaring the name @var{name} of an
5933 initialized variable which is being defined. This macro must output the
5934 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
5935 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
5937 If this macro is not defined, then the variable name is defined in the
5938 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5940 @findex ASM_DECLARE_REGISTER_GLOBAL
5941 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
5942 A C statement (sans semicolon) to output to the stdio stream
5943 @var{stream} any text necessary for claiming a register @var{regno}
5944 for a global variable @var{decl} with name @var{name}.
5946 If you don't define this macro, that is equivalent to defining it to do
5949 @findex ASM_FINISH_DECLARE_OBJECT
5950 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5951 A C statement (sans semicolon) to finish up declaring a variable name
5952 once the compiler has processed its initializer fully and thus has had a
5953 chance to determine the size of an array when controlled by an
5954 initializer. This is used on systems where it's necessary to declare
5955 something about the size of the object.
5957 If you don't define this macro, that is equivalent to defining it to do
5960 @findex ASM_GLOBALIZE_LABEL
5961 @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
5962 A C statement (sans semicolon) to output to the stdio stream
5963 @var{stream} some commands that will make the label @var{name} global;
5964 that is, available for reference from other files. Use the expression
5965 @code{assemble_name (@var{stream}, @var{name})} to output the name
5966 itself; before and after that, output the additional assembler syntax
5967 for making that name global, and a newline.
5969 @findex ASM_WEAKEN_LABEL
5970 @item ASM_WEAKEN_LABEL
5971 A C statement (sans semicolon) to output to the stdio stream
5972 @var{stream} some commands that will make the label @var{name} weak;
5973 that is, available for reference from other files but only used if
5974 no other definition is available. Use the expression
5975 @code{assemble_name (@var{stream}, @var{name})} to output the name
5976 itself; before and after that, output the additional assembler syntax
5977 for making that name weak, and a newline.
5979 If you don't define this macro, GCC will not support weak
5980 symbols and you should not define the @code{SUPPORTS_WEAK} macro.
5982 @findex SUPPORTS_WEAK
5984 A C expression which evaluates to true if the target supports weak symbols.
5986 If you don't define this macro, @file{defaults.h} provides a default
5987 definition. If @code{ASM_WEAKEN_LABEL} is defined, the default
5988 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5989 you want to control weak symbol support with a compiler flag such as
5992 @findex MAKE_DECL_ONE_ONLY (@var{decl})
5993 @item MAKE_DECL_ONE_ONLY
5994 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5995 public symbol such that extra copies in multiple translation units will
5996 be discarded by the linker. Define this macro if your object file
5997 format provides support for this concept, such as the @samp{COMDAT}
5998 section flags in the Microsoft Windows PE/COFF format, and this support
5999 requires changes to @var{decl}, such as putting it in a separate section.
6001 @findex SUPPORTS_ONE_ONLY
6002 @item SUPPORTS_ONE_ONLY
6003 A C expression which evaluates to true if the target supports one-only
6006 If you don't define this macro, @file{varasm.c} provides a default
6007 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6008 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6009 you want to control one-only symbol support with a compiler flag, or if
6010 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6011 be emitted as one-only.
6013 @findex ASM_OUTPUT_EXTERNAL
6014 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6015 A C statement (sans semicolon) to output to the stdio stream
6016 @var{stream} any text necessary for declaring the name of an external
6017 symbol named @var{name} which is referenced in this compilation but
6018 not defined. The value of @var{decl} is the tree node for the
6021 This macro need not be defined if it does not need to output anything.
6022 The GNU assembler and most Unix assemblers don't require anything.
6024 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
6025 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
6026 A C statement (sans semicolon) to output on @var{stream} an assembler
6027 pseudo-op to declare a library function name external. The name of the
6028 library function is given by @var{symref}, which has type @code{rtx} and
6029 is a @code{symbol_ref}.
6031 This macro need not be defined if it does not need to output anything.
6032 The GNU assembler and most Unix assemblers don't require anything.
6034 @findex ASM_OUTPUT_LABELREF
6035 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6036 A C statement (sans semicolon) to output to the stdio stream
6037 @var{stream} a reference in assembler syntax to a label named
6038 @var{name}. This should add @samp{_} to the front of the name, if that
6039 is customary on your operating system, as it is in most Berkeley Unix
6040 systems. This macro is used in @code{assemble_name}.
6042 @ignore @c Seems not to exist anymore.
6043 @findex ASM_OUTPUT_LABELREF_AS_INT
6044 @item ASM_OUTPUT_LABELREF_AS_INT (@var{file}, @var{label})
6045 Define this macro for systems that use the program @code{collect2}.
6046 The definition should be a C statement to output a word containing
6047 a reference to the label @var{label}.
6050 @findex ASM_OUTPUT_SYMBOL_REF
6051 @item ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6052 A C statement (sans semicolon) to output a reference to
6053 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_output}
6054 will be used to output the name of the symbol. This macro may be used
6055 to modify the way a symbol is referenced depending on information
6056 encoded by @code{ENCODE_SECTION_INFO}.
6058 @findex ASM_OUTPUT_INTERNAL_LABEL
6059 @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
6060 A C statement to output to the stdio stream @var{stream} a label whose
6061 name is made from the string @var{prefix} and the number @var{num}.
6063 It is absolutely essential that these labels be distinct from the labels
6064 used for user-level functions and variables. Otherwise, certain programs
6065 will have name conflicts with internal labels.
6067 It is desirable to exclude internal labels from the symbol table of the
6068 object file. Most assemblers have a naming convention for labels that
6069 should be excluded; on many systems, the letter @samp{L} at the
6070 beginning of a label has this effect. You should find out what
6071 convention your system uses, and follow it.
6073 The usual definition of this macro is as follows:
6076 fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
6079 @findex ASM_OUTPUT_DEBUG_LABEL
6080 @item ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6081 A C statement to output to the stdio stream @var{stream} a debug info
6082 label whose name is made from the string @var{prefix} and the number
6083 @var{num}. This is useful for VLIW targets, where debug info labels
6084 may need to be treated differently than branch target labels. On some
6085 systems, branch target labels must be at the beginning of instruction
6086 bundles, but debug info labels can occur in the middle of instruction
6089 If this macro is not defined, then @code{ASM_OUTPUT_INTERNAL_LABEL} will be
6092 @findex ASM_OUTPUT_ALTERNATE_LABEL_NAME
6093 @item ASM_OUTPUT_ALTERNATE_LABEL_NAME (@var{stream}, @var{string})
6094 A C statement to output to the stdio stream @var{stream} the string
6097 The default definition of this macro is as follows:
6100 fprintf (@var{stream}, "%s:\n", LABEL_ALTERNATE_NAME (INSN))
6103 @findex ASM_GENERATE_INTERNAL_LABEL
6104 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6105 A C statement to store into the string @var{string} a label whose name
6106 is made from the string @var{prefix} and the number @var{num}.
6108 This string, when output subsequently by @code{assemble_name}, should
6109 produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
6110 with the same @var{prefix} and @var{num}.
6112 If the string begins with @samp{*}, then @code{assemble_name} will
6113 output the rest of the string unchanged. It is often convenient for
6114 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6115 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6116 to output the string, and may change it. (Of course,
6117 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6118 you should know what it does on your machine.)
6120 @findex ASM_FORMAT_PRIVATE_NAME
6121 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6122 A C expression to assign to @var{outvar} (which is a variable of type
6123 @code{char *}) a newly allocated string made from the string
6124 @var{name} and the number @var{number}, with some suitable punctuation
6125 added. Use @code{alloca} to get space for the string.
6127 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6128 produce an assembler label for an internal static variable whose name is
6129 @var{name}. Therefore, the string must be such as to result in valid
6130 assembler code. The argument @var{number} is different each time this
6131 macro is executed; it prevents conflicts between similarly-named
6132 internal static variables in different scopes.
6134 Ideally this string should not be a valid C identifier, to prevent any
6135 conflict with the user's own symbols. Most assemblers allow periods
6136 or percent signs in assembler symbols; putting at least one of these
6137 between the name and the number will suffice.
6139 @findex ASM_OUTPUT_DEF
6140 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6141 A C statement to output to the stdio stream @var{stream} assembler code
6142 which defines (equates) the symbol @var{name} to have the value @var{value}.
6145 If SET_ASM_OP is defined, a default definition is provided which is
6146 correct for most systems.
6148 @findex ASM_OUTPUT_DEF_FROM_DECLS
6149 @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6150 A C statement to output to the stdio stream @var{stream} assembler code
6151 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6152 to have the value of the tree node @var{decl_of_value}. This macro will
6153 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6154 the tree nodes are available.
6156 @findex ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL
6157 @item ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL (@var{stream}, @var{symbol}, @var{high}, @var{low})
6158 A C statement to output to the stdio stream @var{stream} assembler code
6159 which defines (equates) the symbol @var{symbol} to have a value equal to
6160 the difference of the two symbols @var{high} and @var{low}, i.e.
6161 @var{high} minus @var{low}. GCC guarantees that the symbols @var{high}
6162 and @var{low} are already known by the assembler so that the difference
6163 resolves into a constant.
6166 If SET_ASM_OP is defined, a default definition is provided which is
6167 correct for most systems.
6169 @findex ASM_OUTPUT_WEAK_ALIAS
6170 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6171 A C statement to output to the stdio stream @var{stream} assembler code
6172 which defines (equates) the weak symbol @var{name} to have the value
6175 Define this macro if the target only supports weak aliases; define
6176 ASM_OUTPUT_DEF instead if possible.
6178 @findex OBJC_GEN_METHOD_LABEL
6179 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6180 Define this macro to override the default assembler names used for
6181 Objective C methods.
6183 The default name is a unique method number followed by the name of the
6184 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6185 the category is also included in the assembler name (e.g.@:
6188 These names are safe on most systems, but make debugging difficult since
6189 the method's selector is not present in the name. Therefore, particular
6190 systems define other ways of computing names.
6192 @var{buf} is an expression of type @code{char *} which gives you a
6193 buffer in which to store the name; its length is as long as
6194 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6195 50 characters extra.
6197 The argument @var{is_inst} specifies whether the method is an instance
6198 method or a class method; @var{class_name} is the name of the class;
6199 @var{cat_name} is the name of the category (or NULL if the method is not
6200 in a category); and @var{sel_name} is the name of the selector.
6202 On systems where the assembler can handle quoted names, you can use this
6203 macro to provide more human-readable names.
6206 @node Initialization
6207 @subsection How Initialization Functions Are Handled
6208 @cindex initialization routines
6209 @cindex termination routines
6210 @cindex constructors, output of
6211 @cindex destructors, output of
6213 The compiled code for certain languages includes @dfn{constructors}
6214 (also called @dfn{initialization routines})---functions to initialize
6215 data in the program when the program is started. These functions need
6216 to be called before the program is ``started''---that is to say, before
6217 @code{main} is called.
6219 Compiling some languages generates @dfn{destructors} (also called
6220 @dfn{termination routines}) that should be called when the program
6223 To make the initialization and termination functions work, the compiler
6224 must output something in the assembler code to cause those functions to
6225 be called at the appropriate time. When you port the compiler to a new
6226 system, you need to specify how to do this.
6228 There are two major ways that GCC currently supports the execution of
6229 initialization and termination functions. Each way has two variants.
6230 Much of the structure is common to all four variations.
6232 @findex __CTOR_LIST__
6233 @findex __DTOR_LIST__
6234 The linker must build two lists of these functions---a list of
6235 initialization functions, called @code{__CTOR_LIST__}, and a list of
6236 termination functions, called @code{__DTOR_LIST__}.
6238 Each list always begins with an ignored function pointer (which may hold
6239 0, @minus{}1, or a count of the function pointers after it, depending on
6240 the environment). This is followed by a series of zero or more function
6241 pointers to constructors (or destructors), followed by a function
6242 pointer containing zero.
6244 Depending on the operating system and its executable file format, either
6245 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6246 time and exit time. Constructors are called in reverse order of the
6247 list; destructors in forward order.
