1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000
2 @c Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Target Description Macros
8 @cindex machine description macros
9 @cindex target description macros
10 @cindex macros, target description
11 @cindex @file{tm.h} macros
13 In addition to the file @file{@var{machine}.md}, a machine description
14 includes a C header file conventionally given the name
15 @file{@var{machine}.h}. This header file defines numerous macros
16 that convey the information about the target machine that does not fit
17 into the scheme of the @file{.md} file. The file @file{tm.h} should be
18 a link to @file{@var{machine}.h}. The header file @file{config.h}
19 includes @file{tm.h} and most compiler source files include
23 * Driver:: Controlling how the driver runs the compilation passes.
24 * Run-time Target:: Defining @samp{-m} options like @samp{-m68000} and @samp{-m68020}.
25 * Storage Layout:: Defining sizes and alignments of data.
26 * Type Layout:: Defining sizes and properties of basic user data types.
27 * Registers:: Naming and describing the hardware registers.
28 * Register Classes:: Defining the classes of hardware registers.
29 * Stack and Calling:: Defining which way the stack grows and by how much.
30 * Varargs:: Defining the varargs macros.
31 * Trampolines:: Code set up at run time to enter a nested function.
32 * Library Calls:: Controlling how library routines are implicitly called.
33 * Addressing Modes:: Defining addressing modes valid for memory operands.
34 * Condition Code:: Defining how insns update the condition code.
35 * Costs:: Defining relative costs of different operations.
36 * Sections:: Dividing storage into text, data, and other sections.
37 * PIC:: Macros for position independent code.
38 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
39 * Debugging Info:: Defining the format of debugging output.
40 * Cross-compilation:: Handling floating point for cross-compilers.
41 * Mode Switching:: Insertion of mode-switching instructions.
42 * Misc:: Everything else.
46 @section Controlling the Compilation Driver, @file{gcc}
48 @cindex controlling the compilation driver
50 @c prevent bad page break with this line
51 You can control the compilation driver.
54 @findex SWITCH_TAKES_ARG
55 @item SWITCH_TAKES_ARG (@var{char})
56 A C expression which determines whether the option @samp{-@var{char}}
57 takes arguments. The value should be the number of arguments that
58 option takes--zero, for many options.
60 By default, this macro is defined as
61 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
62 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
63 wish to add additional options which take arguments. Any redefinition
64 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
67 @findex WORD_SWITCH_TAKES_ARG
68 @item WORD_SWITCH_TAKES_ARG (@var{name})
69 A C expression which determines whether the option @samp{-@var{name}}
70 takes arguments. The value should be the number of arguments that
71 option takes--zero, for many options. This macro rather than
72 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
74 By default, this macro is defined as
75 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
76 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
77 wish to add additional options which take arguments. Any redefinition
78 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
81 @findex SWITCH_CURTAILS_COMPILATION
82 @item SWITCH_CURTAILS_COMPILATION (@var{char})
83 A C expression which determines whether the option @samp{-@var{char}}
84 stops compilation before the generation of an executable. The value is
85 boolean, non-zero if the option does stop an executable from being
86 generated, zero otherwise.
88 By default, this macro is defined as
89 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
90 options properly. You need not define
91 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
92 options which affect the generation of an executable. Any redefinition
93 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
94 for additional options.
96 @findex SWITCHES_NEED_SPACES
97 @item SWITCHES_NEED_SPACES
98 A string-valued C expression which enumerates the options for which
99 the linker needs a space between the option and its argument.
101 If this macro is not defined, the default value is @code{""}.
105 A C string constant that tells the GCC driver program options to
106 pass to CPP. It can also specify how to translate options you
107 give to GCC into options for GCC to pass to the CPP.
109 Do not define this macro if it does not need to do anything.
111 @findex NO_BUILTIN_SIZE_TYPE
112 @item NO_BUILTIN_SIZE_TYPE
113 If this macro is defined, the preprocessor will not define the builtin macro
114 @code{__SIZE_TYPE__}. The macro @code{__SIZE_TYPE__} must then be defined
115 by @code{CPP_SPEC} instead.
117 This should be defined if @code{SIZE_TYPE} depends on target dependent flags
118 which are not accessible to the preprocessor. Otherwise, it should not
121 @findex NO_BUILTIN_PTRDIFF_TYPE
122 @item NO_BUILTIN_PTRDIFF_TYPE
123 If this macro is defined, the preprocessor will not define the builtin macro
124 @code{__PTRDIFF_TYPE__}. The macro @code{__PTRDIFF_TYPE__} must then be
125 defined by @code{CPP_SPEC} instead.
127 This should be defined if @code{PTRDIFF_TYPE} depends on target dependent flags
128 which are not accessible to the preprocessor. Otherwise, it should not
131 @findex NO_BUILTIN_WCHAR_TYPE
132 @item NO_BUILTIN_WCHAR_TYPE
133 If this macro is defined, the preprocessor will not define the builtin macro
134 @code{__WCHAR_TYPE__}. The macro @code{__WCHAR_TYPE__} must then be
135 defined by @code{CPP_SPEC} instead.
137 This should be defined if @code{WCHAR_TYPE} depends on target dependent flags
138 which are not accessible to the preprocessor. Otherwise, it should not
141 @findex NO_BUILTIN_WINT_TYPE
142 @item NO_BUILTIN_WINT_TYPE
143 If this macro is defined, the preprocessor will not define the builtin macro
144 @code{__WINT_TYPE__}. The macro @code{__WINT_TYPE__} must then be
145 defined by @code{CPP_SPEC} instead.
147 This should be defined if @code{WINT_TYPE} depends on target dependent flags
148 which are not accessible to the preprocessor. Otherwise, it should not
151 @findex SIGNED_CHAR_SPEC
152 @item SIGNED_CHAR_SPEC
153 A C string constant that tells the GCC driver program options to
154 pass to CPP. By default, this macro is defined to pass the option
155 @samp{-D__CHAR_UNSIGNED__} to CPP if @code{char} will be treated as
156 @code{unsigned char} by @code{cc1}.
158 Do not define this macro unless you need to override the default
163 A C string constant that tells the GCC driver program options to
164 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
166 It can also specify how to translate options you give to GCC into options
167 for GCC to pass to front ends..
169 Do not define this macro if it does not need to do anything.
173 A C string constant that tells the GCC driver program options to
174 pass to @code{cc1plus}. It can also specify how to translate options you
175 give to GCC into options for GCC to pass to the @code{cc1plus}.
177 Do not define this macro if it does not need to do anything.
178 Note that everything defined in CC1_SPEC is already passed to
179 @code{cc1plus} so there is no need to duplicate the contents of
180 CC1_SPEC in CC1PLUS_SPEC.
184 A C string constant that tells the GCC driver program options to
185 pass to the assembler. It can also specify how to translate options
186 you give to GCC into options for GCC to pass to the assembler.
187 See the file @file{sun3.h} for an example of this.
189 Do not define this macro if it does not need to do anything.
191 @findex ASM_FINAL_SPEC
193 A C string constant that tells the GCC driver program how to
194 run any programs which cleanup after the normal assembler.
195 Normally, this is not needed. See the file @file{mips.h} for
198 Do not define this macro if it does not need to do anything.
202 A C string constant that tells the GCC driver program options to
203 pass to the linker. It can also specify how to translate options you
204 give to GCC into options for GCC to pass to the linker.
206 Do not define this macro if it does not need to do anything.
210 Another C string constant used much like @code{LINK_SPEC}. The difference
211 between the two is that @code{LIB_SPEC} is used at the end of the
212 command given to the linker.
214 If this macro is not defined, a default is provided that
215 loads the standard C library from the usual place. See @file{gcc.c}.
219 Another C string constant that tells the GCC driver program
220 how and when to place a reference to @file{libgcc.a} into the
221 linker command line. This constant is placed both before and after
222 the value of @code{LIB_SPEC}.
224 If this macro is not defined, the GCC driver provides a default that
225 passes the string @samp{-lgcc} to the linker unless the @samp{-shared}
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 @samp{-ansi} option is
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 ANSI standard, so they are
556 predefined regardless of whether @samp{-ansi} is specified.
558 For example, on the Sun, one can use the following value:
561 "-Dmc68000 -Dsun -Dunix"
564 The result is to define the macros @code{__mc68000__}, @code{__sun__}
565 and @code{__unix__} unconditionally, and the macros @code{mc68000},
566 @code{sun} and @code{unix} provided @samp{-ansi} is not specified.
568 @findex extern int target_flags
569 @item extern int target_flags;
570 This declaration should be present.
572 @cindex optional hardware or system features
573 @cindex features, optional, in system conventions
575 This series of macros is to allow compiler command arguments to
576 enable or disable the use of optional features of the target machine.
577 For example, one machine description serves both the 68000 and
578 the 68020; a command argument tells the compiler whether it should
579 use 68020-only instructions or not. This command argument works
580 by means of a macro @code{TARGET_68020} that tests a bit in
583 Define a macro @code{TARGET_@var{featurename}} for each such option.
584 Its definition should test a bit in @code{target_flags}. It is
585 recommended that a helper macro @code{TARGET_MASK_@var{featurename}}
586 is defined for each bit-value to test, and used in
587 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
591 #define TARGET_MASK_68020 1
592 #define TARGET_68020 (target_flags & TARGET_MASK_68020)
595 One place where these macros are used is in the condition-expressions
596 of instruction patterns. Note how @code{TARGET_68020} appears
597 frequently in the 68000 machine description file, @file{m68k.md}.
598 Another place they are used is in the definitions of the other
599 macros in the @file{@var{machine}.h} file.
601 @findex TARGET_SWITCHES
602 @item TARGET_SWITCHES
603 This macro defines names of command options to set and clear
604 bits in @code{target_flags}. Its definition is an initializer
605 with a subgrouping for each command option.
607 Each subgrouping contains a string constant, that defines the option
608 name, a number, which contains the bits to set in
609 @code{target_flags}, and a second string which is the description
610 displayed by --help. If the number is negative then the bits specified
611 by the number are cleared instead of being set. If the description
612 string is present but empty, then no help information will be displayed
613 for that option, but it will not count as an undocumented option. The
614 actual option name is made by appending @samp{-m} to the specified name.
616 One of the subgroupings should have a null string. The number in
617 this grouping is the default value for @code{target_flags}. Any
618 target options act starting with that value.
620 Here is an example which defines @samp{-m68000} and @samp{-m68020}
621 with opposite meanings, and picks the latter as the default:
624 #define TARGET_SWITCHES \
625 @{ @{ "68020", TARGET_MASK_68020, "" @}, \
626 @{ "68000", -TARGET_MASK_68020, "Compile for the 68000" @}, \
627 @{ "", TARGET_MASK_68020, "" @}@}
630 @findex TARGET_OPTIONS
632 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
633 options that have values. Its definition is an initializer with a
634 subgrouping for each command option.
636 Each subgrouping contains a string constant, that defines the fixed part
637 of the option name, the address of a variable, and a description string.
638 The variable, type @code{char *}, is set to the variable part of the
639 given option if the fixed part matches. The actual option name is made
640 by appending @samp{-m} to the specified name.
642 Here is an example which defines @samp{-mshort-data-@var{number}}. If the
643 given option is @samp{-mshort-data-512}, the variable @code{m88k_short_data}
644 will be set to the string @code{"512"}.
647 extern char *m88k_short_data;
648 #define TARGET_OPTIONS \
649 @{ @{ "short-data-", &m88k_short_data, "Specify the size of the short data section" @} @}
652 @findex TARGET_VERSION
654 This macro is a C statement to print on @code{stderr} a string
655 describing the particular machine description choice. Every machine
656 description should define @code{TARGET_VERSION}. For example:
660 #define TARGET_VERSION \
661 fprintf (stderr, " (68k, Motorola syntax)");
663 #define TARGET_VERSION \
664 fprintf (stderr, " (68k, MIT syntax)");
668 @findex OVERRIDE_OPTIONS
669 @item OVERRIDE_OPTIONS
670 Sometimes certain combinations of command options do not make sense on
671 a particular target machine. You can define a macro
672 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
673 defined, is executed once just after all the command options have been
676 Don't use this macro to turn on various extra optimizations for
677 @samp{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
679 @findex OPTIMIZATION_OPTIONS
680 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
681 Some machines may desire to change what optimizations are performed for
682 various optimization levels. This macro, if defined, is executed once
683 just after the optimization level is determined and before the remainder
684 of the command options have been parsed. Values set in this macro are
685 used as the default values for the other command line options.
687 @var{level} is the optimization level specified; 2 if @samp{-O2} is
688 specified, 1 if @samp{-O} is specified, and 0 if neither is specified.
690 @var{size} is non-zero if @samp{-Os} is specified and zero otherwise.
692 You should not use this macro to change options that are not
693 machine-specific. These should uniformly selected by the same
694 optimization level on all supported machines. Use this macro to enable
695 machine-specific optimizations.
697 @strong{Do not examine @code{write_symbols} in
698 this macro!} The debugging options are not supposed to alter the
701 @findex CAN_DEBUG_WITHOUT_FP
702 @item CAN_DEBUG_WITHOUT_FP
703 Define this macro if debugging can be performed even without a frame
704 pointer. If this macro is defined, GCC will turn on the
705 @samp{-fomit-frame-pointer} option whenever @samp{-O} is specified.
709 @section Storage Layout
710 @cindex storage layout
712 Note that the definitions of the macros in this table which are sizes or
713 alignments measured in bits do not need to be constant. They can be C
714 expressions that refer to static variables, such as the @code{target_flags}.
715 @xref{Run-time Target}.
718 @findex BITS_BIG_ENDIAN
719 @item BITS_BIG_ENDIAN
720 Define this macro to have the value 1 if the most significant bit in a
721 byte has the lowest number; otherwise define it to have the value zero.
722 This means that bit-field instructions count from the most significant
723 bit. If the machine has no bit-field instructions, then this must still
724 be defined, but it doesn't matter which value it is defined to. This
725 macro need not be a constant.
727 This macro does not affect the way structure fields are packed into
728 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
730 @findex BYTES_BIG_ENDIAN
731 @item BYTES_BIG_ENDIAN
732 Define this macro to have the value 1 if the most significant byte in a
733 word has the lowest number. This macro need not be a constant.
735 @findex WORDS_BIG_ENDIAN
736 @item WORDS_BIG_ENDIAN
737 Define this macro to have the value 1 if, in a multiword object, the
738 most significant word has the lowest number. This applies to both
739 memory locations and registers; GCC fundamentally assumes that the
740 order of words in memory is the same as the order in registers. This
741 macro need not be a constant.
743 @findex LIBGCC2_WORDS_BIG_ENDIAN
744 @item LIBGCC2_WORDS_BIG_ENDIAN
745 Define this macro if WORDS_BIG_ENDIAN is not constant. This must be a
746 constant value with the same meaning as WORDS_BIG_ENDIAN, which will be
747 used only when compiling libgcc2.c. Typically the value will be set
748 based on preprocessor defines.
750 @findex FLOAT_WORDS_BIG_ENDIAN
751 @item FLOAT_WORDS_BIG_ENDIAN
752 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
753 @code{TFmode} floating point numbers are stored in memory with the word
754 containing the sign bit at the lowest address; otherwise define it to
755 have the value 0. This macro need not be a constant.
757 You need not define this macro if the ordering is the same as for
760 @findex BITS_PER_UNIT
762 Define this macro to be the number of bits in an addressable storage
763 unit (byte); normally 8.
765 @findex BITS_PER_WORD
767 Number of bits in a word; normally 32.
769 @findex MAX_BITS_PER_WORD
770 @item MAX_BITS_PER_WORD
771 Maximum number of bits in a word. If this is undefined, the default is
772 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
773 largest value that @code{BITS_PER_WORD} can have at run-time.
775 @findex UNITS_PER_WORD
777 Number of storage units in a word; normally 4.
779 @findex MIN_UNITS_PER_WORD
780 @item MIN_UNITS_PER_WORD
781 Minimum number of units in a word. If this is undefined, the default is
782 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
783 smallest value that @code{UNITS_PER_WORD} can have at run-time.
787 Width of a pointer, in bits. You must specify a value no wider than the
788 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
789 you must define @code{POINTERS_EXTEND_UNSIGNED}.
791 @findex POINTERS_EXTEND_UNSIGNED
792 @item POINTERS_EXTEND_UNSIGNED
793 A C expression whose value is nonzero if pointers that need to be
794 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
795 be zero-extended and zero if they are to be sign-extended.
797 You need not define this macro if the @code{POINTER_SIZE} is equal
798 to the width of @code{Pmode}.
801 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
802 A macro to update @var{m} and @var{unsignedp} when an object whose type
803 is @var{type} and which has the specified mode and signedness is to be
804 stored in a register. This macro is only called when @var{type} is a
807 On most RISC machines, which only have operations that operate on a full
808 register, define this macro to set @var{m} to @code{word_mode} if
809 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
810 cases, only integer modes should be widened because wider-precision
811 floating-point operations are usually more expensive than their narrower
814 For most machines, the macro definition does not change @var{unsignedp}.
815 However, some machines, have instructions that preferentially handle
816 either signed or unsigned quantities of certain modes. For example, on
817 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
818 sign-extend the result to 64 bits. On such machines, set
819 @var{unsignedp} according to which kind of extension is more efficient.
821 Do not define this macro if it would never modify @var{m}.
823 @findex PROMOTE_FUNCTION_ARGS
824 @item PROMOTE_FUNCTION_ARGS
825 Define this macro if the promotion described by @code{PROMOTE_MODE}
826 should also be done for outgoing function arguments.
828 @findex PROMOTE_FUNCTION_RETURN
829 @item PROMOTE_FUNCTION_RETURN
830 Define this macro if the promotion described by @code{PROMOTE_MODE}
831 should also be done for the return value of functions.
833 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
834 promotions done by @code{PROMOTE_MODE}.
836 @findex PROMOTE_FOR_CALL_ONLY
837 @item PROMOTE_FOR_CALL_ONLY
838 Define this macro if the promotion described by @code{PROMOTE_MODE}
839 should @emph{only} be performed for outgoing function arguments or
840 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
841 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
843 @findex PARM_BOUNDARY
845 Normal alignment required for function parameters on the stack, in
846 bits. All stack parameters receive at least this much alignment
847 regardless of data type. On most machines, this is the same as the
850 @findex STACK_BOUNDARY
852 Define this macro if there is a guaranteed alignment for the stack
853 pointer on this machine. The definition is a C expression
854 for the desired alignment (measured in bits). This value is used as a
855 default if PREFERRED_STACK_BOUNDARY is not defined.
857 @findex PREFERRED_STACK_BOUNDARY
858 @item PREFERRED_STACK_BOUNDARY
859 Define this macro if you wish to preserve a certain alignment for
860 the stack pointer. The definition is a C expression
861 for the desired alignment (measured in bits). If STACK_BOUNDARY is
862 also defined, this macro must evaluate to a value equal to or larger
865 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
866 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
867 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
868 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
869 be momentarily unaligned while pushing arguments.
871 @findex FUNCTION_BOUNDARY
872 @item FUNCTION_BOUNDARY
873 Alignment required for a function entry point, in bits.
875 @findex BIGGEST_ALIGNMENT
876 @item BIGGEST_ALIGNMENT
877 Biggest alignment that any data type can require on this machine, in bits.
879 @findex MINIMUM_ATOMIC_ALIGNMENT
880 @item MINIMUM_ATOMIC_ALIGNMENT
881 If defined, the smallest alignment, in bits, that can be given to an
882 object that can be referenced in one operation, without disturbing any
883 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
884 on machines that don't have byte or half-word store operations.
886 @findex BIGGEST_FIELD_ALIGNMENT
887 @item BIGGEST_FIELD_ALIGNMENT
888 Biggest alignment that any structure or union field can require on this
889 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
890 structure and union fields only, unless the field alignment has been set
891 by the @code{__attribute__ ((aligned (@var{n})))} construct.
893 @findex ADJUST_FIELD_ALIGN
894 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
895 An expression for the alignment of a structure field @var{field} if the
896 alignment computed in the usual way is @var{computed}. GCC uses
897 this value instead of the value in @code{BIGGEST_ALIGNMENT} or
898 @code{BIGGEST_FIELD_ALIGNMENT}, if defined, for structure fields only.
900 @findex MAX_OFILE_ALIGNMENT
901 @item MAX_OFILE_ALIGNMENT
902 Biggest alignment supported by the object file format of this machine.
903 Use this macro to limit the alignment which can be specified using the
904 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
905 the default value is @code{BIGGEST_ALIGNMENT}.
907 @findex DATA_ALIGNMENT
908 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
909 If defined, a C expression to compute the alignment for a variables in
910 the static store. @var{type} is the data type, and @var{basic-align} is
911 the alignment that the object would ordinarily have. The value of this
912 macro is used instead of that alignment to align the object.
914 If this macro is not defined, then @var{basic-align} is used.
917 One use of this macro is to increase alignment of medium-size data to
918 make it all fit in fewer cache lines. Another is to cause character
919 arrays to be word-aligned so that @code{strcpy} calls that copy
920 constants to character arrays can be done inline.
922 @findex CONSTANT_ALIGNMENT
923 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
924 If defined, a C expression to compute the alignment given to a constant
925 that is being placed in memory. @var{constant} is the constant and
926 @var{basic-align} is the alignment that the object would ordinarily
927 have. The value of this macro is used instead of that alignment to
930 If this macro is not defined, then @var{basic-align} is used.
932 The typical use of this macro is to increase alignment for string
933 constants to be word aligned so that @code{strcpy} calls that copy
934 constants can be done inline.
936 @findex LOCAL_ALIGNMENT
937 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
938 If defined, a C expression to compute the alignment for a variables in
939 the local store. @var{type} is the data type, and @var{basic-align} is
940 the alignment that the object would ordinarily have. The value of this
941 macro is used instead of that alignment to align the object.
943 If this macro is not defined, then @var{basic-align} is used.
945 One use of this macro is to increase alignment of medium-size data to
946 make it all fit in fewer cache lines.
948 @findex EMPTY_FIELD_BOUNDARY
949 @item EMPTY_FIELD_BOUNDARY
950 Alignment in bits to be given to a structure bit field that follows an
951 empty field such as @code{int : 0;}.
953 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
954 that results from an empty field.
956 @findex STRUCTURE_SIZE_BOUNDARY
957 @item STRUCTURE_SIZE_BOUNDARY
958 Number of bits which any structure or union's size must be a multiple of.
959 Each structure or union's size is rounded up to a multiple of this.
961 If you do not define this macro, the default is the same as
962 @code{BITS_PER_UNIT}.
964 @findex STRICT_ALIGNMENT
965 @item STRICT_ALIGNMENT
966 Define this macro to be the value 1 if instructions will fail to work
967 if given data not on the nominal alignment. If instructions will merely
968 go slower in that case, define this macro as 0.
970 @findex PCC_BITFIELD_TYPE_MATTERS
971 @item PCC_BITFIELD_TYPE_MATTERS
972 Define this if you wish to imitate the way many other C compilers handle
973 alignment of bitfields and the structures that contain them.
975 The behavior is that the type written for a bitfield (@code{int},
976 @code{short}, or other integer type) imposes an alignment for the
977 entire structure, as if the structure really did contain an ordinary
978 field of that type. In addition, the bitfield is placed within the
979 structure so that it would fit within such a field, not crossing a
982 Thus, on most machines, a bitfield whose type is written as @code{int}
983 would not cross a four-byte boundary, and would force four-byte
984 alignment for the whole structure. (The alignment used may not be four
985 bytes; it is controlled by the other alignment parameters.)
987 If the macro is defined, its definition should be a C expression;
988 a nonzero value for the expression enables this behavior.
990 Note that if this macro is not defined, or its value is zero, some
991 bitfields may cross more than one alignment boundary. The compiler can
992 support such references if there are @samp{insv}, @samp{extv}, and
993 @samp{extzv} insns that can directly reference memory.
995 The other known way of making bitfields work is to define
996 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
997 Then every structure can be accessed with fullwords.
999 Unless the machine has bitfield instructions or you define
1000 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1001 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1003 If your aim is to make GCC use the same conventions for laying out
1004 bitfields as are used by another compiler, here is how to investigate
1005 what the other compiler does. Compile and run this program:
1024 printf ("Size of foo1 is %d\n",
1025 sizeof (struct foo1));
1026 printf ("Size of foo2 is %d\n",
1027 sizeof (struct foo2));
1032 If this prints 2 and 5, then the compiler's behavior is what you would
1033 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1035 @findex BITFIELD_NBYTES_LIMITED
1036 @item BITFIELD_NBYTES_LIMITED
1037 Like PCC_BITFIELD_TYPE_MATTERS except that its effect is limited to
1038 aligning a bitfield within the structure.
1040 @findex STRUCT_FORCE_BLK
1041 @item STRUCT_FORCE_BLK (@var{field})
1042 Return 1 if a structure containing @var{field} should be accessed using
1045 Normally, this is not needed. See the file @file{c4x.h} for an example
1046 of how to use this macro to prevent a structure having a floating point
1047 field from being accessed in an integer mode.
1049 @findex ROUND_TYPE_SIZE
1050 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
1051 Define this macro as an expression for the overall size of a type
1052 (given by @var{type} as a tree node) when the size computed in the
1053 usual way is @var{computed} and the alignment is @var{specified}.
1055 The default is to round @var{computed} up to a multiple of @var{specified}.
1057 @findex ROUND_TYPE_SIZE_UNIT
1058 @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified})
1059 Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are
1060 specified in units (bytes). If you define @code{ROUND_TYPE_SIZE},
1061 you must also define this macro and they must be defined consistently
1064 @findex ROUND_TYPE_ALIGN
1065 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1066 Define this macro as an expression for the alignment of a type (given
1067 by @var{type} as a tree node) if the alignment computed in the usual
1068 way is @var{computed} and the alignment explicitly specified was
1071 The default is to use @var{specified} if it is larger; otherwise, use
1072 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1074 @findex MAX_FIXED_MODE_SIZE
1075 @item MAX_FIXED_MODE_SIZE
1076 An integer expression for the size in bits of the largest integer
1077 machine mode that should actually be used. All integer machine modes of
1078 this size or smaller can be used for structures and unions with the
1079 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1080 (DImode)} is assumed.
1082 @findex VECTOR_MODE_SUPPORTED_P
1083 @item VECTOR_MODE_SUPPORTED_P(@var{mode})
1084 Define this macro to be nonzero if the port is prepared to handle insns
1085 involving vector mode @var{mode}. At the very least, it must have move
1086 patterns for this mode.
1088 @findex STACK_SAVEAREA_MODE
1089 @item STACK_SAVEAREA_MODE (@var{save_level})
1090 If defined, an expression of type @code{enum machine_mode} that
1091 specifies the mode of the save area operand of a
1092 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1093 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1094 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1095 having its mode specified.
1097 You need not define this macro if it always returns @code{Pmode}. You
1098 would most commonly define this macro if the
1099 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1102 @findex STACK_SIZE_MODE
1103 @item STACK_SIZE_MODE
1104 If defined, an expression of type @code{enum machine_mode} that
1105 specifies the mode of the size increment operand of an
1106 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1108 You need not define this macro if it always returns @code{word_mode}.
1109 You would most commonly define this macro if the @code{allocate_stack}
1110 pattern needs to support both a 32- and a 64-bit mode.
1112 @findex CHECK_FLOAT_VALUE
1113 @item CHECK_FLOAT_VALUE (@var{mode}, @var{value}, @var{overflow})
1114 A C statement to validate the value @var{value} (of type
1115 @code{double}) for mode @var{mode}. This means that you check whether
1116 @var{value} fits within the possible range of values for mode
1117 @var{mode} on this target machine. The mode @var{mode} is always
1118 a mode of class @code{MODE_FLOAT}. @var{overflow} is nonzero if
1119 the value is already known to be out of range.
1121 If @var{value} is not valid or if @var{overflow} is nonzero, you should
1122 set @var{overflow} to 1 and then assign some valid value to @var{value}.
1123 Allowing an invalid value to go through the compiler can produce
1124 incorrect assembler code which may even cause Unix assemblers to crash.
1126 This macro need not be defined if there is no work for it to do.
1128 @findex TARGET_FLOAT_FORMAT
1129 @item TARGET_FLOAT_FORMAT
1130 A code distinguishing the floating point format of the target machine.
1131 There are three defined values:
1134 @findex IEEE_FLOAT_FORMAT
1135 @item IEEE_FLOAT_FORMAT
1136 This code indicates IEEE floating point. It is the default; there is no
1137 need to define this macro when the format is IEEE.
1139 @findex VAX_FLOAT_FORMAT
1140 @item VAX_FLOAT_FORMAT
1141 This code indicates the peculiar format used on the Vax.
1143 @findex UNKNOWN_FLOAT_FORMAT
1144 @item UNKNOWN_FLOAT_FORMAT
1145 This code indicates any other format.
1148 The value of this macro is compared with @code{HOST_FLOAT_FORMAT}
1149 (@pxref{Config}) to determine whether the target machine has the same
1150 format as the host machine. If any other formats are actually in use on
1151 supported machines, new codes should be defined for them.
1153 The ordering of the component words of floating point values stored in
1154 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN} for the target
1155 machine and @code{HOST_FLOAT_WORDS_BIG_ENDIAN} for the host.
1157 @findex DEFAULT_VTABLE_THUNKS
1158 @item DEFAULT_VTABLE_THUNKS
1159 GCC supports two ways of implementing C++ vtables: traditional or with
1160 so-called ``thunks''. The flag @samp{-fvtable-thunk} chooses between them.
1161 Define this macro to be a C expression for the default value of that flag.
1162 If @code{DEFAULT_VTABLE_THUNKS} is 0, GCC uses the traditional
1163 implementation by default. The ``thunk'' implementation is more efficient
1164 (especially if you have provided an implementation of
1165 @code{ASM_OUTPUT_MI_THUNK}, see @ref{Function Entry}), but is not binary
1166 compatible with code compiled using the traditional implementation.
1167 If you are writing a new port, define @code{DEFAULT_VTABLE_THUNKS} to 1.
1169 If you do not define this macro, the default for @samp{-fvtable-thunk} is 0.
1173 @section Layout of Source Language Data Types
1175 These macros define the sizes and other characteristics of the standard
1176 basic data types used in programs being compiled. Unlike the macros in
1177 the previous section, these apply to specific features of C and related
1178 languages, rather than to fundamental aspects of storage layout.
1181 @findex INT_TYPE_SIZE
1183 A C expression for the size in bits of the type @code{int} on the
1184 target machine. If you don't define this, the default is one word.
1186 @findex MAX_INT_TYPE_SIZE
1187 @item MAX_INT_TYPE_SIZE
1188 Maximum number for the size in bits of the type @code{int} on the target
1189 machine. If this is undefined, the default is @code{INT_TYPE_SIZE}.
1190 Otherwise, it is the constant value that is the largest value that
1191 @code{INT_TYPE_SIZE} can have at run-time. This is used in @code{cpp}.
1193 @findex SHORT_TYPE_SIZE
1194 @item SHORT_TYPE_SIZE
1195 A C expression for the size in bits of the type @code{short} on the
1196 target machine. If you don't define this, the default is half a word.
1197 (If this would be less than one storage unit, it is rounded up to one
1200 @findex LONG_TYPE_SIZE
1201 @item LONG_TYPE_SIZE
1202 A C expression for the size in bits of the type @code{long} on the
1203 target machine. If you don't define this, the default is one word.
1205 @findex MAX_LONG_TYPE_SIZE
1206 @item MAX_LONG_TYPE_SIZE
1207 Maximum number for the size in bits of the type @code{long} on the
1208 target machine. If this is undefined, the default is
1209 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1210 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1213 @findex LONG_LONG_TYPE_SIZE
1214 @item LONG_LONG_TYPE_SIZE
1215 A C expression for the size in bits of the type @code{long long} on the
1216 target machine. If you don't define this, the default is two
1217 words. If you want to support GNU Ada on your machine, the value of this
1218 macro must be at least 64.
1220 @findex CHAR_TYPE_SIZE
1221 @item CHAR_TYPE_SIZE
1222 A C expression for the size in bits of the type @code{char} on the
1223 target machine. If you don't define this, the default is
1224 @code{BITS_PER_UNIT}.
1226 @findex MAX_CHAR_TYPE_SIZE
1227 @item MAX_CHAR_TYPE_SIZE
1228 Maximum number for the size in bits of the type @code{char} on the
1229 target machine. If this is undefined, the default is
1230 @code{CHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1231 largest value that @code{CHAR_TYPE_SIZE} can have at run-time. This is
1234 @findex FLOAT_TYPE_SIZE
1235 @item FLOAT_TYPE_SIZE
1236 A C expression for the size in bits of the type @code{float} on the
1237 target machine. If you don't define this, the default is one word.
1239 @findex DOUBLE_TYPE_SIZE
1240 @item DOUBLE_TYPE_SIZE
1241 A C expression for the size in bits of the type @code{double} on the
1242 target machine. If you don't define this, the default is two
1245 @findex LONG_DOUBLE_TYPE_SIZE
1246 @item LONG_DOUBLE_TYPE_SIZE
1247 A C expression for the size in bits of the type @code{long double} on
1248 the target machine. If you don't define this, the default is two
1251 @findex WIDEST_HARDWARE_FP_SIZE
1252 @item WIDEST_HARDWARE_FP_SIZE
1253 A C expression for the size in bits of the widest floating-point format
1254 supported by the hardware. If you define this macro, you must specify a
1255 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1256 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1259 @findex DEFAULT_SIGNED_CHAR
1260 @item DEFAULT_SIGNED_CHAR
1261 An expression whose value is 1 or 0, according to whether the type
1262 @code{char} should be signed or unsigned by default. The user can
1263 always override this default with the options @samp{-fsigned-char}
1264 and @samp{-funsigned-char}.
1266 @findex DEFAULT_SHORT_ENUMS
1267 @item DEFAULT_SHORT_ENUMS
1268 A C expression to determine whether to give an @code{enum} type
1269 only as many bytes as it takes to represent the range of possible values
1270 of that type. A nonzero value means to do that; a zero value means all
1271 @code{enum} types should be allocated like @code{int}.
1273 If you don't define the macro, the default is 0.
1277 A C expression for a string describing the name of the data type to use
1278 for size values. The typedef name @code{size_t} is defined using the
1279 contents of the string.
1281 The string can contain more than one keyword. If so, separate them with
1282 spaces, and write first any length keyword, then @code{unsigned} if
1283 appropriate, and finally @code{int}. The string must exactly match one
1284 of the data type names defined in the function
1285 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1286 omit @code{int} or change the order---that would cause the compiler to
1289 If you don't define this macro, the default is @code{"long unsigned
1292 @findex PTRDIFF_TYPE
1294 A C expression for a string describing the name of the data type to use
1295 for the result of subtracting two pointers. The typedef name
1296 @code{ptrdiff_t} is defined using the contents of the string. See
1297 @code{SIZE_TYPE} above for more information.
1299 If you don't define this macro, the default is @code{"long int"}.
1303 A C expression for a string describing the name of the data type to use
1304 for wide characters. The typedef name @code{wchar_t} is defined using
1305 the contents of the string. See @code{SIZE_TYPE} above for more
1308 If you don't define this macro, the default is @code{"int"}.
1310 @findex WCHAR_TYPE_SIZE
1311 @item WCHAR_TYPE_SIZE
1312 A C expression for the size in bits of the data type for wide
1313 characters. This is used in @code{cpp}, which cannot make use of
1316 @findex MAX_WCHAR_TYPE_SIZE
1317 @item MAX_WCHAR_TYPE_SIZE
1318 Maximum number for the size in bits of the data type for wide
1319 characters. If this is undefined, the default is
1320 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1321 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1326 A C expression for a string describing the name of the data type to
1327 use for wide characters passed to @code{printf} and returned from
1328 @code{getwc}. The typedef name @code{wint_t} is defined using the
1329 contents of the string. See @code{SIZE_TYPE} above for more
1332 If you don't define this macro, the default is @code{"unsigned int"}.
1336 A C expression for a string describing the name of the data type that
1337 can represent any value of any standard or extended signed integer type.
1338 The typedef name @code{intmax_t} is defined using the contents of the
1339 string. See @code{SIZE_TYPE} above for more information.
1341 If you don't define this macro, the default is the first of
1342 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1343 much precision as @code{long long int}.
1345 @findex UINTMAX_TYPE
1347 A C expression for a string describing the name of the data type that
1348 can represent any value of any standard or extended unsigned integer
1349 type. The typedef name @code{uintmax_t} is defined using the contents
1350 of the string. See @code{SIZE_TYPE} above for more information.
