1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,2002
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 and Functions
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} and a C source file named @file{@var{machine}.c}.
16 The header file defines numerous macros that convey the information
17 about the target machine that does not fit into the scheme of the
18 @file{.md} file. The file @file{tm.h} should be a link to
19 @file{@var{machine}.h}. The header file @file{config.h} includes
20 @file{tm.h} and most compiler source files include @file{config.h}. The
21 source file defines a variable @code{targetm}, which is a structure
22 containing pointers to functions and data relating to the target
23 machine. @file{@var{machine}.c} should also contain their definitions,
24 if they are not defined elsewhere in GCC, and other functions called
25 through the macros defined in the @file{.h} file.
28 * Target Structure:: The @code{targetm} variable.
29 * Driver:: Controlling how the driver runs the compilation passes.
30 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
31 * Per-Function Data:: Defining data structures for per-function information.
32 * Storage Layout:: Defining sizes and alignments of data.
33 * Type Layout:: Defining sizes and properties of basic user data types.
34 * Escape Sequences:: Defining the value of target character escape sequences
35 * Registers:: Naming and describing the hardware registers.
36 * Register Classes:: Defining the classes of hardware registers.
37 * Stack and Calling:: Defining which way the stack grows and by how much.
38 * Varargs:: Defining the varargs macros.
39 * Trampolines:: Code set up at run time to enter a nested function.
40 * Library Calls:: Controlling how library routines are implicitly called.
41 * Addressing Modes:: Defining addressing modes valid for memory operands.
42 * Condition Code:: Defining how insns update the condition code.
43 * Costs:: Defining relative costs of different operations.
44 * Scheduling:: Adjusting the behavior of the instruction scheduler.
45 * Sections:: Dividing storage into text, data, and other sections.
46 * PIC:: Macros for position independent code.
47 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
48 * Debugging Info:: Defining the format of debugging output.
49 * Floating Point:: Handling floating point for cross-compilers.
50 * Mode Switching:: Insertion of mode-switching instructions.
51 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
52 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
53 * Misc:: Everything else.
56 @node Target Structure
57 @section The Global @code{targetm} Variable
59 @cindex target functions
61 @deftypevar {struct gcc_target} targetm
62 The target @file{.c} file must define the global @code{targetm} variable
63 which contains pointers to functions and data relating to the target
64 machine. The variable is declared in @file{target.h};
65 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
66 used to initialize the variable, and macros for the default initializers
67 for elements of the structure. The @file{.c} file should override those
68 macros for which the default definition is inappropriate. For example:
71 #include "target-def.h"
73 /* @r{Initialize the GCC target structure.} */
75 #undef TARGET_COMP_TYPE_ATTRIBUTES
76 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
78 struct gcc_target targetm = TARGET_INITIALIZER;
82 Where a macro should be defined in the @file{.c} file in this manner to
83 form part of the @code{targetm} structure, it is documented below as a
84 ``Target Hook'' with a prototype. Many macros will change in future
85 from being defined in the @file{.h} file to being part of the
86 @code{targetm} structure.
89 @section Controlling the Compilation Driver, @file{gcc}
91 @cindex controlling the compilation driver
93 @c prevent bad page break with this line
94 You can control the compilation driver.
97 @findex SWITCH_TAKES_ARG
98 @item SWITCH_TAKES_ARG (@var{char})
99 A C expression which determines whether the option @option{-@var{char}}
100 takes arguments. The value should be the number of arguments that
101 option takes--zero, for many options.
103 By default, this macro is defined as
104 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
105 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
106 wish to add additional options which take arguments. Any redefinition
107 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
110 @findex WORD_SWITCH_TAKES_ARG
111 @item WORD_SWITCH_TAKES_ARG (@var{name})
112 A C expression which determines whether the option @option{-@var{name}}
113 takes arguments. The value should be the number of arguments that
114 option takes--zero, for many options. This macro rather than
115 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
117 By default, this macro is defined as
118 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
119 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
120 wish to add additional options which take arguments. Any redefinition
121 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
124 @findex SWITCH_CURTAILS_COMPILATION
125 @item SWITCH_CURTAILS_COMPILATION (@var{char})
126 A C expression which determines whether the option @option{-@var{char}}
127 stops compilation before the generation of an executable. The value is
128 boolean, nonzero if the option does stop an executable from being
129 generated, zero otherwise.
131 By default, this macro is defined as
132 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
133 options properly. You need not define
134 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
135 options which affect the generation of an executable. Any redefinition
136 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
137 for additional options.
139 @findex SWITCHES_NEED_SPACES
140 @item SWITCHES_NEED_SPACES
141 A string-valued C expression which enumerates the options for which
142 the linker needs a space between the option and its argument.