6249 The best way to handle static constructors works only for object file
6250 formats which provide arbitrarily-named sections. A section is set
6251 aside for a list of constructors, and another for a list of destructors.
6252 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6253 object file that defines an initialization function also puts a word in
6254 the constructor section to point to that function. The linker
6255 accumulates all these words into one contiguous @samp{.ctors} section.
6256 Termination functions are handled similarly.
6258 To use this method, you need appropriate definitions of the macros
6259 @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}. Usually
6260 you can get them by including @file{svr4.h}.
6262 When arbitrary sections are available, there are two variants, depending
6263 upon how the code in @file{crtstuff.c} is called. On systems that
6264 support an @dfn{init} section which is executed at program startup,
6265 parts of @file{crtstuff.c} are compiled into that section. The
6266 program is linked by the @code{gcc} driver like this:
6269 ld -o @var{output_file} crtbegin.o @dots{} crtend.o -lgcc
6272 The head of a function (@code{__do_global_ctors}) appears in the init
6273 section of @file{crtbegin.o}; the remainder of the function appears in
6274 the init section of @file{crtend.o}. The linker will pull these two
6275 parts of the section together, making a whole function. If any of the
6276 user's object files linked into the middle of it contribute code, then that
6277 code will be executed as part of the body of @code{__do_global_ctors}.
6279 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
6282 If no init section is available, do not define
6283 @code{INIT_SECTION_ASM_OP}. Then @code{__do_global_ctors} is built into
6284 the text section like all other functions, and resides in
6285 @file{libgcc.a}. When GCC compiles any function called @code{main}, it
6286 inserts a procedure call to @code{__main} as the first executable code
6287 after the function prologue. The @code{__main} function, also defined
6288 in @file{libgcc2.c}, simply calls @file{__do_global_ctors}.
6290 In file formats that don't support arbitrary sections, there are again
6291 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
6292 and an `a.out' format must be used. In this case,
6293 @code{ASM_OUTPUT_CONSTRUCTOR} is defined to produce a @code{.stabs}
6294 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
6295 and with the address of the void function containing the initialization
6296 code as its value. The GNU linker recognizes this as a request to add
6297 the value to a ``set''; the values are accumulated, and are eventually
6298 placed in the executable as a vector in the format described above, with
6299 a leading (ignored) count and a trailing zero element.
6300 @code{ASM_OUTPUT_DESTRUCTOR} is handled similarly. Since no init
6301 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
6302 the compilation of @code{main} to call @code{__main} as above, starting
6303 the initialization process.
6305 The last variant uses neither arbitrary sections nor the GNU linker.
6306 This is preferable when you want to do dynamic linking and when using
6307 file formats which the GNU linker does not support, such as `ECOFF'. In
6308 this case, @code{ASM_OUTPUT_CONSTRUCTOR} does not produce an
6309 @code{N_SETT} symbol; initialization and termination functions are
6310 recognized simply by their names. This requires an extra program in the
6311 linkage step, called @code{collect2}. This program pretends to be the
6312 linker, for use with GCC; it does its job by running the ordinary
6313 linker, but also arranges to include the vectors of initialization and
6314 termination functions. These functions are called via @code{__main} as
6317 Choosing among these configuration options has been simplified by a set
6318 of operating-system-dependent files in the @file{config} subdirectory.
6319 These files define all of the relevant parameters. Usually it is
6320 sufficient to include one into your specific machine-dependent
6321 configuration file. These files are:
6325 For operating systems using the `a.out' format.
6328 For operating systems using the `MachO' format.
6331 For System V Release 3 and similar systems using `COFF' format.
6334 For System V Release 4 and similar systems using `ELF' format.
6337 For the VMS operating system.
6341 The following section describes the specific macros that control and
6342 customize the handling of initialization and termination functions.
6345 @node Macros for Initialization
6346 @subsection Macros Controlling Initialization Routines
6348 Here are the macros that control how the compiler handles initialization
6349 and termination functions:
6352 @findex INIT_SECTION_ASM_OP
6353 @item INIT_SECTION_ASM_OP
6354 If defined, a C string constant, including spacing, for the assembler
6355 operation to identify the following data as initialization code. If not
6356 defined, GCC will assume such a section does not exist. When you are
6357 using special sections for initialization and termination functions, this
6358 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6359 run the initialization functions.
6361 @item HAS_INIT_SECTION
6362 @findex HAS_INIT_SECTION
6363 If defined, @code{main} will not call @code{__main} as described above.
6364 This macro should be defined for systems that control the contents of the
6365 init section on a symbol-by-symbol basis, such as OSF/1, and should not
6366 be defined explicitly for systems that support
6367 @code{INIT_SECTION_ASM_OP}.
6369 @item LD_INIT_SWITCH
6370 @findex LD_INIT_SWITCH
6371 If defined, a C string constant for a switch that tells the linker that
6372 the following symbol is an initialization routine.
6374 @item LD_FINI_SWITCH
6375 @findex LD_FINI_SWITCH
6376 If defined, a C string constant for a switch that tells the linker that
6377 the following symbol is a finalization routine.
6380 @findex INVOKE__main
6381 If defined, @code{main} will call @code{__main} despite the presence of
6382 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6383 where the init section is not actually run automatically, but is still
6384 useful for collecting the lists of constructors and destructors.
6386 @item SUPPORTS_INIT_PRIORITY
6387 @findex SUPPORTS_INIT_PRIORITY
6388 If nonzero, the C++ @code{init_priority} attribute is supported and the
6389 compiler should emit instructions to control the order of initialization
6390 of objects. If zero, the compiler will issue an error message upon
6391 encountering an @code{init_priority} attribute.
6393 @item ASM_OUTPUT_CONSTRUCTOR (@var{stream}, @var{name})
6394 @findex ASM_OUTPUT_CONSTRUCTOR
6395 Define this macro as a C statement to output on the stream @var{stream}
6396 the assembler code to arrange to call the function named @var{name} at
6397 initialization time.
6399 Assume that @var{name} is the name of a C function generated
6400 automatically by the compiler. This function takes no arguments. Use
6401 the function @code{assemble_name} to output the name @var{name}; this
6402 performs any system-specific syntactic transformations such as adding an
6405 If you don't define this macro, nothing special is output to arrange to
6406 call the function. This is correct when the function will be called in
6407 some other manner---for example, by means of the @code{collect2} program,
6408 which looks through the symbol table to find these functions by their
6411 @item ASM_OUTPUT_DESTRUCTOR (@var{stream}, @var{name})
6412 @findex ASM_OUTPUT_DESTRUCTOR
6413 This is like @code{ASM_OUTPUT_CONSTRUCTOR} but used for termination
6414 functions rather than initialization functions.
6416 When @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR} are
6417 defined, the initialization routine generated for the generated object
6418 file will have static linkage.
6421 If your system uses @code{collect2} as the means of processing
6422 constructors, then that program normally uses @code{nm} to scan an
6423 object file for constructor functions to be called. On such systems you
6424 must not define @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}
6425 as the object file's initialization routine must have global scope.
6427 On certain kinds of systems, you can define these macros to make
6428 @code{collect2} work faster (and, in some cases, make it work at all):
6431 @findex OBJECT_FORMAT_COFF
6432 @item OBJECT_FORMAT_COFF
6433 Define this macro if the system uses COFF (Common Object File Format)
6434 object files, so that @code{collect2} can assume this format and scan
6435 object files directly for dynamic constructor/destructor functions.
6437 @findex OBJECT_FORMAT_ROSE
6438 @item OBJECT_FORMAT_ROSE
6439 Define this macro if the system uses ROSE format object files, so that
6440 @code{collect2} can assume this format and scan object files directly
6441 for dynamic constructor/destructor functions.
6443 These macros are effective only in a native compiler; @code{collect2} as
6444 part of a cross compiler always uses @code{nm} for the target machine.
6446 @findex REAL_NM_FILE_NAME
6447 @item REAL_NM_FILE_NAME
6448 Define this macro as a C string constant containing the file name to use
6449 to execute @code{nm}. The default is to search the path normally for
6452 If your system supports shared libraries and has a program to list the
6453 dynamic dependencies of a given library or executable, you can define
6454 these macros to enable support for running initialization and
6455 termination functions in shared libraries:
6459 Define this macro to a C string constant containing the name of the
6460 program which lists dynamic dependencies, like @code{"ldd"} under SunOS 4.
6462 @findex PARSE_LDD_OUTPUT
6463 @item PARSE_LDD_OUTPUT (@var{PTR})
6464 Define this macro to be C code that extracts filenames from the output
6465 of the program denoted by @code{LDD_SUFFIX}. @var{PTR} is a variable
6466 of type @code{char *} that points to the beginning of a line of output
6467 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6468 code must advance @var{PTR} to the beginning of the filename on that
6469 line. Otherwise, it must set @var{PTR} to @code{NULL}.
6473 @node Instruction Output
6474 @subsection Output of Assembler Instructions
6476 @c prevent bad page break with this line
6477 This describes assembler instruction output.
6480 @findex REGISTER_NAMES
6481 @item REGISTER_NAMES
6482 A C initializer containing the assembler's names for the machine
6483 registers, each one as a C string constant. This is what translates
6484 register numbers in the compiler into assembler language.
6486 @findex ADDITIONAL_REGISTER_NAMES
6487 @item ADDITIONAL_REGISTER_NAMES
6488 If defined, a C initializer for an array of structures containing a name
6489 and a register number. This macro defines additional names for hard
6490 registers, thus allowing the @code{asm} option in declarations to refer
6491 to registers using alternate names.
6493 @findex ASM_OUTPUT_OPCODE
6494 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6495 Define this macro if you are using an unusual assembler that
6496 requires different names for the machine instructions.
6498 The definition is a C statement or statements which output an
6499 assembler instruction opcode to the stdio stream @var{stream}. The
6500 macro-operand @var{ptr} is a variable of type @code{char *} which
6501 points to the opcode name in its ``internal'' form---the form that is
6502 written in the machine description. The definition should output the
6503 opcode name to @var{stream}, performing any translation you desire, and
6504 increment the variable @var{ptr} to point at the end of the opcode
6505 so that it will not be output twice.
6507 In fact, your macro definition may process less than the entire opcode
6508 name, or more than the opcode name; but if you want to process text
6509 that includes @samp{%}-sequences to substitute operands, you must take
6510 care of the substitution yourself. Just be sure to increment
6511 @var{ptr} over whatever text should not be output normally.
6513 @findex recog_operand
6514 If you need to look at the operand values, they can be found as the
6515 elements of @code{recog_operand}.
6517 If the macro definition does nothing, the instruction is output
6520 @findex FINAL_PRESCAN_INSN
6521 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6522 If defined, a C statement to be executed just prior to the output of
6523 assembler code for @var{insn}, to modify the extracted operands so
6524 they will be output differently.
6526 Here the argument @var{opvec} is the vector containing the operands
6527 extracted from @var{insn}, and @var{noperands} is the number of
6528 elements of the vector which contain meaningful data for this insn.
6529 The contents of this vector are what will be used to convert the insn
6530 template into assembler code, so you can change the assembler output
6531 by changing the contents of the vector.
6533 This macro is useful when various assembler syntaxes share a single
6534 file of instruction patterns; by defining this macro differently, you
6535 can cause a large class of instructions to be output differently (such
6536 as with rearranged operands). Naturally, variations in assembler
6537 syntax affecting individual insn patterns ought to be handled by
6538 writing conditional output routines in those patterns.
6540 If this macro is not defined, it is equivalent to a null statement.
6542 @findex FINAL_PRESCAN_LABEL
6543 @item FINAL_PRESCAN_LABEL
6544 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
6545 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
6546 @var{noperands} will be zero.