1352 If you don't define this macro, the default is the first of
1353 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1354 unsigned int"} that has as much precision as @code{long long unsigned
1357 @findex OBJC_INT_SELECTORS
1358 @item OBJC_INT_SELECTORS
1359 Define this macro if the type of Objective C selectors should be
1362 If this macro is not defined, then selectors should have the type
1363 @code{struct objc_selector *}.
1365 @findex OBJC_SELECTORS_WITHOUT_LABELS
1366 @item OBJC_SELECTORS_WITHOUT_LABELS
1367 Define this macro if the compiler can group all the selectors together
1368 into a vector and use just one label at the beginning of the vector.
1369 Otherwise, the compiler must give each selector its own assembler
1372 On certain machines, it is important to have a separate label for each
1373 selector because this enables the linker to eliminate duplicate selectors.
1377 A C constant expression for the integer value for escape sequence
1382 @findex TARGET_NEWLINE
1385 @itemx TARGET_NEWLINE
1386 C constant expressions for the integer values for escape sequences
1387 @samp{\b}, @samp{\t} and @samp{\n}.
1395 C constant expressions for the integer values for escape sequences
1396 @samp{\v}, @samp{\f} and @samp{\r}.
1400 @section Register Usage
1401 @cindex register usage
1403 This section explains how to describe what registers the target machine
1404 has, and how (in general) they can be used.
1406 The description of which registers a specific instruction can use is
1407 done with register classes; see @ref{Register Classes}. For information
1408 on using registers to access a stack frame, see @ref{Frame Registers}.
1409 For passing values in registers, see @ref{Register Arguments}.
1410 For returning values in registers, see @ref{Scalar Return}.
1413 * Register Basics:: Number and kinds of registers.
1414 * Allocation Order:: Order in which registers are allocated.
1415 * Values in Registers:: What kinds of values each reg can hold.
1416 * Leaf Functions:: Renumbering registers for leaf functions.
1417 * Stack Registers:: Handling a register stack such as 80387.
1420 @node Register Basics
1421 @subsection Basic Characteristics of Registers
1423 @c prevent bad page break with this line
1424 Registers have various characteristics.
1427 @findex FIRST_PSEUDO_REGISTER
1428 @item FIRST_PSEUDO_REGISTER
1429 Number of hardware registers known to the compiler. They receive
1430 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1431 pseudo register's number really is assigned the number
1432 @code{FIRST_PSEUDO_REGISTER}.
1434 @item FIXED_REGISTERS
1435 @findex FIXED_REGISTERS
1436 @cindex fixed register
1437 An initializer that says which registers are used for fixed purposes
1438 all throughout the compiled code and are therefore not available for
1439 general allocation. These would include the stack pointer, the frame
1440 pointer (except on machines where that can be used as a general
1441 register when no frame pointer is needed), the program counter on
1442 machines where that is considered one of the addressable registers,
1443 and any other numbered register with a standard use.
1445 This information is expressed as a sequence of numbers, separated by
1446 commas and surrounded by braces. The @var{n}th number is 1 if
1447 register @var{n} is fixed, 0 otherwise.
1449 The table initialized from this macro, and the table initialized by
1450 the following one, may be overridden at run time either automatically,
1451 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1452 the user with the command options @samp{-ffixed-@var{reg}},
1453 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}}.
1455 @findex CALL_USED_REGISTERS
1456 @item CALL_USED_REGISTERS
1457 @cindex call-used register
1458 @cindex call-clobbered register
1459 @cindex call-saved register
1460 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1461 clobbered (in general) by function calls as well as for fixed
1462 registers. This macro therefore identifies the registers that are not
1463 available for general allocation of values that must live across
1466 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1467 automatically saves it on function entry and restores it on function
1468 exit, if the register is used within the function.
1470 @findex HARD_REGNO_CALL_PART_CLOBBERED
1471 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1472 @cindex call-used register
1473 @cindex call-clobbered register
1474 @cindex call-saved register
1475 A C expression that is non-zero if it is not permissible to store a
1476 value of mode @var{mode} in hard register number @var{regno} across a
1477 call without some part of it being clobbered. For most machines this
1478 macro need not be defined. It is only required for machines that do not
1479 preserve the entire contents of a register across a call.
1481 @findex CONDITIONAL_REGISTER_USAGE
1483 @findex call_used_regs
1484 @item CONDITIONAL_REGISTER_USAGE
1485 Zero or more C statements that may conditionally modify five variables
1486 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1487 (these three are of type @code{char []}), @code{reg_names} (of type
1488 @code{const char * []}) and @code{reg_class_contents} (of type
1489 @code{HARD_REG_SET}).
1490 Before the macro is called @code{fixed_regs}, @code{call_used_regs}
1491 @code{reg_class_contents} and @code{reg_names} have been initialized
1492 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1493 @code{REG_CLASS_CONTENTS} and @code{REGISTER_NAMES}, respectively,
1494 @code{global_regs} has been cleared, and any @samp{-ffixed-@var{reg}},
1495 @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}} command
1496 options have been applied.
1498 This is necessary in case the fixed or call-clobbered registers depend
1501 You need not define this macro if it has no work to do.
1503 @cindex disabling certain registers
1504 @cindex controlling register usage
1505 If the usage of an entire class of registers depends on the target
1506 flags, you may indicate this to GCC by using this macro to modify
1507 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1508 registers in the classes which should not be used by GCC. Also define
1509 the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1510 is called with a letter for a class that shouldn't be used.
1512 (However, if this class is not included in @code{GENERAL_REGS} and all
1513 of the insn patterns whose constraints permit this class are
1514 controlled by target switches, then GCC will automatically avoid using
1515 these registers when the target switches are opposed to them.)
1517 @findex NON_SAVING_SETJMP
1518 @item NON_SAVING_SETJMP
1519 If this macro is defined and has a nonzero value, it means that
1520 @code{setjmp} and related functions fail to save the registers, or that
1521 @code{longjmp} fails to restore them. To compensate, the compiler
1522 avoids putting variables in registers in functions that use
1525 @findex INCOMING_REGNO
1526 @item INCOMING_REGNO (@var{out})
1527 Define this macro if the target machine has register windows. This C
1528 expression returns the register number as seen by the called function
1529 corresponding to the register number @var{out} as seen by the calling
1530 function. Return @var{out} if register number @var{out} is not an
1533 @findex OUTGOING_REGNO
1534 @item OUTGOING_REGNO (@var{in})
1535 Define this macro if the target machine has register windows. This C
1536 expression returns the register number as seen by the calling function
1537 corresponding to the register number @var{in} as seen by the called
1538 function. Return @var{in} if register number @var{in} is not an inbound
1542 @item LOCAL_REGNO (@var{regno})
1543 Define this macro if the target machine has register windows. This C
1544 expression returns true if the register is call-saved but is in the
1545 register window. Unlike most call-saved registers, such registers
1546 need not be explicitly restored on function exit or during non-local
1552 If the program counter has a register number, define this as that
1553 register number. Otherwise, do not define it.
1557 @node Allocation Order
1558 @subsection Order of Allocation of Registers
1559 @cindex order of register allocation
1560 @cindex register allocation order
1562 @c prevent bad page break with this line
1563 Registers are allocated in order.
1566 @findex REG_ALLOC_ORDER
1567 @item REG_ALLOC_ORDER
1568 If defined, an initializer for a vector of integers, containing the
1569 numbers of hard registers in the order in which GCC should prefer
1570 to use them (from most preferred to least).
1572 If this macro is not defined, registers are used lowest numbered first
1573 (all else being equal).
1575 One use of this macro is on machines where the highest numbered
1576 registers must always be saved and the save-multiple-registers
1577 instruction supports only sequences of consecutive registers. On such
1578 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1579 the highest numbered allocable register first.
1581 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1582 @item ORDER_REGS_FOR_LOCAL_ALLOC
1583 A C statement (sans semicolon) to choose the order in which to allocate
1584 hard registers for pseudo-registers local to a basic block.
1586 Store the desired register order in the array @code{reg_alloc_order}.
1587 Element 0 should be the register to allocate first; element 1, the next
1588 register; and so on.
1590 The macro body should not assume anything about the contents of
1591 @code{reg_alloc_order} before execution of the macro.
1593 On most machines, it is not necessary to define this macro.
1596 @node Values in Registers
1597 @subsection How Values Fit in Registers
1599 This section discusses the macros that describe which kinds of values
1600 (specifically, which machine modes) each register can hold, and how many
1601 consecutive registers are needed for a given mode.
1604 @findex HARD_REGNO_NREGS
1605 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1606 A C expression for the number of consecutive hard registers, starting
1607 at register number @var{regno}, required to hold a value of mode
1610 On a machine where all registers are exactly one word, a suitable
1611 definition of this macro is
1614 #define HARD_REGNO_NREGS(REGNO, MODE) \
1615 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1619 @findex ALTER_HARD_SUBREG
1620 @item ALTER_HARD_SUBREG (@var{tgt_mode}, @var{word}, @var{src_mode}, @var{regno})
1621 A C expression that returns an adjusted hard register number for
1624 (subreg:@var{tgt_mode} (reg:@var{src_mode} @var{regno}) @var{word})
1627 This may be needed if the target machine has mixed sized big-endian
1628 registers, like Sparc v9.
1630 @findex HARD_REGNO_MODE_OK
1631 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1632 A C expression that is nonzero if it is permissible to store a value
1633 of mode @var{mode} in hard register number @var{regno} (or in several
1634 registers starting with that one). For a machine where all registers
1635 are equivalent, a suitable definition is
1638 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1641 You need not include code to check for the numbers of fixed registers,
1642 because the allocation mechanism considers them to be always occupied.
1644 @cindex register pairs
1645 On some machines, double-precision values must be kept in even/odd
1646 register pairs. You can implement that by defining this macro to reject
1647 odd register numbers for such modes.
1649 The minimum requirement for a mode to be OK in a register is that the
1650 @samp{mov@var{mode}} instruction pattern support moves between the
1651 register and other hard register in the same class and that moving a
1652 value into the register and back out not alter it.
1654 Since the same instruction used to move @code{word_mode} will work for
1655 all narrower integer modes, it is not necessary on any machine for
1656 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1657 you define patterns @samp{movhi}, etc., to take advantage of this. This
1658 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1659 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1662 Many machines have special registers for floating point arithmetic.
1663 Often people assume that floating point machine modes are allowed only
1664 in floating point registers. This is not true. Any registers that
1665 can hold integers can safely @emph{hold} a floating point machine
1666 mode, whether or not floating arithmetic can be done on it in those
1667 registers. Integer move instructions can be used to move the values.
1669 On some machines, though, the converse is true: fixed-point machine
1670 modes may not go in floating registers. This is true if the floating
1671 registers normalize any value stored in them, because storing a
1672 non-floating value there would garble it. In this case,
1673 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1674 floating registers. But if the floating registers do not automatically
1675 normalize, if you can store any bit pattern in one and retrieve it
1676 unchanged without a trap, then any machine mode may go in a floating
1677 register, so you can define this macro to say so.
1679 The primary significance of special floating registers is rather that
1680 they are the registers acceptable in floating point arithmetic
1681 instructions. However, this is of no concern to
1682 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1683 constraints for those instructions.
1685 On some machines, the floating registers are especially slow to access,
1686 so that it is better to store a value in a stack frame than in such a
1687 register if floating point arithmetic is not being done. As long as the
1688 floating registers are not in class @code{GENERAL_REGS}, they will not
1689 be used unless some pattern's constraint asks for one.
1691 @findex MODES_TIEABLE_P
1692 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
1693 A C expression that is nonzero if a value of mode
1694 @var{mode1} is accessible in mode @var{mode2} without copying.
1696 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
1697 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
1698 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
1699 should be nonzero. If they differ for any @var{r}, you should define
1700 this macro to return zero unless some other mechanism ensures the
1701 accessibility of the value in a narrower mode.
1703 You should define this macro to return nonzero in as many cases as
1704 possible since doing so will allow GCC to perform better register
1707 @findex AVOID_CCMODE_COPIES
1708 @item AVOID_CCMODE_COPIES
1709 Define this macro if the compiler should avoid copies to/from @code{CCmode}
1710 registers. You should only define this macro if support for copying to/from
1711 @code{CCmode} is incomplete.
1714 @node Leaf Functions
1715 @subsection Handling Leaf Functions
1717 @cindex leaf functions
1718 @cindex functions, leaf
1719 On some machines, a leaf function (i.e., one which makes no calls) can run
1720 more efficiently if it does not make its own register window. Often this
1721 means it is required to receive its arguments in the registers where they
1722 are passed by the caller, instead of the registers where they would
1725 The special treatment for leaf functions generally applies only when
1726 other conditions are met; for example, often they may use only those
1727 registers for its own variables and temporaries. We use the term ``leaf
1728 function'' to mean a function that is suitable for this special
1729 handling, so that functions with no calls are not necessarily ``leaf
1732 GCC assigns register numbers before it knows whether the function is
1733 suitable for leaf function treatment. So it needs to renumber the
1734 registers in order to output a leaf function. The following macros
1738 @findex LEAF_REGISTERS
1739 @item LEAF_REGISTERS
1740 Name of a char vector, indexed by hard register number, which
1741 contains 1 for a register that is allowable in a candidate for leaf
1744 If leaf function treatment involves renumbering the registers, then the
1745 registers marked here should be the ones before renumbering---those that
1746 GCC would ordinarily allocate. The registers which will actually be
1747 used in the assembler code, after renumbering, should not be marked with 1
1750 Define this macro only if the target machine offers a way to optimize
1751 the treatment of leaf functions.
1753 @findex LEAF_REG_REMAP
1754 @item LEAF_REG_REMAP (@var{regno})
1755 A C expression whose value is the register number to which @var{regno}
1756 should be renumbered, when a function is treated as a leaf function.
1758 If @var{regno} is a register number which should not appear in a leaf
1759 function before renumbering, then the expression should yield -1, which
1760 will cause the compiler to abort.
1762 Define this macro only if the target machine offers a way to optimize the
1763 treatment of leaf functions, and registers need to be renumbered to do
1767 @findex current_function_is_leaf
1768 @findex current_function_uses_only_leaf_regs
1769 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
1770 treat leaf functions specially. They can test the C variable
1771 @code{current_function_is_leaf} which is nonzero for leaf functions.
1772 @code{current_function_is_leaf} is set prior to local register allocation
1773 and is valid for the remaining compiler passes. They can also test the C
1774 variable @code{current_function_uses_only_leaf_regs} which is nonzero for
1775 leaf functions which only use leaf registers.
1776 @code{current_function_uses_only_leaf_regs} is valid after reload and is
1777 only useful if @code{LEAF_REGISTERS} is defined.
1778 @c changed this to fix overfull. ALSO: why the "it" at the beginning
1779 @c of the next paragraph?! --mew 2feb93
1781 @node Stack Registers
1782 @subsection Registers That Form a Stack
1784 There are special features to handle computers where some of the
1785 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
1786 Stack registers are normally written by pushing onto the stack, and are
1787 numbered relative to the top of the stack.
1789 Currently, GCC can only handle one group of stack-like registers, and
1790 they must be consecutively numbered.
1795 Define this if the machine has any stack-like registers.
1797 @findex FIRST_STACK_REG
1798 @item FIRST_STACK_REG
1799 The number of the first stack-like register. This one is the top
1802 @findex LAST_STACK_REG
1803 @item LAST_STACK_REG
1804 The number of the last stack-like register. This one is the bottom of
1808 @node Register Classes
1809 @section Register Classes
1810 @cindex register class definitions
1811 @cindex class definitions, register
1813 On many machines, the numbered registers are not all equivalent.
1814 For example, certain registers may not be allowed for indexed addressing;
1815 certain registers may not be allowed in some instructions. These machine
1816 restrictions are described to the compiler using @dfn{register classes}.
1818 You define a number of register classes, giving each one a name and saying
1819 which of the registers belong to it. Then you can specify register classes
1820 that are allowed as operands to particular instruction patterns.
1824 In general, each register will belong to several classes. In fact, one
1825 class must be named @code{ALL_REGS} and contain all the registers. Another
1826 class must be named @code{NO_REGS} and contain no registers. Often the
1827 union of two classes will be another class; however, this is not required.
1829 @findex GENERAL_REGS
1830 One of the classes must be named @code{GENERAL_REGS}. There is nothing
1831 terribly special about the name, but the operand constraint letters
1832 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
1833 the same as @code{ALL_REGS}, just define it as a macro which expands
1836 Order the classes so that if class @var{x} is contained in class @var{y}
1837 then @var{x} has a lower class number than @var{y}.
1839 The way classes other than @code{GENERAL_REGS} are specified in operand
1840 constraints is through machine-dependent operand constraint letters.
1841 You can define such letters to correspond to various classes, then use
1842 them in operand constraints.
1844 You should define a class for the union of two classes whenever some
1845 instruction allows both classes. For example, if an instruction allows
1846 either a floating point (coprocessor) register or a general register for a
1847 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
1848 which includes both of them. Otherwise you will get suboptimal code.
1850 You must also specify certain redundant information about the register
1851 classes: for each class, which classes contain it and which ones are
1852 contained in it; for each pair of classes, the largest class contained
1855 When a value occupying several consecutive registers is expected in a
1856 certain class, all the registers used must belong to that class.
1857 Therefore, register classes cannot be used to enforce a requirement for
1858 a register pair to start with an even-numbered register. The way to
1859 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
1861 Register classes used for input-operands of bitwise-and or shift
1862 instructions have a special requirement: each such class must have, for
1863 each fixed-point machine mode, a subclass whose registers can transfer that
1864 mode to or from memory. For example, on some machines, the operations for
1865 single-byte values (@code{QImode}) are limited to certain registers. When
1866 this is so, each register class that is used in a bitwise-and or shift
1867 instruction must have a subclass consisting of registers from which
1868 single-byte values can be loaded or stored. This is so that
1869 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
1872 @findex enum reg_class
1873 @item enum reg_class
1874 An enumeral type that must be defined with all the register class names
1875 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
1876 must be the last register class, followed by one more enumeral value,
1877 @code{LIM_REG_CLASSES}, which is not a register class but rather
1878 tells how many classes there are.
1880 Each register class has a number, which is the value of casting
1881 the class name to type @code{int}. The number serves as an index
1882 in many of the tables described below.
1884 @findex N_REG_CLASSES
1886 The number of distinct register classes, defined as follows:
1889 #define N_REG_CLASSES (int) LIM_REG_CLASSES
1892 @findex REG_CLASS_NAMES
1893 @item REG_CLASS_NAMES
1894 An initializer containing the names of the register classes as C string
1895 constants. These names are used in writing some of the debugging dumps.
1897 @findex REG_CLASS_CONTENTS
1898 @item REG_CLASS_CONTENTS
1899 An initializer containing the contents of the register classes, as integers
1900 which are bit masks. The @var{n}th integer specifies the contents of class
1901 @var{n}. The way the integer @var{mask} is interpreted is that
1902 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
1904 When the machine has more than 32 registers, an integer does not suffice.
1905 Then the integers are replaced by sub-initializers, braced groupings containing
1906 several integers. Each sub-initializer must be suitable as an initializer
1907 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
1908 In this situation, the first integer in each sub-initializer corresponds to
1909 registers 0 through 31, the second integer to registers 32 through 63, and
1912 @findex REGNO_REG_CLASS
1913 @item REGNO_REG_CLASS (@var{regno})
1914 A C expression whose value is a register class containing hard register
1915 @var{regno}. In general there is more than one such class; choose a class
1916 which is @dfn{minimal}, meaning that no smaller class also contains the
1919 @findex BASE_REG_CLASS
1920 @item BASE_REG_CLASS
1921 A macro whose definition is the name of the class to which a valid
1922 base register must belong. A base register is one used in an address
1923 which is the register value plus a displacement.
1925 @findex INDEX_REG_CLASS
1926 @item INDEX_REG_CLASS
1927 A macro whose definition is the name of the class to which a valid
1928 index register must belong. An index register is one used in an
1929 address where its value is either multiplied by a scale factor or
1930 added to another register (as well as added to a displacement).
1932 @findex REG_CLASS_FROM_LETTER
1933 @item REG_CLASS_FROM_LETTER (@var{char})
1934 A C expression which defines the machine-dependent operand constraint
1935 letters for register classes. If @var{char} is such a letter, the
1936 value should be the register class corresponding to it. Otherwise,
1937 the value should be @code{NO_REGS}. The register letter @samp{r},
1938 corresponding to class @code{GENERAL_REGS}, will not be passed
1939 to this macro; you do not need to handle it.
1941 @findex REGNO_OK_FOR_BASE_P
1942 @item REGNO_OK_FOR_BASE_P (@var{num})
1943 A C expression which is nonzero if register number @var{num} is
1944 suitable for use as a base register in operand addresses. It may be
1945 either a suitable hard register or a pseudo register that has been
1946 allocated such a hard register.
1948 @findex REGNO_MODE_OK_FOR_BASE_P
1949 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
1950 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
1951 that expression may examine the mode of the memory reference in
1952 @var{mode}. You should define this macro if the mode of the memory
1953 reference affects whether a register may be used as a base register. If
1954 you define this macro, the compiler will use it instead of
1955 @code{REGNO_OK_FOR_BASE_P}.
1957 @findex REGNO_OK_FOR_INDEX_P
1958 @item REGNO_OK_FOR_INDEX_P (@var{num})
1959 A C expression which is nonzero if register number @var{num} is
1960 suitable for use as an index register in operand addresses. It may be
1961 either a suitable hard register or a pseudo register that has been
1962 allocated such a hard register.
1964 The difference between an index register and a base register is that
1965 the index register may be scaled. If an address involves the sum of
1966 two registers, neither one of them scaled, then either one may be
1967 labeled the ``base'' and the other the ``index''; but whichever
1968 labeling is used must fit the machine's constraints of which registers
1969 may serve in each capacity. The compiler will try both labelings,
1970 looking for one that is valid, and will reload one or both registers
1971 only if neither labeling works.
1973 @findex PREFERRED_RELOAD_CLASS
1974 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
1975 A C expression that places additional restrictions on the register class
1976 to use when it is necessary to copy value @var{x} into a register in class
1977 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
1978 another, smaller class. On many machines, the following definition is
1982 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
1985 Sometimes returning a more restrictive class makes better code. For
1986 example, on the 68000, when @var{x} is an integer constant that is in range
1987 for a @samp{moveq} instruction, the value of this macro is always
1988 @code{DATA_REGS} as long as @var{class} includes the data registers.
1989 Requiring a data register guarantees that a @samp{moveq} will be used.
1991 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
1992 you can force @var{x} into a memory constant. This is useful on
1993 certain machines where immediate floating values cannot be loaded into
1994 certain kinds of registers.
1996 @findex PREFERRED_OUTPUT_RELOAD_CLASS
1997 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
1998 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
1999 input reloads. If you don't define this macro, the default is to use
2000 @var{class}, unchanged.
2002 @findex LIMIT_RELOAD_CLASS
2003 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2004 A C expression that places additional restrictions on the register class
2005 to use when it is necessary to be able to hold a value of mode
2006 @var{mode} in a reload register for which class @var{class} would
2009 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2010 there are certain modes that simply can't go in certain reload classes.
2012 The value is a register class; perhaps @var{class}, or perhaps another,
2015 Don't define this macro unless the target machine has limitations which
2016 require the macro to do something nontrivial.
2018 @findex SECONDARY_RELOAD_CLASS
2019 @findex SECONDARY_INPUT_RELOAD_CLASS
2020 @findex SECONDARY_OUTPUT_RELOAD_CLASS
2021 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2022 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2023 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2024 Many machines have some registers that cannot be copied directly to or
2025 from memory or even from other types of registers. An example is the
2026 @samp{MQ} register, which on most machines, can only be copied to or
2027 from general registers, but not memory. Some machines allow copying all
2028 registers to and from memory, but require a scratch register for stores
2029 to some memory locations (e.g., those with symbolic address on the RT,
2030 and those with certain symbolic address on the Sparc when compiling
2031 PIC). In some cases, both an intermediate and a scratch register are
2034 You should define these macros to indicate to the reload phase that it may
2035 need to allocate at least one register for a reload in addition to the
2036 register to contain the data. Specifically, if copying @var{x} to a
2037 register @var{class} in @var{mode} requires an intermediate register,
2038 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2039 largest register class all of whose registers can be used as
2040 intermediate registers or scratch registers.
2042 If copying a register @var{class} in @var{mode} to @var{x} requires an
2043 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2044 should be defined to return the largest register class required. If the
2045 requirements for input and output reloads are the same, the macro
2046 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2049 The values returned by these macros are often @code{GENERAL_REGS}.
2050 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2051 can be directly copied to or from a register of @var{class} in
2052 @var{mode} without requiring a scratch register. Do not define this
2053 macro if it would always return @code{NO_REGS}.
2055 If a scratch register is required (either with or without an
2056 intermediate register), you should define patterns for
2057 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2058 (@pxref{Standard Names}. These patterns, which will normally be
2059 implemented with a @code{define_expand}, should be similar to the
2060 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2063 Define constraints for the reload register and scratch register that
2064 contain a single register class. If the original reload register (whose
2065 class is @var{class}) can meet the constraint given in the pattern, the
2066 value returned by these macros is used for the class of the scratch
2067 register. Otherwise, two additional reload registers are required.
2068 Their classes are obtained from the constraints in the insn pattern.
2070 @var{x} might be a pseudo-register or a @code{subreg} of a
2071 pseudo-register, which could either be in a hard register or in memory.
2072 Use @code{true_regnum} to find out; it will return -1 if the pseudo is
2073 in memory and the hard register number if it is in a register.
2075 These macros should not be used in the case where a particular class of
2076 registers can only be copied to memory and not to another class of
2077 registers. In that case, secondary reload registers are not needed and
2078 would not be helpful. Instead, a stack location must be used to perform
2079 the copy and the @code{mov@var{m}} pattern should use memory as a
2080 intermediate storage. This case often occurs between floating-point and
2083 @findex SECONDARY_MEMORY_NEEDED
2084 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2085 Certain machines have the property that some registers cannot be copied
2086 to some other registers without using memory. Define this macro on
2087 those machines to be a C expression that is non-zero if objects of mode
2088 @var{m} in registers of @var{class1} can only be copied to registers of
2089 class @var{class2} by storing a register of @var{class1} into memory
2090 and loading that memory location into a register of @var{class2}.
2092 Do not define this macro if its value would always be zero.
2094 @findex SECONDARY_MEMORY_NEEDED_RTX
2095 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2096 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2097 allocates a stack slot for a memory location needed for register copies.
2098 If this macro is defined, the compiler instead uses the memory location
2099 defined by this macro.
2101 Do not define this macro if you do not define
2102 @code{SECONDARY_MEMORY_NEEDED}.
2104 @findex SECONDARY_MEMORY_NEEDED_MODE
2105 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2106 When the compiler needs a secondary memory location to copy between two
2107 registers of mode @var{mode}, it normally allocates sufficient memory to
2108 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2109 load operations in a mode that many bits wide and whose class is the
2110 same as that of @var{mode}.
2112 This is right thing to do on most machines because it ensures that all
2113 bits of the register are copied and prevents accesses to the registers
2114 in a narrower mode, which some machines prohibit for floating-point
2117 However, this default behavior is not correct on some machines, such as
2118 the DEC Alpha, that store short integers in floating-point registers
2119 differently than in integer registers. On those machines, the default
2120 widening will not work correctly and you must define this macro to
2121 suppress that widening in some cases. See the file @file{alpha.h} for
2124 Do not define this macro if you do not define
2125 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2126 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2128 @findex SMALL_REGISTER_CLASSES
2129 @item SMALL_REGISTER_CLASSES
2130 On some machines, it is risky to let hard registers live across arbitrary
2131 insns. Typically, these machines have instructions that require values
2132 to be in specific registers (like an accumulator), and reload will fail
2133 if the required hard register is used for another purpose across such an
2136 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a non-zero
2137 value on these machines. When this macro has a non-zero value, the
2138 compiler will try to minimize the lifetime of hard registers.
2140 It is always safe to define this macro with a non-zero value, but if you
2141 unnecessarily define it, you will reduce the amount of optimizations
2142 that can be performed in some cases. If you do not define this macro
2143 with a non-zero value when it is required, the compiler will run out of
2144 spill registers and print a fatal error message. For most machines, you
2145 should not define this macro at all.
2147 @findex CLASS_LIKELY_SPILLED_P
2148 @item CLASS_LIKELY_SPILLED_P (@var{class})
2149 A C expression whose value is nonzero if pseudos that have been assigned
2150 to registers of class @var{class} would likely be spilled because
2151 registers of @var{class} are needed for spill registers.
2153 The default value of this macro returns 1 if @var{class} has exactly one
2154 register and zero otherwise. On most machines, this default should be
2155 used. Only define this macro to some other expression if pseudos
2156 allocated by @file{local-alloc.c} end up in memory because their hard
2157 registers were needed for spill registers. If this macro returns nonzero
2158 for those classes, those pseudos will only be allocated by
2159 @file{global.c}, which knows how to reallocate the pseudo to another
2160 register. If there would not be another register available for
2161 reallocation, you should not change the definition of this macro since
2162 the only effect of such a definition would be to slow down register
2165 @findex CLASS_MAX_NREGS
2166 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2167 A C expression for the maximum number of consecutive registers
2168 of class @var{class} needed to hold a value of mode @var{mode}.
2170 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2171 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2172 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2173 @var{mode})} for all @var{regno} values in the class @var{class}.
2175 This macro helps control the handling of multiple-word values
2178 @item CLASS_CANNOT_CHANGE_MODE
2179 If defined, a C expression for a class that contains registers for
2180 which the compiler may not change modes arbitrarily.
2182 @item CLASS_CANNOT_CHANGE_MODE_P(@var{from}, @var{to})
2183 A C expression that is true if, for a register in
2184 @code{CLASS_CANNOT_CHANGE_MODE}, the requested mode punning is illegal.
2186 For the example, loading 32-bit integer or floating-point objects into
2187 floating-point registers on the Alpha extends them to 64-bits.
2188 Therefore loading a 64-bit object and then storing it as a 32-bit object
2189 does not store the low-order 32-bits, as would be the case for a normal
2190 register. Therefore, @file{alpha.h} defines @code{CLASS_CANNOT_CHANGE_MODE}
2191 as @code{FLOAT_REGS} and @code{CLASS_CANNOT_CHANGE_MODE_P} restricts
2192 mode changes to same-size modes.
2194 Compare this to IA-64, which extends floating-point values to 82-bits,
2195 and stores 64-bit integers in a different format than 64-bit doubles.
2196 Therefore @code{CLASS_CANNOT_CHANGE_MODE_P} is always true.
2199 Three other special macros describe which operands fit which constraint
2203 @findex CONST_OK_FOR_LETTER_P
2204 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2205 A C expression that defines the machine-dependent operand constraint
2206 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2207 particular ranges of integer values. If @var{c} is one of those
2208 letters, the expression should check that @var{value}, an integer, is in
2209 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2210 not one of those letters, the value should be 0 regardless of
2213 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2214 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2215 A C expression that defines the machine-dependent operand constraint
2216 letters that specify particular ranges of @code{const_double} values
2217 (@samp{G} or @samp{H}).
2219 If @var{c} is one of those letters, the expression should check that
2220 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2221 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2222 letters, the value should be 0 regardless of @var{value}.
2224 @code{const_double} is used for all floating-point constants and for
2225 @code{DImode} fixed-point constants. A given letter can accept either
2226 or both kinds of values. It can use @code{GET_MODE} to distinguish
2227 between these kinds.
2229 @findex EXTRA_CONSTRAINT
2230 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2231 A C expression that defines the optional machine-dependent constraint
2232 letters that can be used to segregate specific types of operands, usually
2233 memory references, for the target machine. Any letter that is not
2234 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER}
2235 may be used. Normally this macro will not be defined.
2237 If it is required for a particular target machine, it should return 1
2238 if @var{value} corresponds to the operand type represented by the
2239 constraint letter @var{c}. If @var{c} is not defined as an extra
2240 constraint, the value returned should be 0 regardless of @var{value}.
2242 For example, on the ROMP, load instructions cannot have their output
2243 in r0 if the memory reference contains a symbolic address. Constraint
2244 letter @samp{Q} is defined as representing a memory address that does
2245 @emph{not} contain a symbolic address. An alternative is specified with
2246 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2247 alternative specifies @samp{m} on the input and a register class that
2248 does not include r0 on the output.
2251 @node Stack and Calling
2252 @section Stack Layout and Calling Conventions
2253 @cindex calling conventions
2255 @c prevent bad page break with this line
2256 This describes the stack layout and calling conventions.
2264 * Register Arguments::
2266 * Aggregate Return::
2275 @subsection Basic Stack Layout
2276 @cindex stack frame layout
2277 @cindex frame layout
2279 @c prevent bad page break with this line
2280 Here is the basic stack layout.
2283 @findex STACK_GROWS_DOWNWARD
2284 @item STACK_GROWS_DOWNWARD
2285 Define this macro if pushing a word onto the stack moves the stack
2286 pointer to a smaller address.
2288 When we say, ``define this macro if @dots{},'' it means that the
2289 compiler checks this macro only with @code{#ifdef} so the precise
2290 definition used does not matter.
2292 @findex FRAME_GROWS_DOWNWARD
2293 @item FRAME_GROWS_DOWNWARD
2294 Define this macro if the addresses of local variable slots are at negative
2295 offsets from the frame pointer.
2297 @findex ARGS_GROW_DOWNWARD
2298 @item ARGS_GROW_DOWNWARD
2299 Define this macro if successive arguments to a function occupy decreasing
2300 addresses on the stack.
2302 @findex STARTING_FRAME_OFFSET
2303 @item STARTING_FRAME_OFFSET
2304 Offset from the frame pointer to the first local variable slot to be allocated.
2306 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2307 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2308 Otherwise, it is found by adding the length of the first slot to the
2309 value @code{STARTING_FRAME_OFFSET}.
2310 @c i'm not sure if the above is still correct.. had to change it to get
2311 @c rid of an overfull. --mew 2feb93
2313 @findex STACK_POINTER_OFFSET
2314 @item STACK_POINTER_OFFSET
2315 Offset from the stack pointer register to the first location at which
2316 outgoing arguments are placed. If not specified, the default value of
2317 zero is used. This is the proper value for most machines.
2319 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2320 the first location at which outgoing arguments are placed.
2322 @findex FIRST_PARM_OFFSET
2323 @item FIRST_PARM_OFFSET (@var{fundecl})
2324 Offset from the argument pointer register to the first argument's
2325 address. On some machines it may depend on the data type of the
2328 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2329 the first argument's address.
2331 @findex STACK_DYNAMIC_OFFSET
2332 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2333 Offset from the stack pointer register to an item dynamically allocated
2334 on the stack, e.g., by @code{alloca}.
2336 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2337 length of the outgoing arguments. The default is correct for most
2338 machines. See @file{function.c} for details.
2340 @findex DYNAMIC_CHAIN_ADDRESS
2341 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2342 A C expression whose value is RTL representing the address in a stack
2343 frame where the pointer to the caller's frame is stored. Assume that
2344 @var{frameaddr} is an RTL expression for the address of the stack frame
2347 If you don't define this macro, the default is to return the value
2348 of @var{frameaddr}---that is, the stack frame address is also the
2349 address of the stack word that points to the previous frame.
2351 @findex SETUP_FRAME_ADDRESSES
2352 @item SETUP_FRAME_ADDRESSES
2353 If defined, a C expression that produces the machine-specific code to
2354 setup the stack so that arbitrary frames can be accessed. For example,
2355 on the Sparc, we must flush all of the register windows to the stack
2356 before we can access arbitrary stack frames. You will seldom need to
2359 @findex BUILTIN_SETJMP_FRAME_VALUE
2360 @item BUILTIN_SETJMP_FRAME_VALUE
2361 If defined, a C expression that contains an rtx that is used to store
2362 the address of the current frame into the built in @code{setjmp} buffer.
2363 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2364 machines. One reason you may need to define this macro is if
2365 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2367 @findex RETURN_ADDR_RTX
2368 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2369 A C expression whose value is RTL representing the value of the return
2370 address for the frame @var{count} steps up from the current frame, after
2371 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2372 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2373 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2375 The value of the expression must always be the correct address when
2376 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2377 determine the return address of other frames.
2379 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2380 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2381 Define this if the return address of a particular stack frame is accessed
2382 from the frame pointer of the previous stack frame.
2384 @findex INCOMING_RETURN_ADDR_RTX
2385 @item INCOMING_RETURN_ADDR_RTX
2386 A C expression whose value is RTL representing the location of the
2387 incoming return address at the beginning of any function, before the
2388 prologue. This RTL is either a @code{REG}, indicating that the return
2389 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2392 You only need to define this macro if you want to support call frame
2393 debugging information like that provided by DWARF 2.