144 If this macro is not defined, the default value is @code{""}.
146 @findex TARGET_OPTION_TRANSLATE_TABLE
147 @item TARGET_OPTION_TRANSLATE_TABLE
148 If defined, a list of pairs of strings, the first of which is a
149 potential command line target to the @file{gcc} driver program, and the
150 second of which is a space-separated (tabs and other whitespace are not
151 supported) list of options with which to replace the first option. The
152 target defining this list is responsible for assuring that the results
153 are valid. Replacement options may not be the @code{--opt} style, they
154 must be the @code{-opt} style. It is the intention of this macro to
155 provide a mechanism for substitution that affects the multilibs chosen,
156 such as one option that enables many options, some of which select
157 multilibs. Example nonsensical definition, where @code{-malt-abi},
158 @code{-EB}, and @code{-mspoo} cause different multilibs to be chosen:
161 #define TARGET_OPTION_TRANSLATE_TABLE \
162 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
163 @{ "-compat", "-EB -malign=4 -mspoo" @}
168 A C string constant that tells the GCC driver program options to
169 pass to CPP@. It can also specify how to translate options you
170 give to GCC into options for GCC to pass to the CPP@.
172 Do not define this macro if it does not need to do anything.
174 @findex CPLUSPLUS_CPP_SPEC
175 @item CPLUSPLUS_CPP_SPEC
176 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
177 than C@. If you do not define this macro, then the value of
178 @code{CPP_SPEC} (if any) will be used instead.
182 A C string constant that tells the GCC driver program options to
183 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
185 It can also specify how to translate options you give to GCC into options
186 for GCC to pass to front ends.
188 Do not define this macro if it does not need to do anything.
192 A C string constant that tells the GCC driver program options to
193 pass to @code{cc1plus}. It can also specify how to translate options you
194 give to GCC into options for GCC to pass to the @code{cc1plus}.
196 Do not define this macro if it does not need to do anything.
197 Note that everything defined in CC1_SPEC is already passed to
198 @code{cc1plus} so there is no need to duplicate the contents of
199 CC1_SPEC in CC1PLUS_SPEC@.
203 A C string constant that tells the GCC driver program options to
204 pass to the assembler. It can also specify how to translate options
205 you give to GCC into options for GCC to pass to the assembler.
206 See the file @file{sun3.h} for an example of this.
208 Do not define this macro if it does not need to do anything.
210 @findex ASM_FINAL_SPEC
212 A C string constant that tells the GCC driver program how to
213 run any programs which cleanup after the normal assembler.
214 Normally, this is not needed. See the file @file{mips.h} for
217 Do not define this macro if it does not need to do anything.
221 A C string constant that tells the GCC driver program options to
222 pass to the linker. It can also specify how to translate options you
223 give to GCC into options for GCC to pass to the linker.
225 Do not define this macro if it does not need to do anything.
229 Another C string constant used much like @code{LINK_SPEC}. The difference
230 between the two is that @code{LIB_SPEC} is used at the end of the
231 command given to the linker.
233 If this macro is not defined, a default is provided that
234 loads the standard C library from the usual place. See @file{gcc.c}.
238 Another C string constant that tells the GCC driver program
239 how and when to place a reference to @file{libgcc.a} into the
240 linker command line. This constant is placed both before and after
241 the value of @code{LIB_SPEC}.
243 If this macro is not defined, the GCC driver provides a default that
244 passes the string @option{-lgcc} to the linker.
246 @findex STARTFILE_SPEC
248 Another C string constant used much like @code{LINK_SPEC}. The
249 difference between the two is that @code{STARTFILE_SPEC} is used at
250 the very beginning of the command given to the linker.
252 If this macro is not defined, a default is provided that loads the
253 standard C startup file from the usual place. See @file{gcc.c}.
257 Another C string constant used much like @code{LINK_SPEC}. The
258 difference between the two is that @code{ENDFILE_SPEC} is used at
259 the very end of the command given to the linker.
261 Do not define this macro if it does not need to do anything.
263 @findex THREAD_MODEL_SPEC
264 @item THREAD_MODEL_SPEC
265 GCC @code{-v} will print the thread model GCC was configured to use.
266 However, this doesn't work on platforms that are multilibbed on thread
267 models, such as AIX 4.3. On such platforms, define
268 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
269 blanks that names one of the recognized thread models. @code{%*}, the
270 default value of this macro, will expand to the value of
271 @code{thread_file} set in @file{config.gcc}.
275 Define this macro to provide additional specifications to put in the
276 @file{specs} file that can be used in various specifications like
279 The definition should be an initializer for an array of structures,
280 containing a string constant, that defines the specification name, and a
281 string constant that provides the specification.