6548 @findex PRINT_OPERAND
6549 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6550 A C compound statement to output to stdio stream @var{stream} the
6551 assembler syntax for an instruction operand @var{x}. @var{x} is an
6554 @var{code} is a value that can be used to specify one of several ways
6555 of printing the operand. It is used when identical operands must be
6556 printed differently depending on the context. @var{code} comes from
6557 the @samp{%} specification that was used to request printing of the
6558 operand. If the specification was just @samp{%@var{digit}} then
6559 @var{code} is 0; if the specification was @samp{%@var{ltr}
6560 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6563 If @var{x} is a register, this macro should print the register's name.
6564 The names can be found in an array @code{reg_names} whose type is
6565 @code{char *[]}. @code{reg_names} is initialized from
6566 @code{REGISTER_NAMES}.
6568 When the machine description has a specification @samp{%@var{punct}}
6569 (a @samp{%} followed by a punctuation character), this macro is called
6570 with a null pointer for @var{x} and the punctuation character for
6573 @findex PRINT_OPERAND_PUNCT_VALID_P
6574 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6575 A C expression which evaluates to true if @var{code} is a valid
6576 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6577 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6578 punctuation characters (except for the standard one, @samp{%}) are used
6581 @findex PRINT_OPERAND_ADDRESS
6582 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6583 A C compound statement to output to stdio stream @var{stream} the
6584 assembler syntax for an instruction operand that is a memory reference
6585 whose address is @var{x}. @var{x} is an RTL expression.
6587 @cindex @code{ENCODE_SECTION_INFO} usage
6588 On some machines, the syntax for a symbolic address depends on the
6589 section that the address refers to. On these machines, define the macro
6590 @code{ENCODE_SECTION_INFO} to store the information into the
6591 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
6593 @findex DBR_OUTPUT_SEQEND
6594 @findex dbr_sequence_length
6595 @item DBR_OUTPUT_SEQEND(@var{file})
6596 A C statement, to be executed after all slot-filler instructions have
6597 been output. If necessary, call @code{dbr_sequence_length} to
6598 determine the number of slots filled in a sequence (zero if not
6599 currently outputting a sequence), to decide how many no-ops to output,
6602 Don't define this macro if it has nothing to do, but it is helpful in
6603 reading assembly output if the extent of the delay sequence is made
6604 explicit (e.g. with white space).
6606 @findex final_sequence
6607 Note that output routines for instructions with delay slots must be
6608 prepared to deal with not being output as part of a sequence (i.e.
6609 when the scheduling pass is not run, or when no slot fillers could be
6610 found.) The variable @code{final_sequence} is null when not
6611 processing a sequence, otherwise it contains the @code{sequence} rtx
6614 @findex REGISTER_PREFIX
6615 @findex LOCAL_LABEL_PREFIX
6616 @findex USER_LABEL_PREFIX
6617 @findex IMMEDIATE_PREFIX
6619 @item REGISTER_PREFIX
6620 @itemx LOCAL_LABEL_PREFIX
6621 @itemx USER_LABEL_PREFIX
6622 @itemx IMMEDIATE_PREFIX
6623 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6624 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6625 @file{final.c}). These are useful when a single @file{md} file must
6626 support multiple assembler formats. In that case, the various @file{tm.h}
6627 files can define these macros differently.
6629 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
6630 @findex ASM_FPRINTF_EXTENSIONS
6631 If defined this macro should expand to a series of @code{case}
6632 statements which will be parsed inside the @code{switch} statement of
6633 the @code{asm_fprintf} function. This allows targets to define extra
6634 printf formats which may useful when generating their assembler
6635 statements. Note that upper case letters are reserved for future
6636 generic extensions to asm_fprintf, and so are not available to target
6637 specific code. The output file is given by the parameter @var{file}.
6638 The varargs input pointer is @var{argptr} and the rest of the format
6639 string, starting the character after the one that is being switched
6640 upon, is pointed to by @var{format}.
6642 @findex ASSEMBLER_DIALECT
6643 @item ASSEMBLER_DIALECT
6644 If your target supports multiple dialects of assembler language (such as
6645 different opcodes), define this macro as a C expression that gives the
6646 numeric index of the assembler language dialect to use, with zero as the
6649 If this macro is defined, you may use constructs of the form
6650 @samp{@{option0|option1|option2@dots{}@}} in the output
6651 templates of patterns (@pxref{Output Template}) or in the first argument
6652 of @code{asm_fprintf}. This construct outputs @samp{option0},
6653 @samp{option1} or @samp{option2}, etc., if the value of
6654 @code{ASSEMBLER_DIALECT} is zero, one or two, etc. Any special
6655 characters within these strings retain their usual meaning.
6657 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6658 @samp{@}} do not have any special meaning when used in templates or
6659 operands to @code{asm_fprintf}.
6661 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6662 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6663 the variations in assembler language syntax with that mechanism. Define
6664 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6665 if the syntax variant are larger and involve such things as different
6666 opcodes or operand order.
6668 @findex ASM_OUTPUT_REG_PUSH
6669 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6670 A C expression to output to @var{stream} some assembler code
6671 which will push hard register number @var{regno} onto the stack.
6672 The code need not be optimal, since this macro is used only when
6675 @findex ASM_OUTPUT_REG_POP
6676 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6677 A C expression to output to @var{stream} some assembler code
6678 which will pop hard register number @var{regno} off of the stack.
6679 The code need not be optimal, since this macro is used only when
6683 @node Dispatch Tables
6684 @subsection Output of Dispatch Tables
6686 @c prevent bad page break with this line
6687 This concerns dispatch tables.
6690 @cindex dispatch table
6691 @findex ASM_OUTPUT_ADDR_DIFF_ELT
6692 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6693 A C statement to output to the stdio stream @var{stream} an assembler
6694 pseudo-instruction to generate a difference between two labels.
6695 @var{value} and @var{rel} are the numbers of two internal labels. The
6696 definitions of these labels are output using
6697 @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
6698 way here. For example,
6701 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6702 @var{value}, @var{rel})
6705 You must provide this macro on machines where the addresses in a
6706 dispatch table are relative to the table's own address. If defined, GNU
6707 CC will also use this macro on all machines when producing PIC.
6708 @var{body} is the body of the ADDR_DIFF_VEC; it is provided so that the
6709 mode and flags can be read.
6711 @findex ASM_OUTPUT_ADDR_VEC_ELT
6712 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6713 This macro should be provided on machines where the addresses
6714 in a dispatch table are absolute.
6716 The definition should be a C statement to output to the stdio stream
6717 @var{stream} an assembler pseudo-instruction to generate a reference to
6718 a label. @var{value} is the number of an internal label whose
6719 definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
6723 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6726 @findex ASM_OUTPUT_CASE_LABEL
6727 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6728 Define this if the label before a jump-table needs to be output
6729 specially. The first three arguments are the same as for
6730 @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
6731 jump-table which follows (a @code{jump_insn} containing an
6732 @code{addr_vec} or @code{addr_diff_vec}).
6734 This feature is used on system V to output a @code{swbeg} statement
6737 If this macro is not defined, these labels are output with
6738 @code{ASM_OUTPUT_INTERNAL_LABEL}.
6740 @findex ASM_OUTPUT_CASE_END
6741 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6742 Define this if something special must be output at the end of a
6743 jump-table. The definition should be a C statement to be executed
6744 after the assembler code for the table is written. It should write
6745 the appropriate code to stdio stream @var{stream}. The argument
6746 @var{table} is the jump-table insn, and @var{num} is the label-number
6747 of the preceding label.
6749 If this macro is not defined, nothing special is output at the end of
6753 @node Exception Region Output
6754 @subsection Assembler Commands for Exception Regions
6756 @c prevent bad page break with this line
6758 This describes commands marking the start and the end of an exception
6762 @findex ASM_OUTPUT_EH_REGION_BEG
6763 @item ASM_OUTPUT_EH_REGION_BEG ()
6764 A C expression to output text to mark the start of an exception region.
6766 This macro need not be defined on most platforms.
6768 @findex ASM_OUTPUT_EH_REGION_END
6769 @item ASM_OUTPUT_EH_REGION_END ()
6770 A C expression to output text to mark the end of an exception region.
6772 This macro need not be defined on most platforms.
6774 @findex EXCEPTION_SECTION
6775 @item EXCEPTION_SECTION ()
6776 A C expression to switch to the section in which the main
6777 exception table is to be placed (@pxref{Sections}). The default is a
6778 section named @code{.gcc_except_table} on machines that support named
6779 sections via @code{ASM_OUTPUT_SECTION_NAME}, otherwise if @samp{-fpic}
6780 or @samp{-fPIC} is in effect, the @code{data_section}, otherwise the
6781 @code{readonly_data_section}.
6783 @findex EH_FRAME_SECTION_ASM_OP
6784 @item EH_FRAME_SECTION_ASM_OP
6785 If defined, a C string constant, including spacing, for the assembler
6786 operation to switch to the section for exception handling frame unwind
6787 information. If not defined, GCC will provide a default definition if the
6788 target supports named sections. @file{crtstuff.c} uses this macro to
6789 switch to the appropriate section.
6791 You should define this symbol if your target supports DWARF 2 frame
6792 unwind information and the default definition does not work.
6794 @findex OMIT_EH_TABLE
6795 @item OMIT_EH_TABLE ()
6796 A C expression that is nonzero if the normal exception table output
6799 This macro need not be defined on most platforms.
6801 @findex EH_TABLE_LOOKUP
6802 @item EH_TABLE_LOOKUP ()
6803 Alternate runtime support for looking up an exception at runtime and
6804 finding the associated handler, if the default method won't work.
6806 This macro need not be defined on most platforms.
6808 @findex DOESNT_NEED_UNWINDER
6809 @item DOESNT_NEED_UNWINDER
6810 A C expression that decides whether or not the current function needs to
6811 have a function unwinder generated for it. See the file @code{except.c}
6812 for details on when to define this, and how.
6814 @findex MASK_RETURN_ADDR
6815 @item MASK_RETURN_ADDR
6816 An rtx used to mask the return address found via RETURN_ADDR_RTX, so
6817 that it does not contain any extraneous set bits in it.
6819 @findex DWARF2_UNWIND_INFO
6820 @item DWARF2_UNWIND_INFO
6821 Define this macro to 0 if your target supports DWARF 2 frame unwind
6822 information, but it does not yet work with exception handling.
6823 Otherwise, if your target supports this information (if it defines
6824 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
6825 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
6828 If this macro is defined to 1, the DWARF 2 unwinder will be the default
6829 exception handling mechanism; otherwise, setjmp/longjmp will be used by
6832 If this macro is defined to anything, the DWARF 2 unwinder will be used
6833 instead of inline unwinders and __unwind_function in the non-setjmp case.
6835 @findex DWARF_CIE_DATA_ALIGNMENT
6836 @item DWARF_CIE_DATA_ALIGNMENT
6837 This macro need only be defined if the target might save registers in the
6838 function prologue at an offset to the stack pointer that is not aligned to
6839 @code{UNITS_PER_WORD}. The definition should be the negative minimum
6840 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
6841 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
6842 the target supports DWARF 2 frame unwind information.
6846 @node Alignment Output
6847 @subsection Assembler Commands for Alignment
6849 @c prevent bad page break with this line
6850 This describes commands for alignment.
6853 @findex LABEL_ALIGN_AFTER_BARRIER
6854 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
6855 The alignment (log base 2) to put in front of @var{label}, which follows
6858 This macro need not be defined if you don't want any special alignment
6859 to be done at such a time. Most machine descriptions do not currently
6862 Unless it's necessary to inspect the @var{label} parameter, it is better
6863 to set the variable @var{align_jumps} in the target's
6864 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6865 selection in @var{align_jumps} in a @code{LABEL_ALIGN_AFTER_BARRIER}
6868 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6869 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6870 The maximum number of bytes to skip when applying
6871 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
6872 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6875 @item LOOP_ALIGN (@var{label})
6876 The alignment (log base 2) to put in front of @var{label}, which follows
6877 a NOTE_INSN_LOOP_BEG note.