2395 If this RTL is a @code{REG}, you should also define
2396 DWARF_FRAME_RETURN_COLUMN to @code{DWARF_FRAME_REGNUM (REGNO)}.
2398 @findex INCOMING_FRAME_SP_OFFSET
2399 @item INCOMING_FRAME_SP_OFFSET
2400 A C expression whose value is an integer giving the offset, in bytes,
2401 from the value of the stack pointer register to the top of the stack
2402 frame at the beginning of any function, before the prologue. The top of
2403 the frame is defined to be the value of the stack pointer in the
2404 previous frame, just before the call instruction.
2406 You only need to define this macro if you want to support call frame
2407 debugging information like that provided by DWARF 2.
2409 @findex ARG_POINTER_CFA_OFFSET
2410 @item ARG_POINTER_CFA_OFFSET (@var{fundecl})
2411 A C expression whose value is an integer giving the offset, in bytes,
2412 from the argument pointer to the canonical frame address (cfa). The
2413 final value should coincide with that calculated by
2414 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2415 during virtual register instantiation.
2417 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2418 which is correct for most machines; in general, the arguments are found
2419 immediately before the stack frame. Note that this is not the case on
2420 some targets that save registers into the caller's frame, such as SPARC
2421 and rs6000, and so such targets need to define this macro.
2423 You only need to define this macro if the default is incorrect, and you
2424 want to support call frame debugging information like that provided by
2429 Define this macro if the stack size for the target is very small. This
2430 has the effect of disabling gcc's builtin @samp{alloca}, though
2431 @samp{__builtin_alloca} is not affected.
2434 @node Stack Checking
2435 @subsection Specifying How Stack Checking is Done
2437 GCC will check that stack references are within the boundaries of
2438 the stack, if the @samp{-fstack-check} is specified, in one of three ways:
2442 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2443 will assume that you have arranged for stack checking to be done at
2444 appropriate places in the configuration files, e.g., in
2445 @code{FUNCTION_PROLOGUE}. GCC will do not other special processing.
2448 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
2449 called @code{check_stack} in your @file{md} file, GCC will call that
2450 pattern with one argument which is the address to compare the stack
2451 value against. You must arrange for this pattern to report an error if
2452 the stack pointer is out of range.
2455 If neither of the above are true, GCC will generate code to periodically
2456 ``probe'' the stack pointer using the values of the macros defined below.
2459 Normally, you will use the default values of these macros, so GCC
2460 will use the third approach.
2463 @findex STACK_CHECK_BUILTIN
2464 @item STACK_CHECK_BUILTIN
2465 A nonzero value if stack checking is done by the configuration files in a
2466 machine-dependent manner. You should define this macro if stack checking
2467 is require by the ABI of your machine or if you would like to have to stack
2468 checking in some more efficient way than GCC's portable approach.
2469 The default value of this macro is zero.
2471 @findex STACK_CHECK_PROBE_INTERVAL
2472 @item STACK_CHECK_PROBE_INTERVAL
2473 An integer representing the interval at which GCC must generate stack
2474 probe instructions. You will normally define this macro to be no larger
2475 than the size of the ``guard pages'' at the end of a stack area. The
2476 default value of 4096 is suitable for most systems.
2478 @findex STACK_CHECK_PROBE_LOAD
2479 @item STACK_CHECK_PROBE_LOAD
2480 A integer which is nonzero if GCC should perform the stack probe
2481 as a load instruction and zero if GCC should use a store instruction.
2482 The default is zero, which is the most efficient choice on most systems.
2484 @findex STACK_CHECK_PROTECT
2485 @item STACK_CHECK_PROTECT
2486 The number of bytes of stack needed to recover from a stack overflow,
2487 for languages where such a recovery is supported. The default value of
2488 75 words should be adequate for most machines.
2490 @findex STACK_CHECK_MAX_FRAME_SIZE
2491 @item STACK_CHECK_MAX_FRAME_SIZE
2492 The maximum size of a stack frame, in bytes. GCC will generate probe
2493 instructions in non-leaf functions to ensure at least this many bytes of
2494 stack are available. If a stack frame is larger than this size, stack
2495 checking will not be reliable and GCC will issue a warning. The
2496 default is chosen so that GCC only generates one instruction on most
2497 systems. You should normally not change the default value of this macro.
2499 @findex STACK_CHECK_FIXED_FRAME_SIZE
2500 @item STACK_CHECK_FIXED_FRAME_SIZE
2501 GCC uses this value to generate the above warning message. It
2502 represents the amount of fixed frame used by a function, not including
2503 space for any callee-saved registers, temporaries and user variables.
2504 You need only specify an upper bound for this amount and will normally
2505 use the default of four words.
2507 @findex STACK_CHECK_MAX_VAR_SIZE
2508 @item STACK_CHECK_MAX_VAR_SIZE
2509 The maximum size, in bytes, of an object that GCC will place in the
2510 fixed area of the stack frame when the user specifies
2511 @samp{-fstack-check}.
2512 GCC computed the default from the values of the above macros and you will
2513 normally not need to override that default.
2517 @node Frame Registers
2518 @subsection Registers That Address the Stack Frame
2520 @c prevent bad page break with this line
2521 This discusses registers that address the stack frame.
2524 @findex STACK_POINTER_REGNUM
2525 @item STACK_POINTER_REGNUM
2526 The register number of the stack pointer register, which must also be a
2527 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2528 the hardware determines which register this is.
2530 @findex FRAME_POINTER_REGNUM
2531 @item FRAME_POINTER_REGNUM
2532 The register number of the frame pointer register, which is used to
2533 access automatic variables in the stack frame. On some machines, the
2534 hardware determines which register this is. On other machines, you can
2535 choose any register you wish for this purpose.
2537 @findex HARD_FRAME_POINTER_REGNUM
2538 @item HARD_FRAME_POINTER_REGNUM
2539 On some machines the offset between the frame pointer and starting
2540 offset of the automatic variables is not known until after register
2541 allocation has been done (for example, because the saved registers are
2542 between these two locations). On those machines, define
2543 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
2544 be used internally until the offset is known, and define
2545 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
2546 used for the frame pointer.
2548 You should define this macro only in the very rare circumstances when it
2549 is not possible to calculate the offset between the frame pointer and
2550 the automatic variables until after register allocation has been
2551 completed. When this macro is defined, you must also indicate in your
2552 definition of @code{ELIMINABLE_REGS} how to eliminate
2553 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
2554 or @code{STACK_POINTER_REGNUM}.
2556 Do not define this macro if it would be the same as
2557 @code{FRAME_POINTER_REGNUM}.
2559 @findex ARG_POINTER_REGNUM
2560 @item ARG_POINTER_REGNUM
2561 The register number of the arg pointer register, which is used to access
2562 the function's argument list. On some machines, this is the same as the
2563 frame pointer register. On some machines, the hardware determines which
2564 register this is. On other machines, you can choose any register you
2565 wish for this purpose. If this is not the same register as the frame
2566 pointer register, then you must mark it as a fixed register according to
2567 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2568 (@pxref{Elimination}).
2570 @findex RETURN_ADDRESS_POINTER_REGNUM
2571 @item RETURN_ADDRESS_POINTER_REGNUM
2572 The register number of the return address pointer register, which is used to
2573 access the current function's return address from the stack. On some
2574 machines, the return address is not at a fixed offset from the frame
2575 pointer or stack pointer or argument pointer. This register can be defined
2576 to point to the return address on the stack, and then be converted by
2577 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
2579 Do not define this macro unless there is no other way to get the return
2580 address from the stack.
2582 @findex STATIC_CHAIN_REGNUM
2583 @findex STATIC_CHAIN_INCOMING_REGNUM
2584 @item STATIC_CHAIN_REGNUM
2585 @itemx STATIC_CHAIN_INCOMING_REGNUM
2586 Register numbers used for passing a function's static chain pointer. If
2587 register windows are used, the register number as seen by the called
2588 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
2589 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
2590 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
2591 not be defined.@refill
2593 The static chain register need not be a fixed register.
2595 If the static chain is passed in memory, these macros should not be
2596 defined; instead, the next two macros should be defined.
2598 @findex STATIC_CHAIN
2599 @findex STATIC_CHAIN_INCOMING
2601 @itemx STATIC_CHAIN_INCOMING
2602 If the static chain is passed in memory, these macros provide rtx giving
2603 @code{mem} expressions that denote where they are stored.
2604 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
2605 as seen by the calling and called functions, respectively. Often the former
2606 will be at an offset from the stack pointer and the latter at an offset from
2607 the frame pointer.@refill
2609 @findex stack_pointer_rtx
2610 @findex frame_pointer_rtx
2611 @findex arg_pointer_rtx
2612 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
2613 @code{arg_pointer_rtx} will have been initialized prior to the use of these
2614 macros and should be used to refer to those items.
2616 If the static chain is passed in a register, the two previous macros should
2621 @subsection Eliminating Frame Pointer and Arg Pointer
2623 @c prevent bad page break with this line
2624 This is about eliminating the frame pointer and arg pointer.
2627 @findex FRAME_POINTER_REQUIRED
2628 @item FRAME_POINTER_REQUIRED
2629 A C expression which is nonzero if a function must have and use a frame
2630 pointer. This expression is evaluated in the reload pass. If its value is
2631 nonzero the function will have a frame pointer.
2633 The expression can in principle examine the current function and decide
2634 according to the facts, but on most machines the constant 0 or the
2635 constant 1 suffices. Use 0 when the machine allows code to be generated
2636 with no frame pointer, and doing so saves some time or space. Use 1
2637 when there is no possible advantage to avoiding a frame pointer.
2639 In certain cases, the compiler does not know how to produce valid code
2640 without a frame pointer. The compiler recognizes those cases and
2641 automatically gives the function a frame pointer regardless of what
2642 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
2645 In a function that does not require a frame pointer, the frame pointer
2646 register can be allocated for ordinary usage, unless you mark it as a
2647 fixed register. See @code{FIXED_REGISTERS} for more information.
2649 @findex INITIAL_FRAME_POINTER_OFFSET
2650 @findex get_frame_size
2651 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
2652 A C statement to store in the variable @var{depth-var} the difference
2653 between the frame pointer and the stack pointer values immediately after
2654 the function prologue. The value would be computed from information
2655 such as the result of @code{get_frame_size ()} and the tables of
2656 registers @code{regs_ever_live} and @code{call_used_regs}.
2658 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
2659 need not be defined. Otherwise, it must be defined even if
2660 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
2661 case, you may set @var{depth-var} to anything.
2663 @findex ELIMINABLE_REGS
2664 @item ELIMINABLE_REGS
2665 If defined, this macro specifies a table of register pairs used to
2666 eliminate unneeded registers that point into the stack frame. If it is not
2667 defined, the only elimination attempted by the compiler is to replace
2668 references to the frame pointer with references to the stack pointer.
2670 The definition of this macro is a list of structure initializations, each
2671 of which specifies an original and replacement register.
2673 On some machines, the position of the argument pointer is not known until
2674 the compilation is completed. In such a case, a separate hard register
2675 must be used for the argument pointer. This register can be eliminated by
2676 replacing it with either the frame pointer or the argument pointer,
2677 depending on whether or not the frame pointer has been eliminated.
2679 In this case, you might specify:
2681 #define ELIMINABLE_REGS \
2682 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
2683 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
2684 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
2687 Note that the elimination of the argument pointer with the stack pointer is
2688 specified first since that is the preferred elimination.
2690 @findex CAN_ELIMINATE
2691 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
2692 A C expression that returns non-zero if the compiler is allowed to try
2693 to replace register number @var{from-reg} with register number
2694 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
2695 is defined, and will usually be the constant 1, since most of the cases
2696 preventing register elimination are things that the compiler already
2699 @findex INITIAL_ELIMINATION_OFFSET
2700 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
2701 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
2702 specifies the initial difference between the specified pair of
2703 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
2706 @findex LONGJMP_RESTORE_FROM_STACK
2707 @item LONGJMP_RESTORE_FROM_STACK
2708 Define this macro if the @code{longjmp} function restores registers from
2709 the stack frames, rather than from those saved specifically by
2710 @code{setjmp}. Certain quantities must not be kept in registers across
2711 a call to @code{setjmp} on such machines.
2714 @node Stack Arguments
2715 @subsection Passing Function Arguments on the Stack
2716 @cindex arguments on stack
2717 @cindex stack arguments
2719 The macros in this section control how arguments are passed
2720 on the stack. See the following section for other macros that
2721 control passing certain arguments in registers.
2724 @findex PROMOTE_PROTOTYPES
2725 @item PROMOTE_PROTOTYPES
2726 A C expression whose value is nonzero if an argument declared in
2727 a prototype as an integral type smaller than @code{int} should
2728 actually be passed as an @code{int}. In addition to avoiding
2729 errors in certain cases of mismatch, it also makes for better
2730 code on certain machines. If the macro is not defined in target
2731 header files, it defaults to 0.
2735 A C expression. If nonzero, push insns will be used to pass
2737 If the target machine does not have a push instruction, set it to zero.
2738 That directs GCC to use an alternate strategy: to
2739 allocate the entire argument block and then store the arguments into
2740 it. When PUSH_ARGS is nonzero, PUSH_ROUNDING must be defined too.
2741 On some machines, the definition
2743 @findex PUSH_ROUNDING
2744 @item PUSH_ROUNDING (@var{npushed})
2745 A C expression that is the number of bytes actually pushed onto the
2746 stack when an instruction attempts to push @var{npushed} bytes.
2747 @findex PUSH_ROUNDING
2748 @item PUSH_ROUNDING (@var{npushed})
2749 A C expression that is the number of bytes actually pushed onto the
2750 stack when an instruction attempts to push @var{npushed} bytes.
2752 On some machines, the definition
2755 #define PUSH_ROUNDING(BYTES) (BYTES)
2759 will suffice. But on other machines, instructions that appear
2760 to push one byte actually push two bytes in an attempt to maintain
2761 alignment. Then the definition should be
2764 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
2767 @findex ACCUMULATE_OUTGOING_ARGS
2768 @findex current_function_outgoing_args_size
2769 @item ACCUMULATE_OUTGOING_ARGS
2770 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
2771 will be computed and placed into the variable
2772 @code{current_function_outgoing_args_size}. No space will be pushed
2773 onto the stack for each call; instead, the function prologue should
2774 increase the stack frame size by this amount.
2776 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
2779 @findex REG_PARM_STACK_SPACE
2780 @item REG_PARM_STACK_SPACE (@var{fndecl})
2781 Define this macro if functions should assume that stack space has been
2782 allocated for arguments even when their values are passed in
2785 The value of this macro is the size, in bytes, of the area reserved for
2786 arguments passed in registers for the function represented by @var{fndecl},
2787 which can be zero if GCC is calling a library function.
2789 This space can be allocated by the caller, or be a part of the
2790 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
2792 @c above is overfull. not sure what to do. --mew 5feb93 did
2793 @c something, not sure if it looks good. --mew 10feb93
2795 @findex MAYBE_REG_PARM_STACK_SPACE
2796 @findex FINAL_REG_PARM_STACK_SPACE
2797 @item MAYBE_REG_PARM_STACK_SPACE
2798 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
2799 Define these macros in addition to the one above if functions might
2800 allocate stack space for arguments even when their values are passed
2801 in registers. These should be used when the stack space allocated
2802 for arguments in registers is not a simple constant independent of the
2803 function declaration.
2805 The value of the first macro is the size, in bytes, of the area that
2806 we should initially assume would be reserved for arguments passed in registers.
2808 The value of the second macro is the actual size, in bytes, of the area
2809 that will be reserved for arguments passed in registers. This takes two
2810 arguments: an integer representing the number of bytes of fixed sized
2811 arguments on the stack, and a tree representing the number of bytes of
2812 variable sized arguments on the stack.
2814 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
2815 called for libcall functions, the current function, or for a function
2816 being called when it is known that such stack space must be allocated.
2817 In each case this value can be easily computed.
2819 When deciding whether a called function needs such stack space, and how
2820 much space to reserve, GCC uses these two macros instead of
2821 @code{REG_PARM_STACK_SPACE}.
2823 @findex OUTGOING_REG_PARM_STACK_SPACE
2824 @item OUTGOING_REG_PARM_STACK_SPACE
2825 Define this if it is the responsibility of the caller to allocate the area
2826 reserved for arguments passed in registers.
2828 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
2829 whether the space for these arguments counts in the value of
2830 @code{current_function_outgoing_args_size}.
2832 @findex STACK_PARMS_IN_REG_PARM_AREA
2833 @item STACK_PARMS_IN_REG_PARM_AREA
2834 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
2835 stack parameters don't skip the area specified by it.
2836 @c i changed this, makes more sens and it should have taken care of the
2837 @c overfull.. not as specific, tho. --mew 5feb93
2839 Normally, when a parameter is not passed in registers, it is placed on the
2840 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
2841 suppresses this behavior and causes the parameter to be passed on the
2842 stack in its natural location.
2844 @findex RETURN_POPS_ARGS
2845 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
2846 A C expression that should indicate the number of bytes of its own
2847 arguments that a function pops on returning, or 0 if the
2848 function pops no arguments and the caller must therefore pop them all
2849 after the function returns.
2851 @var{fundecl} is a C variable whose value is a tree node that describes
2852 the function in question. Normally it is a node of type
2853 @code{FUNCTION_DECL} that describes the declaration of the function.
2854 From this you can obtain the DECL_MACHINE_ATTRIBUTES of the function.
2856 @var{funtype} is a C variable whose value is a tree node that
2857 describes the function in question. Normally it is a node of type
2858 @code{FUNCTION_TYPE} that describes the data type of the function.
2859 From this it is possible to obtain the data types of the value and
2860 arguments (if known).
2862 When a call to a library function is being considered, @var{fundecl}
2863 will contain an identifier node for the library function. Thus, if
2864 you need to distinguish among various library functions, you can do so
2865 by their names. Note that ``library function'' in this context means
2866 a function used to perform arithmetic, whose name is known specially
2867 in the compiler and was not mentioned in the C code being compiled.
2869 @var{stack-size} is the number of bytes of arguments passed on the
2870 stack. If a variable number of bytes is passed, it is zero, and
2871 argument popping will always be the responsibility of the calling function.
2873 On the Vax, all functions always pop their arguments, so the definition
2874 of this macro is @var{stack-size}. On the 68000, using the standard
2875 calling convention, no functions pop their arguments, so the value of
2876 the macro is always 0 in this case. But an alternative calling
2877 convention is available in which functions that take a fixed number of
2878 arguments pop them but other functions (such as @code{printf}) pop
2879 nothing (the caller pops all). When this convention is in use,
2880 @var{funtype} is examined to determine whether a function takes a fixed
2881 number of arguments.
2884 @node Register Arguments
2885 @subsection Passing Arguments in Registers
2886 @cindex arguments in registers
2887 @cindex registers arguments
2889 This section describes the macros which let you control how various
2890 types of arguments are passed in registers or how they are arranged in
2894 @findex FUNCTION_ARG
2895 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2896 A C expression that controls whether a function argument is passed
2897 in a register, and which register.
2899 The arguments are @var{cum}, which summarizes all the previous
2900 arguments; @var{mode}, the machine mode of the argument; @var{type},
2901 the data type of the argument as a tree node or 0 if that is not known
2902 (which happens for C support library functions); and @var{named},
2903 which is 1 for an ordinary argument and 0 for nameless arguments that
2904 correspond to @samp{@dots{}} in the called function's prototype.
2906 The value of the expression is usually either a @code{reg} RTX for the
2907 hard register in which to pass the argument, or zero to pass the
2908 argument on the stack.
2910 For machines like the Vax and 68000, where normally all arguments are
2911 pushed, zero suffices as a definition.
2913 The value of the expression can also be a @code{parallel} RTX. This is
2914 used when an argument is passed in multiple locations. The mode of the
2915 of the @code{parallel} should be the mode of the entire argument. The
2916 @code{parallel} holds any number of @code{expr_list} pairs; each one
2917 describes where part of the argument is passed. In each
2918 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
2919 register in which to pass this part of the argument, and the mode of the
2920 register RTX indicates how large this part of the argument is. The
2921 second operand of the @code{expr_list} is a @code{const_int} which gives
2922 the offset in bytes into the entire argument of where this part starts.
2923 As a special exception the first @code{expr_list} in the @code{parallel}
2924 RTX may have a first operand of zero. This indicates that the entire
2925 argument is also stored on the stack.
2927 @cindex @file{stdarg.h} and register arguments
2928 The usual way to make the ANSI library @file{stdarg.h} work on a machine
2929 where some arguments are usually passed in registers, is to cause
2930 nameless arguments to be passed on the stack instead. This is done
2931 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
2933 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
2934 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
2935 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
2936 in the definition of this macro to determine if this argument is of a
2937 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
2938 is not defined and @code{FUNCTION_ARG} returns non-zero for such an
2939 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
2940 defined, the argument will be computed in the stack and then loaded into
2943 @findex MUST_PASS_IN_STACK
2944 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
2945 Define as a C expression that evaluates to nonzero if we do not know how
2946 to pass TYPE solely in registers. The file @file{expr.h} defines a
2947 definition that is usually appropriate, refer to @file{expr.h} for additional
2950 @findex FUNCTION_INCOMING_ARG
2951 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2952 Define this macro if the target machine has ``register windows'', so
2953 that the register in which a function sees an arguments is not
2954 necessarily the same as the one in which the caller passed the
2957 For such machines, @code{FUNCTION_ARG} computes the register in which
2958 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
2959 be defined in a similar fashion to tell the function being called
2960 where the arguments will arrive.
2962 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
2963 serves both purposes.@refill
2965 @findex FUNCTION_ARG_PARTIAL_NREGS
2966 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
2967 A C expression for the number of words, at the beginning of an
2968 argument, that must be put in registers. The value must be zero for
2969 arguments that are passed entirely in registers or that are entirely
2970 pushed on the stack.
2972 On some machines, certain arguments must be passed partially in
2973 registers and partially in memory. On these machines, typically the
2974 first @var{n} words of arguments are passed in registers, and the rest
2975 on the stack. If a multi-word argument (a @code{double} or a
2976 structure) crosses that boundary, its first few words must be passed
2977 in registers and the rest must be pushed. This macro tells the
2978 compiler when this occurs, and how many of the words should go in
2981 @code{FUNCTION_ARG} for these arguments should return the first
2982 register to be used by the caller for this argument; likewise
2983 @code{FUNCTION_INCOMING_ARG}, for the called function.
2985 @findex FUNCTION_ARG_PASS_BY_REFERENCE
2986 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2987 A C expression that indicates when an argument must be passed by reference.
2988 If nonzero for an argument, a copy of that argument is made in memory and a
2989 pointer to the argument is passed instead of the argument itself.
2990 The pointer is passed in whatever way is appropriate for passing a pointer
2993 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
2994 definition of this macro might be
2996 #define FUNCTION_ARG_PASS_BY_REFERENCE\
2997 (CUM, MODE, TYPE, NAMED) \
2998 MUST_PASS_IN_STACK (MODE, TYPE)
3000 @c this is *still* too long. --mew 5feb93
3002 @findex FUNCTION_ARG_CALLEE_COPIES
3003 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3004 If defined, a C expression that indicates when it is the called function's
3005 responsibility to make a copy of arguments passed by invisible reference.
3006 Normally, the caller makes a copy and passes the address of the copy to the
3007 routine being called. When FUNCTION_ARG_CALLEE_COPIES is defined and is
3008 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3009 ``live'' value. The called function must not modify this value. If it can be
3010 determined that the value won't be modified, it need not make a copy;
3011 otherwise a copy must be made.
3013 @findex CUMULATIVE_ARGS
3014 @item CUMULATIVE_ARGS
3015 A C type for declaring a variable that is used as the first argument of
3016 @code{FUNCTION_ARG} and other related values. For some target machines,
3017 the type @code{int} suffices and can hold the number of bytes of
3020 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3021 arguments that have been passed on the stack. The compiler has other
3022 variables to keep track of that. For target machines on which all
3023 arguments are passed on the stack, there is no need to store anything in
3024 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3025 should not be empty, so use @code{int}.
3027 @findex INIT_CUMULATIVE_ARGS
3028 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
3029 A C statement (sans semicolon) for initializing the variable @var{cum}
3030 for the state at the beginning of the argument list. The variable has
3031 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
3032 for the data type of the function which will receive the args, or 0
3033 if the args are to a compiler support library function. The value of
3034 @var{indirect} is nonzero when processing an indirect call, for example
3035 a call through a function pointer. The value of @var{indirect} is zero
3036 for a call to an explicitly named function, a library function call, or when
3037 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3040 When processing a call to a compiler support library function,
3041 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3042 contains the name of the function, as a string. @var{libname} is 0 when
3043 an ordinary C function call is being processed. Thus, each time this
3044 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3045 never both of them at once.
3047 @findex INIT_CUMULATIVE_INCOMING_ARGS
3048 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3049 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3050 finding the arguments for the function being compiled. If this macro is
3051 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3053 The value passed for @var{libname} is always 0, since library routines
3054 with special calling conventions are never compiled with GCC. The
3055 argument @var{libname} exists for symmetry with
3056 @code{INIT_CUMULATIVE_ARGS}.
3057 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3058 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3060 @findex FUNCTION_ARG_ADVANCE
3061 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3062 A C statement (sans semicolon) to update the summarizer variable
3063 @var{cum} to advance past an argument in the argument list. The
3064 values @var{mode}, @var{type} and @var{named} describe that argument.
3065 Once this is done, the variable @var{cum} is suitable for analyzing
3066 the @emph{following} argument with @code{FUNCTION_ARG}, etc.@refill
3068 This macro need not do anything if the argument in question was passed
3069 on the stack. The compiler knows how to track the amount of stack space
3070 used for arguments without any special help.
3072 @findex FUNCTION_ARG_PADDING
3073 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3074 If defined, a C expression which determines whether, and in which direction,
3075 to pad out an argument with extra space. The value should be of type
3076 @code{enum direction}: either @code{upward} to pad above the argument,
3077 @code{downward} to pad below, or @code{none} to inhibit padding.
3079 The @emph{amount} of padding is always just enough to reach the next
3080 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3083 This macro has a default definition which is right for most systems.
3084 For little-endian machines, the default is to pad upward. For
3085 big-endian machines, the default is to pad downward for an argument of
3086 constant size shorter than an @code{int}, and upward otherwise.
3088 @findex PAD_VARARGS_DOWN
3089 @item PAD_VARARGS_DOWN
3090 If defined, a C expression which determines whether the default
3091 implementation of va_arg will attempt to pad down before reading the
3092 next argument, if that argument is smaller than its aligned space as
3093 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3094 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3096 @findex FUNCTION_ARG_BOUNDARY
3097 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3098 If defined, a C expression that gives the alignment boundary, in bits,
3099 of an argument with the specified mode and type. If it is not defined,
3100 @code{PARM_BOUNDARY} is used for all arguments.
3102 @findex FUNCTION_ARG_REGNO_P
3103 @item FUNCTION_ARG_REGNO_P (@var{regno})
3104 A C expression that is nonzero if @var{regno} is the number of a hard
3105 register in which function arguments are sometimes passed. This does
3106 @emph{not} include implicit arguments such as the static chain and
3107 the structure-value address. On many machines, no registers can be
3108 used for this purpose since all function arguments are pushed on the
3111 @findex LOAD_ARGS_REVERSED
3112 @item LOAD_ARGS_REVERSED
3113 If defined, the order in which arguments are loaded into their
3114 respective argument registers is reversed so that the last
3115 argument is loaded first. This macro only affects arguments
3116 passed in registers.
3121 @subsection How Scalar Function Values Are Returned
3122 @cindex return values in registers
3123 @cindex values, returned by functions
3124 @cindex scalars, returned as values
3126 This section discusses the macros that control returning scalars as
3127 values---values that can fit in registers.
3130 @findex TRADITIONAL_RETURN_FLOAT
3131 @item TRADITIONAL_RETURN_FLOAT
3132 Define this macro if @samp{-traditional} should not cause functions
3133 declared to return @code{float} to convert the value to @code{double}.
3135 @findex FUNCTION_VALUE
3136 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3137 A C expression to create an RTX representing the place where a
3138 function returns a value of data type @var{valtype}. @var{valtype} is
3139 a tree node representing a data type. Write @code{TYPE_MODE
3140 (@var{valtype})} to get the machine mode used to represent that type.
3141 On many machines, only the mode is relevant. (Actually, on most
3142 machines, scalar values are returned in the same place regardless of
3145 The value of the expression is usually a @code{reg} RTX for the hard
3146 register where the return value is stored. The value can also be a
3147 @code{parallel} RTX, if the return value is in multiple places. See
3148 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3150 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3151 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3154 If the precise function being called is known, @var{func} is a tree
3155 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3156 pointer. This makes it possible to use a different value-returning
3157 convention for specific functions when all their calls are
3160 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3161 types, because these are returned in another way. See
3162 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3164 @findex FUNCTION_OUTGOING_VALUE
3165 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3166 Define this macro if the target machine has ``register windows''
3167 so that the register in which a function returns its value is not
3168 the same as the one in which the caller sees the value.
3170 For such machines, @code{FUNCTION_VALUE} computes the register in which
3171 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3172 defined in a similar fashion to tell the function where to put the
3175 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3176 @code{FUNCTION_VALUE} serves both purposes.@refill
3178 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3179 aggregate data types, because these are returned in another way. See
3180 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3182 @findex LIBCALL_VALUE
3183 @item LIBCALL_VALUE (@var{mode})
3184 A C expression to create an RTX representing the place where a library
3185 function returns a value of mode @var{mode}. If the precise function
3186 being called is known, @var{func} is a tree node
3187 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3188 pointer. This makes it possible to use a different value-returning
3189 convention for specific functions when all their calls are
3192 Note that ``library function'' in this context means a compiler
3193 support routine, used to perform arithmetic, whose name is known
3194 specially by the compiler and was not mentioned in the C code being
3197 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3198 data types, because none of the library functions returns such types.
3200 @findex FUNCTION_VALUE_REGNO_P
3201 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3202 A C expression that is nonzero if @var{regno} is the number of a hard
3203 register in which the values of called function may come back.
3205 A register whose use for returning values is limited to serving as the
3206 second of a pair (for a value of type @code{double}, say) need not be
3207 recognized by this macro. So for most machines, this definition
3211 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3214 If the machine has register windows, so that the caller and the called
3215 function use different registers for the return value, this macro
3216 should recognize only the caller's register numbers.
3218 @findex APPLY_RESULT_SIZE
3219 @item APPLY_RESULT_SIZE
3220 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3221 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3222 saving and restoring an arbitrary return value.
3225 @node Aggregate Return
3226 @subsection How Large Values Are Returned
3227 @cindex aggregates as return values
3228 @cindex large return values
3229 @cindex returning aggregate values
3230 @cindex structure value address
3232 When a function value's mode is @code{BLKmode} (and in some other
3233 cases), the value is not returned according to @code{FUNCTION_VALUE}
3234 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3235 block of memory in which the value should be stored. This address
3236 is called the @dfn{structure value address}.
3238 This section describes how to control returning structure values in
3242 @findex RETURN_IN_MEMORY
3243 @item RETURN_IN_MEMORY (@var{type})
3244 A C expression which can inhibit the returning of certain function
3245 values in registers, based on the type of value. A nonzero value says
3246 to return the function value in memory, just as large structures are
3247 always returned. Here @var{type} will be a C expression of type
3248 @code{tree}, representing the data type of the value.
3250 Note that values of mode @code{BLKmode} must be explicitly handled
3251 by this macro. Also, the option @samp{-fpcc-struct-return}
3252 takes effect regardless of this macro. On most systems, it is
3253 possible to leave the macro undefined; this causes a default
3254 definition to be used, whose value is the constant 1 for @code{BLKmode}
3255 values, and 0 otherwise.
3257 Do not use this macro to indicate that structures and unions should always
3258 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3261 @findex DEFAULT_PCC_STRUCT_RETURN
3262 @item DEFAULT_PCC_STRUCT_RETURN
3263 Define this macro to be 1 if all structure and union return values must be
3264 in memory. Since this results in slower code, this should be defined
3265 only if needed for compatibility with other compilers or with an ABI.
3266 If you define this macro to be 0, then the conventions used for structure
3267 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3269 If not defined, this defaults to the value 1.
3271 @findex STRUCT_VALUE_REGNUM
3272 @item STRUCT_VALUE_REGNUM
3273 If the structure value address is passed in a register, then
3274 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3276 @findex STRUCT_VALUE
3278 If the structure value address is not passed in a register, define
3279 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3280 where the address is passed. If it returns 0, the address is passed as
3281 an ``invisible'' first argument.
3283 @findex STRUCT_VALUE_INCOMING_REGNUM
3284 @item STRUCT_VALUE_INCOMING_REGNUM
3285 On some architectures the place where the structure value address
3286 is found by the called function is not the same place that the
3287 caller put it. This can be due to register windows, or it could
3288 be because the function prologue moves it to a different place.
3290 If the incoming location of the structure value address is in a
3291 register, define this macro as the register number.
3293 @findex STRUCT_VALUE_INCOMING
3294 @item STRUCT_VALUE_INCOMING
3295 If the incoming location is not a register, then you should define
3296 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3297 called function should find the value. If it should find the value on
3298 the stack, define this to create a @code{mem} which refers to the frame
3299 pointer. A definition of 0 means that the address is passed as an
3300 ``invisible'' first argument.
3302 @findex PCC_STATIC_STRUCT_RETURN
3303 @item PCC_STATIC_STRUCT_RETURN
3304 Define this macro if the usual system convention on the target machine
3305 for returning structures and unions is for the called function to return
3306 the address of a static variable containing the value.
3308 Do not define this if the usual system convention is for the caller to
3309 pass an address to the subroutine.
3311 This macro has effect in @samp{-fpcc-struct-return} mode, but it does
3312 nothing when you use @samp{-freg-struct-return} mode.
3316 @subsection Caller-Saves Register Allocation
3318 If you enable it, GCC can save registers around function calls. This
3319 makes it possible to use call-clobbered registers to hold variables that
3320 must live across calls.
3323 @findex DEFAULT_CALLER_SAVES
3324 @item DEFAULT_CALLER_SAVES
3325 Define this macro if function calls on the target machine do not preserve
3326 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3327 for all registers. When defined, this macro enables @samp{-fcaller-saves}
3328 by default for all optimization levels. It has no effect for optimization
3329 levels 2 and higher, where @samp{-fcaller-saves} is the default.
3331 @findex CALLER_SAVE_PROFITABLE
3332 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3333 A C expression to determine whether it is worthwhile to consider placing
3334 a pseudo-register in a call-clobbered hard register and saving and
3335 restoring it around each function call. The expression should be 1 when
3336 this is worth doing, and 0 otherwise.
3338 If you don't define this macro, a default is used which is good on most
3339 machines: @code{4 * @var{calls} < @var{refs}}.
3341 @findex HARD_REGNO_CALLER_SAVE_MODE
3342 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3343 A C expression specifying which mode is required for saving @var{nregs}
3344 of a pseudo-register in call-clobbered hard register @var{regno}. If
3345 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3346 returned. For most machines this macro need not be defined since GCC
3347 will select the smallest suitable mode.
3350 @node Function Entry
3351 @subsection Function Entry and Exit
3352 @cindex function entry and exit
3356 This section describes the macros that output function entry
3357 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3360 @findex FUNCTION_PROLOGUE
3361 @item FUNCTION_PROLOGUE (@var{file}, @var{size})
3362 A C compound statement that outputs the assembler code for entry to a
3363 function. The prologue is responsible for setting up the stack frame,
3364 initializing the frame pointer register, saving registers that must be
3365 saved, and allocating @var{size} additional bytes of storage for the
3366 local variables. @var{size} is an integer. @var{file} is a stdio
3367 stream to which the assembler code should be output.
3369 The label for the beginning of the function need not be output by this
3370 macro. That has already been done when the macro is run.
3372 @findex regs_ever_live
3373 To determine which registers to save, the macro can refer to the array
3374 @code{regs_ever_live}: element @var{r} is nonzero if hard register
3375 @var{r} is used anywhere within the function. This implies the function
3376 prologue should save register @var{r}, provided it is not one of the
3377 call-used registers. (@code{FUNCTION_EPILOGUE} must likewise use
3378 @code{regs_ever_live}.)
3380 On machines that have ``register windows'', the function entry code does
3381 not save on the stack the registers that are in the windows, even if
3382 they are supposed to be preserved by function calls; instead it takes
3383 appropriate steps to ``push'' the register stack, if any non-call-used
3384 registers are used in the function.