6879 This macro need not be defined if you don't want any special alignment
6880 to be done at such a time. Most machine descriptions do not currently
6883 Unless it's necessary to inspect the @var{label} parameter, it is better
6884 to set the variable @var{align_loops} in the target's
6885 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6886 selection in @var{align_loops} in a @code{LOOP_ALIGN} implementation.
6888 @findex LOOP_ALIGN_MAX_SKIP
6889 @item LOOP_ALIGN_MAX_SKIP
6890 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
6891 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6894 @item LABEL_ALIGN (@var{label})
6895 The alignment (log base 2) to put in front of @var{label}.
6896 If LABEL_ALIGN_AFTER_BARRIER / LOOP_ALIGN specify a different alignment,
6897 the maximum of the specified values is used.
6899 Unless it's necessary to inspect the @var{label} parameter, it is better
6900 to set the variable @var{align_labels} in the target's
6901 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6902 selection in @var{align_labels} in a @code{LABEL_ALIGN} implementation.
6904 @findex LABEL_ALIGN_MAX_SKIP
6905 @item LABEL_ALIGN_MAX_SKIP
6906 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
6907 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6909 @findex ASM_OUTPUT_SKIP
6910 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6911 A C statement to output to the stdio stream @var{stream} an assembler
6912 instruction to advance the location counter by @var{nbytes} bytes.
6913 Those bytes should be zero when loaded. @var{nbytes} will be a C
6914 expression of type @code{int}.
6916 @findex ASM_NO_SKIP_IN_TEXT
6917 @item ASM_NO_SKIP_IN_TEXT
6918 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6919 text section because it fails to put zeros in the bytes that are skipped.
6920 This is true on many Unix systems, where the pseudo--op to skip bytes
6921 produces no-op instructions rather than zeros when used in the text
6924 @findex ASM_OUTPUT_ALIGN
6925 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6926 A C statement to output to the stdio stream @var{stream} an assembler
6927 command to advance the location counter to a multiple of 2 to the
6928 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
6930 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
6931 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
6932 A C statement to output to the stdio stream @var{stream} an assembler
6933 command to advance the location counter to a multiple of 2 to the
6934 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
6935 satisfy the alignment request. @var{power} and @var{max_skip} will be
6936 a C expression of type @code{int}.
6940 @node Debugging Info
6941 @section Controlling Debugging Information Format
6943 @c prevent bad page break with this line
6944 This describes how to specify debugging information.
6947 * All Debuggers:: Macros that affect all debugging formats uniformly.
6948 * DBX Options:: Macros enabling specific options in DBX format.
6949 * DBX Hooks:: Hook macros for varying DBX format.
6950 * File Names and DBX:: Macros controlling output of file names in DBX format.
6951 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
6955 @subsection Macros Affecting All Debugging Formats
6957 @c prevent bad page break with this line
6958 These macros affect all debugging formats.
6961 @findex DBX_REGISTER_NUMBER
6962 @item DBX_REGISTER_NUMBER (@var{regno})
6963 A C expression that returns the DBX register number for the compiler
6964 register number @var{regno}. In simple cases, the value of this
6965 expression may be @var{regno} itself. But sometimes there are some
6966 registers that the compiler knows about and DBX does not, or vice
6967 versa. In such cases, some register may need to have one number in
6968 the compiler and another for DBX.
6970 If two registers have consecutive numbers inside GCC, and they can be
6971 used as a pair to hold a multiword value, then they @emph{must} have
6972 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
6973 Otherwise, debuggers will be unable to access such a pair, because they
6974 expect register pairs to be consecutive in their own numbering scheme.
6976 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
6977 does not preserve register pairs, then what you must do instead is
6978 redefine the actual register numbering scheme.
6980 @findex DEBUGGER_AUTO_OFFSET
6981 @item DEBUGGER_AUTO_OFFSET (@var{x})
6982 A C expression that returns the integer offset value for an automatic
6983 variable having address @var{x} (an RTL expression). The default
6984 computation assumes that @var{x} is based on the frame-pointer and
6985 gives the offset from the frame-pointer. This is required for targets
6986 that produce debugging output for DBX or COFF-style debugging output
6987 for SDB and allow the frame-pointer to be eliminated when the
6988 @samp{-g} options is used.
6990 @findex DEBUGGER_ARG_OFFSET
6991 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6992 A C expression that returns the integer offset value for an argument
6993 having address @var{x} (an RTL expression). The nominal offset is
6996 @findex PREFERRED_DEBUGGING_TYPE
6997 @item PREFERRED_DEBUGGING_TYPE
6998 A C expression that returns the type of debugging output GCC should
6999 produce when the user specifies just @samp{-g}. Define
7000 this if you have arranged for GCC to support more than one format of
7001 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7002 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG}, and
7005 When the user specifies @samp{-ggdb}, GCC normally also uses the
7006 value of this macro to select the debugging output format, but with two
7007 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7008 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7009 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7010 defined, GCC uses @code{DBX_DEBUG}.
7012 The value of this macro only affects the default debugging output; the
7013 user can always get a specific type of output by using @samp{-gstabs},
7014 @samp{-gcoff}, @samp{-gdwarf-1}, @samp{-gdwarf-2}, or @samp{-gxcoff}.
7018 @subsection Specific Options for DBX Output
7020 @c prevent bad page break with this line
7021 These are specific options for DBX output.
7024 @findex DBX_DEBUGGING_INFO
7025 @item DBX_DEBUGGING_INFO
7026 Define this macro if GCC should produce debugging output for DBX
7027 in response to the @samp{-g} option.
7029 @findex XCOFF_DEBUGGING_INFO
7030 @item XCOFF_DEBUGGING_INFO
7031 Define this macro if GCC should produce XCOFF format debugging output
7032 in response to the @samp{-g} option. This is a variant of DBX format.
7034 @findex DEFAULT_GDB_EXTENSIONS
7035 @item DEFAULT_GDB_EXTENSIONS
7036 Define this macro to control whether GCC should by default generate
7037 GDB's extended version of DBX debugging information (assuming DBX-format
7038 debugging information is enabled at all). If you don't define the
7039 macro, the default is 1: always generate the extended information
7040 if there is any occasion to.
7042 @findex DEBUG_SYMS_TEXT
7043 @item DEBUG_SYMS_TEXT
7044 Define this macro if all @code{.stabs} commands should be output while
7045 in the text section.
7047 @findex ASM_STABS_OP
7049 A C string constant, including spacing, naming the assembler pseudo op to
7050 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7051 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7052 applies only to DBX debugging information format.
7054 @findex ASM_STABD_OP
7056 A C string constant, including spacing, naming the assembler pseudo op to
7057 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7058 value is the current location. If you don't define this macro,
7059 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7062 @findex ASM_STABN_OP
7064 A C string constant, including spacing, naming the assembler pseudo op to
7065 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7066 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7067 macro applies only to DBX debugging information format.
7069 @findex DBX_NO_XREFS
7071 Define this macro if DBX on your system does not support the construct
7072 @samp{xs@var{tagname}}. On some systems, this construct is used to
7073 describe a forward reference to a structure named @var{tagname}.
7074 On other systems, this construct is not supported at all.
7076 @findex DBX_CONTIN_LENGTH
7077 @item DBX_CONTIN_LENGTH
7078 A symbol name in DBX-format debugging information is normally
7079 continued (split into two separate @code{.stabs} directives) when it
7080 exceeds a certain length (by default, 80 characters). On some
7081 operating systems, DBX requires this splitting; on others, splitting
7082 must not be done. You can inhibit splitting by defining this macro
7083 with the value zero. You can override the default splitting-length by
7084 defining this macro as an expression for the length you desire.
7086 @findex DBX_CONTIN_CHAR
7087 @item DBX_CONTIN_CHAR
7088 Normally continuation is indicated by adding a @samp{\} character to
7089 the end of a @code{.stabs} string when a continuation follows. To use
7090 a different character instead, define this macro as a character
7091 constant for the character you want to use. Do not define this macro
7092 if backslash is correct for your system.
7094 @findex DBX_STATIC_STAB_DATA_SECTION
7095 @item DBX_STATIC_STAB_DATA_SECTION
7096 Define this macro if it is necessary to go to the data section before
7097 outputting the @samp{.stabs} pseudo-op for a non-global static
7100 @findex DBX_TYPE_DECL_STABS_CODE
7101 @item DBX_TYPE_DECL_STABS_CODE
7102 The value to use in the ``code'' field of the @code{.stabs} directive
7103 for a typedef. The default is @code{N_LSYM}.
7105 @findex DBX_STATIC_CONST_VAR_CODE
7106 @item DBX_STATIC_CONST_VAR_CODE
7107 The value to use in the ``code'' field of the @code{.stabs} directive
7108 for a static variable located in the text section. DBX format does not
7109 provide any ``right'' way to do this. The default is @code{N_FUN}.
7111 @findex DBX_REGPARM_STABS_CODE
7112 @item DBX_REGPARM_STABS_CODE
7113 The value to use in the ``code'' field of the @code{.stabs} directive
7114 for a parameter passed in registers. DBX format does not provide any
7115 ``right'' way to do this. The default is @code{N_RSYM}.
7117 @findex DBX_REGPARM_STABS_LETTER
7118 @item DBX_REGPARM_STABS_LETTER
7119 The letter to use in DBX symbol data to identify a symbol as a parameter
7120 passed in registers. DBX format does not customarily provide any way to
7121 do this. The default is @code{'P'}.
7123 @findex DBX_MEMPARM_STABS_LETTER
7124 @item DBX_MEMPARM_STABS_LETTER
7125 The letter to use in DBX symbol data to identify a symbol as a stack
7126 parameter. The default is @code{'p'}.
7128 @findex DBX_FUNCTION_FIRST
7129 @item DBX_FUNCTION_FIRST
7130 Define this macro if the DBX information for a function and its
7131 arguments should precede the assembler code for the function. Normally,
7132 in DBX format, the debugging information entirely follows the assembler
7135 @findex DBX_LBRAC_FIRST
7136 @item DBX_LBRAC_FIRST
7137 Define this macro if the @code{N_LBRAC} symbol for a block should
7138 precede the debugging information for variables and functions defined in
7139 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
7142 @findex DBX_BLOCKS_FUNCTION_RELATIVE
7143 @item DBX_BLOCKS_FUNCTION_RELATIVE
7144 Define this macro if the value of a symbol describing the scope of a
7145 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7146 of the enclosing function. Normally, GNU C uses an absolute address.
7148 @findex DBX_USE_BINCL
7150 Define this macro if GNU C should generate @code{N_BINCL} and
7151 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7152 macro also directs GNU C to output a type number as a pair of a file
7153 number and a type number within the file. Normally, GNU C does not
7154 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7155 number for a type number.
7159 @subsection Open-Ended Hooks for DBX Format
7161 @c prevent bad page break with this line
7162 These are hooks for DBX format.
7165 @findex DBX_OUTPUT_LBRAC
7166 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7167 Define this macro to say how to output to @var{stream} the debugging
7168 information for the start of a scope level for variable names. The
7169 argument @var{name} is the name of an assembler symbol (for use with
7170 @code{assemble_name}) whose value is the address where the scope begins.
7172 @findex DBX_OUTPUT_RBRAC
7173 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7174 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7176 @findex DBX_OUTPUT_ENUM
7177 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
7178 Define this macro if the target machine requires special handling to
7179 output an enumeration type. The definition should be a C statement
7180 (sans semicolon) to output the appropriate information to @var{stream}
7181 for the type @var{type}.
7183 @findex DBX_OUTPUT_FUNCTION_END
7184 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7185 Define this macro if the target machine requires special output at the
7186 end of the debugging information for a function. The definition should
7187 be a C statement (sans semicolon) to output the appropriate information
7188 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7191 @findex DBX_OUTPUT_STANDARD_TYPES
7192 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7193 Define this macro if you need to control the order of output of the
7194 standard data types at the beginning of compilation. The argument
7195 @var{syms} is a @code{tree} which is a chain of all the predefined
7196 global symbols, including names of data types.