3386 @findex frame_pointer_needed
3387 On machines where functions may or may not have frame-pointers, the
3388 function entry code must vary accordingly; it must set up the frame
3389 pointer if one is wanted, and not otherwise. To determine whether a
3390 frame pointer is in wanted, the macro can refer to the variable
3391 @code{frame_pointer_needed}. The variable's value will be 1 at run
3392 time in a function that needs a frame pointer. @xref{Elimination}.
3394 The function entry code is responsible for allocating any stack space
3395 required for the function. This stack space consists of the regions
3396 listed below. In most cases, these regions are allocated in the
3397 order listed, with the last listed region closest to the top of the
3398 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
3399 the highest address if it is not defined). You can use a different order
3400 for a machine if doing so is more convenient or required for
3401 compatibility reasons. Except in cases where required by standard
3402 or by a debugger, there is no reason why the stack layout used by GCC
3403 need agree with that used by other compilers for a machine.
3407 @findex current_function_pretend_args_size
3408 A region of @code{current_function_pretend_args_size} bytes of
3409 uninitialized space just underneath the first argument arriving on the
3410 stack. (This may not be at the very start of the allocated stack region
3411 if the calling sequence has pushed anything else since pushing the stack
3412 arguments. But usually, on such machines, nothing else has been pushed
3413 yet, because the function prologue itself does all the pushing.) This
3414 region is used on machines where an argument may be passed partly in
3415 registers and partly in memory, and, in some cases to support the
3416 features in @file{varargs.h} and @file{stdargs.h}.
3419 An area of memory used to save certain registers used by the function.
3420 The size of this area, which may also include space for such things as
3421 the return address and pointers to previous stack frames, is
3422 machine-specific and usually depends on which registers have been used
3423 in the function. Machines with register windows often do not require
3427 A region of at least @var{size} bytes, possibly rounded up to an allocation
3428 boundary, to contain the local variables of the function. On some machines,
3429 this region and the save area may occur in the opposite order, with the
3430 save area closer to the top of the stack.
3433 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3434 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3435 @code{current_function_outgoing_args_size} bytes to be used for outgoing
3436 argument lists of the function. @xref{Stack Arguments}.
3439 Normally, it is necessary for the macros @code{FUNCTION_PROLOGUE} and
3440 @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C
3441 variable @code{current_function_is_leaf} is nonzero for such a function.
3443 @findex EXIT_IGNORE_STACK
3444 @item EXIT_IGNORE_STACK
3445 Define this macro as a C expression that is nonzero if the return
3446 instruction or the function epilogue ignores the value of the stack
3447 pointer; in other words, if it is safe to delete an instruction to
3448 adjust the stack pointer before a return from the function.
3450 Note that this macro's value is relevant only for functions for which
3451 frame pointers are maintained. It is never safe to delete a final
3452 stack adjustment in a function that has no frame pointer, and the
3453 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3455 @findex EPILOGUE_USES
3456 @item EPILOGUE_USES (@var{regno})
3457 Define this macro as a C expression that is nonzero for registers that are
3458 used by the epilogue or the @samp{return} pattern. The stack and frame
3459 pointer registers are already be assumed to be used as needed.
3461 @findex FUNCTION_EPILOGUE
3462 @item FUNCTION_EPILOGUE (@var{file}, @var{size})
3463 A C compound statement that outputs the assembler code for exit from a
3464 function. The epilogue is responsible for restoring the saved
3465 registers and stack pointer to their values when the function was
3466 called, and returning control to the caller. This macro takes the
3467 same arguments as the macro @code{FUNCTION_PROLOGUE}, and the
3468 registers to restore are determined from @code{regs_ever_live} and
3469 @code{CALL_USED_REGISTERS} in the same way.
3471 On some machines, there is a single instruction that does all the work
3472 of returning from the function. On these machines, give that
3473 instruction the name @samp{return} and do not define the macro
3474 @code{FUNCTION_EPILOGUE} at all.
3476 Do not define a pattern named @samp{return} if you want the
3477 @code{FUNCTION_EPILOGUE} to be used. If you want the target switches
3478 to control whether return instructions or epilogues are used, define a
3479 @samp{return} pattern with a validity condition that tests the target
3480 switches appropriately. If the @samp{return} pattern's validity
3481 condition is false, epilogues will be used.
3483 On machines where functions may or may not have frame-pointers, the
3484 function exit code must vary accordingly. Sometimes the code for these
3485 two cases is completely different. To determine whether a frame pointer
3486 is wanted, the macro can refer to the variable
3487 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
3488 a function that needs a frame pointer.
3490 Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
3491 treat leaf functions specially. The C variable @code{current_function_is_leaf}
3492 is nonzero for such a function. @xref{Leaf Functions}.
3494 On some machines, some functions pop their arguments on exit while
3495 others leave that for the caller to do. For example, the 68020 when
3496 given @samp{-mrtd} pops arguments in functions that take a fixed
3497 number of arguments.
3499 @findex current_function_pops_args
3500 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
3501 functions pop their own arguments. @code{FUNCTION_EPILOGUE} needs to
3502 know what was decided. The variable that is called
3503 @code{current_function_pops_args} is the number of bytes of its
3504 arguments that a function should pop. @xref{Scalar Return}.
3505 @c what is the "its arguments" in the above sentence referring to, pray
3506 @c tell? --mew 5feb93
3508 @findex DELAY_SLOTS_FOR_EPILOGUE
3509 @item DELAY_SLOTS_FOR_EPILOGUE
3510 Define this macro if the function epilogue contains delay slots to which
3511 instructions from the rest of the function can be ``moved''. The
3512 definition should be a C expression whose value is an integer
3513 representing the number of delay slots there.
3515 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
3516 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
3517 A C expression that returns 1 if @var{insn} can be placed in delay
3518 slot number @var{n} of the epilogue.
3520 The argument @var{n} is an integer which identifies the delay slot now
3521 being considered (since different slots may have different rules of
3522 eligibility). It is never negative and is always less than the number
3523 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
3524 If you reject a particular insn for a given delay slot, in principle, it
3525 may be reconsidered for a subsequent delay slot. Also, other insns may
3526 (at least in principle) be considered for the so far unfilled delay
3529 @findex current_function_epilogue_delay_list
3530 @findex final_scan_insn
3531 The insns accepted to fill the epilogue delay slots are put in an RTL
3532 list made with @code{insn_list} objects, stored in the variable
3533 @code{current_function_epilogue_delay_list}. The insn for the first
3534 delay slot comes first in the list. Your definition of the macro
3535 @code{FUNCTION_EPILOGUE} should fill the delay slots by outputting the
3536 insns in this list, usually by calling @code{final_scan_insn}.
3538 You need not define this macro if you did not define
3539 @code{DELAY_SLOTS_FOR_EPILOGUE}.
3541 @findex ASM_OUTPUT_MI_THUNK
3542 @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function})
3543 A C compound statement that outputs the assembler code for a thunk
3544 function, used to implement C++ virtual function calls with multiple
3545 inheritance. The thunk acts as a wrapper around a virtual function,
3546 adjusting the implicit object parameter before handing control off to
3549 First, emit code to add the integer @var{delta} to the location that
3550 contains the incoming first argument. Assume that this argument
3551 contains a pointer, and is the one used to pass the @code{this} pointer
3552 in C++. This is the incoming argument @emph{before} the function prologue,
3553 e.g. @samp{%o0} on a sparc. The addition must preserve the values of
3554 all other incoming arguments.
3556 After the addition, emit code to jump to @var{function}, which is a
3557 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
3558 not touch the return address. Hence returning from @var{FUNCTION} will
3559 return to whoever called the current @samp{thunk}.
3561 The effect must be as if @var{function} had been called directly with
3562 the adjusted first argument. This macro is responsible for emitting all
3563 of the code for a thunk function; @code{FUNCTION_PROLOGUE} and
3564 @code{FUNCTION_EPILOGUE} are not invoked.
3566 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
3567 have already been extracted from it.) It might possibly be useful on
3568 some targets, but probably not.
3570 If you do not define this macro, the target-independent code in the C++
3571 frontend will generate a less efficient heavyweight thunk that calls
3572 @var{function} instead of jumping to it. The generic approach does
3573 not support varargs.
3577 @subsection Generating Code for Profiling
3578 @cindex profiling, code generation
3580 These macros will help you generate code for profiling.
3583 @findex FUNCTION_PROFILER
3584 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
3585 A C statement or compound statement to output to @var{file} some
3586 assembler code to call the profiling subroutine @code{mcount}.
3589 The details of how @code{mcount} expects to be called are determined by
3590 your operating system environment, not by GCC. To figure them out,
3591 compile a small program for profiling using the system's installed C
3592 compiler and look at the assembler code that results.
3594 Older implementations of @code{mcount} expect the address of a counter
3595 variable to be loaded into some register. The name of this variable is
3596 @samp{LP} followed by the number @var{labelno}, so you would generate
3597 the name using @samp{LP%d} in a @code{fprintf}.
3599 @findex NO_PROFILE_COUNTERS
3600 @item NO_PROFILE_COUNTERS
3601 Define this macro if the @code{mcount} subroutine on your system does
3602 not need a counter variable allocated for each function. This is true
3603 for almost all modern implementations. If you define this macro, you
3604 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
3606 @findex PROFILE_BEFORE_PROLOGUE
3607 @item PROFILE_BEFORE_PROLOGUE
3608 Define this macro if the code for function profiling should come before
3609 the function prologue. Normally, the profiling code comes after.
3611 @findex FUNCTION_BLOCK_PROFILER
3612 @vindex profile_block_flag
3613 @item FUNCTION_BLOCK_PROFILER (@var{file}, @var{labelno})
3614 A C statement or compound statement to output to @var{file} some
3615 assembler code to initialize basic-block profiling for the current
3616 object module. The global compile flag @code{profile_block_flag}
3617 distinguishes two profile modes.
3620 @findex __bb_init_func
3621 @item profile_block_flag != 2
3622 Output code to call the subroutine @code{__bb_init_func} once per
3623 object module, passing it as its sole argument the address of a block
3624 allocated in the object module.
3626 The name of the block is a local symbol made with this statement:
3629 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3632 Of course, since you are writing the definition of
3633 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3634 can take a short cut in the definition of this macro and use the name
3635 that you know will result.
3637 The first word of this block is a flag which will be nonzero if the
3638 object module has already been initialized. So test this word first,
3639 and do not call @code{__bb_init_func} if the flag is
3640 nonzero. BLOCK_OR_LABEL contains a unique number which may be used to
3641 generate a label as a branch destination when @code{__bb_init_func}
3644 Described in assembler language, the code to be output looks like:
3654 @findex __bb_init_trace_func
3655 @item profile_block_flag == 2
3656 Output code to call the subroutine @code{__bb_init_trace_func}
3657 and pass two parameters to it. The first parameter is the same as
3658 for @code{__bb_init_func}. The second parameter is the number of the
3659 first basic block of the function as given by BLOCK_OR_LABEL. Note
3660 that @code{__bb_init_trace_func} has to be called, even if the object
3661 module has been initialized already.
3663 Described in assembler language, the code to be output looks like:
3666 parameter2 <- BLOCK_OR_LABEL
3667 call __bb_init_trace_func
3671 @findex BLOCK_PROFILER
3672 @vindex profile_block_flag
3673 @item BLOCK_PROFILER (@var{file}, @var{blockno})
3674 A C statement or compound statement to output to @var{file} some
3675 assembler code to increment the count associated with the basic
3676 block number @var{blockno}. The global compile flag
3677 @code{profile_block_flag} distinguishes two profile modes.
3680 @item profile_block_flag != 2
3681 Output code to increment the counter directly. Basic blocks are
3682 numbered separately from zero within each compilation. The count
3683 associated with block number @var{blockno} is at index
3684 @var{blockno} in a vector of words; the name of this array is a local
3685 symbol made with this statement:
3688 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 2);
3691 @c This paragraph is the same as one a few paragraphs up.
3692 @c That is not an error.
3693 Of course, since you are writing the definition of
3694 @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
3695 can take a short cut in the definition of this macro and use the name
3696 that you know will result.
3698 Described in assembler language, the code to be output looks like:
3701 inc (LPBX2+4*BLOCKNO)
3705 @findex __bb_trace_func
3706 @item profile_block_flag == 2
3707 Output code to initialize the global structure @code{__bb} and
3708 call the function @code{__bb_trace_func}, which will increment the
3711 @code{__bb} consists of two words. In the first word, the current
3712 basic block number, as given by BLOCKNO, has to be stored. In
3713 the second word, the address of a block allocated in the object
3714 module has to be stored. The address is given by the label created
3715 with this statement:
3718 ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
3721 Described in assembler language, the code to be output looks like:
3723 move BLOCKNO -> (__bb)
3724 move LPBX0 -> (__bb+4)
3725 call __bb_trace_func
3729 @findex FUNCTION_BLOCK_PROFILER_EXIT
3730 @findex __bb_trace_ret
3731 @vindex profile_block_flag
3732 @item FUNCTION_BLOCK_PROFILER_EXIT (@var{file})
3733 A C statement or compound statement to output to @var{file}
3734 assembler code to call function @code{__bb_trace_ret}. The
3735 assembler code should only be output
3736 if the global compile flag @code{profile_block_flag} == 2. This
3737 macro has to be used at every place where code for returning from
3738 a function is generated (e.g. @code{FUNCTION_EPILOGUE}). Although
3739 you have to write the definition of @code{FUNCTION_EPILOGUE}
3740 as well, you have to define this macro to tell the compiler, that
3741 the proper call to @code{__bb_trace_ret} is produced.
3743 @findex MACHINE_STATE_SAVE
3744 @findex __bb_init_trace_func
3745 @findex __bb_trace_func
3746 @findex __bb_trace_ret
3747 @item MACHINE_STATE_SAVE (@var{id})
3748 A C statement or compound statement to save all registers, which may
3749 be clobbered by a function call, including condition codes. The
3750 @code{asm} statement will be mostly likely needed to handle this
3751 task. Local labels in the assembler code can be concatenated with the
3752 string @var{id}, to obtain a unique label name.
3754 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3755 @code{FUNCTION_EPILOGUE} must be saved in the macros
3756 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3757 @code{BLOCK_PROFILER} prior calling @code{__bb_init_trace_func},
3758 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3760 @findex MACHINE_STATE_RESTORE
3761 @findex __bb_init_trace_func
3762 @findex __bb_trace_func
3763 @findex __bb_trace_ret
3764 @item MACHINE_STATE_RESTORE (@var{id})
3765 A C statement or compound statement to restore all registers, including
3766 condition codes, saved by @code{MACHINE_STATE_SAVE}.
3768 Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or
3769 @code{FUNCTION_EPILOGUE} must be restored in the macros
3770 @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and
3771 @code{BLOCK_PROFILER} after calling @code{__bb_init_trace_func},
3772 @code{__bb_trace_ret} and @code{__bb_trace_func} respectively.
3774 @findex BLOCK_PROFILER_CODE
3775 @item BLOCK_PROFILER_CODE
3776 A C function or functions which are needed in the library to
3777 support block profiling.
3781 @subsection Permitting inlining of functions with attributes
3784 By default if a function has a target specific attribute attached to it,
3785 it will not be inlined. This behaviour can be overridden if the target
3786 defines the @samp{FUNCTION_ATTRIBUTE_INLINABLE_P} macro. This macro
3787 takes one argument, a @samp{DECL} describing the function. It should
3788 return non-zero if the function can be inlined, otherwise it should
3792 @subsection Permitting tail calls to functions
3794 @cindex sibling calls
3797 @findex FUNCTION_OK_FOR_SIBCALL
3798 @item FUNCTION_OK_FOR_SIBCALL (@var{decl})
3799 A C expression that evaluates to true if it is ok to perform a sibling
3802 It is not uncommon for limitations of calling conventions to prevent
3803 tail calls to functions outside the current unit of translation, or
3804 during PIC compilation. Use this macro to enforce these restrictions,
3805 as the @code{sibcall} md pattern can not fail, or fall over to a
3810 @section Implementing the Varargs Macros
3811 @cindex varargs implementation
3813 GCC comes with an implementation of @file{varargs.h} and
3814 @file{stdarg.h} that work without change on machines that pass arguments
3815 on the stack. Other machines require their own implementations of
3816 varargs, and the two machine independent header files must have
3817 conditionals to include it.
3819 ANSI @file{stdarg.h} differs from traditional @file{varargs.h} mainly in
3820 the calling convention for @code{va_start}. The traditional
3821 implementation takes just one argument, which is the variable in which
3822 to store the argument pointer. The ANSI implementation of
3823 @code{va_start} takes an additional second argument. The user is
3824 supposed to write the last named argument of the function here.
3826 However, @code{va_start} should not use this argument. The way to find
3827 the end of the named arguments is with the built-in functions described
3831 @findex __builtin_saveregs
3832 @item __builtin_saveregs ()
3833 Use this built-in function to save the argument registers in memory so
3834 that the varargs mechanism can access them. Both ANSI and traditional
3835 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3836 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
3838 On some machines, @code{__builtin_saveregs} is open-coded under the
3839 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
3840 it calls a routine written in assembler language, found in
3843 Code generated for the call to @code{__builtin_saveregs} appears at the
3844 beginning of the function, as opposed to where the call to
3845 @code{__builtin_saveregs} is written, regardless of what the code is.
3846 This is because the registers must be saved before the function starts
3847 to use them for its own purposes.
3848 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3851 @findex __builtin_args_info
3852 @item __builtin_args_info (@var{category})
3853 Use this built-in function to find the first anonymous arguments in
3856 In general, a machine may have several categories of registers used for
3857 arguments, each for a particular category of data types. (For example,
3858 on some machines, floating-point registers are used for floating-point
3859 arguments while other arguments are passed in the general registers.)
3860 To make non-varargs functions use the proper calling convention, you
3861 have defined the @code{CUMULATIVE_ARGS} data type to record how many
3862 registers in each category have been used so far
3864 @code{__builtin_args_info} accesses the same data structure of type
3865 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
3866 with it, with @var{category} specifying which word to access. Thus, the
3867 value indicates the first unused register in a given category.
3869 Normally, you would use @code{__builtin_args_info} in the implementation
3870 of @code{va_start}, accessing each category just once and storing the
3871 value in the @code{va_list} object. This is because @code{va_list} will
3872 have to update the values, and there is no way to alter the
3873 values accessed by @code{__builtin_args_info}.
3875 @findex __builtin_next_arg
3876 @item __builtin_next_arg (@var{lastarg})
3877 This is the equivalent of @code{__builtin_args_info}, for stack
3878 arguments. It returns the address of the first anonymous stack
3879 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
3880 returns the address of the location above the first anonymous stack
3881 argument. Use it in @code{va_start} to initialize the pointer for
3882 fetching arguments from the stack. Also use it in @code{va_start} to
3883 verify that the second parameter @var{lastarg} is the last named argument
3884 of the current function.
3886 @findex __builtin_classify_type
3887 @item __builtin_classify_type (@var{object})
3888 Since each machine has its own conventions for which data types are
3889 passed in which kind of register, your implementation of @code{va_arg}
3890 has to embody these conventions. The easiest way to categorize the
3891 specified data type is to use @code{__builtin_classify_type} together
3892 with @code{sizeof} and @code{__alignof__}.
3894 @code{__builtin_classify_type} ignores the value of @var{object},
3895 considering only its data type. It returns an integer describing what
3896 kind of type that is---integer, floating, pointer, structure, and so on.
3898 The file @file{typeclass.h} defines an enumeration that you can use to
3899 interpret the values of @code{__builtin_classify_type}.
3902 These machine description macros help implement varargs:
3905 @findex EXPAND_BUILTIN_SAVEREGS
3906 @item EXPAND_BUILTIN_SAVEREGS ()
3907 If defined, is a C expression that produces the machine-specific code
3908 for a call to @code{__builtin_saveregs}. This code will be moved to the
3909 very beginning of the function, before any parameter access are made.
3910 The return value of this function should be an RTX that contains the
3911 value to use as the return of @code{__builtin_saveregs}.
3913 @findex SETUP_INCOMING_VARARGS
3914 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
3915 This macro offers an alternative to using @code{__builtin_saveregs} and
3916 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
3917 anonymous register arguments into the stack so that all the arguments
3918 appear to have been passed consecutively on the stack. Once this is
3919 done, you can use the standard implementation of varargs that works for
3920 machines that pass all their arguments on the stack.
3922 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
3923 structure, containing the values that are obtained after processing the
3924 named arguments. The arguments @var{mode} and @var{type} describe the
3925 last named argument---its machine mode and its data type as a tree node.
3927 The macro implementation should do two things: first, push onto the
3928 stack all the argument registers @emph{not} used for the named
3929 arguments, and second, store the size of the data thus pushed into the
3930 @code{int}-valued variable whose name is supplied as the argument
3931 @var{pretend_args_size}. The value that you store here will serve as
3932 additional offset for setting up the stack frame.
3934 Because you must generate code to push the anonymous arguments at
3935 compile time without knowing their data types,
3936 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
3937 a single category of argument register and use it uniformly for all data
3940 If the argument @var{second_time} is nonzero, it means that the
3941 arguments of the function are being analyzed for the second time. This
3942 happens for an inline function, which is not actually compiled until the
3943 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
3944 not generate any instructions in this case.
3946 @findex STRICT_ARGUMENT_NAMING
3947 @item STRICT_ARGUMENT_NAMING
3948 Define this macro to be a nonzero value if the location where a function
3949 argument is passed depends on whether or not it is a named argument.
3951 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
3952 is set for varargs and stdarg functions. If this macro returns a
3953 nonzero value, the @var{named} argument is always true for named
3954 arguments, and false for unnamed arguments. If it returns a value of
3955 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
3956 are treated as named. Otherwise, all named arguments except the last
3957 are treated as named.
3959 You need not define this macro if it always returns zero.
3961 @findex PRETEND_OUTGOING_VARARGS_NAMED
3962 @item PRETEND_OUTGOING_VARARGS_NAMED
3963 If you need to conditionally change ABIs so that one works with
3964 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
3965 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
3966 defined, then define this macro to return nonzero if
3967 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
3968 Otherwise, you should not define this macro.
3972 @section Trampolines for Nested Functions
3973 @cindex trampolines for nested functions
3974 @cindex nested functions, trampolines for
3976 A @dfn{trampoline} is a small piece of code that is created at run time
3977 when the address of a nested function is taken. It normally resides on
3978 the stack, in the stack frame of the containing function. These macros
3979 tell GCC how to generate code to allocate and initialize a
3982 The instructions in the trampoline must do two things: load a constant
3983 address into the static chain register, and jump to the real address of
3984 the nested function. On CISC machines such as the m68k, this requires
3985 two instructions, a move immediate and a jump. Then the two addresses
3986 exist in the trampoline as word-long immediate operands. On RISC
3987 machines, it is often necessary to load each address into a register in
3988 two parts. Then pieces of each address form separate immediate
3991 The code generated to initialize the trampoline must store the variable
3992 parts---the static chain value and the function address---into the
3993 immediate operands of the instructions. On a CISC machine, this is
3994 simply a matter of copying each address to a memory reference at the
3995 proper offset from the start of the trampoline. On a RISC machine, it
3996 may be necessary to take out pieces of the address and store them
4000 @findex TRAMPOLINE_TEMPLATE
4001 @item TRAMPOLINE_TEMPLATE (@var{file})
4002 A C statement to output, on the stream @var{file}, assembler code for a
4003 block of data that contains the constant parts of a trampoline. This
4004 code should not include a label---the label is taken care of
4007 If you do not define this macro, it means no template is needed
4008 for the target. Do not define this macro on systems where the block move
4009 code to copy the trampoline into place would be larger than the code
4010 to generate it on the spot.
4012 @findex TRAMPOLINE_SECTION
4013 @item TRAMPOLINE_SECTION
4014 The name of a subroutine to switch to the section in which the
4015 trampoline template is to be placed (@pxref{Sections}). The default is
4016 a value of @samp{readonly_data_section}, which places the trampoline in
4017 the section containing read-only data.
4019 @findex TRAMPOLINE_SIZE
4020 @item TRAMPOLINE_SIZE
4021 A C expression for the size in bytes of the trampoline, as an integer.
4023 @findex TRAMPOLINE_ALIGNMENT
4024 @item TRAMPOLINE_ALIGNMENT
4025 Alignment required for trampolines, in bits.
4027 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4028 is used for aligning trampolines.
4030 @findex INITIALIZE_TRAMPOLINE
4031 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4032 A C statement to initialize the variable parts of a trampoline.
4033 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4034 an RTX for the address of the nested function; @var{static_chain} is an
4035 RTX for the static chain value that should be passed to the function
4038 @findex ALLOCATE_TRAMPOLINE
4039 @item ALLOCATE_TRAMPOLINE (@var{fp})
4040 A C expression to allocate run-time space for a trampoline. The
4041 expression value should be an RTX representing a memory reference to the
4042 space for the trampoline.
4044 @cindex @code{FUNCTION_EPILOGUE} and trampolines
4045 @cindex @code{FUNCTION_PROLOGUE} and trampolines
4046 If this macro is not defined, by default the trampoline is allocated as
4047 a stack slot. This default is right for most machines. The exceptions
4048 are machines where it is impossible to execute instructions in the stack
4049 area. On such machines, you may have to implement a separate stack,
4050 using this macro in conjunction with @code{FUNCTION_PROLOGUE} and
4051 @code{FUNCTION_EPILOGUE}.
4053 @var{fp} points to a data structure, a @code{struct function}, which
4054 describes the compilation status of the immediate containing function of
4055 the function which the trampoline is for. Normally (when
4056 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
4057 trampoline is in the stack frame of this containing function. Other
4058 allocation strategies probably must do something analogous with this
4062 Implementing trampolines is difficult on many machines because they have
4063 separate instruction and data caches. Writing into a stack location
4064 fails to clear the memory in the instruction cache, so when the program
4065 jumps to that location, it executes the old contents.
4067 Here are two possible solutions. One is to clear the relevant parts of
4068 the instruction cache whenever a trampoline is set up. The other is to
4069 make all trampolines identical, by having them jump to a standard
4070 subroutine. The former technique makes trampoline execution faster; the
4071 latter makes initialization faster.
4073 To clear the instruction cache when a trampoline is initialized, define
4074 the following macros which describe the shape of the cache.
4077 @findex INSN_CACHE_SIZE
4078 @item INSN_CACHE_SIZE
4079 The total size in bytes of the cache.
4081 @findex INSN_CACHE_LINE_WIDTH
4082 @item INSN_CACHE_LINE_WIDTH
4083 The length in bytes of each cache line. The cache is divided into cache
4084 lines which are disjoint slots, each holding a contiguous chunk of data
4085 fetched from memory. Each time data is brought into the cache, an
4086 entire line is read at once. The data loaded into a cache line is
4087 always aligned on a boundary equal to the line size.
4089 @findex INSN_CACHE_DEPTH
4090 @item INSN_CACHE_DEPTH
4091 The number of alternative cache lines that can hold any particular memory
4095 Alternatively, if the machine has system calls or instructions to clear
4096 the instruction cache directly, you can define the following macro.
4099 @findex CLEAR_INSN_CACHE
4100 @item CLEAR_INSN_CACHE (@var{BEG}, @var{END})
4101 If defined, expands to a C expression clearing the @emph{instruction
4102 cache} in the specified interval. If it is not defined, and the macro
4103 INSN_CACHE_SIZE is defined, some generic code is generated to clear the
4104 cache. The definition of this macro would typically be a series of
4105 @code{asm} statements. Both @var{BEG} and @var{END} are both pointer
4109 To use a standard subroutine, define the following macro. In addition,
4110 you must make sure that the instructions in a trampoline fill an entire
4111 cache line with identical instructions, or else ensure that the
4112 beginning of the trampoline code is always aligned at the same point in
4113 its cache line. Look in @file{m68k.h} as a guide.
4116 @findex TRANSFER_FROM_TRAMPOLINE
4117 @item TRANSFER_FROM_TRAMPOLINE
4118 Define this macro if trampolines need a special subroutine to do their
4119 work. The macro should expand to a series of @code{asm} statements
4120 which will be compiled with GCC. They go in a library function named
4121 @code{__transfer_from_trampoline}.
4123 If you need to avoid executing the ordinary prologue code of a compiled
4124 C function when you jump to the subroutine, you can do so by placing a
4125 special label of your own in the assembler code. Use one @code{asm}
4126 statement to generate an assembler label, and another to make the label
4127 global. Then trampolines can use that label to jump directly to your
4128 special assembler code.
4132 @section Implicit Calls to Library Routines
4133 @cindex library subroutine names
4134 @cindex @file{libgcc.a}
4136 @c prevent bad page break with this line
4137 Here is an explanation of implicit calls to library routines.
4140 @findex MULSI3_LIBCALL
4141 @item MULSI3_LIBCALL
4142 A C string constant giving the name of the function to call for
4143 multiplication of one signed full-word by another. If you do not
4144 define this macro, the default name is used, which is @code{__mulsi3},
4145 a function defined in @file{libgcc.a}.
4147 @findex DIVSI3_LIBCALL
4148 @item DIVSI3_LIBCALL
4149 A C string constant giving the name of the function to call for
4150 division of one signed full-word by another. If you do not define
4151 this macro, the default name is used, which is @code{__divsi3}, a
4152 function defined in @file{libgcc.a}.
4154 @findex UDIVSI3_LIBCALL
4155 @item UDIVSI3_LIBCALL
4156 A C string constant giving the name of the function to call for
4157 division of one unsigned full-word by another. If you do not define
4158 this macro, the default name is used, which is @code{__udivsi3}, a
4159 function defined in @file{libgcc.a}.
4161 @findex MODSI3_LIBCALL
4162 @item MODSI3_LIBCALL
4163 A C string constant giving the name of the function to call for the
4164 remainder in division of one signed full-word by another. If you do
4165 not define this macro, the default name is used, which is
4166 @code{__modsi3}, a function defined in @file{libgcc.a}.
4168 @findex UMODSI3_LIBCALL
4169 @item UMODSI3_LIBCALL
4170 A C string constant giving the name of the function to call for the
4171 remainder in division of one unsigned full-word by another. If you do
4172 not define this macro, the default name is used, which is
4173 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4175 @findex MULDI3_LIBCALL
4176 @item MULDI3_LIBCALL
4177 A C string constant giving the name of the function to call for
4178 multiplication of one signed double-word by another. If you do not
4179 define this macro, the default name is used, which is @code{__muldi3},
4180 a function defined in @file{libgcc.a}.
4182 @findex DIVDI3_LIBCALL
4183 @item DIVDI3_LIBCALL
4184 A C string constant giving the name of the function to call for
4185 division of one signed double-word by another. If you do not define
4186 this macro, the default name is used, which is @code{__divdi3}, a
4187 function defined in @file{libgcc.a}.
4189 @findex UDIVDI3_LIBCALL
4190 @item UDIVDI3_LIBCALL
4191 A C string constant giving the name of the function to call for
4192 division of one unsigned full-word by another. If you do not define
4193 this macro, the default name is used, which is @code{__udivdi3}, a
4194 function defined in @file{libgcc.a}.
4196 @findex MODDI3_LIBCALL
4197 @item MODDI3_LIBCALL
4198 A C string constant giving the name of the function to call for the
4199 remainder in division of one signed double-word by another. If you do
4200 not define this macro, the default name is used, which is
4201 @code{__moddi3}, a function defined in @file{libgcc.a}.
4203 @findex UMODDI3_LIBCALL
4204 @item UMODDI3_LIBCALL
4205 A C string constant giving the name of the function to call for the
4206 remainder in division of one unsigned full-word by another. If you do
4207 not define this macro, the default name is used, which is
4208 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4210 @findex INIT_TARGET_OPTABS
4211 @item INIT_TARGET_OPTABS
4212 Define this macro as a C statement that declares additional library
4213 routines renames existing ones. @code{init_optabs} calls this macro after
4214 initializing all the normal library routines.
4216 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4217 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4218 Define this macro as a C statement that returns nonzero if a call to
4219 the floating point comparison library function will return a boolean
4220 value that indicates the result of the comparison. It should return
4221 zero if one of gcc's own libgcc functions is called.
4223 Most ports don't need to define this macro.
4226 @cindex @code{EDOM}, implicit usage
4228 The value of @code{EDOM} on the target machine, as a C integer constant
4229 expression. If you don't define this macro, GCC does not attempt to
4230 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4231 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4234 If you do not define @code{TARGET_EDOM}, then compiled code reports
4235 domain errors by calling the library function and letting it report the
4236 error. If mathematical functions on your system use @code{matherr} when
4237 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4238 that @code{matherr} is used normally.
4240 @findex GEN_ERRNO_RTX
4241 @cindex @code{errno}, implicit usage
4243 Define this macro as a C expression to create an rtl expression that
4244 refers to the global ``variable'' @code{errno}. (On certain systems,
4245 @code{errno} may not actually be a variable.) If you don't define this
4246 macro, a reasonable default is used.
4248 @findex TARGET_MEM_FUNCTIONS
4249 @cindex @code{bcopy}, implicit usage
4250 @cindex @code{memcpy}, implicit usage
4251 @cindex @code{bzero}, implicit usage
4252 @cindex @code{memset}, implicit usage
4253 @item TARGET_MEM_FUNCTIONS
4254 Define this macro if GCC should generate calls to the System V
4255 (and ANSI C) library functions @code{memcpy} and @code{memset}
4256 rather than the BSD functions @code{bcopy} and @code{bzero}.
4258 @findex LIBGCC_NEEDS_DOUBLE
4259 @item LIBGCC_NEEDS_DOUBLE
4260 Define this macro if only @code{float} arguments cannot be passed to
4261 library routines (so they must be converted to @code{double}). This
4262 macro affects both how library calls are generated and how the library
4263 routines in @file{libgcc1.c} accept their arguments. It is useful on
4264 machines where floating and fixed point arguments are passed
4265 differently, such as the i860.
4267 @findex FLOAT_ARG_TYPE
4268 @item FLOAT_ARG_TYPE
4269 Define this macro to override the type used by the library routines to
4270 pick up arguments of type @code{float}. (By default, they use a union
4271 of @code{float} and @code{int}.)
4273 The obvious choice would be @code{float}---but that won't work with
4274 traditional C compilers that expect all arguments declared as @code{float}
4275 to arrive as @code{double}. To avoid this conversion, the library routines
4276 ask for the value as some other type and then treat it as a @code{float}.
4278 On some systems, no other type will work for this. For these systems,
4279 you must use @code{LIBGCC_NEEDS_DOUBLE} instead, to force conversion of
4280 the values @code{double} before they are passed.
4283 @item FLOATIFY (@var{passed-value})
4284 Define this macro to override the way library routines redesignate a
4285 @code{float} argument as a @code{float} instead of the type it was
4286 passed as. The default is an expression which takes the @code{float}
4289 @findex FLOAT_VALUE_TYPE
4290 @item FLOAT_VALUE_TYPE
4291 Define this macro to override the type used by the library routines to
4292 return values that ought to have type @code{float}. (By default, they
4295 The obvious choice would be @code{float}---but that won't work with
4296 traditional C compilers gratuitously convert values declared as
4297 @code{float} into @code{double}.
4300 @item INTIFY (@var{float-value})
4301 Define this macro to override the way the value of a
4302 @code{float}-returning library routine should be packaged in order to
4303 return it. These functions are actually declared to return type
4304 @code{FLOAT_VALUE_TYPE} (normally @code{int}).
4306 These values can't be returned as type @code{float} because traditional
4307 C compilers would gratuitously convert the value to a @code{double}.
4309 A local variable named @code{intify} is always available when the macro
4310 @code{INTIFY} is used. It is a union of a @code{float} field named
4311 @code{f} and a field named @code{i} whose type is
4312 @code{FLOAT_VALUE_TYPE} or @code{int}.
4314 If you don't define this macro, the default definition works by copying
4315 the value through that union.
4317 @findex nongcc_SI_type
4318 @item nongcc_SI_type
4319 Define this macro as the name of the data type corresponding to
4320 @code{SImode} in the system's own C compiler.
4322 You need not define this macro if that type is @code{long int}, as it usually
4325 @findex nongcc_word_type
4326 @item nongcc_word_type
4327 Define this macro as the name of the data type corresponding to the
4328 word_mode in the system's own C compiler.
4330 You need not define this macro if that type is @code{long int}, as it usually
4333 @findex perform_@dots{}
4334 @item perform_@dots{}
4335 Define these macros to supply explicit C statements to carry out various
4336 arithmetic operations on types @code{float} and @code{double} in the
4337 library routines in @file{libgcc1.c}. See that file for a full list
4338 of these macros and their arguments.