7198 Normally, DBX output starts with definitions of the types for integers
7199 and characters, followed by all the other predefined types of the
7200 particular language in no particular order.
7202 On some machines, it is necessary to output different particular types
7203 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
7204 those symbols in the necessary order. Any predefined types that you
7205 don't explicitly output will be output afterward in no particular order.
7207 Be careful not to define this macro so that it works only for C. There
7208 are no global variables to access most of the built-in types, because
7209 another language may have another set of types. The way to output a
7210 particular type is to look through @var{syms} to see if you can find it.
7216 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7217 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
7219 dbxout_symbol (decl);
7225 This does nothing if the expected type does not exist.
7227 See the function @code{init_decl_processing} in @file{c-decl.c} to find
7228 the names to use for all the built-in C types.
7230 Here is another way of finding a particular type:
7232 @c this is still overfull. --mew 10feb93
7236 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7237 if (TREE_CODE (decl) == TYPE_DECL
7238 && (TREE_CODE (TREE_TYPE (decl))
7240 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
7241 && TYPE_UNSIGNED (TREE_TYPE (decl)))
7243 /* @r{This must be @code{unsigned short}.} */
7244 dbxout_symbol (decl);
7250 @findex NO_DBX_FUNCTION_END
7251 @item NO_DBX_FUNCTION_END
7252 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7253 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extention construct.
7254 On those machines, define this macro to turn this feature off without
7255 disturbing the rest of the gdb extensions.
7259 @node File Names and DBX
7260 @subsection File Names in DBX Format
7262 @c prevent bad page break with this line
7263 This describes file names in DBX format.
7266 @findex DBX_WORKING_DIRECTORY
7267 @item DBX_WORKING_DIRECTORY
7268 Define this if DBX wants to have the current directory recorded in each
7271 Note that the working directory is always recorded if GDB extensions are
7274 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
7275 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7276 A C statement to output DBX debugging information to the stdio stream
7277 @var{stream} which indicates that file @var{name} is the main source
7278 file---the file specified as the input file for compilation.
7279 This macro is called only once, at the beginning of compilation.
7281 This macro need not be defined if the standard form of output
7282 for DBX debugging information is appropriate.
7284 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
7285 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
7286 A C statement to output DBX debugging information to the stdio stream
7287 @var{stream} which indicates that the current directory during
7288 compilation is named @var{name}.
7290 This macro need not be defined if the standard form of output
7291 for DBX debugging information is appropriate.
7293 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
7294 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7295 A C statement to output DBX debugging information at the end of
7296 compilation of the main source file @var{name}.
7298 If you don't define this macro, nothing special is output at the end
7299 of compilation, which is correct for most machines.
7301 @findex DBX_OUTPUT_SOURCE_FILENAME
7302 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7303 A C statement to output DBX debugging information to the stdio stream
7304 @var{stream} which indicates that file @var{name} is the current source
7305 file. This output is generated each time input shifts to a different
7306 source file as a result of @samp{#include}, the end of an included file,
7307 or a @samp{#line} command.
7309 This macro need not be defined if the standard form of output
7310 for DBX debugging information is appropriate.
7315 @subsection Macros for SDB and DWARF Output
7317 @c prevent bad page break with this line
7318 Here are macros for SDB and DWARF output.
7321 @findex SDB_DEBUGGING_INFO
7322 @item SDB_DEBUGGING_INFO
7323 Define this macro if GCC should produce COFF-style debugging output
7324 for SDB in response to the @samp{-g} option.
7326 @findex DWARF_DEBUGGING_INFO
7327 @item DWARF_DEBUGGING_INFO
7328 Define this macro if GCC should produce dwarf format debugging output
7329 in response to the @samp{-g} option.
7331 @findex DWARF2_DEBUGGING_INFO
7332 @item DWARF2_DEBUGGING_INFO
7333 Define this macro if GCC should produce dwarf version 2 format
7334 debugging output in response to the @samp{-g} option.
7336 To support optional call frame debugging information, you must also
7337 define @code{INCOMING_RETURN_ADDR_RTX} and either set
7338 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
7339 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
7340 as appropriate from @code{FUNCTION_PROLOGUE} if you don't.
7342 @findex DWARF2_FRAME_INFO
7343 @item DWARF2_FRAME_INFO
7344 Define this macro to a nonzero value if GCC should always output
7345 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
7346 (@pxref{Exception Region Output} is nonzero, GCC will output this
7347 information not matter how you define @code{DWARF2_FRAME_INFO}.
7349 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
7350 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
7351 Define this macro if the linker does not work with Dwarf version 2.
7352 Normally, if the user specifies only @samp{-ggdb} GCC will use Dwarf
7353 version 2 if available; this macro disables this. See the description
7354 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
7356 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
7357 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
7358 By default, the Dwarf 2 debugging information generator will generate a
7359 label to mark the beginning of the text section. If it is better simply
7360 to use the name of the text section itself, rather than an explicit label,
7361 to indicate the beginning of the text section, define this macro to zero.
7363 @findex DWARF2_ASM_LINE_DEBUG_INFO
7364 @item DWARF2_ASM_LINE_DEBUG_INFO
7365 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
7366 line debug info sections. This will result in much more compact line number
7367 tables, and hence is desirable if it works.
7369 @findex PUT_SDB_@dots{}
7370 @item PUT_SDB_@dots{}
7371 Define these macros to override the assembler syntax for the special
7372 SDB assembler directives. See @file{sdbout.c} for a list of these
7373 macros and their arguments. If the standard syntax is used, you need
7374 not define them yourself.
7378 Some assemblers do not support a semicolon as a delimiter, even between
7379 SDB assembler directives. In that case, define this macro to be the
7380 delimiter to use (usually @samp{\n}). It is not necessary to define
7381 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
7384 @findex SDB_GENERATE_FAKE
7385 @item SDB_GENERATE_FAKE
7386 Define this macro to override the usual method of constructing a dummy
7387 name for anonymous structure and union types. See @file{sdbout.c} for
7390 @findex SDB_ALLOW_UNKNOWN_REFERENCES
7391 @item SDB_ALLOW_UNKNOWN_REFERENCES
7392 Define this macro to allow references to unknown structure,
7393 union, or enumeration tags to be emitted. Standard COFF does not
7394 allow handling of unknown references, MIPS ECOFF has support for
7397 @findex SDB_ALLOW_FORWARD_REFERENCES
7398 @item SDB_ALLOW_FORWARD_REFERENCES
7399 Define this macro to allow references to structure, union, or
7400 enumeration tags that have not yet been seen to be handled. Some
7401 assemblers choke if forward tags are used, while some require it.
7404 @node Cross-compilation
7405 @section Cross Compilation and Floating Point
7406 @cindex cross compilation and floating point
7407 @cindex floating point and cross compilation
7409 While all modern machines use 2's complement representation for integers,
7410 there are a variety of representations for floating point numbers. This
7411 means that in a cross-compiler the representation of floating point numbers
7412 in the compiled program may be different from that used in the machine
7413 doing the compilation.
7416 Because different representation systems may offer different amounts of
7417 range and precision, the cross compiler cannot safely use the host
7418 machine's floating point arithmetic. Therefore, floating point constants
7419 must be represented in the target machine's format. This means that the
7420 cross compiler cannot use @code{atof} to parse a floating point constant;
7421 it must have its own special routine to use instead. Also, constant
7422 folding must emulate the target machine's arithmetic (or must not be done
7425 The macros in the following table should be defined only if you are cross
7426 compiling between different floating point formats.
7428 Otherwise, don't define them. Then default definitions will be set up which
7429 use @code{double} as the data type, @code{==} to test for equality, etc.
7431 You don't need to worry about how many times you use an operand of any
7432 of these macros. The compiler never uses operands which have side effects.
7435 @findex REAL_VALUE_TYPE
7436 @item REAL_VALUE_TYPE
7437 A macro for the C data type to be used to hold a floating point value
7438 in the target machine's format. Typically this would be a
7439 @code{struct} containing an array of @code{int}.
7441 @findex REAL_VALUES_EQUAL
7442 @item REAL_VALUES_EQUAL (@var{x}, @var{y})
7443 A macro for a C expression which compares for equality the two values,
7444 @var{x} and @var{y}, both of type @code{REAL_VALUE_TYPE}.
7446 @findex REAL_VALUES_LESS
7447 @item REAL_VALUES_LESS (@var{x}, @var{y})
7448 A macro for a C expression which tests whether @var{x} is less than
7449 @var{y}, both values being of type @code{REAL_VALUE_TYPE} and
7450 interpreted as floating point numbers in the target machine's
7453 @findex REAL_VALUE_LDEXP
7455 @item REAL_VALUE_LDEXP (@var{x}, @var{scale})
7456 A macro for a C expression which performs the standard library
7457 function @code{ldexp}, but using the target machine's floating point
7458 representation. Both @var{x} and the value of the expression have
7459 type @code{REAL_VALUE_TYPE}. The second argument, @var{scale}, is an
7462 @findex REAL_VALUE_FIX
7463 @item REAL_VALUE_FIX (@var{x})
7464 A macro whose definition is a C expression to convert the target-machine
7465 floating point value @var{x} to a signed integer. @var{x} has type
7466 @code{REAL_VALUE_TYPE}.
7468 @findex REAL_VALUE_UNSIGNED_FIX
7469 @item REAL_VALUE_UNSIGNED_FIX (@var{x})
7470 A macro whose definition is a C expression to convert the target-machine
7471 floating point value @var{x} to an unsigned integer. @var{x} has type
7472 @code{REAL_VALUE_TYPE}.
7474 @findex REAL_VALUE_RNDZINT
7475 @item REAL_VALUE_RNDZINT (@var{x})
7476 A macro whose definition is a C expression to round the target-machine
7477 floating point value @var{x} towards zero to an integer value (but still
7478 as a floating point number). @var{x} has type @code{REAL_VALUE_TYPE},
7479 and so does the value.
7481 @findex REAL_VALUE_UNSIGNED_RNDZINT
7482 @item REAL_VALUE_UNSIGNED_RNDZINT (@var{x})
7483 A macro whose definition is a C expression to round the target-machine
7484 floating point value @var{x} towards zero to an unsigned integer value
7485 (but still represented as a floating point number). @var{x} has type
7486 @code{REAL_VALUE_TYPE}, and so does the value.
7488 @findex REAL_VALUE_ATOF
7489 @item REAL_VALUE_ATOF (@var{string}, @var{mode})
7490 A macro for a C expression which converts @var{string}, an expression of
7491 type @code{char *}, into a floating point number in the target machine's
7492 representation for mode @var{mode}. The value has type
7493 @code{REAL_VALUE_TYPE}.
7495 @findex REAL_INFINITY
7497 Define this macro if infinity is a possible floating point value, and
7498 therefore division by 0 is legitimate.
7500 @findex REAL_VALUE_ISINF
7502 @item REAL_VALUE_ISINF (@var{x})
7503 A macro for a C expression which determines whether @var{x}, a floating
7504 point value, is infinity. The value has type @code{int}.
7505 By default, this is defined to call @code{isinf}.
7507 @findex REAL_VALUE_ISNAN
7509 @item REAL_VALUE_ISNAN (@var{x})
7510 A macro for a C expression which determines whether @var{x}, a floating
7511 point value, is a ``nan'' (not-a-number). The value has type
7512 @code{int}. By default, this is defined to call @code{isnan}.
7515 @cindex constant folding and floating point
7516 Define the following additional macros if you want to make floating
7517 point constant folding work while cross compiling. If you don't
7518 define them, cross compilation is still possible, but constant folding
7519 will not happen for floating point values.
7522 @findex REAL_ARITHMETIC
7523 @item REAL_ARITHMETIC (@var{output}, @var{code}, @var{x}, @var{y})
7524 A macro for a C statement which calculates an arithmetic operation of
7525 the two floating point values @var{x} and @var{y}, both of type
7526 @code{REAL_VALUE_TYPE} in the target machine's representation, to
7527 produce a result of the same type and representation which is stored
7528 in @var{output} (which will be a variable).
7530 The operation to be performed is specified by @var{code}, a tree code
7531 which will always be one of the following: @code{PLUS_EXPR},
7532 @code{MINUS_EXPR}, @code{MULT_EXPR}, @code{RDIV_EXPR},
7533 @code{MAX_EXPR}, @code{MIN_EXPR}.@refill
7535 @cindex overflow while constant folding
7536 The expansion of this macro is responsible for checking for overflow.
7537 If overflow happens, the macro expansion should execute the statement
7538 @code{return 0;}, which indicates the inability to perform the
7539 arithmetic operation requested.
7541 @findex REAL_VALUE_NEGATE
7542 @item REAL_VALUE_NEGATE (@var{x})
7543 A macro for a C expression which returns the negative of the floating
7544 point value @var{x}. Both @var{x} and the value of the expression
7545 have type @code{REAL_VALUE_TYPE} and are in the target machine's
7546 floating point representation.
7548 There is no way for this macro to report overflow, since overflow
7549 can't happen in the negation operation.
7551 @findex REAL_VALUE_TRUNCATE
7552 @item REAL_VALUE_TRUNCATE (@var{mode}, @var{x})
7553 A macro for a C expression which converts the floating point value
7554 @var{x} to mode @var{mode}.
7556 Both @var{x} and the value of the expression are in the target machine's
7557 floating point representation and have type @code{REAL_VALUE_TYPE}.
7558 However, the value should have an appropriate bit pattern to be output
7559 properly as a floating constant whose precision accords with mode
7562 There is no way for this macro to report overflow.
7564 @findex REAL_VALUE_TO_INT
7565 @item REAL_VALUE_TO_INT (@var{low}, @var{high}, @var{x})
7566 A macro for a C expression which converts a floating point value
7567 @var{x} into a double-precision integer which is then stored into
7568 @var{low} and @var{high}, two variables of type @var{int}.
7570 @item REAL_VALUE_FROM_INT (@var{x}, @var{low}, @var{high}, @var{mode})
7571 @findex REAL_VALUE_FROM_INT
7572 A macro for a C expression which converts a double-precision integer
7573 found in @var{low} and @var{high}, two variables of type @var{int},
7574 into a floating point value which is then stored into @var{x}.
7575 The value is in the target machine's representation for mode @var{mode}
7576 and has the type @code{REAL_VALUE_TYPE}.
7579 @node Mode Switching
7580 @section Mode Switching Instructions
7581 @cindex mode switching
7582 The following macros control mode switching optimizations:
7585 @findex OPTIMIZE_MODE_SWITCHING
7586 @item OPTIMIZE_MODE_SWITCHING (@var{entity})
7587 Define this macro if the port needs extra instructions inserted for mode
7588 switching in an optimizing compilation.
7590 For an example, the SH4 can perform both single and double precision
7591 floating point operations, but to perform a single precision operation,
7592 the FPSCR PR bit has to be cleared, while for a double precision
7593 operation, this bit has to be set. Changing the PR bit requires a general
7594 purpose register as a scratch register, hence these FPSCR sets have to
7595 be inserted before reload, i.e. you can't put this into instruction emitting
7596 or MACHINE_DEPENDENT_REORG.
7598 You can have multiple entities that are mode-switched, and select at run time
7599 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
7600 return non-zero for any @var{entity} that that needs mode-switching.
7601 If you define this macro, you also have to define
7602 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
7603 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
7604 @code{NORMAL_MODE} is optional.
7606 @findex NUM_MODES_FOR_MODE_SWITCHING
7607 @item NUM_MODES_FOR_MODE_SWITCHING
7608 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
7609 initializer for an array of integers. Each initializer element
7610 N refers to an entity that needs mode switching, and specifies the number
7611 of different modes that might need to be set for this entity.
7612 The position of the initializer in the initializer - starting counting at
7613 zero - determines the integer that is used to refer to the mode-switched
7615 In macros that take mode arguments / yield a mode result, modes are
7616 represented as numbers 0 .. N - 1. N is used to specify that no mode
7617 switch is needed / supplied.
7620 @item MODE_NEEDED (@var{entity}, @var{insn})
7621 @var{entity} is an integer specifying a mode-switched entity. If
7622 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
7623 return an integer value not larger than the corresponding element in
7624 NUM_MODES_FOR_MODE_SWITCHING, to denote the mode that @var{entity} must
7625 be switched into prior to the execution of INSN.
7628 @item NORMAL_MODE (@var{entity})
7629 If this macro is defined, it is evaluated for every @var{entity} that needs
7630 mode switching. It should evaluate to an integer, which is a mode that
7631 @var{entity} is assumed to be switched to at function entry and exit.
7633 @findex MODE_PRIORITY_TO_MODE
7634 @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
7635 This macro specifies the order in which modes for ENTITY are processed.
7636 0 is the highest priority, NUM_MODES_FOR_MODE_SWITCHING[ENTITY] - 1 the
7637 lowest. The value of the macro should be an integer designating a mode
7638 for ENTITY. For any fixed @var{entity}, @code{mode_priority_to_mode}
7639 (@var{entity}, @var{n}) shall be a bijection in 0 ..
7640 @code{num_modes_for_mode_switching}[@var{entity}] - 1 .
7642 @findex EMIT_MODE_SET
7643 @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
7644 Generate one or more insns to set @var{entity} to @var{mode}.
7645 @var{hard_reg_live} is the set of hard registers live at the point where
7646 the insn(s) are to be inserted.
7650 @section Miscellaneous Parameters
7651 @cindex parameters, miscellaneous
7653 @c prevent bad page break with this line
7654 Here are several miscellaneous parameters.
7657 @item PREDICATE_CODES
7658 @findex PREDICATE_CODES
7659 Define this if you have defined special-purpose predicates in the file
7660 @file{@var{machine}.c}. This macro is called within an initializer of an
7661 array of structures. The first field in the structure is the name of a
7662 predicate and the second field is an array of rtl codes. For each
7663 predicate, list all rtl codes that can be in expressions matched by the
7664 predicate. The list should have a trailing comma. Here is an example
7665 of two entries in the list for a typical RISC machine:
7668 #define PREDICATE_CODES \
7669 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
7670 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
7673 Defining this macro does not affect the generated code (however,
7674 incorrect definitions that omit an rtl code that may be matched by the
7675 predicate can cause the compiler to malfunction). Instead, it allows
7676 the table built by @file{genrecog} to be more compact and efficient,
7677 thus speeding up the compiler. The most important predicates to include
7678 in the list specified by this macro are those used in the most insn
7681 For each predicate function named in @var{PREDICATE_CODES}, a
7682 declaration will be generated in @file{insn-codes.h}.
7684 @item SPECIAL_MODE_PREDICATES
7685 @findex SPECIAL_MODE_PREDICATES
7686 Define this if you have special predicates that know special things
7687 about modes. Genrecog will warn about certain forms of
7688 @code{match_operand} without a mode; if the operand predicate is
7689 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
7692 Here is an example from the IA-32 port (@code{ext_register_operand}
7693 specially checks for @code{HImode} or @code{SImode} in preparation
7694 for a byte extraction from @code{%ah} etc.).
7697 #define SPECIAL_MODE_PREDICATES \
7698 "ext_register_operand",
7701 @findex CASE_VECTOR_MODE
7702 @item CASE_VECTOR_MODE
7703 An alias for a machine mode name. This is the machine mode that
7704 elements of a jump-table should have.
7706 @findex CASE_VECTOR_SHORTEN_MODE
7707 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7708 Optional: return the preferred mode for an @code{addr_diff_vec}
7709 when the minimum and maximum offset are known. If you define this,
7710 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7711 To make this work, you also have to define INSN_ALIGN and
7712 make the alignment for @code{addr_diff_vec} explicit.
7713 The @var{body} argument is provided so that the offset_unsigned and scale
7714 flags can be updated.
7716 @findex CASE_VECTOR_PC_RELATIVE
7717 @item CASE_VECTOR_PC_RELATIVE
7718 Define this macro to be a C expression to indicate when jump-tables
7719 should contain relative addresses. If jump-tables never contain
7720 relative addresses, then you need not define this macro.
7722 @findex CASE_DROPS_THROUGH
7723 @item CASE_DROPS_THROUGH
7724 Define this if control falls through a @code{case} insn when the index
7725 value is out of range. This means the specified default-label is
7726 actually ignored by the @code{case} insn proper.
7728 @findex CASE_VALUES_THRESHOLD
7729 @item CASE_VALUES_THRESHOLD
7730 Define this to be the smallest number of different values for which it
7731 is best to use a jump-table instead of a tree of conditional branches.
7732 The default is four for machines with a @code{casesi} instruction and
7733 five otherwise. This is best for most machines.
7735 @findex WORD_REGISTER_OPERATIONS
7736 @item WORD_REGISTER_OPERATIONS
7737 Define this macro if operations between registers with integral mode
7738 smaller than a word are always performed on the entire register.
7739 Most RISC machines have this property and most CISC machines do not.
7741 @findex LOAD_EXTEND_OP
7742 @item LOAD_EXTEND_OP (@var{mode})
7743 Define this macro to be a C expression indicating when insns that read
7744 memory in @var{mode}, an integral mode narrower than a word, set the
7745 bits outside of @var{mode} to be either the sign-extension or the
7746 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7747 of @var{mode} for which the
7748 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7749 @code{NIL} for other modes.
7751 This macro is not called with @var{mode} non-integral or with a width
7752 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7753 value in this case. Do not define this macro if it would always return
7754 @code{NIL}. On machines where this macro is defined, you will normally
7755 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7757 @findex SHORT_IMMEDIATES_SIGN_EXTEND
7758 @item SHORT_IMMEDIATES_SIGN_EXTEND
7759 Define this macro if loading short immediate values into registers sign
7762 @findex IMPLICIT_FIX_EXPR
7763 @item IMPLICIT_FIX_EXPR
7764 An alias for a tree code that should be used by default for conversion
7765 of floating point values to fixed point. Normally,
7766 @code{FIX_ROUND_EXPR} is used.@refill
7768 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
7769 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
7770 Define this macro if the same instructions that convert a floating
7771 point number to a signed fixed point number also convert validly to an
7774 @findex EASY_DIV_EXPR
7776 An alias for a tree code that is the easiest kind of division to
7777 compile code for in the general case. It may be
7778 @code{TRUNC_DIV_EXPR}, @code{FLOOR_DIV_EXPR}, @code{CEIL_DIV_EXPR} or
7779 @code{ROUND_DIV_EXPR}. These four division operators differ in how
7780 they round the result to an integer. @code{EASY_DIV_EXPR} is used
7781 when it is permissible to use any of those kinds of division and the
7782 choice should be made on the basis of efficiency.@refill
7786 The maximum number of bytes that a single instruction can move quickly
7787 between memory and registers or between two memory locations.
7789 @findex MAX_MOVE_MAX
7791 The maximum number of bytes that a single instruction can move quickly
7792 between memory and registers or between two memory locations. If this
7793 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7794 constant value that is the largest value that @code{MOVE_MAX} can have
7797 @findex SHIFT_COUNT_TRUNCATED
7798 @item SHIFT_COUNT_TRUNCATED
7799 A C expression that is nonzero if on this machine the number of bits
7800 actually used for the count of a shift operation is equal to the number
7801 of bits needed to represent the size of the object being shifted. When
7802 this macro is non-zero, the compiler will assume that it is safe to omit
7803 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7804 truncates the count of a shift operation. On machines that have
7805 instructions that act on bitfields at variable positions, which may
7806 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7807 also enables deletion of truncations of the values that serve as
7808 arguments to bitfield instructions.
7810 If both types of instructions truncate the count (for shifts) and
7811 position (for bitfield operations), or if no variable-position bitfield
7812 instructions exist, you should define this macro.