4340 On most machines, you don't need to define any of these macros, because
4341 the C compiler that comes with the system takes care of doing them.
4343 @findex NEXT_OBJC_RUNTIME
4344 @item NEXT_OBJC_RUNTIME
4345 Define this macro to generate code for Objective C message sending using
4346 the calling convention of the NeXT system. This calling convention
4347 involves passing the object, the selector and the method arguments all
4348 at once to the method-lookup library function.
4350 The default calling convention passes just the object and the selector
4351 to the lookup function, which returns a pointer to the method.
4354 @node Addressing Modes
4355 @section Addressing Modes
4356 @cindex addressing modes
4358 @c prevent bad page break with this line
4359 This is about addressing modes.
4362 @findex HAVE_PRE_INCREMENT
4363 @findex HAVE_PRE_DECREMENT
4364 @findex HAVE_POST_INCREMENT
4365 @findex HAVE_POST_DECREMENT
4366 @item HAVE_PRE_INCREMENT
4367 @itemx HAVE_PRE_DECREMENT
4368 @itemx HAVE_POST_INCREMENT
4369 @itemx HAVE_POST_DECREMENT
4370 A C expression that is non-zero if the machine supports pre-increment,
4371 pre-decrement, post-increment, or post-decrement addressing respectively.
4373 @findex HAVE_POST_MODIFY_DISP
4374 @findex HAVE_PRE_MODIFY_DISP
4375 @item HAVE_PRE_MODIFY_DISP
4376 @itemx HAVE_POST_MODIFY_DISP
4377 A C expression that is non-zero if the machine supports pre- or
4378 post-address side-effect generation involving constants other than
4379 the size of the memory operand.
4381 @findex HAVE_POST_MODIFY_REG
4382 @findex HAVE_PRE_MODIFY_REG
4383 @item HAVE_PRE_MODIFY_REG
4384 @itemx HAVE_POST_MODIFY_REG
4385 A C expression that is non-zero if the machine supports pre- or
4386 post-address side-effect generation involving a register displacement.
4388 @findex CONSTANT_ADDRESS_P
4389 @item CONSTANT_ADDRESS_P (@var{x})
4390 A C expression that is 1 if the RTX @var{x} is a constant which
4391 is a valid address. On most machines, this can be defined as
4392 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4393 in which constant addresses are supported.
4396 @code{CONSTANT_P} accepts integer-values expressions whose values are
4397 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4398 @code{high} expressions and @code{const} arithmetic expressions, in
4399 addition to @code{const_int} and @code{const_double} expressions.
4401 @findex MAX_REGS_PER_ADDRESS
4402 @item MAX_REGS_PER_ADDRESS
4403 A number, the maximum number of registers that can appear in a valid
4404 memory address. Note that it is up to you to specify a value equal to
4405 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4408 @findex GO_IF_LEGITIMATE_ADDRESS
4409 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4410 A C compound statement with a conditional @code{goto @var{label};}
4411 executed if @var{x} (an RTX) is a legitimate memory address on the
4412 target machine for a memory operand of mode @var{mode}.
4414 It usually pays to define several simpler macros to serve as
4415 subroutines for this one. Otherwise it may be too complicated to
4418 This macro must exist in two variants: a strict variant and a
4419 non-strict one. The strict variant is used in the reload pass. It
4420 must be defined so that any pseudo-register that has not been
4421 allocated a hard register is considered a memory reference. In
4422 contexts where some kind of register is required, a pseudo-register
4423 with no hard register must be rejected.
4425 The non-strict variant is used in other passes. It must be defined to
4426 accept all pseudo-registers in every context where some kind of
4427 register is required.
4429 @findex REG_OK_STRICT
4430 Compiler source files that want to use the strict variant of this
4431 macro define the macro @code{REG_OK_STRICT}. You should use an
4432 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4433 in that case and the non-strict variant otherwise.
4435 Subroutines to check for acceptable registers for various purposes (one
4436 for base registers, one for index registers, and so on) are typically
4437 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4438 Then only these subroutine macros need have two variants; the higher
4439 levels of macros may be the same whether strict or not.@refill
4441 Normally, constant addresses which are the sum of a @code{symbol_ref}
4442 and an integer are stored inside a @code{const} RTX to mark them as
4443 constant. Therefore, there is no need to recognize such sums
4444 specifically as legitimate addresses. Normally you would simply
4445 recognize any @code{const} as legitimate.
4447 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4448 sums that are not marked with @code{const}. It assumes that a naked
4449 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4450 naked constant sums as illegitimate addresses, so that none of them will
4451 be given to @code{PRINT_OPERAND_ADDRESS}.
4453 @cindex @code{ENCODE_SECTION_INFO} and address validation
4454 On some machines, whether a symbolic address is legitimate depends on
4455 the section that the address refers to. On these machines, define the
4456 macro @code{ENCODE_SECTION_INFO} to store the information into the
4457 @code{symbol_ref}, and then check for it here. When you see a
4458 @code{const}, you will have to look inside it to find the
4459 @code{symbol_ref} in order to determine the section. @xref{Assembler
4462 @findex saveable_obstack
4463 The best way to modify the name string is by adding text to the
4464 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4465 the new name in @code{saveable_obstack}. You will have to modify
4466 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4467 output the name accordingly, and define @code{STRIP_NAME_ENCODING} to
4468 access the original name string.
4470 You can check the information stored here into the @code{symbol_ref} in
4471 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4472 @code{PRINT_OPERAND_ADDRESS}.
4474 @findex REG_OK_FOR_BASE_P
4475 @item REG_OK_FOR_BASE_P (@var{x})
4476 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4477 RTX) is valid for use as a base register. For hard registers, it
4478 should always accept those which the hardware permits and reject the
4479 others. Whether the macro accepts or rejects pseudo registers must be
4480 controlled by @code{REG_OK_STRICT} as described above. This usually
4481 requires two variant definitions, of which @code{REG_OK_STRICT}
4482 controls the one actually used.
4484 @findex REG_MODE_OK_FOR_BASE_P
4485 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4486 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4487 that expression may examine the mode of the memory reference in
4488 @var{mode}. You should define this macro if the mode of the memory
4489 reference affects whether a register may be used as a base register. If
4490 you define this macro, the compiler will use it instead of
4491 @code{REG_OK_FOR_BASE_P}.
4493 @findex REG_OK_FOR_INDEX_P
4494 @item REG_OK_FOR_INDEX_P (@var{x})
4495 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4496 RTX) is valid for use as an index register.
4498 The difference between an index register and a base register is that
4499 the index register may be scaled. If an address involves the sum of
4500 two registers, neither one of them scaled, then either one may be
4501 labeled the ``base'' and the other the ``index''; but whichever
4502 labeling is used must fit the machine's constraints of which registers
4503 may serve in each capacity. The compiler will try both labelings,
4504 looking for one that is valid, and will reload one or both registers
4505 only if neither labeling works.
4507 @findex FIND_BASE_TERM
4508 @item FIND_BASE_TERM (@var{x})
4509 A C expression to determine the base term of address @var{x}.
4510 This macro is used in only one place: `find_base_term' in alias.c.
4512 It is always safe for this macro to not be defined. It exists so
4513 that alias analysis can understand machine-dependent addresses.
4515 The typical use of this macro is to handle addresses containing
4516 a label_ref or symbol_ref within an UNSPEC.
4518 @findex LEGITIMIZE_ADDRESS
4519 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4520 A C compound statement that attempts to replace @var{x} with a valid
4521 memory address for an operand of mode @var{mode}. @var{win} will be a
4522 C statement label elsewhere in the code; the macro definition may use
4525 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4529 to avoid further processing if the address has become legitimate.
4531 @findex break_out_memory_refs
4532 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4533 and @var{oldx} will be the operand that was given to that function to produce
4536 The code generated by this macro should not alter the substructure of
4537 @var{x}. If it transforms @var{x} into a more legitimate form, it
4538 should assign @var{x} (which will always be a C variable) a new value.
4540 It is not necessary for this macro to come up with a legitimate
4541 address. The compiler has standard ways of doing so in all cases. In
4542 fact, it is safe for this macro to do nothing. But often a
4543 machine-dependent strategy can generate better code.
4545 @findex LEGITIMIZE_RELOAD_ADDRESS
4546 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4547 A C compound statement that attempts to replace @var{x}, which is an address
4548 that needs reloading, with a valid memory address for an operand of mode
4549 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4550 It is not necessary to define this macro, but it might be useful for
4551 performance reasons.
4553 For example, on the i386, it is sometimes possible to use a single
4554 reload register instead of two by reloading a sum of two pseudo
4555 registers into a register. On the other hand, for number of RISC
4556 processors offsets are limited so that often an intermediate address
4557 needs to be generated in order to address a stack slot. By defining
4558 LEGITIMIZE_RELOAD_ADDRESS appropriately, the intermediate addresses
4559 generated for adjacent some stack slots can be made identical, and thus
4562 @emph{Note}: This macro should be used with caution. It is necessary
4563 to know something of how reload works in order to effectively use this,
4564 and it is quite easy to produce macros that build in too much knowledge
4565 of reload internals.
4567 @emph{Note}: This macro must be able to reload an address created by a
4568 previous invocation of this macro. If it fails to handle such addresses
4569 then the compiler may generate incorrect code or abort.
4572 The macro definition should use @code{push_reload} to indicate parts that
4573 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4574 suitable to be passed unaltered to @code{push_reload}.
4576 The code generated by this macro must not alter the substructure of
4577 @var{x}. If it transforms @var{x} into a more legitimate form, it
4578 should assign @var{x} (which will always be a C variable) a new value.
4579 This also applies to parts that you change indirectly by calling
4582 @findex strict_memory_address_p
4583 The macro definition may use @code{strict_memory_address_p} to test if
4584 the address has become legitimate.
4587 If you want to change only a part of @var{x}, one standard way of doing
4588 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4589 single level of rtl. Thus, if the part to be changed is not at the
4590 top level, you'll need to replace first the top leve
4591 It is not necessary for this macro to come up with a legitimate
4592 address; but often a machine-dependent strategy can generate better code.
4594 @findex GO_IF_MODE_DEPENDENT_ADDRESS
4595 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4596 A C statement or compound statement with a conditional @code{goto
4597 @var{label};} executed if memory address @var{x} (an RTX) can have
4598 different meanings depending on the machine mode of the memory
4599 reference it is used for or if the address is valid for some modes
4602 Autoincrement and autodecrement addresses typically have mode-dependent
4603 effects because the amount of the increment or decrement is the size
4604 of the operand being addressed. Some machines have other mode-dependent
4605 addresses. Many RISC machines have no mode-dependent addresses.
4607 You may assume that @var{addr} is a valid address for the machine.
4609 @findex LEGITIMATE_CONSTANT_P
4610 @item LEGITIMATE_CONSTANT_P (@var{x})
4611 A C expression that is nonzero if @var{x} is a legitimate constant for
4612 an immediate operand on the target machine. You can assume that
4613 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4614 @samp{1} is a suitable definition for this macro on machines where
4615 anything @code{CONSTANT_P} is valid.@refill
4618 @node Condition Code
4619 @section Condition Code Status
4620 @cindex condition code status
4622 @c prevent bad page break with this line
4623 This describes the condition code status.
4626 The file @file{conditions.h} defines a variable @code{cc_status} to
4627 describe how the condition code was computed (in case the interpretation of
4628 the condition code depends on the instruction that it was set by). This
4629 variable contains the RTL expressions on which the condition code is
4630 currently based, and several standard flags.
4632 Sometimes additional machine-specific flags must be defined in the machine
4633 description header file. It can also add additional machine-specific
4634 information by defining @code{CC_STATUS_MDEP}.
4637 @findex CC_STATUS_MDEP
4638 @item CC_STATUS_MDEP
4639 C code for a data type which is used for declaring the @code{mdep}
4640 component of @code{cc_status}. It defaults to @code{int}.
4642 This macro is not used on machines that do not use @code{cc0}.
4644 @findex CC_STATUS_MDEP_INIT
4645 @item CC_STATUS_MDEP_INIT
4646 A C expression to initialize the @code{mdep} field to ``empty''.
4647 The default definition does nothing, since most machines don't use
4648 the field anyway. If you want to use the field, you should probably
4649 define this macro to initialize it.
4651 This macro is not used on machines that do not use @code{cc0}.
4653 @findex NOTICE_UPDATE_CC
4654 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4655 A C compound statement to set the components of @code{cc_status}
4656 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4657 this macro's responsibility to recognize insns that set the condition
4658 code as a byproduct of other activity as well as those that explicitly
4661 This macro is not used on machines that do not use @code{cc0}.
4663 If there are insns that do not set the condition code but do alter
4664 other machine registers, this macro must check to see whether they
4665 invalidate the expressions that the condition code is recorded as
4666 reflecting. For example, on the 68000, insns that store in address
4667 registers do not set the condition code, which means that usually
4668 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4669 insns. But suppose that the previous insn set the condition code
4670 based on location @samp{a4@@(102)} and the current insn stores a new
4671 value in @samp{a4}. Although the condition code is not changed by
4672 this, it will no longer be true that it reflects the contents of
4673 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4674 @code{cc_status} in this case to say that nothing is known about the
4675 condition code value.
4677 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4678 with the results of peephole optimization: insns whose patterns are
4679 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4680 constants which are just the operands. The RTL structure of these
4681 insns is not sufficient to indicate what the insns actually do. What
4682 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4683 @code{CC_STATUS_INIT}.
4685 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4686 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4687 @samp{cc}. This avoids having detailed information about patterns in
4688 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4690 @findex EXTRA_CC_MODES
4691 @item EXTRA_CC_MODES
4692 A list of additional modes for condition code values in registers
4693 (@pxref{Jump Patterns}). This macro should expand to a sequence of
4694 calls of the macro @code{CC} separated by white space. @code{CC} takes
4695 two arguments. The first is the enumeration name of the mode, which
4696 should begin with @samp{CC} and end with @samp{mode}. The second is a C
4697 string giving the printable name of the mode; it should be the same as
4698 the first argument, but with the trailing @samp{mode} removed.
4700 You should only define this macro if additional modes are required.
4702 A sample definition of @code{EXTRA_CC_MODES} is:
4704 #define EXTRA_CC_MODES \
4705 CC(CC_NOOVmode, "CC_NOOV") \
4706 CC(CCFPmode, "CCFP") \
4707 CC(CCFPEmode, "CCFPE")
4710 @findex SELECT_CC_MODE
4711 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4712 Returns a mode from class @code{MODE_CC} to be used when comparison
4713 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
4714 example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
4715 @pxref{Jump Patterns} for a description of the reason for this
4719 #define SELECT_CC_MODE(OP,X,Y) \
4720 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4721 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
4722 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4723 || GET_CODE (X) == NEG) \
4724 ? CC_NOOVmode : CCmode))
4727 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
4729 @findex CANONICALIZE_COMPARISON
4730 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
4731 On some machines not all possible comparisons are defined, but you can
4732 convert an invalid comparison into a valid one. For example, the Alpha
4733 does not have a @code{GT} comparison, but you can use an @code{LT}
4734 comparison instead and swap the order of the operands.
4736 On such machines, define this macro to be a C statement to do any
4737 required conversions. @var{code} is the initial comparison code
4738 and @var{op0} and @var{op1} are the left and right operands of the
4739 comparison, respectively. You should modify @var{code}, @var{op0}, and
4740 @var{op1} as required.
4742 GCC will not assume that the comparison resulting from this macro is
4743 valid but will see if the resulting insn matches a pattern in the
4746 You need not define this macro if it would never change the comparison
4749 @findex REVERSIBLE_CC_MODE
4750 @item REVERSIBLE_CC_MODE (@var{mode})
4751 A C expression whose value is one if it is always safe to reverse a
4752 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4753 can ever return @var{mode} for a floating-point inequality comparison,
4754 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4756 You need not define this macro if it would always returns zero or if the
4757 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4758 For example, here is the definition used on the Sparc, where floating-point
4759 inequality comparisons are always given @code{CCFPEmode}:
4762 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
4768 @section Describing Relative Costs of Operations
4769 @cindex costs of instructions
4770 @cindex relative costs
4771 @cindex speed of instructions
4773 These macros let you describe the relative speed of various operations
4774 on the target machine.
4778 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
4779 A part of a C @code{switch} statement that describes the relative costs
4780 of constant RTL expressions. It must contain @code{case} labels for
4781 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
4782 @code{label_ref} and @code{const_double}. Each case must ultimately
4783 reach a @code{return} statement to return the relative cost of the use
4784 of that kind of constant value in an expression. The cost may depend on
4785 the precise value of the constant, which is available for examination in
4786 @var{x}, and the rtx code of the expression in which it is contained,
4787 found in @var{outer_code}.
4789 @var{code} is the expression code---redundant, since it can be
4790 obtained with @code{GET_CODE (@var{x})}.
4793 @findex COSTS_N_INSNS
4794 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4795 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
4796 This can be used, for example, to indicate how costly a multiply
4797 instruction is. In writing this macro, you can use the construct
4798 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
4799 instructions. @var{outer_code} is the code of the expression in which
4800 @var{x} is contained.
4802 This macro is optional; do not define it if the default cost assumptions
4803 are adequate for the target machine.
4805 @findex DEFAULT_RTX_COSTS
4806 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
4807 This macro, if defined, is called for any case not handled by the
4808 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
4809 to put case labels into the macro, but the code, or any functions it
4810 calls, must assume that the RTL in @var{x} could be of any type that has
4811 not already been handled. The arguments are the same as for
4812 @code{RTX_COSTS}, and the macro should execute a return statement giving
4813 the cost of any RTL expressions that it can handle. The default cost
4814 calculation is used for any RTL for which this macro does not return a
4817 This macro is optional; do not define it if the default cost assumptions
4818 are adequate for the target machine.
4820 @findex ADDRESS_COST
4821 @item ADDRESS_COST (@var{address})
4822 An expression giving the cost of an addressing mode that contains
4823 @var{address}. If not defined, the cost is computed from
4824 the @var{address} expression and the @code{CONST_COSTS} values.
4826 For most CISC machines, the default cost is a good approximation of the
4827 true cost of the addressing mode. However, on RISC machines, all
4828 instructions normally have the same length and execution time. Hence
4829 all addresses will have equal costs.
4831 In cases where more than one form of an address is known, the form with
4832 the lowest cost will be used. If multiple forms have the same, lowest,
4833 cost, the one that is the most complex will be used.
4835 For example, suppose an address that is equal to the sum of a register
4836 and a constant is used twice in the same basic block. When this macro
4837 is not defined, the address will be computed in a register and memory
4838 references will be indirect through that register. On machines where
4839 the cost of the addressing mode containing the sum is no higher than
4840 that of a simple indirect reference, this will produce an additional
4841 instruction and possibly require an additional register. Proper
4842 specification of this macro eliminates this overhead for such machines.
4844 Similar use of this macro is made in strength reduction of loops.
4846 @var{address} need not be valid as an address. In such a case, the cost
4847 is not relevant and can be any value; invalid addresses need not be
4848 assigned a different cost.
4850 On machines where an address involving more than one register is as
4851 cheap as an address computation involving only one register, defining
4852 @code{ADDRESS_COST} to reflect this can cause two registers to be live
4853 over a region of code where only one would have been if
4854 @code{ADDRESS_COST} were not defined in that manner. This effect should
4855 be considered in the definition of this macro. Equivalent costs should
4856 probably only be given to addresses with different numbers of registers
4857 on machines with lots of registers.
4859 This macro will normally either not be defined or be defined as a
4862 @findex REGISTER_MOVE_COST
4863 @item REGISTER_MOVE_COST (@var{from}, @var{to})
4864 A C expression for the cost of moving data from a register in class
4865 @var{from} to one in class @var{to}. The classes are expressed using
4866 the enumeration values such as @code{GENERAL_REGS}. A value of 2 is the
4867 default; other values are interpreted relative to that.
4869 It is not required that the cost always equal 2 when @var{from} is the
4870 same as @var{to}; on some machines it is expensive to move between
4871 registers if they are not general registers.
4873 If reload sees an insn consisting of a single @code{set} between two
4874 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4875 classes returns a value of 2, reload does not check to ensure that the
4876 constraints of the insn are met. Setting a cost of other than 2 will
4877 allow reload to verify that the constraints are met. You should do this
4878 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4880 @findex MEMORY_MOVE_COST
4881 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4882 A C expression for the cost of moving data of mode @var{mode} between a
4883 register of class @var{class} and memory; @var{in} is zero if the value
4884 is to be written to memory, non-zero if it is to be read in. This cost
4885 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
4886 registers and memory is more expensive than between two registers, you
4887 should define this macro to express the relative cost.
4889 If you do not define this macro, GCC uses a default cost of 4 plus
4890 the cost of copying via a secondary reload register, if one is
4891 needed. If your machine requires a secondary reload register to copy
4892 between memory and a register of @var{class} but the reload mechanism is
4893 more complex than copying via an intermediate, define this macro to
4894 reflect the actual cost of the move.
4896 GCC defines the function @code{memory_move_secondary_cost} if
4897 secondary reloads are needed. It computes the costs due to copying via
4898 a secondary register. If your machine copies from memory using a
4899 secondary register in the conventional way but the default base value of
4900 4 is not correct for your machine, define this macro to add some other
4901 value to the result of that function. The arguments to that function
4902 are the same as to this macro.
4906 A C expression for the cost of a branch instruction. A value of 1 is
4907 the default; other values are interpreted relative to that.
4910 Here are additional macros which do not specify precise relative costs,
4911 but only that certain actions are more expensive than GCC would
4915 @findex SLOW_BYTE_ACCESS
4916 @item SLOW_BYTE_ACCESS
4917 Define this macro as a C expression which is nonzero if accessing less
4918 than a word of memory (i.e. a @code{char} or a @code{short}) is no
4919 faster than accessing a word of memory, i.e., if such access
4920 require more than one instruction or if there is no difference in cost
4921 between byte and (aligned) word loads.
4923 When this macro is not defined, the compiler will access a field by
4924 finding the smallest containing object; when it is defined, a fullword
4925 load will be used if alignment permits. Unless bytes accesses are
4926 faster than word accesses, using word accesses is preferable since it
4927 may eliminate subsequent memory access if subsequent accesses occur to
4928 other fields in the same word of the structure, but to different bytes.
4930 @findex SLOW_ZERO_EXTEND
4931 @item SLOW_ZERO_EXTEND
4932 Define this macro if zero-extension (of a @code{char} or @code{short}
4933 to an @code{int}) can be done faster if the destination is a register
4934 that is known to be zero.
4936 If you define this macro, you must have instruction patterns that
4937 recognize RTL structures like this:
4940 (set (strict_low_part (subreg:QI (reg:SI @dots{}) 0)) @dots{})
4944 and likewise for @code{HImode}.
4946 @findex SLOW_UNALIGNED_ACCESS
4947 @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
4948 Define this macro to be the value 1 if memory accesses described by the
4949 @var{mode} and @var{alignment} parameters have a cost many times greater
4950 than aligned accesses, for example if they are emulated in a trap
4953 When this macro is non-zero, the compiler will act as if
4954 @code{STRICT_ALIGNMENT} were non-zero when generating code for block
4955 moves. This can cause significantly more instructions to be produced.
4956 Therefore, do not set this macro non-zero if unaligned accesses only add a
4957 cycle or two to the time for a memory access.
4959 If the value of this macro is always zero, it need not be defined. If
4960 this macro is defined, it should produce a non-zero value when
4961 @code{STRICT_ALIGNMENT} is non-zero.
4963 @findex DONT_REDUCE_ADDR
4964 @item DONT_REDUCE_ADDR
4965 Define this macro to inhibit strength reduction of memory addresses.
4966 (On some machines, such strength reduction seems to do harm rather
4971 The threshold of number of scalar memory-to-memory move insns, @emph{below}
4972 which a sequence of insns should be generated instead of a
4973 string move insn or a library call. Increasing the value will always
4974 make code faster, but eventually incurs high cost in increased code size.
4976 Note that on machines where the corresponding move insn is a
4977 @code{define_expand} that emits a sequence of insns, this macro counts
4978 the number of such sequences.
4980 If you don't define this, a reasonable default is used.
4982 @findex MOVE_BY_PIECES_P
4983 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
4984 A C expression used to determine whether @code{move_by_pieces} will be used to
4985 copy a chunk of memory, or whether some other block move mechanism
4986 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
4987 than @code{MOVE_RATIO}.
4989 @findex MOVE_MAX_PIECES
4990 @item MOVE_MAX_PIECES
4991 A C expression used by @code{move_by_pieces} to determine the largest unit
4992 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
4994 @findex USE_LOAD_POST_INCREMENT
4995 @item USE_LOAD_POST_INCREMENT (@var{mode})
4996 A C expression used to determine whether a load postincrement is a good
4997 thing to use for a given mode. Defaults to the value of
4998 @code{HAVE_POST_INCREMENT}.
5000 @findex USE_LOAD_POST_DECREMENT
5001 @item USE_LOAD_POST_DECREMENT (@var{mode})
5002 A C expression used to determine whether a load postdecrement is a good
5003 thing to use for a given mode. Defaults to the value of
5004 @code{HAVE_POST_DECREMENT}.
5006 @findex USE_LOAD_PRE_INCREMENT
5007 @item USE_LOAD_PRE_INCREMENT (@var{mode})
5008 A C expression used to determine whether a load preincrement is a good
5009 thing to use for a given mode. Defaults to the value of
5010 @code{HAVE_PRE_INCREMENT}.
5012 @findex USE_LOAD_PRE_DECREMENT
5013 @item USE_LOAD_PRE_DECREMENT (@var{mode})
5014 A C expression used to determine whether a load predecrement is a good
5015 thing to use for a given mode. Defaults to the value of
5016 @code{HAVE_PRE_DECREMENT}.
5018 @findex USE_STORE_POST_INCREMENT
5019 @item USE_STORE_POST_INCREMENT (@var{mode})
5020 A C expression used to determine whether a store postincrement is a good
5021 thing to use for a given mode. Defaults to the value of
5022 @code{HAVE_POST_INCREMENT}.
5024 @findex USE_STORE_POST_DECREMENT
5025 @item USE_STORE_POST_DECREMENT (@var{mode})
5026 A C expression used to determine whether a store postdeccrement is a good
5027 thing to use for a given mode. Defaults to the value of
5028 @code{HAVE_POST_DECREMENT}.
5030 @findex USE_STORE_PRE_INCREMENT
5031 @item USE_STORE_PRE_INCREMENT (@var{mode})
5032 This macro is used to determine whether a store preincrement is a good
5033 thing to use for a given mode. Defaults to the value of
5034 @code{HAVE_PRE_INCREMENT}.
5036 @findex USE_STORE_PRE_DECREMENT
5037 @item USE_STORE_PRE_DECREMENT (@var{mode})
5038 This macro is used to determine whether a store predecrement is a good
5039 thing to use for a given mode. Defaults to the value of
5040 @code{HAVE_PRE_DECREMENT}.
5042 @findex NO_FUNCTION_CSE
5043 @item NO_FUNCTION_CSE
5044 Define this macro if it is as good or better to call a constant
5045 function address than to call an address kept in a register.
5047 @findex NO_RECURSIVE_FUNCTION_CSE
5048 @item NO_RECURSIVE_FUNCTION_CSE
5049 Define this macro if it is as good or better for a function to call
5050 itself with an explicit address than to call an address kept in a
5054 @item ADJUST_COST (@var{insn}, @var{link}, @var{dep_insn}, @var{cost})
5055 A C statement (sans semicolon) to update the integer variable @var{cost}
5056 based on the relationship between @var{insn} that is dependent on
5057 @var{dep_insn} through the dependence @var{link}. The default is to
5058 make no adjustment to @var{cost}. This can be used for example to
5059 specify to the scheduler that an output- or anti-dependence does not
5060 incur the same cost as a data-dependence.
5062 @findex ADJUST_PRIORITY
5063 @item ADJUST_PRIORITY (@var{insn})
5064 A C statement (sans semicolon) to update the integer scheduling
5065 priority @code{INSN_PRIORITY(@var{insn})}. Reduce the priority
5066 to execute the @var{insn} earlier, increase the priority to execute
5067 @var{insn} later. Do not define this macro if you do not need to
5068 adjust the scheduling priorities of insns.
5072 @section Dividing the Output into Sections (Texts, Data, @dots{})
5073 @c the above section title is WAY too long. maybe cut the part between
5074 @c the (...)? --mew 10feb93
5076 An object file is divided into sections containing different types of
5077 data. In the most common case, there are three sections: the @dfn{text
5078 section}, which holds instructions and read-only data; the @dfn{data
5079 section}, which holds initialized writable data; and the @dfn{bss
5080 section}, which holds uninitialized data. Some systems have other kinds
5083 The compiler must tell the assembler when to switch sections. These
5084 macros control what commands to output to tell the assembler this. You
5085 can also define additional sections.
5088 @findex TEXT_SECTION_ASM_OP
5089 @item TEXT_SECTION_ASM_OP
5090 A C expression whose value is a string, including spacing, containing the
5091 assembler operation that should precede instructions and read-only data.
5092 Normally @code{"\t.text"} is right.
5094 @findex DATA_SECTION_ASM_OP
5095 @item DATA_SECTION_ASM_OP
5096 A C expression whose value is a string, including spacing, containing the
5097 assembler operation to identify the following data as writable initialized
5098 data. Normally @code{"\t.data"} is right.
5100 @findex SHARED_SECTION_ASM_OP
5101 @item SHARED_SECTION_ASM_OP
5102 If defined, a C expression whose value is a string, including spacing,
5103 containing the assembler operation to identify the following data as
5104 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5106 @findex BSS_SECTION_ASM_OP
5107 @item BSS_SECTION_ASM_OP
5108 If defined, a C expression whose value is a string, including spacing,
5109 containing the assembler operation to identify the following data as
5110 uninitialized global data. If not defined, and neither
5111 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5112 uninitialized global data will be output in the data section if
5113 @samp{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5116 @findex SHARED_BSS_SECTION_ASM_OP
5117 @item SHARED_BSS_SECTION_ASM_OP
5118 If defined, a C expression whose value is a string, including spacing,
5119 containing the assembler operation to identify the following data as
5120 uninitialized global shared data. If not defined, and
5121 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5123 @findex INIT_SECTION_ASM_OP
5124 @item INIT_SECTION_ASM_OP
5125 If defined, a C expression whose value is a string, including spacing,
5126 containing the assembler operation to identify the following data as
5127 initialization code. If not defined, GCC will assume such a section does
5130 @findex FINI_SECTION_ASM_OP
5131 @item FINI_SECTION_ASM_OP
5132 If defined, a C expression whose value is a string, including spacing,
5133 containing the assembler operation to identify the following data as
5134 finalization code. If not defined, GCC will assume such a section does
5137 @findex CRT_CALL_STATIC_FUNCTION
5138 @item CRT_CALL_STATIC_FUNCTION
5139 If defined, a C statement that calls the function named as the sole
5140 argument of this macro. This is used in @file{crtstuff.c} if
5141 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls to
5142 initialization and finalization functions from the init and fini
5143 sections. By default, this macro is a simple function call. Some
5144 ports need hand-crafted assembly code to avoid dependencies on
5145 registers initialized in the function prologue or to ensure that
5146 constant pools don't end up too far way in the text section.
5148 @findex EXTRA_SECTIONS
5151 @item EXTRA_SECTIONS
5152 A list of names for sections other than the standard two, which are
5153 @code{in_text} and @code{in_data}. You need not define this macro
5154 on a system with no other sections (that GCC needs to use).
5156 @findex EXTRA_SECTION_FUNCTIONS
5157 @findex text_section
5158 @findex data_section
5159 @item EXTRA_SECTION_FUNCTIONS
5160 One or more functions to be defined in @file{varasm.c}. These
5161 functions should do jobs analogous to those of @code{text_section} and
5162 @code{data_section}, for your additional sections. Do not define this
5163 macro if you do not define @code{EXTRA_SECTIONS}.
5165 @findex READONLY_DATA_SECTION
5166 @item READONLY_DATA_SECTION
5167 On most machines, read-only variables, constants, and jump tables are
5168 placed in the text section. If this is not the case on your machine,
5169 this macro should be defined to be the name of a function (either
5170 @code{data_section} or a function defined in @code{EXTRA_SECTIONS}) that
5171 switches to the section to be used for read-only items.
5173 If these items should be placed in the text section, this macro should
5176 @findex SELECT_SECTION
5177 @item SELECT_SECTION (@var{exp}, @var{reloc})
5178 A C statement or statements to switch to the appropriate section for
5179 output of @var{exp}. You can assume that @var{exp} is either a
5180 @code{VAR_DECL} node or a constant of some sort. @var{reloc}
5181 indicates whether the initial value of @var{exp} requires link-time
5182 relocations. Select the section by calling @code{text_section} or one
5183 of the alternatives for other sections.
5185 Do not define this macro if you put all read-only variables and
5186 constants in the read-only data section (usually the text section).
5188 @findex SELECT_RTX_SECTION
5189 @item SELECT_RTX_SECTION (@var{mode}, @var{rtx})
5190 A C statement or statements to switch to the appropriate section for
5191 output of @var{rtx} in mode @var{mode}. You can assume that @var{rtx}
5192 is some kind of constant in RTL. The argument @var{mode} is redundant
5193 except in the case of a @code{const_int} rtx. Select the section by
5194 calling @code{text_section} or one of the alternatives for other
5197 Do not define this macro if you put all constants in the read-only
5200 @findex JUMP_TABLES_IN_TEXT_SECTION
5201 @item JUMP_TABLES_IN_TEXT_SECTION
5202 Define this macro to be an expression with a non-zero value if jump
5203 tables (for @code{tablejump} insns) should be output in the text
5204 section, along with the assembler instructions. Otherwise, the
5205 readonly data section is used.
5207 This macro is irrelevant if there is no separate readonly data section.
5209 @findex ENCODE_SECTION_INFO
5210 @item ENCODE_SECTION_INFO (@var{decl})
5211 Define this macro if references to a symbol must be treated differently
5212 depending on something about the variable or function named by the
5213 symbol (such as what section it is in).
5215 The macro definition, if any, is executed immediately after the rtl for
5216 @var{decl} has been created and stored in @code{DECL_RTL (@var{decl})}.
5217 The value of the rtl will be a @code{mem} whose address is a
5220 @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO}
5221 The usual thing for this macro to do is to record a flag in the
5222 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
5223 modified name string in the @code{symbol_ref} (if one bit is not enough
5226 @findex STRIP_NAME_ENCODING
5227 @item STRIP_NAME_ENCODING (@var{var}, @var{sym_name})
5228 Decode @var{sym_name} and store the real name part in @var{var}, sans
5229 the characters that encode section info. Define this macro if
5230 @code{ENCODE_SECTION_INFO} alters the symbol's name string.
5232 @findex UNIQUE_SECTION_P
5233 @item UNIQUE_SECTION_P (@var{decl})
5234 A C expression which evaluates to true if @var{decl} should be placed
5235 into a unique section for some target-specific reason. If you do not
5236 define this macro, the default is @samp{0}. Note that the flag
5237 @samp{-ffunction-sections} will also cause functions to be placed into
5240 @findex UNIQUE_SECTION
5241 @item UNIQUE_SECTION (@var{decl}, @var{reloc})
5242 A C statement to build up a unique section name, expressed as a
5243 STRING_CST node, and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5244 @var{reloc} indicates whether the initial value of @var{exp} requires
5245 link-time relocations. If you do not define this macro, GCC will use
5246 the symbol name prefixed by @samp{.} as the section name. Note - this
5247 macro can now be called for unitialised data items as well as
5248 initialised data and functions.
5252 @section Position Independent Code
5253 @cindex position independent code
5256 This section describes macros that help implement generation of position
5257 independent code. Simply defining these macros is not enough to
5258 generate valid PIC; you must also add support to the macros
5259 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5260 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5261 @samp{movsi} to do something appropriate when the source operand
5262 contains a symbolic address. You may also need to alter the handling of
5263 switch statements so that they use relative addresses.
5264 @c i rearranged the order of the macros above to try to force one of
5265 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5268 @findex PIC_OFFSET_TABLE_REGNUM
5269 @item PIC_OFFSET_TABLE_REGNUM
5270 The register number of the register used to address a table of static
5271 data addresses in memory. In some cases this register is defined by a
5272 processor's ``application binary interface'' (ABI). When this macro
5273 is defined, RTL is generated for this register once, as with the stack
5274 pointer and frame pointer registers. If this macro is not defined, it
5275 is up to the machine-dependent files to allocate such a register (if
5278 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5279 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5280 Define this macro if the register defined by
5281 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
5282 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5284 @findex FINALIZE_PIC
5286 By generating position-independent code, when two different programs (A
5287 and B) share a common library (libC.a), the text of the library can be
5288 shared whether or not the library is linked at the same address for both
5289 programs. In some of these environments, position-independent code
5290 requires not only the use of different addressing modes, but also
5291 special code to enable the use of these addressing modes.
5293 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
5294 codes once the function is being compiled into assembly code, but not
5295 before. (It is not done before, because in the case of compiling an
5296 inline function, it would lead to multiple PIC prologues being
5297 included in functions which used inline functions and were compiled to
5300 @findex LEGITIMATE_PIC_OPERAND_P
5301 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
5302 A C expression that is nonzero if @var{x} is a legitimate immediate
5303 operand on the target machine when generating position independent code.
5304 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5305 check this. You can also assume @var{flag_pic} is true, so you need not
5306 check it either. You need not define this macro if all constants
5307 (including @code{SYMBOL_REF}) can be immediate operands when generating
5308 position independent code.
5311 @node Assembler Format
5312 @section Defining the Output Assembler Language
5314 This section describes macros whose principal purpose is to describe how
5315 to write instructions in assembler language--rather than what the
5319 * File Framework:: Structural information for the assembler file.
5320 * Data Output:: Output of constants (numbers, strings, addresses).
5321 * Uninitialized Data:: Output of uninitialized variables.
5322 * Label Output:: Output and generation of labels.
5323 * Initialization:: General principles of initialization
5324 and termination routines.
5325 * Macros for Initialization::
5326 Specific macros that control the handling of
5327 initialization and termination routines.
5328 * Instruction Output:: Output of actual instructions.
5329 * Dispatch Tables:: Output of jump tables.
5330 * Exception Region Output:: Output of exception region code.
5331 * Alignment Output:: Pseudo ops for alignment and skipping data.
5334 @node File Framework
5335 @subsection The Overall Framework of an Assembler File
5336 @cindex assembler format
5337 @cindex output of assembler code
5339 @c prevent bad page break with this line
5340 This describes the overall framework of an assembler file.
5343 @findex ASM_FILE_START
5344 @item ASM_FILE_START (@var{stream})
5345 A C expression which outputs to the stdio stream @var{stream}
5346 some appropriate text to go at the start of an assembler file.
5348 Normally this macro is defined to output a line containing
5349 @samp{#NO_APP}, which is a comment that has no effect on most
5350 assemblers but tells the GNU assembler that it can save time by not
5351 checking for certain assembler constructs.
5353 On systems that use SDB, it is necessary to output certain commands;
5354 see @file{attasm.h}.
5356 @findex ASM_FILE_END
5357 @item ASM_FILE_END (@var{stream})
5358 A C expression which outputs to the stdio stream @var{stream}
5359 some appropriate text to go at the end of an assembler file.
5361 If this macro is not defined, the default is to output nothing
5362 special at the end of the file. Most systems don't require any
5365 On systems that use SDB, it is necessary to output certain commands;
5366 see @file{attasm.h}.
5368 @findex ASM_IDENTIFY_GCC
5369 @item ASM_IDENTIFY_GCC (@var{file})
5370 A C statement to output assembler commands which will identify
5371 the object file as having been compiled with GCC (or another
5374 If you don't define this macro, the string @samp{gcc_compiled.:}
5375 is output. This string is calculated to define a symbol which,
5376 on BSD systems, will never be defined for any other reason.
5377 GDB checks for the presence of this symbol when reading the
5378 symbol table of an executable.
5380 On non-BSD systems, you must arrange communication with GDB in
5381 some other fashion. If GDB is not used on your system, you can
5382 define this macro with an empty body.
5384 @findex ASM_COMMENT_START
5385 @item ASM_COMMENT_START
5386 A C string constant describing how to begin a comment in the target
5387 assembler language. The compiler assumes that the comment will end at
5388 the end of the line.
5392 A C string constant for text to be output before each @code{asm}
5393 statement or group of consecutive ones. Normally this is
5394 @code{"#APP"}, which is a comment that has no effect on most
5395 assemblers but tells the GNU assembler that it must check the lines
5396 that follow for all valid assembler constructs.
5400 A C string constant for text to be output after each @code{asm}
5401 statement or group of consecutive ones. Normally this is
5402 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5403 time-saving assumptions that are valid for ordinary compiler output.
5405 @findex ASM_OUTPUT_SOURCE_FILENAME
5406 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5407 A C statement to output COFF information or DWARF debugging information
5408 which indicates that filename @var{name} is the current source file to
5409 the stdio stream @var{stream}.
5411 This macro need not be defined if the standard form of output
5412 for the file format in use is appropriate.
5414 @findex OUTPUT_QUOTED_STRING
5415 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5416 A C statement to output the string @var{string} to the stdio stream
5417 @var{stream}. If you do not call the function @code{output_quoted_string}
5418 in your config files, GCC will only call it to output filenames to
5419 the assembler source. So you can use it to canonicalize the format
5420 of the filename using this macro.
5422 @findex ASM_OUTPUT_SOURCE_LINE
5423 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
5424 A C statement to output DBX or SDB debugging information before code
5425 for line number @var{line} of the current source file to the
5426 stdio stream @var{stream}.
5428 This macro need not be defined if the standard form of debugging
5429 information for the debugger in use is appropriate.
5431 @findex ASM_OUTPUT_IDENT
5432 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
5433 A C statement to output something to the assembler file to handle a
5434 @samp{#ident} directive containing the text @var{string}. If this
5435 macro is not defined, nothing is output for a @samp{#ident} directive.
5437 @findex ASM_OUTPUT_SECTION_NAME
5438 @item ASM_OUTPUT_SECTION_NAME (@var{stream}, @var{decl}, @var{name}, @var{reloc})
5439 A C statement to output something to the assembler file to switch to section
5440 @var{name} for object @var{decl} which is either a @code{FUNCTION_DECL}, a
5441 @code{VAR_DECL} or @code{NULL_TREE}. @var{reloc}
5442 indicates whether the initial value of @var{exp} requires link-time
5443 relocations. Some target formats do not support
5444 arbitrary sections. Do not define this macro in such cases.
5446 At present this macro is only used to support section attributes.
5447 When this macro is undefined, section attributes are disabled.
5449 @findex OBJC_PROLOGUE
5451 A C statement to output any assembler statements which are required to
5452 precede any Objective C object definitions or message sending. The
5453 statement is executed only when compiling an Objective C program.
5458 @subsection Output of Data
5460 @c prevent bad page break with this line
5461 This describes data output.
5464 @findex ASM_OUTPUT_LONG_DOUBLE
5465 @findex ASM_OUTPUT_DOUBLE
5466 @findex ASM_OUTPUT_FLOAT
5467 @item ASM_OUTPUT_LONG_DOUBLE (@var{stream}, @var{value})
5468 @itemx ASM_OUTPUT_DOUBLE (@var{stream}, @var{value})
5469 @itemx ASM_OUTPUT_FLOAT (@var{stream}, @var{value})
5470 @itemx ASM_OUTPUT_THREE_QUARTER_FLOAT (@var{stream}, @var{value})
5471 @itemx ASM_OUTPUT_SHORT_FLOAT (@var{stream}, @var{value})
5472 @itemx ASM_OUTPUT_BYTE_FLOAT (@var{stream}, @var{value})
5473 A C statement to output to the stdio stream @var{stream} an assembler
5474 instruction to assemble a floating-point constant of @code{TFmode},
5475 @code{DFmode}, @code{SFmode}, @code{TQFmode}, @code{HFmode}, or
5476 @code{QFmode}, respectively, whose value is @var{value}. @var{value}
5477 will be a C expression of type @code{REAL_VALUE_TYPE}. Macros such as
5478 @code{REAL_VALUE_TO_TARGET_DOUBLE} are useful for writing these
5481 @findex ASM_OUTPUT_QUADRUPLE_INT
5482 @findex ASM_OUTPUT_DOUBLE_INT
5483 @findex ASM_OUTPUT_INT
5484 @findex ASM_OUTPUT_SHORT
5485 @findex ASM_OUTPUT_CHAR
5486 @findex output_addr_const
5487 @item ASM_OUTPUT_QUADRUPLE_INT (@var{stream}, @var{exp})
5488 @itemx ASM_OUTPUT_DOUBLE_INT (@var{stream}, @var{exp})
5489 @itemx ASM_OUTPUT_INT (@var{stream}, @var{exp})
5490 @itemx ASM_OUTPUT_SHORT (@var{stream}, @var{exp})
5491 @itemx ASM_OUTPUT_CHAR (@var{stream}, @var{exp})
5492 A C statement to output to the stdio stream @var{stream} an assembler
5493 instruction to assemble an integer of 16, 8, 4, 2 or 1 bytes,
5494 respectively, whose value is @var{value}. The argument @var{exp} will
5495 be an RTL expression which represents a constant value. Use
5496 @samp{output_addr_const (@var{stream}, @var{exp})} to output this value
5497 as an assembler expression.@refill
5499 For sizes larger than @code{UNITS_PER_WORD}, if the action of a macro
5500 would be identical to repeatedly calling the macro corresponding to
5501 a size of @code{UNITS_PER_WORD}, once for each word, you need not define
5504 @findex ASM_OUTPUT_BYTE
5505 @item ASM_OUTPUT_BYTE (@var{stream}, @var{value})
5506 A C statement to output to the stdio stream @var{stream} an assembler
5507 instruction to assemble a single byte containing the number @var{value}.
5511 A C string constant, including spacing, giving the pseudo-op to use for a
5512 sequence of single-byte constants. If this macro is not defined, the
5513 default is @code{"\t.byte\t"}.
5515 @findex ASM_OUTPUT_ASCII
5516 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5517 A C statement to output to the stdio stream @var{stream} an assembler
5518 instruction to assemble a string constant containing the @var{len}
5519 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5520 @code{char *} and @var{len} a C expression of type @code{int}.
5522 If the assembler has a @code{.ascii} pseudo-op as found in the
5523 Berkeley Unix assembler, do not define the macro
5524 @code{ASM_OUTPUT_ASCII}.
5526 @findex CONSTANT_POOL_BEFORE_FUNCTION
5527 @item CONSTANT_POOL_BEFORE_FUNCTION
5528 You may define this macro as a C expression. You should define the
5529 expression to have a non-zero value if GCC should output the constant
5530 pool for a function before the code for the function, or a zero value if
5531 GCC should output the constant pool after the function. If you do
5532 not define this macro, the usual case, GCC will output the constant
5533 pool before the function.
5535 @findex ASM_OUTPUT_POOL_PROLOGUE
5536 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
5537 A C statement to output assembler commands to define the start of the
5538 constant pool for a function. @var{funname} is a string giving
5539 the name of the function. Should the return type of the function
5540 be required, it can be obtained via @var{fundecl}. @var{size}
5541 is the size, in bytes, of the constant pool that will be written
5542 immediately after this call.
5544 If no constant-pool prefix is required, the usual case, this macro need
5547 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
5548 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5549 A C statement (with or without semicolon) to output a constant in the
5550 constant pool, if it needs special treatment. (This macro need not do
5551 anything for RTL expressions that can be output normally.)
5553 The argument @var{file} is the standard I/O stream to output the
5554 assembler code on. @var{x} is the RTL expression for the constant to
5555 output, and @var{mode} is the machine mode (in case @var{x} is a
5556 @samp{const_int}). @var{align} is the required alignment for the value
5557 @var{x}; you should output an assembler directive to force this much
5560 The argument @var{labelno} is a number to use in an internal label for
5561 the address of this pool entry. The definition of this macro is
5562 responsible for outputting the label definition at the proper place.
5563 Here is how to do this:
5566 ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
5569 When you output a pool entry specially, you should end with a
5570 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5571 entry from being output a second time in the usual manner.
5573 You need not define this macro if it would do nothing.
5575 @findex CONSTANT_AFTER_FUNCTION_P
5576 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
5577 Define this macro as a C expression which is nonzero if the constant
5578 @var{exp}, of type @code{tree}, should be output after the code for a
5579 function. The compiler will normally output all constants before the
5580 function; you need not define this macro if this is OK.
5582 @findex ASM_OUTPUT_POOL_EPILOGUE
5583 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5584 A C statement to output assembler commands to at the end of the constant
5585 pool for a function. @var{funname} is a string giving the name of the
5586 function. Should the return type of the function be required, you can
5587 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5588 constant pool that GCC wrote immediately before this call.
5590 If no constant-pool epilogue is required, the usual case, you need not
5593 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
5594 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
5595 Define this macro as a C expression which is nonzero if @var{C} is
5596 used as a logical line separator by the assembler.
5598 If you do not define this macro, the default is that only
5599 the character @samp{;} is treated as a logical line separator.
5602 @findex ASM_OPEN_PAREN
5603 @findex ASM_CLOSE_PAREN
5604 @item ASM_OPEN_PAREN
5605 @itemx ASM_CLOSE_PAREN
5606 These macros are defined as C string constant, describing the syntax
5607 in the assembler for grouping arithmetic expressions. The following
5608 definitions are correct for most assemblers:
5611 #define ASM_OPEN_PAREN "("
5612 #define ASM_CLOSE_PAREN ")"
5616 These macros are provided by @file{real.h} for writing the definitions
5617 of @code{ASM_OUTPUT_DOUBLE} and the like:
5620 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5621 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5622 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5623 @findex REAL_VALUE_TO_TARGET_SINGLE
5624 @findex REAL_VALUE_TO_TARGET_DOUBLE
5625 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
5626 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
5627 floating point representation, and store its bit pattern in the array of
5628 @code{long int} whose address is @var{l}. The number of elements in the
5629 output array is determined by the size of the desired target floating
5630 point data type: 32 bits of it go in each @code{long int} array
5631 element. Each array element holds 32 bits of the result, even if
5632 @code{long int} is wider than 32 bits on the host machine.
5634 The array element values are designed so that you can print them out
5635 using @code{fprintf} in the order they should appear in the target
5638 @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
5639 @findex REAL_VALUE_TO_DECIMAL
5640 This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
5641 decimal number and stores it as a string into @var{string}.
5642 You must pass, as @var{string}, the address of a long enough block
5643 of space to hold the result.
5645 The argument @var{format} is a @code{printf}-specification that serves
5646 as a suggestion for how to format the output string.
5649 @node Uninitialized Data
5650 @subsection Output of Uninitialized Variables
5652 Each of the macros in this section is used to do the whole job of
5653 outputting a single uninitialized variable.
5656 @findex ASM_OUTPUT_COMMON
5657 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5658 A C statement (sans semicolon) to output to the stdio stream
5659 @var{stream} the assembler definition of a common-label named
5660 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5661 is the size rounded up to whatever alignment the caller wants.
5663 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5664 output the name itself; before and after that, output the additional
5665 assembler syntax for defining the name, and a newline.
5667 This macro controls how the assembler definitions of uninitialized
5668 common global variables are output.
5670 @findex ASM_OUTPUT_ALIGNED_COMMON
5671 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5672 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5673 separate, explicit argument. If you define this macro, it is used in
5674 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5675 handling the required alignment of the variable. The alignment is specified
5676 as the number of bits.
5678 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
5679 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5680 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5681 variable to be output, if there is one, or @code{NULL_TREE} if there
5682 is no corresponding variable. If you define this macro, GCC will use it
5683 in place of both @code{ASM_OUTPUT_COMMON} and
5684 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5685 the variable's decl in order to chose what to output.
5687 @findex ASM_OUTPUT_SHARED_COMMON
5688 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5689 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
5690 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
5693 @findex ASM_OUTPUT_BSS
5694 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5695 A C statement (sans semicolon) to output to the stdio stream
5696 @var{stream} the assembler definition of uninitialized global @var{decl} named
5697 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5698 is the size rounded up to whatever alignment the caller wants.
5700 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
5701 defining this macro. If unable, use the expression
5702 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5703 before and after that, output the additional assembler syntax for defining
5704 the name, and a newline.
5706 This macro controls how the assembler definitions of uninitialized global
5707 variables are output. This macro exists to properly support languages like
5708 @code{c++} which do not have @code{common} data. However, this macro currently
5709 is not defined for all targets. If this macro and
5710 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
5711 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
5712 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
5714 @findex ASM_OUTPUT_ALIGNED_BSS
5715 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5716 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
5717 separate, explicit argument. If you define this macro, it is used in
5718 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
5719 handling the required alignment of the variable. The alignment is specified
5720 as the number of bits.
5722 Try to use function @code{asm_output_aligned_bss} defined in file
5723 @file{varasm.c} when defining this macro.
5725 @findex ASM_OUTPUT_SHARED_BSS
5726 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
5727 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
5728 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
5731 @findex ASM_OUTPUT_LOCAL
5732 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5733 A C statement (sans semicolon) to output to the stdio stream
5734 @var{stream} the assembler definition of a local-common-label named
5735 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5736 is the size rounded up to whatever alignment the caller wants.
5738 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5739 output the name itself; before and after that, output the additional
5740 assembler syntax for defining the name, and a newline.
5742 This macro controls how the assembler definitions of uninitialized
5743 static variables are output.
5745 @findex ASM_OUTPUT_ALIGNED_LOCAL
5746 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5747 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5748 separate, explicit argument. If you define this macro, it is used in
5749 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5750 handling the required alignment of the variable. The alignment is specified
5751 as the number of bits.
5753 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
5754 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5755 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5756 variable to be output, if there is one, or @code{NULL_TREE} if there
5757 is no corresponding variable. If you define this macro, GCC will use it
5758 in place of both @code{ASM_OUTPUT_DECL} and
5759 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5760 the variable's decl in order to chose what to output.
5762 @findex ASM_OUTPUT_SHARED_LOCAL
5763 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5764 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
5765 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
5770 @subsection Output and Generation of Labels
5772 @c prevent bad page break with this line
5773 This is about outputting labels.
5776 @findex ASM_OUTPUT_LABEL
5777 @findex assemble_name
5778 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5779 A C statement (sans semicolon) to output to the stdio stream
5780 @var{stream} the assembler definition of a label named @var{name}.
5781 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5782 output the name itself; before and after that, output the additional
5783 assembler syntax for defining the name, and a newline.
5785 @findex ASM_DECLARE_FUNCTION_NAME
5786 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5787 A C statement (sans semicolon) to output to the stdio stream
5788 @var{stream} any text necessary for declaring the name @var{name} of a
5789 function which is being defined. This macro is responsible for
5790 outputting the label definition (perhaps using
5791 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
5792 @code{FUNCTION_DECL} tree node representing the function.
5794 If this macro is not defined, then the function name is defined in the
5795 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5797 @findex ASM_DECLARE_FUNCTION_SIZE
5798 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5799 A C statement (sans semicolon) to output to the stdio stream
5800 @var{stream} any text necessary for declaring the size of a function
5801 which is being defined. The argument @var{name} is the name of the
5802 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5803 representing the function.
5805 If this macro is not defined, then the function size is not defined.
5807 @findex ASM_DECLARE_OBJECT_NAME
5808 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5809 A C statement (sans semicolon) to output to the stdio stream
5810 @var{stream} any text necessary for declaring the name @var{name} of an
5811 initialized variable which is being defined. This macro must output the
5812 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
5813 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
5815 If this macro is not defined, then the variable name is defined in the
5816 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5818 @findex ASM_DECLARE_REGISTER_GLOBAL
5819 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
5820 A C statement (sans semicolon) to output to the stdio stream
5821 @var{stream} any text necessary for claiming a register @var{regno}
5822 for a global variable @var{decl} with name @var{name}.
5824 If you don't define this macro, that is equivalent to defining it to do
5827 @findex ASM_FINISH_DECLARE_OBJECT
5828 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5829 A C statement (sans semicolon) to finish up declaring a variable name
5830 once the compiler has processed its initializer fully and thus has had a
5831 chance to determine the size of an array when controlled by an
5832 initializer. This is used on systems where it's necessary to declare
5833 something about the size of the object.
5835 If you don't define this macro, that is equivalent to defining it to do
5838 @findex ASM_GLOBALIZE_LABEL
5839 @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
5840 A C statement (sans semicolon) to output to the stdio stream
5841 @var{stream} some commands that will make the label @var{name} global;
5842 that is, available for reference from other files. Use the expression
5843 @code{assemble_name (@var{stream}, @var{name})} to output the name
5844 itself; before and after that, output the additional assembler syntax
5845 for making that name global, and a newline.
5847 @findex ASM_WEAKEN_LABEL
5848 @item ASM_WEAKEN_LABEL
5849 A C statement (sans semicolon) to output to the stdio stream
5850 @var{stream} some commands that will make the label @var{name} weak;
5851 that is, available for reference from other files but only used if
5852 no other definition is available. Use the expression
5853 @code{assemble_name (@var{stream}, @var{name})} to output the name
5854 itself; before and after that, output the additional assembler syntax
5855 for making that name weak, and a newline.
5857 If you don't define this macro, GCC will not support weak
5858 symbols and you should not define the @code{SUPPORTS_WEAK} macro.
5860 @findex SUPPORTS_WEAK
5862 A C expression which evaluates to true if the target supports weak symbols.
5864 If you don't define this macro, @file{defaults.h} provides a default
5865 definition. If @code{ASM_WEAKEN_LABEL} is defined, the default
5866 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5867 you want to control weak symbol support with a compiler flag such as
5870 @findex MAKE_DECL_ONE_ONLY (@var{decl})
5871 @item MAKE_DECL_ONE_ONLY
5872 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5873 public symbol such that extra copies in multiple translation units will
5874 be discarded by the linker. Define this macro if your object file
5875 format provides support for this concept, such as the @samp{COMDAT}
5876 section flags in the Microsoft Windows PE/COFF format, and this support
5877 requires changes to @var{decl}, such as putting it in a separate section.
5879 @findex SUPPORTS_ONE_ONLY
5880 @item SUPPORTS_ONE_ONLY
5881 A C expression which evaluates to true if the target supports one-only
5884 If you don't define this macro, @file{varasm.c} provides a default
5885 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
5886 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5887 you want to control one-only symbol support with a compiler flag, or if
5888 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
5889 be emitted as one-only.
5891 @findex ASM_OUTPUT_EXTERNAL
5892 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
5893 A C statement (sans semicolon) to output to the stdio stream
5894 @var{stream} any text necessary for declaring the name of an external
5895 symbol named @var{name} which is referenced in this compilation but
5896 not defined. The value of @var{decl} is the tree node for the
5899 This macro need not be defined if it does not need to output anything.
5900 The GNU assembler and most Unix assemblers don't require anything.
5902 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
5903 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
5904 A C statement (sans semicolon) to output on @var{stream} an assembler
5905 pseudo-op to declare a library function name external. The name of the
5906 library function is given by @var{symref}, which has type @code{rtx} and
5907 is a @code{symbol_ref}.
5909 This macro need not be defined if it does not need to output anything.
5910 The GNU assembler and most Unix assemblers don't require anything.
5912 @findex ASM_OUTPUT_LABELREF
5913 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
5914 A C statement (sans semicolon) to output to the stdio stream
5915 @var{stream} a reference in assembler syntax to a label named
5916 @var{name}. This should add @samp{_} to the front of the name, if that
5917 is customary on your operating system, as it is in most Berkeley Unix
5918 systems. This macro is used in @code{assemble_name}.
5920 @ignore @c Seems not to exist anymore.
5921 @findex ASM_OUTPUT_LABELREF_AS_INT
5922 @item ASM_OUTPUT_LABELREF_AS_INT (@var{file}, @var{label})
5923 Define this macro for systems that use the program @code{collect2}.
5924 The definition should be a C statement to output a word containing
5925 a reference to the label @var{label}.
5928 @findex ASM_OUTPUT_INTERNAL_LABEL
5929 @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
5930 A C statement to output to the stdio stream @var{stream} a label whose
5931 name is made from the string @var{prefix} and the number @var{num}.
5933 It is absolutely essential that these labels be distinct from the labels
5934 used for user-level functions and variables. Otherwise, certain programs
5935 will have name conflicts with internal labels.
5937 It is desirable to exclude internal labels from the symbol table of the
5938 object file. Most assemblers have a naming convention for labels that
5939 should be excluded; on many systems, the letter @samp{L} at the
5940 beginning of a label has this effect. You should find out what
5941 convention your system uses, and follow it.
5943 The usual definition of this macro is as follows:
5946 fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
5949 @findex ASM_OUTPUT_ALTERNATE_LABEL_NAME
5950 @item ASM_OUTPUT_ALTERNATE_LABEL_NAME (@var{stream}, @var{string})
5951 A C statement to output to the stdio stream @var{stream} the string
5954 The default definition of this macro is as follows:
5957 fprintf (@var{stream}, "%s:\n", LABEL_ALTERNATE_NAME (INSN))
5960 @findex ASM_GENERATE_INTERNAL_LABEL
5961 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
5962 A C statement to store into the string @var{string} a label whose name
5963 is made from the string @var{prefix} and the number @var{num}.
5965 This string, when output subsequently by @code{assemble_name}, should
5966 produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
5967 with the same @var{prefix} and @var{num}.
5969 If the string begins with @samp{*}, then @code{assemble_name} will
5970 output the rest of the string unchanged. It is often convenient for
5971 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
5972 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
5973 to output the string, and may change it. (Of course,
5974 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
5975 you should know what it does on your machine.)
5977 @findex ASM_FORMAT_PRIVATE_NAME
5978 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
5979 A C expression to assign to @var{outvar} (which is a variable of type
5980 @code{char *}) a newly allocated string made from the string
5981 @var{name} and the number @var{number}, with some suitable punctuation
5982 added. Use @code{alloca} to get space for the string.
5984 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
5985 produce an assembler label for an internal static variable whose name is
5986 @var{name}. Therefore, the string must be such as to result in valid
5987 assembler code. The argument @var{number} is different each time this
5988 macro is executed; it prevents conflicts between similarly-named
5989 internal static variables in different scopes.
5991 Ideally this string should not be a valid C identifier, to prevent any
5992 conflict with the user's own symbols. Most assemblers allow periods
5993 or percent signs in assembler symbols; putting at least one of these
5994 between the name and the number will suffice.
5996 @findex ASM_OUTPUT_DEF
5997 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
5998 A C statement to output to the stdio stream @var{stream} assembler code
5999 which defines (equates) the symbol @var{name} to have the value @var{value}.
6002 If SET_ASM_OP is defined, a default definition is provided which is
6003 correct for most systems.
6005 @findex ASM_OUTPUT_DEF_FROM_DECLS
6006 @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6007 A C statement to output to the stdio stream @var{stream} assembler code
6008 which defines (equates) the symbol whoes tree node is @var{decl_of_name}
6009 to have the value of the tree node @var{decl_of_value}. This macro will
6010 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6011 the tree nodes are available.
6013 @findex ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL
6014 @item ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL (@var{stream}, @var{symbol}, @var{high}, @var{low})
6015 A C statement to output to the stdio stream @var{stream} assembler code
6016 which defines (equates) the symbol @var{symbol} to have a value equal to
6017 the difference of the two symbols @var{high} and @var{low}, i.e.
6018 @var{high} minus @var{low}. GCC guarantees that the symbols @var{high}
6019 and @var{low} are already known by the assembler so that the difference
6020 resolves into a constant.
6023 If SET_ASM_OP is defined, a default definition is provided which is
6024 correct for most systems.
6026 @findex ASM_OUTPUT_WEAK_ALIAS
6027 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6028 A C statement to output to the stdio stream @var{stream} assembler code
6029 which defines (equates) the weak symbol @var{name} to have the value
6032 Define this macro if the target only supports weak aliases; define
6033 ASM_OUTPUT_DEF instead if possible.
6035 @findex OBJC_GEN_METHOD_LABEL
6036 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6037 Define this macro to override the default assembler names used for
6038 Objective C methods.
6040 The default name is a unique method number followed by the name of the
6041 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6042 the category is also included in the assembler name (e.g.@:
6045 These names are safe on most systems, but make debugging difficult since
6046 the method's selector is not present in the name. Therefore, particular
6047 systems define other ways of computing names.
6049 @var{buf} is an expression of type @code{char *} which gives you a
6050 buffer in which to store the name; its length is as long as
6051 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6052 50 characters extra.
6054 The argument @var{is_inst} specifies whether the method is an instance
6055 method or a class method; @var{class_name} is the name of the class;
6056 @var{cat_name} is the name of the category (or NULL if the method is not
6057 in a category); and @var{sel_name} is the name of the selector.
6059 On systems where the assembler can handle quoted names, you can use this
6060 macro to provide more human-readable names.
6063 @node Initialization
6064 @subsection How Initialization Functions Are Handled
6065 @cindex initialization routines
6066 @cindex termination routines
6067 @cindex constructors, output of
6068 @cindex destructors, output of
6070 The compiled code for certain languages includes @dfn{constructors}
6071 (also called @dfn{initialization routines})---functions to initialize
6072 data in the program when the program is started. These functions need
6073 to be called before the program is ``started''---that is to say, before
6074 @code{main} is called.
6076 Compiling some languages generates @dfn{destructors} (also called
6077 @dfn{termination routines}) that should be called when the program
6080 To make the initialization and termination functions work, the compiler
6081 must output something in the assembler code to cause those functions to
6082 be called at the appropriate time. When you port the compiler to a new
6083 system, you need to specify how to do this.
6085 There are two major ways that GCC currently supports the execution of
6086 initialization and termination functions. Each way has two variants.
6087 Much of the structure is common to all four variations.
6089 @findex __CTOR_LIST__
6090 @findex __DTOR_LIST__
6091 The linker must build two lists of these functions---a list of
6092 initialization functions, called @code{__CTOR_LIST__}, and a list of
6093 termination functions, called @code{__DTOR_LIST__}.
6095 Each list always begins with an ignored function pointer (which may hold
6096 0, @minus{}1, or a count of the function pointers after it, depending on
6097 the environment). This is followed by a series of zero or more function
6098 pointers to constructors (or destructors), followed by a function
6099 pointer containing zero.
6101 Depending on the operating system and its executable file format, either
6102 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6103 time and exit time. Constructors are called in reverse order of the
6104 list; destructors in forward order.
6106 The best way to handle static constructors works only for object file
6107 formats which provide arbitrarily-named sections. A section is set
6108 aside for a list of constructors, and another for a list of destructors.
6109 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6110 object file that defines an initialization function also puts a word in
6111 the constructor section to point to that function. The linker
6112 accumulates all these words into one contiguous @samp{.ctors} section.
6113 Termination functions are handled similarly.
6115 To use this method, you need appropriate definitions of the macros
6116 @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}. Usually
6117 you can get them by including @file{svr4.h}.
6119 When arbitrary sections are available, there are two variants, depending
6120 upon how the code in @file{crtstuff.c} is called. On systems that
6121 support an @dfn{init} section which is executed at program startup,
6122 parts of @file{crtstuff.c} are compiled into that section. The
6123 program is linked by the @code{gcc} driver like this:
6126 ld -o @var{output_file} crtbegin.o @dots{} crtend.o -lgcc
6129 The head of a function (@code{__do_global_ctors}) appears in the init
6130 section of @file{crtbegin.o}; the remainder of the function appears in
6131 the init section of @file{crtend.o}. The linker will pull these two
6132 parts of the section together, making a whole function. If any of the
6133 user's object files linked into the middle of it contribute code, then that
6134 code will be executed as part of the body of @code{__do_global_ctors}.
6136 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
6139 If no init section is available, do not define
6140 @code{INIT_SECTION_ASM_OP}. Then @code{__do_global_ctors} is built into
6141 the text section like all other functions, and resides in
6142 @file{libgcc.a}. When GCC compiles any function called @code{main}, it
6143 inserts a procedure call to @code{__main} as the first executable code
6144 after the function prologue. The @code{__main} function, also defined
6145 in @file{libgcc2.c}, simply calls @file{__do_global_ctors}.
6147 In file formats that don't support arbitrary sections, there are again
6148 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
6149 and an `a.out' format must be used. In this case,
6150 @code{ASM_OUTPUT_CONSTRUCTOR} is defined to produce a @code{.stabs}
6151 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
6152 and with the address of the void function containing the initialization
6153 code as its value. The GNU linker recognizes this as a request to add
6154 the value to a ``set''; the values are accumulated, and are eventually
6155 placed in the executable as a vector in the format described above, with
6156 a leading (ignored) count and a trailing zero element.
6157 @code{ASM_OUTPUT_DESTRUCTOR} is handled similarly. Since no init
6158 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
6159 the compilation of @code{main} to call @code{__main} as above, starting
6160 the initialization process.
6162 The last variant uses neither arbitrary sections nor the GNU linker.
6163 This is preferable when you want to do dynamic linking and when using
6164 file formats which the GNU linker does not support, such as `ECOFF'. In
6165 this case, @code{ASM_OUTPUT_CONSTRUCTOR} does not produce an
6166 @code{N_SETT} symbol; initialization and termination functions are
6167 recognized simply by their names. This requires an extra program in the
6168 linkage step, called @code{collect2}. This program pretends to be the
6169 linker, for use with GCC; it does its job by running the ordinary
6170 linker, but also arranges to include the vectors of initialization and
6171 termination functions. These functions are called via @code{__main} as
6174 Choosing among these configuration options has been simplified by a set
6175 of operating-system-dependent files in the @file{config} subdirectory.
6176 These files define all of the relevant parameters. Usually it is
6177 sufficient to include one into your specific machine-dependent
6178 configuration file. These files are:
6182 For operating systems using the `a.out' format.
6185 For operating systems using the `MachO' format.
6188 For System V Release 3 and similar systems using `COFF' format.
6191 For System V Release 4 and similar systems using `ELF' format.
6194 For the VMS operating system.
6198 The following section describes the specific macros that control and
6199 customize the handling of initialization and termination functions.
6202 @node Macros for Initialization
6203 @subsection Macros Controlling Initialization Routines
6205 Here are the macros that control how the compiler handles initialization
6206 and termination functions:
6209 @findex INIT_SECTION_ASM_OP
6210 @item INIT_SECTION_ASM_OP
6211 If defined, a C string constant, including spacing, for the assembler
6212 operation to identify the following data as initialization code. If not
6213 defined, GCC will assume such a section does not exist. When you are
6214 using special sections for initialization and termination functions, this
6215 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6216 run the initialization functions.
6218 @item HAS_INIT_SECTION
6219 @findex HAS_INIT_SECTION
6220 If defined, @code{main} will not call @code{__main} as described above.
6221 This macro should be defined for systems that control the contents of the
6222 init section on a symbol-by-symbol basis, such as OSF/1, and should not
6223 be defined explicitly for systems that support
6224 @code{INIT_SECTION_ASM_OP}.
6226 @item LD_INIT_SWITCH
6227 @findex LD_INIT_SWITCH
6228 If defined, a C string constant for a switch that tells the linker that
6229 the following symbol is an initialization routine.
6231 @item LD_FINI_SWITCH
6232 @findex LD_FINI_SWITCH
6233 If defined, a C string constant for a switch that tells the linker that
6234 the following symbol is a finalization routine.
6237 @findex INVOKE__main
6238 If defined, @code{main} will call @code{__main} despite the presence of
6239 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6240 where the init section is not actually run automatically, but is still
6241 useful for collecting the lists of constructors and destructors.
6243 @item ASM_OUTPUT_CONSTRUCTOR (@var{stream}, @var{name})
6244 @findex ASM_OUTPUT_CONSTRUCTOR
6245 Define this macro as a C statement to output on the stream @var{stream}
6246 the assembler code to arrange to call the function named @var{name} at
6247 initialization time.
6249 Assume that @var{name} is the name of a C function generated
6250 automatically by the compiler. This function takes no arguments. Use
6251 the function @code{assemble_name} to output the name @var{name}; this
6252 performs any system-specific syntactic transformations such as adding an
6255 If you don't define this macro, nothing special is output to arrange to
6256 call the function. This is correct when the function will be called in
6257 some other manner---for example, by means of the @code{collect2} program,
6258 which looks through the symbol table to find these functions by their
6261 @item ASM_OUTPUT_DESTRUCTOR (@var{stream}, @var{name})
6262 @findex ASM_OUTPUT_DESTRUCTOR
6263 This is like @code{ASM_OUTPUT_CONSTRUCTOR} but used for termination
6264 functions rather than initialization functions.
6266 When @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR} are
6267 defined, the initializaiton routine generated for the generated object
6268 file will have static linkage.
6271 If your system uses @code{collect2} as the means of processing
6272 constructors, then that program normally uses @code{nm} to scan an
6273 object file for constructor functions to be called. On such systems you
6274 must not define @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}
6275 as the object file's initialization routine must have global scope.