7814 However, on some machines, such as the 80386 and the 680x0, truncation
7815 only applies to shift operations and not the (real or pretended)
7816 bitfield operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7817 such machines. Instead, add patterns to the @file{md} file that include
7818 the implied truncation of the shift instructions.
7820 You need not define this macro if it would always have the value of zero.
7822 @findex TRULY_NOOP_TRUNCATION
7823 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
7824 A C expression which is nonzero if on this machine it is safe to
7825 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
7826 bits (where @var{outprec} is smaller than @var{inprec}) by merely
7827 operating on it as if it had only @var{outprec} bits.
7829 On many machines, this expression can be 1.
7831 @c rearranged this, removed the phrase "it is reported that". this was
7832 @c to fix an overfull hbox. --mew 10feb93
7833 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
7834 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
7835 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
7836 such cases may improve things.
7838 @findex STORE_FLAG_VALUE
7839 @item STORE_FLAG_VALUE
7840 A C expression describing the value returned by a comparison operator
7841 with an integral mode and stored by a store-flag instruction
7842 (@samp{s@var{cond}}) when the condition is true. This description must
7843 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
7844 comparison operators whose results have a @code{MODE_INT} mode.
7846 A value of 1 or -1 means that the instruction implementing the
7847 comparison operator returns exactly 1 or -1 when the comparison is true
7848 and 0 when the comparison is false. Otherwise, the value indicates
7849 which bits of the result are guaranteed to be 1 when the comparison is
7850 true. This value is interpreted in the mode of the comparison
7851 operation, which is given by the mode of the first operand in the
7852 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
7853 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7856 If @code{STORE_FLAG_VALUE} is neither 1 or -1, the compiler will
7857 generate code that depends only on the specified bits. It can also
7858 replace comparison operators with equivalent operations if they cause
7859 the required bits to be set, even if the remaining bits are undefined.
7860 For example, on a machine whose comparison operators return an
7861 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7862 @samp{0x80000000}, saying that just the sign bit is relevant, the
7866 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7873 (ashift:SI @var{x} (const_int @var{n}))
7877 where @var{n} is the appropriate shift count to move the bit being
7878 tested into the sign bit.
7880 There is no way to describe a machine that always sets the low-order bit
7881 for a true value, but does not guarantee the value of any other bits,
7882 but we do not know of any machine that has such an instruction. If you
7883 are trying to port GCC to such a machine, include an instruction to
7884 perform a logical-and of the result with 1 in the pattern for the
7885 comparison operators and let us know
7887 (@pxref{Bug Reporting,,How to Report Bugs}).
7890 (@pxref{Bug Reporting,,How to Report Bugs,gcc.info,Using GCC}).
7893 Often, a machine will have multiple instructions that obtain a value
7894 from a comparison (or the condition codes). Here are rules to guide the
7895 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7900 Use the shortest sequence that yields a valid definition for
7901 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7902 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7903 comparison operators to do so because there may be opportunities to
7904 combine the normalization with other operations.
7907 For equal-length sequences, use a value of 1 or -1, with -1 being
7908 slightly preferred on machines with expensive jumps and 1 preferred on
7912 As a second choice, choose a value of @samp{0x80000001} if instructions
7913 exist that set both the sign and low-order bits but do not define the
7917 Otherwise, use a value of @samp{0x80000000}.
7920 Many machines can produce both the value chosen for
7921 @code{STORE_FLAG_VALUE} and its negation in the same number of
7922 instructions. On those machines, you should also define a pattern for
7923 those cases, e.g., one matching
7926 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7929 Some machines can also perform @code{and} or @code{plus} operations on
7930 condition code values with less instructions than the corresponding
7931 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
7932 machines, define the appropriate patterns. Use the names @code{incscc}
7933 and @code{decscc}, respectively, for the patterns which perform
7934 @code{plus} or @code{minus} operations on condition code values. See
7935 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
7936 find such instruction sequences on other machines.
7938 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7941 @findex FLOAT_STORE_FLAG_VALUE
7942 @item FLOAT_STORE_FLAG_VALUE (@var{mode})
7943 A C expression that gives a non-zero @code{REAL_VALUE_TYPE} value that is
7944 returned when comparison operators with floating-point results are true.
7945 Define this macro on machine that have comparison operations that return
7946 floating-point values. If there are no such operations, do not define
7951 An alias for the machine mode for pointers. On most machines, define
7952 this to be the integer mode corresponding to the width of a hardware
7953 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7954 On some machines you must define this to be one of the partial integer
7955 modes, such as @code{PSImode}.
7957 The width of @code{Pmode} must be at least as large as the value of
7958 @code{POINTER_SIZE}. If it is not equal, you must define the macro
7959 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7962 @findex FUNCTION_MODE
7964 An alias for the machine mode used for memory references to functions
7965 being called, in @code{call} RTL expressions. On most machines this
7966 should be @code{QImode}.
7968 @findex INTEGRATE_THRESHOLD
7969 @item INTEGRATE_THRESHOLD (@var{decl})
7970 A C expression for the maximum number of instructions above which the
7971 function @var{decl} should not be inlined. @var{decl} is a
7972 @code{FUNCTION_DECL} node.
7974 The default definition of this macro is 64 plus 8 times the number of
7975 arguments that the function accepts. Some people think a larger
7976 threshold should be used on RISC machines.
7978 @findex SCCS_DIRECTIVE
7979 @item SCCS_DIRECTIVE
7980 Define this if the preprocessor should ignore @code{#sccs} directives
7981 and print no error message.
7983 @findex NO_IMPLICIT_EXTERN_C
7984 @item NO_IMPLICIT_EXTERN_C
7985 Define this macro if the system header files support C++ as well as C.
7986 This macro inhibits the usual method of using system header files in
7987 C++, which is to pretend that the file's contents are enclosed in
7988 @samp{extern "C" @{@dots{}@}}.
7990 @findex HANDLE_PRAGMA
7991 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
7992 This macro is no longer supported. You must use
7993 @code{REGISTER_TARGET_PRAGMAS} instead.
7995 @findex REGISTER_TARGET_PRAGMAS
7998 @item REGISTER_TARGET_PRAGMAS (@var{pfile})
7999 Define this macro if you want to implement any target-specific pragmas.
8000 If defined, it is a C expression which makes a series of calls to the
8001 @code{cpp_register_pragma} and/or @code{cpp_register_pragma_space}
8002 functions. The @var{pfile} argument is the first argument to supply to
8003 these functions. The macro may also do setup required for the pragmas.
8005 The primary reason to define this macro is to provide compatibility with
8006 other compilers for the same target. In general, we discourage
8007 definition of target-specific pragmas for GCC.
8009 If the pragma can be implemented by attributes then the macro
8010 @samp{INSERT_ATTRIBUTES} might be a useful one to define as well.
8012 Preprocessor macros that appear on pragma lines are not expanded. All
8013 @samp{#pragma} directives that do not match any registered pragma are
8014 silently ignored, unless the user specifies @samp{-Wunknown-pragmas}.
8016 @deftypefun void cpp_register_pragma (cpp_reader *@var{pfile}, const char *@var{space}, const char *@var{name}, void (*@var{callback}) (cpp_reader *))
8018 Each call to @code{cpp_register_pragma} establishes one pragma. The
8019 @var{callback} routine will be called when the preprocessor encounters a
8023 #pragma [@var{space}] @var{name} @dots{}
8026 @var{space} must have been the subject of a previous call to
8027 @code{cpp_register_pragma_space}, or else be a null pointer. The
8028 callback routine receives @var{pfile} as its first argument, but must
8029 not use it for anything (this may change in the future). It may read
8030 any text after the @var{name} by making calls to @code{c_lex}. Text
8031 which is not read by the callback will be silently ignored.
8033 Note that both @var{space} and @var{name} are case sensitive.
8035 For an example use of this routine, see @file{c4x.h} and the callback
8036 routines defined in @file{c4x.c}.
8038 Note that the use of @code{c_lex} is specific to the C and C++
8039 compilers. It will not work in the Java or Fortran compilers, or any
8040 other language compilers for that matter. Thus if @code{c_lex} is going
8041 to be called from target-specific code, it must only be done so when
8042 building hte C and C++ compilers. This can be done by defining the
8043 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8044 target entry in the @code{config.gcc} file. These variables should name
8045 the target-specific, language-specific object file which contains the
8046 code that uses @code{c_lex}. Note it will also be necessary to add a
8047 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8048 how to build this object file.
8051 @deftypefun void cpp_register_pragma_space (cpp_reader *@var{pfile}, const char *@var{space})
8052 This routine establishes a namespace for pragmas, which will be
8053 registered by subsequent calls to @code{cpp_register_pragma}. For
8054 example, pragmas defined by the C standard are in the @samp{STDC}
8055 namespace, and pragmas specific to GCC are in the @samp{GCC} namespace.
8057 For an example use of this routine in a target header, see @file{v850.h}.
8060 @findex HANDLE_SYSV_PRAGMA
8063 @item HANDLE_SYSV_PRAGMA
8064 Define this macro (to a value of 1) if you want the System V style
8065 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8066 [=<value>]} to be supported by gcc.
8068 The pack pragma specifies the maximum alignment (in bytes) of fields
8069 within a structure, in much the same way as the @samp{__aligned__} and
8070 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8071 the behaviour to the default.
8073 The weak pragma only works if @code{SUPPORTS_WEAK} and
8074 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8075 of specifically named weak labels, optionally with a value.
8077 @findex HANDLE_PRAGMA_PACK_PUSH_POP
8080 @item HANDLE_PRAGMA_PACK_PUSH_POP
8081 Define this macro (to a value of 1) if you want to support the Win32
8082 style pragmas @samp{#pragma pack(push,<n>)} and @samp{#pragma
8083 pack(pop)}. The pack(push,<n>) pragma specifies the maximum alignment
8084 (in bytes) of fields within a structure, in much the same way as the
8085 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8086 pack value of zero resets the behaviour to the default. Successive
8087 invocations of this pragma cause the previous values to be stacked, so
8088 that invocations of @samp{#pragma pack(pop)} will return to the previous
8091 @findex VALID_MACHINE_DECL_ATTRIBUTE
8092 @item VALID_MACHINE_DECL_ATTRIBUTE (@var{decl}, @var{attributes}, @var{identifier}, @var{args})
8093 If defined, a C expression whose value is nonzero if @var{identifier} with
8094 arguments @var{args} is a valid machine specific attribute for @var{decl}.
8095 The attributes in @var{attributes} have previously been assigned to @var{decl}.
8097 @findex VALID_MACHINE_TYPE_ATTRIBUTE
8098 @item VALID_MACHINE_TYPE_ATTRIBUTE (@var{type}, @var{attributes}, @var{identifier}, @var{args})
8099 If defined, a C expression whose value is nonzero if @var{identifier} with
8100 arguments @var{args} is a valid machine specific attribute for @var{type}.
8101 The attributes in @var{attributes} have previously been assigned to @var{type}.
8103 @findex COMP_TYPE_ATTRIBUTES
8104 @item COMP_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
8105 If defined, a C expression whose value is zero if the attributes on
8106 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8107 and two if they are nearly compatible (which causes a warning to be
8110 @findex SET_DEFAULT_TYPE_ATTRIBUTES
8111 @item SET_DEFAULT_TYPE_ATTRIBUTES (@var{type})
8112 If defined, a C statement that assigns default attributes to
8113 newly defined @var{type}.
8115 @findex MERGE_MACHINE_TYPE_ATTRIBUTES
8116 @item MERGE_MACHINE_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
8117 Define this macro if the merging of type attributes needs special handling.
8118 If defined, the result is a list of the combined TYPE_ATTRIBUTES of
8119 @var{type1} and @var{type2}. It is assumed that comptypes has already been
8120 called and returned 1.