6277 On certain kinds of systems, you can define these macros to make
6278 @code{collect2} work faster (and, in some cases, make it work at all):
6281 @findex OBJECT_FORMAT_COFF
6282 @item OBJECT_FORMAT_COFF
6283 Define this macro if the system uses COFF (Common Object File Format)
6284 object files, so that @code{collect2} can assume this format and scan
6285 object files directly for dynamic constructor/destructor functions.
6287 @findex OBJECT_FORMAT_ROSE
6288 @item OBJECT_FORMAT_ROSE
6289 Define this macro if the system uses ROSE format object files, so that
6290 @code{collect2} can assume this format and scan object files directly
6291 for dynamic constructor/destructor functions.
6293 These macros are effective only in a native compiler; @code{collect2} as
6294 part of a cross compiler always uses @code{nm} for the target machine.
6296 @findex REAL_NM_FILE_NAME
6297 @item REAL_NM_FILE_NAME
6298 Define this macro as a C string constant containing the file name to use
6299 to execute @code{nm}. The default is to search the path normally for
6302 If your system supports shared libraries and has a program to list the
6303 dynamic dependencies of a given library or executable, you can define
6304 these macros to enable support for running initialization and
6305 termination functions in shared libraries:
6309 Define this macro to a C string constant containing the name of the
6310 program which lists dynamic dependencies, like @code{"ldd"} under SunOS 4.
6312 @findex PARSE_LDD_OUTPUT
6313 @item PARSE_LDD_OUTPUT (@var{PTR})
6314 Define this macro to be C code that extracts filenames from the output
6315 of the program denoted by @code{LDD_SUFFIX}. @var{PTR} is a variable
6316 of type @code{char *} that points to the beginning of a line of output
6317 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6318 code must advance @var{PTR} to the beginning of the filename on that
6319 line. Otherwise, it must set @var{PTR} to @code{NULL}.
6323 @node Instruction Output
6324 @subsection Output of Assembler Instructions
6326 @c prevent bad page break with this line
6327 This describes assembler instruction output.
6330 @findex REGISTER_NAMES
6331 @item REGISTER_NAMES
6332 A C initializer containing the assembler's names for the machine
6333 registers, each one as a C string constant. This is what translates
6334 register numbers in the compiler into assembler language.
6336 @findex ADDITIONAL_REGISTER_NAMES
6337 @item ADDITIONAL_REGISTER_NAMES
6338 If defined, a C initializer for an array of structures containing a name
6339 and a register number. This macro defines additional names for hard
6340 registers, thus allowing the @code{asm} option in declarations to refer
6341 to registers using alternate names.
6343 @findex ASM_OUTPUT_OPCODE
6344 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6345 Define this macro if you are using an unusual assembler that
6346 requires different names for the machine instructions.
6348 The definition is a C statement or statements which output an
6349 assembler instruction opcode to the stdio stream @var{stream}. The
6350 macro-operand @var{ptr} is a variable of type @code{char *} which
6351 points to the opcode name in its ``internal'' form---the form that is
6352 written in the machine description. The definition should output the
6353 opcode name to @var{stream}, performing any translation you desire, and
6354 increment the variable @var{ptr} to point at the end of the opcode
6355 so that it will not be output twice.
6357 In fact, your macro definition may process less than the entire opcode
6358 name, or more than the opcode name; but if you want to process text
6359 that includes @samp{%}-sequences to substitute operands, you must take
6360 care of the substitution yourself. Just be sure to increment
6361 @var{ptr} over whatever text should not be output normally.
6363 @findex recog_operand
6364 If you need to look at the operand values, they can be found as the
6365 elements of @code{recog_operand}.
6367 If the macro definition does nothing, the instruction is output
6370 @findex FINAL_PRESCAN_INSN
6371 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6372 If defined, a C statement to be executed just prior to the output of
6373 assembler code for @var{insn}, to modify the extracted operands so
6374 they will be output differently.
6376 Here the argument @var{opvec} is the vector containing the operands
6377 extracted from @var{insn}, and @var{noperands} is the number of
6378 elements of the vector which contain meaningful data for this insn.
6379 The contents of this vector are what will be used to convert the insn
6380 template into assembler code, so you can change the assembler output
6381 by changing the contents of the vector.
6383 This macro is useful when various assembler syntaxes share a single
6384 file of instruction patterns; by defining this macro differently, you
6385 can cause a large class of instructions to be output differently (such
6386 as with rearranged operands). Naturally, variations in assembler
6387 syntax affecting individual insn patterns ought to be handled by
6388 writing conditional output routines in those patterns.
6390 If this macro is not defined, it is equivalent to a null statement.
6392 @findex FINAL_PRESCAN_LABEL
6393 @item FINAL_PRESCAN_LABEL
6394 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
6395 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
6396 @var{noperands} will be zero.
6398 @findex PRINT_OPERAND
6399 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6400 A C compound statement to output to stdio stream @var{stream} the
6401 assembler syntax for an instruction operand @var{x}. @var{x} is an
6404 @var{code} is a value that can be used to specify one of several ways
6405 of printing the operand. It is used when identical operands must be
6406 printed differently depending on the context. @var{code} comes from
6407 the @samp{%} specification that was used to request printing of the
6408 operand. If the specification was just @samp{%@var{digit}} then
6409 @var{code} is 0; if the specification was @samp{%@var{ltr}
6410 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6413 If @var{x} is a register, this macro should print the register's name.
6414 The names can be found in an array @code{reg_names} whose type is
6415 @code{char *[]}. @code{reg_names} is initialized from
6416 @code{REGISTER_NAMES}.
6418 When the machine description has a specification @samp{%@var{punct}}
6419 (a @samp{%} followed by a punctuation character), this macro is called
6420 with a null pointer for @var{x} and the punctuation character for
6423 @findex PRINT_OPERAND_PUNCT_VALID_P
6424 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6425 A C expression which evaluates to true if @var{code} is a valid
6426 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6427 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6428 punctuation characters (except for the standard one, @samp{%}) are used
6431 @findex PRINT_OPERAND_ADDRESS
6432 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6433 A C compound statement to output to stdio stream @var{stream} the
6434 assembler syntax for an instruction operand that is a memory reference
6435 whose address is @var{x}. @var{x} is an RTL expression.
6437 @cindex @code{ENCODE_SECTION_INFO} usage
6438 On some machines, the syntax for a symbolic address depends on the
6439 section that the address refers to. On these machines, define the macro
6440 @code{ENCODE_SECTION_INFO} to store the information into the
6441 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
6443 @findex DBR_OUTPUT_SEQEND
6444 @findex dbr_sequence_length
6445 @item DBR_OUTPUT_SEQEND(@var{file})
6446 A C statement, to be executed after all slot-filler instructions have
6447 been output. If necessary, call @code{dbr_sequence_length} to
6448 determine the number of slots filled in a sequence (zero if not
6449 currently outputting a sequence), to decide how many no-ops to output,
6452 Don't define this macro if it has nothing to do, but it is helpful in
6453 reading assembly output if the extent of the delay sequence is made
6454 explicit (e.g. with white space).
6456 @findex final_sequence
6457 Note that output routines for instructions with delay slots must be
6458 prepared to deal with not being output as part of a sequence (i.e.
6459 when the scheduling pass is not run, or when no slot fillers could be
6460 found.) The variable @code{final_sequence} is null when not
6461 processing a sequence, otherwise it contains the @code{sequence} rtx
6464 @findex REGISTER_PREFIX
6465 @findex LOCAL_LABEL_PREFIX
6466 @findex USER_LABEL_PREFIX
6467 @findex IMMEDIATE_PREFIX
6469 @item REGISTER_PREFIX
6470 @itemx LOCAL_LABEL_PREFIX
6471 @itemx USER_LABEL_PREFIX
6472 @itemx IMMEDIATE_PREFIX
6473 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6474 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6475 @file{final.c}). These are useful when a single @file{md} file must
6476 support multiple assembler formats. In that case, the various @file{tm.h}
6477 files can define these macros differently.
6479 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
6480 @findex ASM_FPRINTF_EXTENSIONS
6481 If defiend this macro should expand to a series of @code{case}
6482 statements which will be parsed inside the @code{switch} statement of
6483 the @code{asm_fprintf} function. This allows targets to define extra
6484 printf formats which may useful when generating their assembler
6485 statements. Noet that upper case letters are reserved for future
6486 generic extensions to asm_fprintf, and so are not available to target
6487 specific code. The output file is given by the parameter @var{file}.
6488 The varargs input pointer is @var{argptr} and the rest of the format
6489 string, starting the character after the one that is being switched
6490 upon, is pointed to by @var{format}.
6492 @findex ASSEMBLER_DIALECT
6493 @item ASSEMBLER_DIALECT
6494 If your target supports multiple dialects of assembler language (such as
6495 different opcodes), define this macro as a C expression that gives the
6496 numeric index of the assembler language dialect to use, with zero as the
6499 If this macro is defined, you may use constructs of the form
6500 @samp{@{option0|option1|option2@dots{}@}} in the output
6501 templates of patterns (@pxref{Output Template}) or in the first argument
6502 of @code{asm_fprintf}. This construct outputs @samp{option0},
6503 @samp{option1} or @samp{option2}, etc., if the value of
6504 @code{ASSEMBLER_DIALECT} is zero, one or two, etc. Any special
6505 characters within these strings retain their usual meaning.
6507 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6508 @samp{@}} do not have any special meaning when used in templates or
6509 operands to @code{asm_fprintf}.
6511 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6512 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6513 the variations in assembler language syntax with that mechanism. Define
6514 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6515 if the syntax variant are larger and involve such things as different
6516 opcodes or operand order.
6518 @findex ASM_OUTPUT_REG_PUSH
6519 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6520 A C expression to output to @var{stream} some assembler code
6521 which will push hard register number @var{regno} onto the stack.
6522 The code need not be optimal, since this macro is used only when
6525 @findex ASM_OUTPUT_REG_POP
6526 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6527 A C expression to output to @var{stream} some assembler code
6528 which will pop hard register number @var{regno} off of the stack.
6529 The code need not be optimal, since this macro is used only when
6533 @node Dispatch Tables
6534 @subsection Output of Dispatch Tables
6536 @c prevent bad page break with this line
6537 This concerns dispatch tables.
6540 @cindex dispatch table
6541 @findex ASM_OUTPUT_ADDR_DIFF_ELT
6542 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6543 A C statement to output to the stdio stream @var{stream} an assembler
6544 pseudo-instruction to generate a difference between two labels.
6545 @var{value} and @var{rel} are the numbers of two internal labels. The
6546 definitions of these labels are output using
6547 @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
6548 way here. For example,
6551 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6552 @var{value}, @var{rel})
6555 You must provide this macro on machines where the addresses in a
6556 dispatch table are relative to the table's own address. If defined, GNU
6557 CC will also use this macro on all machines when producing PIC.
6558 @var{body} is the body of the ADDR_DIFF_VEC; it is provided so that the
6559 mode and flags can be read.
6561 @findex ASM_OUTPUT_ADDR_VEC_ELT
6562 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6563 This macro should be provided on machines where the addresses
6564 in a dispatch table are absolute.
6566 The definition should be a C statement to output to the stdio stream
6567 @var{stream} an assembler pseudo-instruction to generate a reference to
6568 a label. @var{value} is the number of an internal label whose
6569 definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
6573 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6576 @findex ASM_OUTPUT_CASE_LABEL
6577 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6578 Define this if the label before a jump-table needs to be output
6579 specially. The first three arguments are the same as for
6580 @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
6581 jump-table which follows (a @code{jump_insn} containing an
6582 @code{addr_vec} or @code{addr_diff_vec}).
6584 This feature is used on system V to output a @code{swbeg} statement
6587 If this macro is not defined, these labels are output with
6588 @code{ASM_OUTPUT_INTERNAL_LABEL}.
6590 @findex ASM_OUTPUT_CASE_END
6591 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6592 Define this if something special must be output at the end of a
6593 jump-table. The definition should be a C statement to be executed
6594 after the assembler code for the table is written. It should write
6595 the appropriate code to stdio stream @var{stream}. The argument
6596 @var{table} is the jump-table insn, and @var{num} is the label-number
6597 of the preceding label.
6599 If this macro is not defined, nothing special is output at the end of
6603 @node Exception Region Output
6604 @subsection Assembler Commands for Exception Regions
6606 @c prevent bad page break with this line
6608 This describes commands marking the start and the end of an exception
6612 @findex ASM_OUTPUT_EH_REGION_BEG
6613 @item ASM_OUTPUT_EH_REGION_BEG ()
6614 A C expression to output text to mark the start of an exception region.
6616 This macro need not be defined on most platforms.
6618 @findex ASM_OUTPUT_EH_REGION_END
6619 @item ASM_OUTPUT_EH_REGION_END ()
6620 A C expression to output text to mark the end of an exception region.
6622 This macro need not be defined on most platforms.
6624 @findex EXCEPTION_SECTION
6625 @item EXCEPTION_SECTION ()
6626 A C expression to switch to the section in which the main
6627 exception table is to be placed (@pxref{Sections}). The default is a
6628 section named @code{.gcc_except_table} on machines that support named
6629 sections via @code{ASM_OUTPUT_SECTION_NAME}, otherwise if @samp{-fpic}
6630 or @samp{-fPIC} is in effect, the @code{data_section}, otherwise the
6631 @code{readonly_data_section}.
6633 @findex EH_FRAME_SECTION_ASM_OP
6634 @item EH_FRAME_SECTION_ASM_OP
6635 If defined, a C string constant, including spacing, for the assembler
6636 operation to switch to the section for exception handling frame unwind
6637 information. If not defined, GCC will provide a default definition if the
6638 target supports named sections. @file{crtstuff.c} uses this macro to
6639 switch to the appropriate section.
6641 You should define this symbol if your target supports DWARF 2 frame
6642 unwind information and the default definition does not work.
6644 @findex OMIT_EH_TABLE
6645 @item OMIT_EH_TABLE ()
6646 A C expression that is nonzero if the normal exception table output
6649 This macro need not be defined on most platforms.
6651 @findex EH_TABLE_LOOKUP
6652 @item EH_TABLE_LOOKUP ()
6653 Alternate runtime support for looking up an exception at runtime and
6654 finding the associated handler, if the default method won't work.
6656 This macro need not be defined on most platforms.
6658 @findex DOESNT_NEED_UNWINDER
6659 @item DOESNT_NEED_UNWINDER
6660 A C expression that decides whether or not the current function needs to
6661 have a function unwinder generated for it. See the file @code{except.c}
6662 for details on when to define this, and how.
6664 @findex MASK_RETURN_ADDR
6665 @item MASK_RETURN_ADDR
6666 An rtx used to mask the return address found via RETURN_ADDR_RTX, so
6667 that it does not contain any extraneous set bits in it.
6669 @findex DWARF2_UNWIND_INFO
6670 @item DWARF2_UNWIND_INFO
6671 Define this macro to 0 if your target supports DWARF 2 frame unwind
6672 information, but it does not yet work with exception handling.
6673 Otherwise, if your target supports this information (if it defines
6674 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
6675 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
6678 If this macro is defined to 1, the DWARF 2 unwinder will be the default
6679 exception handling mechanism; otherwise, setjmp/longjmp will be used by
6682 If this macro is defined to anything, the DWARF 2 unwinder will be used
6683 instead of inline unwinders and __unwind_function in the non-setjmp case.
6687 @node Alignment Output
6688 @subsection Assembler Commands for Alignment
6690 @c prevent bad page break with this line
6691 This describes commands for alignment.
6694 @findex LABEL_ALIGN_AFTER_BARRIER
6695 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
6696 The alignment (log base 2) to put in front of @var{label}, which follows
6699 This macro need not be defined if you don't want any special alignment
6700 to be done at such a time. Most machine descriptions do not currently
6703 Unless it's necessary to inspect the @var{label} parameter, it is better
6704 to set the variable @var{align_jumps} in the target's
6705 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6706 selection in @var{align_jumps} in a @code{LABEL_ALIGN_AFTER_BARRIER}
6709 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6710 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6711 The maximum number of bytes to skip when applying
6712 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
6713 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6716 @item LOOP_ALIGN (@var{label})
6717 The alignment (log base 2) to put in front of @var{label}, which follows
6718 a NOTE_INSN_LOOP_BEG note.
6720 This macro need not be defined if you don't want any special alignment
6721 to be done at such a time. Most machine descriptions do not currently
6724 Unless it's necessary to inspect the @var{label} parameter, it is better
6725 to set the variable @var{align_loops} in the target's
6726 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6727 selection in @var{align_loops} in a @code{LOOP_ALIGN} implementation.
6729 @findex LOOP_ALIGN_MAX_SKIP
6730 @item LOOP_ALIGN_MAX_SKIP
6731 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
6732 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6735 @item LABEL_ALIGN (@var{label})
6736 The alignment (log base 2) to put in front of @var{label}.
6737 If LABEL_ALIGN_AFTER_BARRIER / LOOP_ALIGN specify a different alignment,
6738 the maximum of the specified values is used.
6740 Unless it's necessary to inspect the @var{label} parameter, it is better
6741 to set the variable @var{align_labels} in the target's
6742 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honour the user's
6743 selection in @var{align_labels} in a @code{LABEL_ALIGN} implementation.
6745 @findex LABEL_ALIGN_MAX_SKIP
6746 @item LABEL_ALIGN_MAX_SKIP
6747 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
6748 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
6750 @findex ASM_OUTPUT_SKIP
6751 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6752 A C statement to output to the stdio stream @var{stream} an assembler
6753 instruction to advance the location counter by @var{nbytes} bytes.
6754 Those bytes should be zero when loaded. @var{nbytes} will be a C
6755 expression of type @code{int}.
6757 @findex ASM_NO_SKIP_IN_TEXT
6758 @item ASM_NO_SKIP_IN_TEXT
6759 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6760 text section because it fails to put zeros in the bytes that are skipped.
6761 This is true on many Unix systems, where the pseudo--op to skip bytes
6762 produces no-op instructions rather than zeros when used in the text
6765 @findex ASM_OUTPUT_ALIGN
6766 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6767 A C statement to output to the stdio stream @var{stream} an assembler
6768 command to advance the location counter to a multiple of 2 to the
6769 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
6771 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
6772 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
6773 A C statement to output to the stdio stream @var{stream} an assembler
6774 command to advance the location counter to a multiple of 2 to the
6775 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
6776 satisfy the alignment request. @var{power} and @var{max_skip} will be
6777 a C expression of type @code{int}.
6781 @node Debugging Info
6782 @section Controlling Debugging Information Format
6784 @c prevent bad page break with this line
6785 This describes how to specify debugging information.
6788 * All Debuggers:: Macros that affect all debugging formats uniformly.
6789 * DBX Options:: Macros enabling specific options in DBX format.
6790 * DBX Hooks:: Hook macros for varying DBX format.
6791 * File Names and DBX:: Macros controlling output of file names in DBX format.
6792 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
6796 @subsection Macros Affecting All Debugging Formats
6798 @c prevent bad page break with this line
6799 These macros affect all debugging formats.
6802 @findex DBX_REGISTER_NUMBER
6803 @item DBX_REGISTER_NUMBER (@var{regno})
6804 A C expression that returns the DBX register number for the compiler
6805 register number @var{regno}. In simple cases, the value of this
6806 expression may be @var{regno} itself. But sometimes there are some
6807 registers that the compiler knows about and DBX does not, or vice
6808 versa. In such cases, some register may need to have one number in
6809 the compiler and another for DBX.
6811 If two registers have consecutive numbers inside GCC, and they can be
6812 used as a pair to hold a multiword value, then they @emph{must} have
6813 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
6814 Otherwise, debuggers will be unable to access such a pair, because they
6815 expect register pairs to be consecutive in their own numbering scheme.
6817 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
6818 does not preserve register pairs, then what you must do instead is
6819 redefine the actual register numbering scheme.
6821 @findex DEBUGGER_AUTO_OFFSET
6822 @item DEBUGGER_AUTO_OFFSET (@var{x})
6823 A C expression that returns the integer offset value for an automatic
6824 variable having address @var{x} (an RTL expression). The default
6825 computation assumes that @var{x} is based on the frame-pointer and
6826 gives the offset from the frame-pointer. This is required for targets
6827 that produce debugging output for DBX or COFF-style debugging output
6828 for SDB and allow the frame-pointer to be eliminated when the
6829 @samp{-g} options is used.
6831 @findex DEBUGGER_ARG_OFFSET
6832 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6833 A C expression that returns the integer offset value for an argument
6834 having address @var{x} (an RTL expression). The nominal offset is
6837 @findex PREFERRED_DEBUGGING_TYPE
6838 @item PREFERRED_DEBUGGING_TYPE
6839 A C expression that returns the type of debugging output GCC should
6840 produce when the user specifies just @samp{-g}. Define
6841 this if you have arranged for GCC to support more than one format of
6842 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
6843 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG}, and
6846 When the user specifies @samp{-ggdb}, GCC normally also uses the
6847 value of this macro to select the debugging output format, but with two
6848 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
6849 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
6850 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
6851 defined, GCC uses @code{DBX_DEBUG}.
6853 The value of this macro only affects the default debugging output; the
6854 user can always get a specific type of output by using @samp{-gstabs},
6855 @samp{-gcoff}, @samp{-gdwarf-1}, @samp{-gdwarf-2}, or @samp{-gxcoff}.
6859 @subsection Specific Options for DBX Output
6861 @c prevent bad page break with this line
6862 These are specific options for DBX output.
6865 @findex DBX_DEBUGGING_INFO
6866 @item DBX_DEBUGGING_INFO
6867 Define this macro if GCC should produce debugging output for DBX
6868 in response to the @samp{-g} option.
6870 @findex XCOFF_DEBUGGING_INFO
6871 @item XCOFF_DEBUGGING_INFO
6872 Define this macro if GCC should produce XCOFF format debugging output
6873 in response to the @samp{-g} option. This is a variant of DBX format.
6875 @findex DEFAULT_GDB_EXTENSIONS
6876 @item DEFAULT_GDB_EXTENSIONS
6877 Define this macro to control whether GCC should by default generate
6878 GDB's extended version of DBX debugging information (assuming DBX-format
6879 debugging information is enabled at all). If you don't define the
6880 macro, the default is 1: always generate the extended information
6881 if there is any occasion to.
6883 @findex DEBUG_SYMS_TEXT
6884 @item DEBUG_SYMS_TEXT
6885 Define this macro if all @code{.stabs} commands should be output while
6886 in the text section.
6888 @findex ASM_STABS_OP
6890 A C string constant, including spacing, naming the assembler pseudo op to
6891 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
6892 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
6893 applies only to DBX debugging information format.
6895 @findex ASM_STABD_OP
6897 A C string constant, including spacing, naming the assembler pseudo op to
6898 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
6899 value is the current location. If you don't define this macro,
6900 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
6903 @findex ASM_STABN_OP
6905 A C string constant, including spacing, naming the assembler pseudo op to
6906 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
6907 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
6908 macro applies only to DBX debugging information format.
6910 @findex DBX_NO_XREFS
6912 Define this macro if DBX on your system does not support the construct
6913 @samp{xs@var{tagname}}. On some systems, this construct is used to
6914 describe a forward reference to a structure named @var{tagname}.
6915 On other systems, this construct is not supported at all.
6917 @findex DBX_CONTIN_LENGTH
6918 @item DBX_CONTIN_LENGTH
6919 A symbol name in DBX-format debugging information is normally
6920 continued (split into two separate @code{.stabs} directives) when it
6921 exceeds a certain length (by default, 80 characters). On some
6922 operating systems, DBX requires this splitting; on others, splitting
6923 must not be done. You can inhibit splitting by defining this macro
6924 with the value zero. You can override the default splitting-length by
6925 defining this macro as an expression for the length you desire.
6927 @findex DBX_CONTIN_CHAR
6928 @item DBX_CONTIN_CHAR
6929 Normally continuation is indicated by adding a @samp{\} character to
6930 the end of a @code{.stabs} string when a continuation follows. To use
6931 a different character instead, define this macro as a character
6932 constant for the character you want to use. Do not define this macro
6933 if backslash is correct for your system.
6935 @findex DBX_STATIC_STAB_DATA_SECTION
6936 @item DBX_STATIC_STAB_DATA_SECTION
6937 Define this macro if it is necessary to go to the data section before
6938 outputting the @samp{.stabs} pseudo-op for a non-global static
6941 @findex DBX_TYPE_DECL_STABS_CODE
6942 @item DBX_TYPE_DECL_STABS_CODE
6943 The value to use in the ``code'' field of the @code{.stabs} directive
6944 for a typedef. The default is @code{N_LSYM}.
6946 @findex DBX_STATIC_CONST_VAR_CODE
6947 @item DBX_STATIC_CONST_VAR_CODE
6948 The value to use in the ``code'' field of the @code{.stabs} directive
6949 for a static variable located in the text section. DBX format does not
6950 provide any ``right'' way to do this. The default is @code{N_FUN}.
6952 @findex DBX_REGPARM_STABS_CODE
6953 @item DBX_REGPARM_STABS_CODE
6954 The value to use in the ``code'' field of the @code{.stabs} directive
6955 for a parameter passed in registers. DBX format does not provide any
6956 ``right'' way to do this. The default is @code{N_RSYM}.
6958 @findex DBX_REGPARM_STABS_LETTER
6959 @item DBX_REGPARM_STABS_LETTER
6960 The letter to use in DBX symbol data to identify a symbol as a parameter
6961 passed in registers. DBX format does not customarily provide any way to
6962 do this. The default is @code{'P'}.
6964 @findex DBX_MEMPARM_STABS_LETTER
6965 @item DBX_MEMPARM_STABS_LETTER
6966 The letter to use in DBX symbol data to identify a symbol as a stack
6967 parameter. The default is @code{'p'}.
6969 @findex DBX_FUNCTION_FIRST
6970 @item DBX_FUNCTION_FIRST
6971 Define this macro if the DBX information for a function and its
6972 arguments should precede the assembler code for the function. Normally,
6973 in DBX format, the debugging information entirely follows the assembler
6976 @findex DBX_LBRAC_FIRST
6977 @item DBX_LBRAC_FIRST
6978 Define this macro if the @code{N_LBRAC} symbol for a block should
6979 precede the debugging information for variables and functions defined in
6980 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
6983 @findex DBX_BLOCKS_FUNCTION_RELATIVE
6984 @item DBX_BLOCKS_FUNCTION_RELATIVE
6985 Define this macro if the value of a symbol describing the scope of a
6986 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
6987 of the enclosing function. Normally, GNU C uses an absolute address.
6989 @findex DBX_USE_BINCL
6991 Define this macro if GNU C should generate @code{N_BINCL} and
6992 @code{N_EINCL} stabs for included header files, as on Sun systems. This
6993 macro also directs GNU C to output a type number as a pair of a file
6994 number and a type number within the file. Normally, GNU C does not
6995 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
6996 number for a type number.
7000 @subsection Open-Ended Hooks for DBX Format
7002 @c prevent bad page break with this line
7003 These are hooks for DBX format.
7006 @findex DBX_OUTPUT_LBRAC
7007 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7008 Define this macro to say how to output to @var{stream} the debugging
7009 information for the start of a scope level for variable names. The
7010 argument @var{name} is the name of an assembler symbol (for use with
7011 @code{assemble_name}) whose value is the address where the scope begins.
7013 @findex DBX_OUTPUT_RBRAC
7014 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7015 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7017 @findex DBX_OUTPUT_ENUM
7018 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
7019 Define this macro if the target machine requires special handling to
7020 output an enumeration type. The definition should be a C statement
7021 (sans semicolon) to output the appropriate information to @var{stream}
7022 for the type @var{type}.
7024 @findex DBX_OUTPUT_FUNCTION_END
7025 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7026 Define this macro if the target machine requires special output at the
7027 end of the debugging information for a function. The definition should
7028 be a C statement (sans semicolon) to output the appropriate information
7029 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7032 @findex DBX_OUTPUT_STANDARD_TYPES
7033 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7034 Define this macro if you need to control the order of output of the
7035 standard data types at the beginning of compilation. The argument
7036 @var{syms} is a @code{tree} which is a chain of all the predefined
7037 global symbols, including names of data types.
7039 Normally, DBX output starts with definitions of the types for integers
7040 and characters, followed by all the other predefined types of the
7041 particular language in no particular order.
7043 On some machines, it is necessary to output different particular types
7044 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
7045 those symbols in the necessary order. Any predefined types that you
7046 don't explicitly output will be output afterward in no particular order.
7048 Be careful not to define this macro so that it works only for C. There
7049 are no global variables to access most of the built-in types, because
7050 another language may have another set of types. The way to output a
7051 particular type is to look through @var{syms} to see if you can find it.
7057 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7058 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
7060 dbxout_symbol (decl);
7066 This does nothing if the expected type does not exist.
7068 See the function @code{init_decl_processing} in @file{c-decl.c} to find
7069 the names to use for all the built-in C types.
7071 Here is another way of finding a particular type:
7073 @c this is still overfull. --mew 10feb93
7077 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7078 if (TREE_CODE (decl) == TYPE_DECL
7079 && (TREE_CODE (TREE_TYPE (decl))
7081 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
7082 && TYPE_UNSIGNED (TREE_TYPE (decl)))
7084 /* @r{This must be @code{unsigned short}.} */
7085 dbxout_symbol (decl);
7091 @findex NO_DBX_FUNCTION_END
7092 @item NO_DBX_FUNCTION_END
7093 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7094 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extention construct.
7095 On those machines, define this macro to turn this feature off without
7096 disturbing the rest of the gdb extensions.
7100 @node File Names and DBX
7101 @subsection File Names in DBX Format
7103 @c prevent bad page break with this line
7104 This describes file names in DBX format.
7107 @findex DBX_WORKING_DIRECTORY
7108 @item DBX_WORKING_DIRECTORY
7109 Define this if DBX wants to have the current directory recorded in each
7112 Note that the working directory is always recorded if GDB extensions are
7115 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
7116 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7117 A C statement to output DBX debugging information to the stdio stream
7118 @var{stream} which indicates that file @var{name} is the main source
7119 file---the file specified as the input file for compilation.
7120 This macro is called only once, at the beginning of compilation.
7122 This macro need not be defined if the standard form of output
7123 for DBX debugging information is appropriate.
7125 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
7126 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
7127 A C statement to output DBX debugging information to the stdio stream
7128 @var{stream} which indicates that the current directory during
7129 compilation is named @var{name}.
7131 This macro need not be defined if the standard form of output
7132 for DBX debugging information is appropriate.
7134 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
7135 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7136 A C statement to output DBX debugging information at the end of
7137 compilation of the main source file @var{name}.
7139 If you don't define this macro, nothing special is output at the end
7140 of compilation, which is correct for most machines.
7142 @findex DBX_OUTPUT_SOURCE_FILENAME
7143 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7144 A C statement to output DBX debugging information to the stdio stream
7145 @var{stream} which indicates that file @var{name} is the current source
7146 file. This output is generated each time input shifts to a different
7147 source file as a result of @samp{#include}, the end of an included file,
7148 or a @samp{#line} command.
7150 This macro need not be defined if the standard form of output
7151 for DBX debugging information is appropriate.
7156 @subsection Macros for SDB and DWARF Output
7158 @c prevent bad page break with this line
7159 Here are macros for SDB and DWARF output.
7162 @findex SDB_DEBUGGING_INFO
7163 @item SDB_DEBUGGING_INFO
7164 Define this macro if GCC should produce COFF-style debugging output
7165 for SDB in response to the @samp{-g} option.
7167 @findex DWARF_DEBUGGING_INFO
7168 @item DWARF_DEBUGGING_INFO
7169 Define this macro if GCC should produce dwarf format debugging output
7170 in response to the @samp{-g} option.
7172 @findex DWARF2_DEBUGGING_INFO
7173 @item DWARF2_DEBUGGING_INFO
7174 Define this macro if GCC should produce dwarf version 2 format
7175 debugging output in response to the @samp{-g} option.
7177 To support optional call frame debugging information, you must also
7178 define @code{INCOMING_RETURN_ADDR_RTX} and either set
7179 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
7180 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
7181 as appropriate from @code{FUNCTION_PROLOGUE} if you don't.
7183 @findex DWARF2_FRAME_INFO
7184 @item DWARF2_FRAME_INFO
7185 Define this macro to a nonzero value if GCC should always output
7186 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
7187 (@pxref{Exception Region Output} is nonzero, GCC will output this
7188 information not matter how you define @code{DWARF2_FRAME_INFO}.
7190 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
7191 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
7192 Define this macro if the linker does not work with Dwarf version 2.
7193 Normally, if the user specifies only @samp{-ggdb} GCC will use Dwarf
7194 version 2 if available; this macro disables this. See the description
7195 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
7197 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
7198 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
7199 By default, the Dwarf 2 debugging information generator will generate a
7200 label to mark the beginning of the text section. If it is better simply
7201 to use the name of the text section itself, rather than an explicit label,
7202 to indicate the beginning of the text section, define this macro to zero.
7204 @findex DWARF2_ASM_LINE_DEBUG_INFO
7205 @item DWARF2_ASM_LINE_DEBUG_INFO
7206 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
7207 line debug info sections. This will result in much more compact line number
7208 tables, and hence is desirable if it works.
7210 @findex PUT_SDB_@dots{}
7211 @item PUT_SDB_@dots{}
7212 Define these macros to override the assembler syntax for the special
7213 SDB assembler directives. See @file{sdbout.c} for a list of these
7214 macros and their arguments. If the standard syntax is used, you need
7215 not define them yourself.
7219 Some assemblers do not support a semicolon as a delimiter, even between
7220 SDB assembler directives. In that case, define this macro to be the
7221 delimiter to use (usually @samp{\n}). It is not necessary to define
7222 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
7225 @findex SDB_GENERATE_FAKE
7226 @item SDB_GENERATE_FAKE
7227 Define this macro to override the usual method of constructing a dummy
7228 name for anonymous structure and union types. See @file{sdbout.c} for
7231 @findex SDB_ALLOW_UNKNOWN_REFERENCES
7232 @item SDB_ALLOW_UNKNOWN_REFERENCES
7233 Define this macro to allow references to unknown structure,
7234 union, or enumeration tags to be emitted. Standard COFF does not
7235 allow handling of unknown references, MIPS ECOFF has support for
7238 @findex SDB_ALLOW_FORWARD_REFERENCES
7239 @item SDB_ALLOW_FORWARD_REFERENCES
7240 Define this macro to allow references to structure, union, or
7241 enumeration tags that have not yet been seen to be handled. Some
7242 assemblers choke if forward tags are used, while some require it.
7245 @node Cross-compilation
7246 @section Cross Compilation and Floating Point
7247 @cindex cross compilation and floating point
7248 @cindex floating point and cross compilation
7250 While all modern machines use 2's complement representation for integers,
7251 there are a variety of representations for floating point numbers. This
7252 means that in a cross-compiler the representation of floating point numbers
7253 in the compiled program may be different from that used in the machine
7254 doing the compilation.
7257 Because different representation systems may offer different amounts of
7258 range and precision, the cross compiler cannot safely use the host
7259 machine's floating point arithmetic. Therefore, floating point constants
7260 must be represented in the target machine's format. This means that the
7261 cross compiler cannot use @code{atof} to parse a floating point constant;
7262 it must have its own special routine to use instead. Also, constant
7263 folding must emulate the target machine's arithmetic (or must not be done
7266 The macros in the following table should be defined only if you are cross
7267 compiling between different floating point formats.
7269 Otherwise, don't define them. Then default definitions will be set up which
7270 use @code{double} as the data type, @code{==} to test for equality, etc.
7272 You don't need to worry about how many times you use an operand of any
7273 of these macros. The compiler never uses operands which have side effects.
7276 @findex REAL_VALUE_TYPE
7277 @item REAL_VALUE_TYPE
7278 A macro for the C data type to be used to hold a floating point value
7279 in the target machine's format. Typically this would be a
7280 @code{struct} containing an array of @code{int}.
7282 @findex REAL_VALUES_EQUAL
7283 @item REAL_VALUES_EQUAL (@var{x}, @var{y})
7284 A macro for a C expression which compares for equality the two values,
7285 @var{x} and @var{y}, both of type @code{REAL_VALUE_TYPE}.
7287 @findex REAL_VALUES_LESS
7288 @item REAL_VALUES_LESS (@var{x}, @var{y})
7289 A macro for a C expression which tests whether @var{x} is less than
7290 @var{y}, both values being of type @code{REAL_VALUE_TYPE} and
7291 interpreted as floating point numbers in the target machine's
7294 @findex REAL_VALUE_LDEXP
7296 @item REAL_VALUE_LDEXP (@var{x}, @var{scale})
7297 A macro for a C expression which performs the standard library
7298 function @code{ldexp}, but using the target machine's floating point
7299 representation. Both @var{x} and the value of the expression have
7300 type @code{REAL_VALUE_TYPE}. The second argument, @var{scale}, is an
7303 @findex REAL_VALUE_FIX
7304 @item REAL_VALUE_FIX (@var{x})
7305 A macro whose definition is a C expression to convert the target-machine
7306 floating point value @var{x} to a signed integer. @var{x} has type
7307 @code{REAL_VALUE_TYPE}.
7309 @findex REAL_VALUE_UNSIGNED_FIX
7310 @item REAL_VALUE_UNSIGNED_FIX (@var{x})
7311 A macro whose definition is a C expression to convert the target-machine
7312 floating point value @var{x} to an unsigned integer. @var{x} has type
7313 @code{REAL_VALUE_TYPE}.
7315 @findex REAL_VALUE_RNDZINT
7316 @item REAL_VALUE_RNDZINT (@var{x})
7317 A macro whose definition is a C expression to round the target-machine
7318 floating point value @var{x} towards zero to an integer value (but still
7319 as a floating point number). @var{x} has type @code{REAL_VALUE_TYPE},
7320 and so does the value.
7322 @findex REAL_VALUE_UNSIGNED_RNDZINT
7323 @item REAL_VALUE_UNSIGNED_RNDZINT (@var{x})
7324 A macro whose definition is a C expression to round the target-machine
7325 floating point value @var{x} towards zero to an unsigned integer value
7326 (but still represented as a floating point number). @var{x} has type
7327 @code{REAL_VALUE_TYPE}, and so does the value.
7329 @findex REAL_VALUE_ATOF
7330 @item REAL_VALUE_ATOF (@var{string}, @var{mode})
7331 A macro for a C expression which converts @var{string}, an expression of
7332 type @code{char *}, into a floating point number in the target machine's
7333 representation for mode @var{mode}. The value has type
7334 @code{REAL_VALUE_TYPE}.
7336 @findex REAL_INFINITY
7338 Define this macro if infinity is a possible floating point value, and
7339 therefore division by 0 is legitimate.
7341 @findex REAL_VALUE_ISINF
7343 @item REAL_VALUE_ISINF (@var{x})
7344 A macro for a C expression which determines whether @var{x}, a floating
7345 point value, is infinity. The value has type @code{int}.
7346 By default, this is defined to call @code{isinf}.
7348 @findex REAL_VALUE_ISNAN
7350 @item REAL_VALUE_ISNAN (@var{x})
7351 A macro for a C expression which determines whether @var{x}, a floating
7352 point value, is a ``nan'' (not-a-number). The value has type
7353 @code{int}. By default, this is defined to call @code{isnan}.
7356 @cindex constant folding and floating point
7357 Define the following additional macros if you want to make floating
7358 point constant folding work while cross compiling. If you don't
7359 define them, cross compilation is still possible, but constant folding
7360 will not happen for floating point values.
7363 @findex REAL_ARITHMETIC
7364 @item REAL_ARITHMETIC (@var{output}, @var{code}, @var{x}, @var{y})
7365 A macro for a C statement which calculates an arithmetic operation of
7366 the two floating point values @var{x} and @var{y}, both of type
7367 @code{REAL_VALUE_TYPE} in the target machine's representation, to
7368 produce a result of the same type and representation which is stored
7369 in @var{output} (which will be a variable).
7371 The operation to be performed is specified by @var{code}, a tree code
7372 which will always be one of the following: @code{PLUS_EXPR},
7373 @code{MINUS_EXPR}, @code{MULT_EXPR}, @code{RDIV_EXPR},
7374 @code{MAX_EXPR}, @code{MIN_EXPR}.@refill
7376 @cindex overflow while constant folding
7377 The expansion of this macro is responsible for checking for overflow.
7378 If overflow happens, the macro expansion should execute the statement
7379 @code{return 0;}, which indicates the inability to perform the
7380 arithmetic operation requested.
7382 @findex REAL_VALUE_NEGATE
7383 @item REAL_VALUE_NEGATE (@var{x})
7384 A macro for a C expression which returns the negative of the floating
7385 point value @var{x}. Both @var{x} and the value of the expression
7386 have type @code{REAL_VALUE_TYPE} and are in the target machine's
7387 floating point representation.
7389 There is no way for this macro to report overflow, since overflow
7390 can't happen in the negation operation.
7392 @findex REAL_VALUE_TRUNCATE
7393 @item REAL_VALUE_TRUNCATE (@var{mode}, @var{x})
7394 A macro for a C expression which converts the floating point value
7395 @var{x} to mode @var{mode}.
7397 Both @var{x} and the value of the expression are in the target machine's
7398 floating point representation and have type @code{REAL_VALUE_TYPE}.
7399 However, the value should have an appropriate bit pattern to be output
7400 properly as a floating constant whose precision accords with mode
7403 There is no way for this macro to report overflow.
7405 @findex REAL_VALUE_TO_INT
7406 @item REAL_VALUE_TO_INT (@var{low}, @var{high}, @var{x})
7407 A macro for a C expression which converts a floating point value
7408 @var{x} into a double-precision integer which is then stored into
7409 @var{low} and @var{high}, two variables of type @var{int}.
7411 @item REAL_VALUE_FROM_INT (@var{x}, @var{low}, @var{high}, @var{mode})
7412 @findex REAL_VALUE_FROM_INT
7413 A macro for a C expression which converts a double-precision integer
7414 found in @var{low} and @var{high}, two variables of type @var{int},
7415 into a floating point value which is then stored into @var{x}.
7416 The value is in the target machine's representation for mode @var{mode}
7417 and has the type @code{REAL_VALUE_TYPE}.
7420 @node Mode Switching
7421 @section Mode Switching Instructions
7422 @cindex mode switching
7423 The following macros control mode switching optimizations:
7426 @findex OPTIMIZE_MODE_SWITCHING
7427 @item OPTIMIZE_MODE_SWITCHING (@var{entity})
7428 Define this macro if the port needs extra instructions inserted for mode
7429 switching in an optimizing compilation.
7431 For an example, the SH4 can perform both single and double precision
7432 floating point operations, but to perform a single precision operation,
7433 the FPSCR PR bit has to be cleared, while for a double precision
7434 operation, this bit has to be set. Changing the PR bit requires a general
7435 purpose register as a scratch register, hence these FPSCR sets have to
7436 be inserted before reload, i.e. you can't put this into instruction emitting
7437 or MACHINE_DEPENDENT_REORG.
7439 You can have multiple entities that are mode-switched, and select at run time
7440 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
7441 return non-zero for any @var{entity} that that needs mode-switching.
7442 If you define this macro, you also have to define
7443 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
7444 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
7445 @code{NORMAL_MODE} is optional.
7447 @findex NUM_MODES_FOR_MODE_SWITCHING
7448 @item NUM_MODES_FOR_MODE_SWITCHING
7449 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
7450 initializer for an array of integers. Each initializer element
7451 N refers to an entity that needs mode switching, and specifies the number
7452 of different modes that might need to be set for this entity.
7453 The position of the initializer in the initializer - starting counting at
7454 zero - determines the integer that is used to refer to the mode-switched
7456 In macros that take mode arguments / yield a mode result, modes are
7457 represented as numbers 0 .. N - 1. N is used to specify that no mode
7458 switch is needed / supplied.
7461 @item MODE_NEEDED (@var{entity}, @var{insn})
7462 @var{entity} is an integer specifying a mode-switched entity. If
7463 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
7464 return an integer value not larger than the corresponding element in
7465 NUM_MODES_FOR_MODE_SWITCHING, to denote the mode that @var{entity} must
7466 be switched into prior to the execution of INSN.
7469 @item NORMAL_MODE (@var{entity})
7470 If this macro is defined, it is evaluated for every @var{entity} that needs
7471 mode switching. It should evaluate to an integer, which is a mode that
7472 @var{entity} is assumed to be switched to at function entry and exit.
7474 @findex MODE_PRIORITY_TO_MODE
7475 @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
7476 This macro specifies the order in which modes for ENTITY are processed.
7477 0 is the highest priority, NUM_MODES_FOR_MODE_SWITCHING[ENTITY] - 1 the
7478 lowest. The value of the macro should be an integer designating a mode
7479 for ENTITY. For any fixed @var{entity}, @code{mode_priority_to_mode}
7480 (@var{entity}, @var{n}) shall be a bijection in 0 ..
7481 @code{num_modes_for_mode_switching}[@var{entity}] - 1 .
7483 @findex EMIT_MODE_SET
7484 @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
7485 Generate one or more insns to set @var{entity} to @var{mode}.
7486 @var{hard_reg_live} is the set of hard registers live at the point where
7487 the insn(s) are to be inserted.
7491 @section Miscellaneous Parameters
7492 @cindex parameters, miscellaneous
7494 @c prevent bad page break with this line
7495 Here are several miscellaneous parameters.
7498 @item PREDICATE_CODES
7499 @findex PREDICATE_CODES
7500 Define this if you have defined special-purpose predicates in the file
7501 @file{@var{machine}.c}. This macro is called within an initializer of an
7502 array of structures. The first field in the structure is the name of a
7503 predicate and the second field is an array of rtl codes. For each
7504 predicate, list all rtl codes that can be in expressions matched by the
7505 predicate. The list should have a trailing comma. Here is an example
7506 of two entries in the list for a typical RISC machine:
7509 #define PREDICATE_CODES \
7510 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
7511 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
7514 Defining this macro does not affect the generated code (however,
7515 incorrect definitions that omit an rtl code that may be matched by the
7516 predicate can cause the compiler to malfunction). Instead, it allows
7517 the table built by @file{genrecog} to be more compact and efficient,
7518 thus speeding up the compiler. The most important predicates to include
7519 in the list specified by this macro are those used in the most insn
7522 @item SPECIAL_MODE_PREDICATES
7523 @findex SPECIAL_MODE_PREDICATES
7524 Define this if you have special predicates that know special things
7525 about modes. Genrecog will warn about certain forms of
7526 @code{match_operand} without a mode; if the operand predicate is
7527 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
7530 Here is an example from the IA-32 port (@code{ext_register_operand}
7531 specially checks for @code{HImode} or @code{SImode} in preparation
7532 for a byte extraction from @code{%ah} etc.).
7535 #define SPECIAL_MODE_PREDICATES \
7536 "ext_register_operand",
7539 @findex CASE_VECTOR_MODE
7540 @item CASE_VECTOR_MODE
7541 An alias for a machine mode name. This is the machine mode that
7542 elements of a jump-table should have.
7544 @findex CASE_VECTOR_SHORTEN_MODE
7545 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7546 Optional: return the preferred mode for an @code{addr_diff_vec}
7547 when the minimum and maximum offset are known. If you define this,
7548 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7549 To make this work, you also have to define INSN_ALIGN and
7550 make the alignment for @code{addr_diff_vec} explicit.
7551 The @var{body} argument is provided so that the offset_unsigned and scale
7552 flags can be updated.
7554 @findex CASE_VECTOR_PC_RELATIVE
7555 @item CASE_VECTOR_PC_RELATIVE
7556 Define this macro to be a C expression to indicate when jump-tables
7557 should contain relative addresses. If jump-tables never contain
7558 relative addresses, then you need not define this macro.
7560 @findex CASE_DROPS_THROUGH
7561 @item CASE_DROPS_THROUGH
7562 Define this if control falls through a @code{case} insn when the index
7563 value is out of range. This means the specified default-label is
7564 actually ignored by the @code{case} insn proper.
7566 @findex CASE_VALUES_THRESHOLD
7567 @item CASE_VALUES_THRESHOLD
7568 Define this to be the smallest number of different values for which it
7569 is best to use a jump-table instead of a tree of conditional branches.
7570 The default is four for machines with a @code{casesi} instruction and
7571 five otherwise. This is best for most machines.
7573 @findex WORD_REGISTER_OPERATIONS
7574 @item WORD_REGISTER_OPERATIONS
7575 Define this macro if operations between registers with integral mode
7576 smaller than a word are always performed on the entire register.
7577 Most RISC machines have this property and most CISC machines do not.
7579 @findex LOAD_EXTEND_OP
7580 @item LOAD_EXTEND_OP (@var{mode})
7581 Define this macro to be a C expression indicating when insns that read
7582 memory in @var{mode}, an integral mode narrower than a word, set the
7583 bits outside of @var{mode} to be either the sign-extension or the
7584 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7585 of @var{mode} for which the
7586 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7587 @code{NIL} for other modes.
7589 This macro is not called with @var{mode} non-integral or with a width
7590 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7591 value in this case. Do not define this macro if it would always return
7592 @code{NIL}. On machines where this macro is defined, you will normally
7593 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7595 @findex SHORT_IMMEDIATES_SIGN_EXTEND
7596 @item SHORT_IMMEDIATES_SIGN_EXTEND
7597 Define this macro if loading short immediate values into registers sign
7600 @findex IMPLICIT_FIX_EXPR
7601 @item IMPLICIT_FIX_EXPR
7602 An alias for a tree code that should be used by default for conversion
7603 of floating point values to fixed point. Normally,
7604 @code{FIX_ROUND_EXPR} is used.@refill
7606 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
7607 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
7608 Define this macro if the same instructions that convert a floating
7609 point number to a signed fixed point number also convert validly to an
7612 @findex EASY_DIV_EXPR
7614 An alias for a tree code that is the easiest kind of division to
7615 compile code for in the general case. It may be
7616 @code{TRUNC_DIV_EXPR}, @code{FLOOR_DIV_EXPR}, @code{CEIL_DIV_EXPR} or
7617 @code{ROUND_DIV_EXPR}. These four division operators differ in how
7618 they round the result to an integer. @code{EASY_DIV_EXPR} is used
7619 when it is permissible to use any of those kinds of division and the
7620 choice should be made on the basis of efficiency.@refill
7624 The maximum number of bytes that a single instruction can move quickly
7625 between memory and registers or between two memory locations.
7627 @findex MAX_MOVE_MAX
7629 The maximum number of bytes that a single instruction can move quickly
7630 between memory and registers or between two memory locations. If this
7631 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7632 constant value that is the largest value that @code{MOVE_MAX} can have
7635 @findex SHIFT_COUNT_TRUNCATED
7636 @item SHIFT_COUNT_TRUNCATED
7637 A C expression that is nonzero if on this machine the number of bits
7638 actually used for the count of a shift operation is equal to the number
7639 of bits needed to represent the size of the object being shifted. When
7640 this macro is non-zero, the compiler will assume that it is safe to omit
7641 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7642 truncates the count of a shift operation. On machines that have
7643 instructions that act on bitfields at variable positions, which may
7644 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7645 also enables deletion of truncations of the values that serve as
7646 arguments to bitfield instructions.
7648 If both types of instructions truncate the count (for shifts) and
7649 position (for bitfield operations), or if no variable-position bitfield
7650 instructions exist, you should define this macro.
7652 However, on some machines, such as the 80386 and the 680x0, truncation
7653 only applies to shift operations and not the (real or pretended)
7654 bitfield operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7655 such machines. Instead, add patterns to the @file{md} file that include
7656 the implied truncation of the shift instructions.
7658 You need not define this macro if it would always have the value of zero.
7660 @findex TRULY_NOOP_TRUNCATION
7661 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
7662 A C expression which is nonzero if on this machine it is safe to
7663 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
7664 bits (where @var{outprec} is smaller than @var{inprec}) by merely
7665 operating on it as if it had only @var{outprec} bits.
7667 On many machines, this expression can be 1.
7669 @c rearranged this, removed the phrase "it is reported that". this was
7670 @c to fix an overfull hbox. --mew 10feb93
7671 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
7672 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
7673 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
7674 such cases may improve things.
7676 @findex STORE_FLAG_VALUE
7677 @item STORE_FLAG_VALUE
7678 A C expression describing the value returned by a comparison operator
7679 with an integral mode and stored by a store-flag instruction
7680 (@samp{s@var{cond}}) when the condition is true. This description must
7681 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
7682 comparison operators whose results have a @code{MODE_INT} mode.
7684 A value of 1 or -1 means that the instruction implementing the
7685 comparison operator returns exactly 1 or -1 when the comparison is true
7686 and 0 when the comparison is false. Otherwise, the value indicates
7687 which bits of the result are guaranteed to be 1 when the comparison is
7688 true. This value is interpreted in the mode of the comparison
7689 operation, which is given by the mode of the first operand in the
7690 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
7691 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7694 If @code{STORE_FLAG_VALUE} is neither 1 or -1, the compiler will
7695 generate code that depends only on the specified bits. It can also
7696 replace comparison operators with equivalent operations if they cause
7697 the required bits to be set, even if the remaining bits are undefined.
7698 For example, on a machine whose comparison operators return an
7699 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7700 @samp{0x80000000}, saying that just the sign bit is relevant, the
7704 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7711 (ashift:SI @var{x} (const_int @var{n}))
7715 where @var{n} is the appropriate shift count to move the bit being
7716 tested into the sign bit.
7718 There is no way to describe a machine that always sets the low-order bit
7719 for a true value, but does not guarantee the value of any other bits,
7720 but we do not know of any machine that has such an instruction. If you
7721 are trying to port GCC to such a machine, include an instruction to
7722 perform a logical-and of the result with 1 in the pattern for the
7723 comparison operators and let us know
7725 (@pxref{Bug Reporting,,How to Report Bugs}).
7728 (@pxref{Bug Reporting,,How to Report Bugs,gcc.info,Using GCC}).
7731 Often, a machine will have multiple instructions that obtain a value
7732 from a comparison (or the condition codes). Here are rules to guide the
7733 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7738 Use the shortest sequence that yields a valid definition for
7739 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7740 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7741 comparison operators to do so because there may be opportunities to
7742 combine the normalization with other operations.
7745 For equal-length sequences, use a value of 1 or -1, with -1 being
7746 slightly preferred on machines with expensive jumps and 1 preferred on
7750 As a second choice, choose a value of @samp{0x80000001} if instructions
7751 exist that set both the sign and low-order bits but do not define the
7755 Otherwise, use a value of @samp{0x80000000}.
7758 Many machines can produce both the value chosen for
7759 @code{STORE_FLAG_VALUE} and its negation in the same number of
7760 instructions. On those machines, you should also define a pattern for
7761 those cases, e.g., one matching
7764 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7767 Some machines can also perform @code{and} or @code{plus} operations on
7768 condition code values with less instructions than the corresponding
7769 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
7770 machines, define the appropriate patterns. Use the names @code{incscc}
7771 and @code{decscc}, respectively, for the patterns which perform
7772 @code{plus} or @code{minus} operations on condition code values. See
7773 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
7774 find such instruction sequences on other machines.
7776 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7779 @findex FLOAT_STORE_FLAG_VALUE
7780 @item FLOAT_STORE_FLAG_VALUE (@var{mode})
7781 A C expression that gives a non-zero @code{REAL_VALUE_TYPE} value that is
7782 returned when comparison operators with floating-point results are true.
7783 Define this macro on machine that have comparison operations that return
7784 floating-point values. If there are no such operations, do not define
7789 An alias for the machine mode for pointers. On most machines, define
7790 this to be the integer mode corresponding to the width of a hardware
7791 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7792 On some machines you must define this to be one of the partial integer
7793 modes, such as @code{PSImode}.
7795 The width of @code{Pmode} must be at least as large as the value of
7796 @code{POINTER_SIZE}. If it is not equal, you must define the macro
7797 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7800 @findex FUNCTION_MODE
7802 An alias for the machine mode used for memory references to functions
7803 being called, in @code{call} RTL expressions. On most machines this
7804 should be @code{QImode}.
7806 @findex INTEGRATE_THRESHOLD
7807 @item INTEGRATE_THRESHOLD (@var{decl})
7808 A C expression for the maximum number of instructions above which the
7809 function @var{decl} should not be inlined. @var{decl} is a
7810 @code{FUNCTION_DECL} node.
7812 The default definition of this macro is 64 plus 8 times the number of
7813 arguments that the function accepts. Some people think a larger
7814 threshold should be used on RISC machines.
7816 @findex SCCS_DIRECTIVE
7817 @item SCCS_DIRECTIVE
7818 Define this if the preprocessor should ignore @code{#sccs} directives
7819 and print no error message.
7821 @findex NO_IMPLICIT_EXTERN_C
7822 @item NO_IMPLICIT_EXTERN_C
7823 Define this macro if the system header files support C++ as well as C.
7824 This macro inhibits the usual method of using system header files in
7825 C++, which is to pretend that the file's contents are enclosed in
7826 @samp{extern "C" @{@dots{}@}}.
7828 @findex HANDLE_PRAGMA
7829 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
7830 This macro is no longer supported. You must use
7831 @code{REGISTER_TARGET_PRAGMAS} instead.
7833 @findex REGISTER_TARGET_PRAGMAS
7836 @item REGISTER_TARGET_PRAGMAS (@var{pfile})
7837 Define this macro if you want to implement any target-specific pragmas.
7838 If defined, it is a C expression which makes a series of calls to the
7839 @code{cpp_register_pragma} and/or @code{cpp_register_pragma_space}
7840 functions. The @var{pfile} argument is the first argument to supply to
7841 these functions. The macro may also do setup required for the pragmas.
7843 The primary reason to define this macro is to provide compatibility with
7844 other compilers for the same target. In general, we discourage
7845 definition of target-specific pragmas for GCC.
7847 If the pragma can be implemented by atttributes then the macro
7848 @samp{INSERT_ATTRIBUTES} might be a useful one to define as well.
7850 Preprocessor macros that appear on pragma lines are not expanded. All
7851 @samp{#pragma} directives that do not match any registered pragma are
7852 silently ignored, unless the user specifies @samp{-Wunknown-pragmas}.
7854 @deftypefun void cpp_register_pragma (cpp_reader *@var{pfile}, const char *@var{space}, const char *@var{name}, void (*@var{callback}) (cpp_reader *))
7856 Each call to @code{cpp_register_pragma} establishes one pragma. The
7857 @var{callback} routine will be called when the preprocessor encounters a
7861 #pragma [@var{space}] @var{name} @dots{}
7864 @var{space} must have been the subject of a previous call to
7865 @code{cpp_register_pragma_space}, or else be a null pointer. The
7866 callback routine receives @var{pfile} as its first argument, but must
7867 not use it for anything (this may change in the future). It may read
7868 any text after the @var{name} by making calls to @code{c_lex}. Text
7869 which is not read by the callback will be silently ignored.
7871 Note that both @var{space} and @var{name} are case sensitive.
7873 For an example use of this routine, see @file{c4x.h} and the callback
7874 routines defined in @file{c4x.c}.
7877 @deftypefun void cpp_register_pragma_space (cpp_reader *@var{pfile}, const char *@var{space})
7878 This routine establishes a namespace for pragmas, which will be
7879 registered by subsequent calls to @code{cpp_register_pragma}. For
7880 example, pragmas defined by the C standard are in the @samp{STDC}
7881 namespace, and pragmas specific to GCC are in the @samp{GCC} namespace.
7883 For an example use of this routine in a target header, see @file{v850.h}.
7886 @findex HANDLE_SYSV_PRAGMA
7889 @item HANDLE_SYSV_PRAGMA
7890 Define this macro (to a value of 1) if you want the System V style
7891 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
7892 [=<value>]} to be supported by gcc.
7894 The pack pragma specifies the maximum alignment (in bytes) of fields
7895 within a structure, in much the same way as the @samp{__aligned__} and
7896 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
7897 the behaviour to the default.
7899 The weak pragma only works if @code{SUPPORTS_WEAK} and
7900 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
7901 of specifically named weak labels, optionally with a value.
7903 @findex HANDLE_PRAGMA_PACK_PUSH_POP
7906 @item HANDLE_PRAGMA_PACK_PUSH_POP
7907 Define this macro (to a value of 1) if you want to support the Win32
7908 style pragmas @samp{#pragma pack(push,<n>)} and @samp{#pragma
7909 pack(pop)}. The pack(push,<n>) pragma specifies the maximum alignment
7910 (in bytes) of fields within a structure, in much the same way as the
7911 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
7912 pack value of zero resets the behaviour to the default. Successive
7913 invocations of this pragma cause the previous values to be stacked, so
7914 that invocations of @samp{#pragma pack(pop)} will return to the previous
7917 @findex VALID_MACHINE_DECL_ATTRIBUTE
7918 @item VALID_MACHINE_DECL_ATTRIBUTE (@var{decl}, @var{attributes}, @var{identifier}, @var{args})
7919 If defined, a C expression whose value is nonzero if @var{identifier} with
7920 arguments @var{args} is a valid machine specific attribute for @var{decl}.
7921 The attributes in @var{attributes} have previously been assigned to @var{decl}.
7923 @findex VALID_MACHINE_TYPE_ATTRIBUTE
7924 @item VALID_MACHINE_TYPE_ATTRIBUTE (@var{type}, @var{attributes}, @var{identifier}, @var{args})
7925 If defined, a C expression whose value is nonzero if @var{identifier} with
7926 arguments @var{args} is a valid machine specific attribute for @var{type}.
7927 The attributes in @var{attributes} have previously been assigned to @var{type}.
7929 @findex COMP_TYPE_ATTRIBUTES
7930 @item COMP_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
7931 If defined, a C expression whose value is zero if the attributes on
7932 @var{type1} and @var{type2} are incompatible, one if they are compatible,
7933 and two if they are nearly compatible (which causes a warning to be
7936 @findex SET_DEFAULT_TYPE_ATTRIBUTES
7937 @item SET_DEFAULT_TYPE_ATTRIBUTES (@var{type})
7938 If defined, a C statement that assigns default attributes to
7939 newly defined @var{type}.
7941 @findex MERGE_MACHINE_TYPE_ATTRIBUTES
7942 @item MERGE_MACHINE_TYPE_ATTRIBUTES (@var{type1}, @var{type2})
7943 Define this macro if the merging of type attributes needs special handling.
7944 If defined, the result is a list of the combined TYPE_ATTRIBUTES of
7945 @var{type1} and @var{type2}. It is assumed that comptypes has already been
7946 called and returned 1.
7948 @findex MERGE_MACHINE_DECL_ATTRIBUTES
7949 @item MERGE_MACHINE_DECL_ATTRIBUTES (@var{olddecl}, @var{newdecl})
7950 Define this macro if the merging of decl attributes needs special handling.
7951 If defined, the result is a list of the combined DECL_MACHINE_ATTRIBUTES of
7952 @var{olddecl} and @var{newdecl}. @var{newdecl} is a duplicate declaration
7953 of @var{olddecl}. Examples of when this is needed are when one attribute
7954 overrides another, or when an attribute is nullified by a subsequent
7957 @findex INSERT_ATTRIBUTES
7958 @item INSERT_ATTRIBUTES (@var{node}, @var{attr_ptr}, @var{prefix_ptr})
7959 Define this macro if you want to be able to add attributes to a decl
7960 when it is being created. This is normally useful for backends which
7961 wish to implement a pragma by using the attributes which correspond to
7962 the pragma's effect. The @var{node} argument is the decl which is being
7963 created. The @var{attr_ptr} argument is a pointer to the attribute list
7964 for this decl. The @var{prefix_ptr} is a pointer to the list of
7965 attributes that have appeared after the specifiers and modifiers of the
7966 declaration, but before the declaration proper.
7968 @findex SET_DEFAULT_DECL_ATTRIBUTES
7969 @item SET_DEFAULT_DECL_ATTRIBUTES (@var{decl}, @var{attributes})
7970 If defined, a C statement that assigns default attributes to
7971 newly defined @var{decl}.
7973 @findex DOLLARS_IN_IDENTIFIERS
7974 @item DOLLARS_IN_IDENTIFIERS
7975 Define this macro to control use of the character @samp{$} in identifier
7976 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
7977 1 is the default; there is no need to define this macro in that case.
7978 This macro controls the compiler proper; it does not affect the preprocessor.
7980 @findex NO_DOLLAR_IN_LABEL
7981 @item NO_DOLLAR_IN_LABEL
7982 Define this macro if the assembler does not accept the character
7983 @samp{$} in label names. By default constructors and destructors in
7984 G++ have @samp{$} in the identifiers. If this macro is defined,
7985 @samp{.} is used instead.
7987 @findex NO_DOT_IN_LABEL
7988 @item NO_DOT_IN_LABEL
7989 Define this macro if the assembler does not accept the character
7990 @samp{.} in label names. By default constructors and destructors in G++
7991 have names that use @samp{.}. If this macro is defined, these names
7992 are rewritten to avoid @samp{.}.
7994 @findex DEFAULT_MAIN_RETURN
7995 @item DEFAULT_MAIN_RETURN
7996 Define this macro if the target system expects every program's @code{main}
7997 function to return a standard ``success'' value by default (if no other
7998 value is explicitly returned).
8000 The definition should be a C statement (sans semicolon) to generate the
8001 appropriate rtl instructions. It is used only when compiling the end of
8006 Define this if the target system lacks the function @code{atexit}
8007 from the ANSI C standard. If this macro is defined, a default definition
8008 will be provided to support C++. If @code{ON_EXIT} is not defined,
8009 a default @code{exit} function will also be provided.
8013 Define this macro if the target has another way to implement atexit
8014 functionality without replacing @code{exit}. For instance, SunOS 4 has
8015 a similar @code{on_exit} library function.
8017 The definition should be a functional macro which can be used just like
8018 the @code{atexit} function.
8022 Define this if your @code{exit} function needs to do something
8023 besides calling an external function @code{_cleanup} before
8024 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
8025 only needed if neither @code{HAVE_ATEXIT} nor
8026 @code{INIT_SECTION_ASM_OP} are defined.
8028 @findex INSN_SETS_ARE_DELAYED
8029 @item INSN_SETS_ARE_DELAYED (@var{insn})
8030 Define this macro as a C expression that is nonzero if it is safe for the
8031 delay slot scheduler to place instructions in the delay slot of @var{insn},
8032 even if they appear to use a resource set or clobbered in @var{insn}.
8033 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8034 every @code{call_insn} has this behavior. On machines where some @code{insn}
8035 or @code{jump_insn} is really a function call and hence has this behavior,
8036 you should define this macro.
8038 You need not define this macro if it would always return zero.
8040 @findex INSN_REFERENCES_ARE_DELAYED
8041 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
8042 Define this macro as a C expression that is nonzero if it is safe for the
8043 delay slot scheduler to place instructions in the delay slot of @var{insn},
8044 even if they appear to set or clobber a resource referenced in @var{insn}.
8045 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8046 some @code{insn} or @code{jump_insn} is really a function call and its operands
8047 are registers whose use is actually in the subroutine it calls, you should
8048 define this macro. Doing so allows the delay slot scheduler to move
8049 instructions which copy arguments into the argument registers into the delay
8052 You need not define this macro if it would always return zero.
8054 @findex MACHINE_DEPENDENT_REORG
8055 @item MACHINE_DEPENDENT_REORG (@var{insn})
8056 In rare cases, correct code generation requires extra machine
8057 dependent processing between the second jump optimization pass and
8058 delayed branch scheduling. On those machines, define this macro as a C
8059 statement to act on the code starting at @var{insn}.
8061 @findex MULTIPLE_SYMBOL_SPACES
8062 @item MULTIPLE_SYMBOL_SPACES
8063 Define this macro if in some cases global symbols from one translation
8064 unit may not be bound to undefined symbols in another translation unit
8065 without user intervention. For instance, under Microsoft Windows
8066 symbols must be explicitly imported from shared libraries (DLLs).
8068 @findex MD_ASM_CLOBBERS
8069 @item MD_ASM_CLOBBERS
8070 A C statement that adds to @var{CLOBBERS} @code{STRING_CST} trees for
8071 any hard regs the port wishes to automatically clobber for all asms.
8075 A C expression that returns how many instructions can be issued at the
8076 same time if the machine is a superscalar machine.
8078 @findex MD_SCHED_INIT
8079 @item MD_SCHED_INIT (@var{file}, @var{verbose})
8080 A C statement which is executed by the scheduler at the
8081 beginning of each block of instructions that are to be scheduled.
8082 @var{file} is either a null pointer, or a stdio stream to write any
8083 debug output to. @var{verbose} is the verbose level provided by
8084 @samp{-fsched-verbose-}@var{n}.
8086 @findex MD_SCHED_REORDER
8087 @item MD_SCHED_REORDER (@var{file}, @var{verbose}, @var{ready}, @var{n_ready}, @var{clock}, @var{can_issue_more})
8088 A C statement which is executed by the scheduler after it
8089 has scheduled the ready list to allow the machine description to reorder
8090 it (for example to combine two small instructions together on
8091 @samp{VLIW} machines). @var{file} is either a null pointer, or a stdio
8092 stream to write any debug output to. @var{verbose} is the verbose level
8093 provided by @samp{-fsched-verbose-}@var{n}. @var{ready} is a pointer to
8094 the ready list of instructions that are ready to be scheduled.
8095 @var{n_ready} is the number of elements in the ready list. The
8096 scheduler reads the ready list in reverse order, starting with
8097 @var{ready}[@var{n_ready}-1] and going to @var{ready}[0]. @var{clock}
8098 is the timer tick of the scheduler. @var{can_issue_more} is an output
8099 parameter that is set to the number of insns that can issue this clock;
8100 normally this is just @code{issue_rate}.
8102 @findex MD_SCHED_VARIABLE_ISSUE
8103 @item MD_SCHED_VARIABLE_ISSUE (@var{file}, @var{verbose}, @var{insn}, @var{more})
8104 A C statement which is executed by the scheduler after it
8105 has scheduled an insn from the ready list. @var{file} is either a null
8106 pointer, or a stdio stream to write any debug output to. @var{verbose}
8107 is the verbose level provided by @samp{-fsched-verbose-}@var{n}.
8108 @var{insn} is the instruction that was scheduled. @var{more} is the
8109 number of instructions that can be issued in the current cycle. The
8110 @samp{MD_SCHED_VARIABLE_ISSUE} macro is responsible for updating the
8111 value of @var{more} (typically by @var{more}--).
8113 @findex MAX_INTEGER_COMPUTATION_MODE
8114 @item MAX_INTEGER_COMPUTATION_MODE
8115 Define this to the largest integer machine mode which can be used for
8116 operations other than load, store and copy operations.
8118 You need only define this macro if the target holds values larger than
8119 @code{word_mode} in general purpose registers. Most targets should not define
8122 @findex MATH_LIBRARY
8124 Define this macro as a C string constant for the linker argument to link
8125 in the system math library, or @samp{""} if the target does not have a
8126 separate math library.
8128 You need only define this macro if the default of @samp{"-lm"} is wrong.
8130 @findex LIBRARY_PATH_ENV
8131 @item LIBRARY_PATH_ENV
8132 Define this macro as a C string constant for the environment variable that
8133 specifies where the linker should look for libraries.
8135 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
8138 @findex TARGET_HAS_F_SETLKW
8139 @item TARGET_HAS_F_SETLKW
8140 Define this macro if the target supports file locking with fcntl / F_SETLKW.
8141 Note that this functionality is part of POSIX.
8142 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
8143 to use file locking when exiting a program, which avoids race conditions
8144 if the program has forked.
8146 @findex MAX_CONDITIONAL_EXECUTE
8147 @item MAX_CONDITIONAL_EXECUTE
8149 A C expression for the maximum number of instructions to execute via
8150 conditional execution instructions instead of a branch. A value of
8151 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8152 1 if it does use cc0.
8154 @findex IFCVT_MODIFY_TESTS
8155 @item IFCVT_MODIFY_TESTS
8156 A C expression to modify the tests in @code{TRUE_EXPR}, and
8157 @code{FALSE_EXPPR} for use in converting insns in @code{TEST_BB},
8158 @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB} basic blocks to
8159 conditional execution. Set either @code{TRUE_EXPR} or @code{FALSE_EXPR}
8160 to a null pointer if the tests cannot be converted.
8162 @findex IFCVT_MODIFY_INSN
8163 @item IFCVT_MODIFY_INSN
8164 A C expression to modify the @code{PATTERN} of an @code{INSN} that is to
8165 be converted to conditional execution format.
8167 @findex IFCVT_MODIFY_FINAL
8168 @item IFCVT_MODIFY_FINAL
8169 A C expression to perform any final machine dependent modifications in
8170 converting code to conditional execution in the basic blocks
8171 @code{TEST_BB}, @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB}.
8173 @findex IFCVT_MODIFY_CANCEL
8174 @item IFCVT_MODIFY_CANCEL
8175 A C expression to cancel any machine dependent modifications in
8176 converting code to conditional execution in the basic blocks
8177 @code{TEST_BB}, @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB}.