8122 @findex MERGE_MACHINE_DECL_ATTRIBUTES
8123 @item MERGE_MACHINE_DECL_ATTRIBUTES (@var{olddecl}, @var{newdecl})
8124 Define this macro if the merging of decl attributes needs special handling.
8125 If defined, the result is a list of the combined DECL_MACHINE_ATTRIBUTES of
8126 @var{olddecl} and @var{newdecl}. @var{newdecl} is a duplicate declaration
8127 of @var{olddecl}. Examples of when this is needed are when one attribute
8128 overrides another, or when an attribute is nullified by a subsequent
8131 @findex INSERT_ATTRIBUTES
8132 @item INSERT_ATTRIBUTES (@var{node}, @var{attr_ptr}, @var{prefix_ptr})
8133 Define this macro if you want to be able to add attributes to a decl
8134 when it is being created. This is normally useful for backends which
8135 wish to implement a pragma by using the attributes which correspond to
8136 the pragma's effect. The @var{node} argument is the decl which is being
8137 created. The @var{attr_ptr} argument is a pointer to the attribute list
8138 for this decl. The @var{prefix_ptr} is a pointer to the list of
8139 attributes that have appeared after the specifiers and modifiers of the
8140 declaration, but before the declaration proper.
8142 @findex SET_DEFAULT_DECL_ATTRIBUTES
8143 @item SET_DEFAULT_DECL_ATTRIBUTES (@var{decl}, @var{attributes})
8144 If defined, a C statement that assigns default attributes to
8145 newly defined @var{decl}.
8147 @findex DOLLARS_IN_IDENTIFIERS
8148 @item DOLLARS_IN_IDENTIFIERS
8149 Define this macro to control use of the character @samp{$} in identifier
8150 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
8151 1 is the default; there is no need to define this macro in that case.
8152 This macro controls the compiler proper; it does not affect the preprocessor.
8154 @findex NO_DOLLAR_IN_LABEL
8155 @item NO_DOLLAR_IN_LABEL
8156 Define this macro if the assembler does not accept the character
8157 @samp{$} in label names. By default constructors and destructors in
8158 G++ have @samp{$} in the identifiers. If this macro is defined,
8159 @samp{.} is used instead.
8161 @findex NO_DOT_IN_LABEL
8162 @item NO_DOT_IN_LABEL
8163 Define this macro if the assembler does not accept the character
8164 @samp{.} in label names. By default constructors and destructors in G++
8165 have names that use @samp{.}. If this macro is defined, these names
8166 are rewritten to avoid @samp{.}.
8168 @findex DEFAULT_MAIN_RETURN
8169 @item DEFAULT_MAIN_RETURN
8170 Define this macro if the target system expects every program's @code{main}
8171 function to return a standard ``success'' value by default (if no other
8172 value is explicitly returned).
8174 The definition should be a C statement (sans semicolon) to generate the
8175 appropriate rtl instructions. It is used only when compiling the end of
8180 Define this if the target system lacks the function @code{atexit}
8181 from the ISO C standard. If this macro is defined, a default definition
8182 will be provided to support C++. If @code{ON_EXIT} is not defined,
8183 a default @code{exit} function will also be provided.
8187 Define this macro if the target has another way to implement atexit
8188 functionality without replacing @code{exit}. For instance, SunOS 4 has
8189 a similar @code{on_exit} library function.
8191 The definition should be a functional macro which can be used just like
8192 the @code{atexit} function.
8196 Define this if your @code{exit} function needs to do something
8197 besides calling an external function @code{_cleanup} before
8198 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
8199 only needed if @code{NEED_ATEXIT} is defined and @code{ON_EXIT} is not
8202 @findex INSN_SETS_ARE_DELAYED
8203 @item INSN_SETS_ARE_DELAYED (@var{insn})
8204 Define this macro as a C expression that is nonzero if it is safe for the
8205 delay slot scheduler to place instructions in the delay slot of @var{insn},
8206 even if they appear to use a resource set or clobbered in @var{insn}.
8207 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8208 every @code{call_insn} has this behavior. On machines where some @code{insn}
8209 or @code{jump_insn} is really a function call and hence has this behavior,
8210 you should define this macro.
8212 You need not define this macro if it would always return zero.
8214 @findex INSN_REFERENCES_ARE_DELAYED
8215 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
8216 Define this macro as a C expression that is nonzero if it is safe for the
8217 delay slot scheduler to place instructions in the delay slot of @var{insn},
8218 even if they appear to set or clobber a resource referenced in @var{insn}.
8219 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8220 some @code{insn} or @code{jump_insn} is really a function call and its operands
8221 are registers whose use is actually in the subroutine it calls, you should
8222 define this macro. Doing so allows the delay slot scheduler to move
8223 instructions which copy arguments into the argument registers into the delay
8226 You need not define this macro if it would always return zero.
8228 @findex MACHINE_DEPENDENT_REORG
8229 @item MACHINE_DEPENDENT_REORG (@var{insn})
8230 In rare cases, correct code generation requires extra machine
8231 dependent processing between the second jump optimization pass and
8232 delayed branch scheduling. On those machines, define this macro as a C
8233 statement to act on the code starting at @var{insn}.
8235 @findex MULTIPLE_SYMBOL_SPACES
8236 @item MULTIPLE_SYMBOL_SPACES
8237 Define this macro if in some cases global symbols from one translation
8238 unit may not be bound to undefined symbols in another translation unit
8239 without user intervention. For instance, under Microsoft Windows
8240 symbols must be explicitly imported from shared libraries (DLLs).
8242 @findex MD_ASM_CLOBBERS
8243 @item MD_ASM_CLOBBERS
8244 A C statement that adds to @var{CLOBBERS} @code{STRING_CST} trees for
8245 any hard regs the port wishes to automatically clobber for all asms.
8249 A C expression that returns how many instructions can be issued at the
8250 same time if the machine is a superscalar machine.
8252 @findex MD_SCHED_INIT
8253 @item MD_SCHED_INIT (@var{file}, @var{verbose}, @var{max_ready})
8254 A C statement which is executed by the scheduler at the
8255 beginning of each block of instructions that are to be scheduled.
8256 @var{file} is either a null pointer, or a stdio stream to write any
8257 debug output to. @var{verbose} is the verbose level provided by
8258 @samp{-fsched-verbose-}@var{n}. @var{max_ready} is the maximum number
8259 of insns in the current scheduling region that can be live at the same
8260 time. This can be used to allocate scratch space if it is needed.
8262 @findex MD_SCHED_FINISH
8263 @item MD_SCHED_FINISH (@var{file}, @var{verbose})
8264 A C statement which is executed by the scheduler at the end of each block
8265 of instructions that are to be scheduled. It can be used to perform
8266 cleanup of any actions done by the other scheduling macros.
8267 @var{file} is either a null pointer, or a stdio stream to write any
8268 debug output to. @var{verbose} is the verbose level provided by
8269 @samp{-fsched-verbose-}@var{n}.
8271 @findex MD_SCHED_REORDER
8272 @item MD_SCHED_REORDER (@var{file}, @var{verbose}, @var{ready}, @var{n_ready}, @var{clock}, @var{can_issue_more})
8273 A C statement which is executed by the scheduler after it
8274 has scheduled the ready list to allow the machine description to reorder
8275 it (for example to combine two small instructions together on
8276 @samp{VLIW} machines). @var{file} is either a null pointer, or a stdio
8277 stream to write any debug output to. @var{verbose} is the verbose level
8278 provided by @samp{-fsched-verbose-}@var{n}. @var{ready} is a pointer to
8279 the ready list of instructions that are ready to be scheduled.
8280 @var{n_ready} is the number of elements in the ready list. The
8281 scheduler reads the ready list in reverse order, starting with
8282 @var{ready}[@var{n_ready}-1] and going to @var{ready}[0]. @var{clock}
8283 is the timer tick of the scheduler. @var{can_issue_more} is an output
8284 parameter that is set to the number of insns that can issue this clock;
8285 normally this is just @code{issue_rate}. See also @samp{MD_SCHED_REORDER2}.
8287 @findex MD_SCHED_REORDER2
8288 @item MD_SCHED_REORDER2 (@var{file}, @var{verbose}, @var{ready}, @var{n_ready}, @var{clock}, @var{can_issue_more})
8289 Like @samp{MD_SCHED_REORDER}, but called at a different time. While the
8290 @samp{MD_SCHED_REORDER} macro is called whenever the scheduler starts a
8291 new cycle, this macro is used immediately after @samp{MD_SCHED_VARIABLE_ISSUE}
8292 is called; it can reorder the ready list and set @var{can_issue_more} to
8293 determine whether there are more insns to be scheduled in the same cycle.
8294 Defining this macro can be useful if there are frequent situations where
8295 scheduling one insn causes other insns to become ready in the same cycle,
8296 these other insns can then be taken into account properly.
8298 @findex MD_SCHED_VARIABLE_ISSUE
8299 @item MD_SCHED_VARIABLE_ISSUE (@var{file}, @var{verbose}, @var{insn}, @var{more})
8300 A C statement which is executed by the scheduler after it
8301 has scheduled an insn from the ready list. @var{file} is either a null
8302 pointer, or a stdio stream to write any debug output to. @var{verbose}
8303 is the verbose level provided by @samp{-fsched-verbose-}@var{n}.
8304 @var{insn} is the instruction that was scheduled. @var{more} is the
8305 number of instructions that can be issued in the current cycle. The
8306 @samp{MD_SCHED_VARIABLE_ISSUE} macro is responsible for updating the
8307 value of @var{more} (typically by @var{more}--).
8309 @findex MAX_INTEGER_COMPUTATION_MODE
8310 @item MAX_INTEGER_COMPUTATION_MODE
8311 Define this to the largest integer machine mode which can be used for
8312 operations other than load, store and copy operations.
8314 You need only define this macro if the target holds values larger than
8315 @code{word_mode} in general purpose registers. Most targets should not define
8318 @findex MATH_LIBRARY
8320 Define this macro as a C string constant for the linker argument to link
8321 in the system math library, or @samp{""} if the target does not have a
8322 separate math library.
8324 You need only define this macro if the default of @samp{"-lm"} is wrong.
8326 @findex LIBRARY_PATH_ENV
8327 @item LIBRARY_PATH_ENV
8328 Define this macro as a C string constant for the environment variable that
8329 specifies where the linker should look for libraries.
8331 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
8334 @findex TARGET_HAS_F_SETLKW
8335 @item TARGET_HAS_F_SETLKW
8336 Define this macro if the target supports file locking with fcntl / F_SETLKW.
8337 Note that this functionality is part of POSIX.
8338 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
8339 to use file locking when exiting a program, which avoids race conditions
8340 if the program has forked.
8342 @findex MAX_CONDITIONAL_EXECUTE
8343 @item MAX_CONDITIONAL_EXECUTE
8345 A C expression for the maximum number of instructions to execute via
8346 conditional execution instructions instead of a branch. A value of
8347 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8348 1 if it does use cc0.
8350 @findex IFCVT_MODIFY_TESTS
8351 @item IFCVT_MODIFY_TESTS
8352 A C expression to modify the tests in @code{TRUE_EXPR}, and
8353 @code{FALSE_EXPPR} for use in converting insns in @code{TEST_BB},
8354 @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB} basic blocks to
8355 conditional execution. Set either @code{TRUE_EXPR} or @code{FALSE_EXPR}
8356 to a null pointer if the tests cannot be converted.
8358 @findex IFCVT_MODIFY_INSN
8359 @item IFCVT_MODIFY_INSN
8360 A C expression to modify the @code{PATTERN} of an @code{INSN} that is to
8361 be converted to conditional execution format.
8363 @findex IFCVT_MODIFY_FINAL
8364 @item IFCVT_MODIFY_FINAL
8365 A C expression to perform any final machine dependent modifications in
8366 converting code to conditional execution in the basic blocks
8367 @code{TEST_BB}, @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB}.
8369 @findex IFCVT_MODIFY_CANCEL
8370 @item IFCVT_MODIFY_CANCEL
8371 A C expression to cancel any machine dependent modifications in
8372 converting code to conditional execution in the basic blocks
8373 @code{TEST_BB}, @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB}.