283 Do not define this macro if it does not need to do anything.
285 @code{EXTRA_SPECS} is useful when an architecture contains several
286 related targets, which have various @code{@dots{}_SPECS} which are similar
287 to each other, and the maintainer would like one central place to keep
290 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
291 define either @code{_CALL_SYSV} when the System V calling sequence is
292 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
295 The @file{config/rs6000/rs6000.h} target file defines:
298 #define EXTRA_SPECS \
299 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
301 #define CPP_SYS_DEFAULT ""
304 The @file{config/rs6000/sysv.h} target file defines:
308 "%@{posix: -D_POSIX_SOURCE @} \
309 %@{mcall-sysv: -D_CALL_SYSV @} %@{mcall-aix: -D_CALL_AIX @} \
310 %@{!mcall-sysv: %@{!mcall-aix: %(cpp_sysv_default) @}@} \
311 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
313 #undef CPP_SYSV_DEFAULT
314 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
317 while the @file{config/rs6000/eabiaix.h} target file defines
318 @code{CPP_SYSV_DEFAULT} as:
321 #undef CPP_SYSV_DEFAULT
322 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
325 @findex LINK_LIBGCC_SPECIAL
326 @item LINK_LIBGCC_SPECIAL
327 Define this macro if the driver program should find the library
328 @file{libgcc.a} itself and should not pass @option{-L} options to the
329 linker. If you do not define this macro, the driver program will pass
330 the argument @option{-lgcc} to tell the linker to do the search and will
331 pass @option{-L} options to it.
333 @findex LINK_LIBGCC_SPECIAL_1
334 @item LINK_LIBGCC_SPECIAL_1
335 Define this macro if the driver program should find the library
336 @file{libgcc.a}. If you do not define this macro, the driver program will pass
337 the argument @option{-lgcc} to tell the linker to do the search.
338 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
339 not affect @option{-L} options.
341 @findex LINK_GCC_C_SEQUENCE_SPEC
342 @item LINK_GCC_C_SEQUENCE_SPEC
343 The sequence in which libgcc and libc are specified to the linker.
344 By default this is @code{%G %L %G}.
346 @findex LINK_COMMAND_SPEC
347 @item LINK_COMMAND_SPEC
348 A C string constant giving the complete command line need to execute the
349 linker. When you do this, you will need to update your port each time a
350 change is made to the link command line within @file{gcc.c}. Therefore,
351 define this macro only if you need to completely redefine the command
352 line for invoking the linker and there is no other way to accomplish
353 the effect you need. Overriding this macro may be avoidable by overriding
354 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
356 @findex LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
357 @item LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
358 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
359 directories from linking commands. Do not give it a nonzero value if
360 removing duplicate search directories changes the linker's semantics.
362 @findex MULTILIB_DEFAULTS
363 @item MULTILIB_DEFAULTS
364 Define this macro as a C expression for the initializer of an array of
365 string to tell the driver program which options are defaults for this
366 target and thus do not need to be handled specially when using
367 @code{MULTILIB_OPTIONS}.
369 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
370 the target makefile fragment or if none of the options listed in
371 @code{MULTILIB_OPTIONS} are set by default.
372 @xref{Target Fragment}.
374 @findex RELATIVE_PREFIX_NOT_LINKDIR
375 @item RELATIVE_PREFIX_NOT_LINKDIR
376 Define this macro to tell @code{gcc} that it should only translate
377 a @option{-B} prefix into a @option{-L} linker option if the prefix
378 indicates an absolute file name.
380 @findex STANDARD_EXEC_PREFIX
381 @item STANDARD_EXEC_PREFIX
382 Define this macro as a C string constant if you wish to override the
383 standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
384 try when searching for the executable files of the compiler.
386 @findex MD_EXEC_PREFIX
388 If defined, this macro is an additional prefix to try after
389 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
390 when the @option{-b} option is used, or the compiler is built as a cross
391 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
392 to the list of directories used to find the assembler in @file{configure.in}.
394 @findex STANDARD_STARTFILE_PREFIX
395 @item STANDARD_STARTFILE_PREFIX
396 Define this macro as a C string constant if you wish to override the
397 standard choice of @file{/usr/local/lib/} as the default prefix to
398 try when searching for startup files such as @file{crt0.o}.
400 @findex MD_STARTFILE_PREFIX
401 @item MD_STARTFILE_PREFIX
402 If defined, this macro supplies an additional prefix to try after the
403 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
404 @option{-b} option is used, or when the compiler is built as a cross
407 @findex MD_STARTFILE_PREFIX_1
408 @item MD_STARTFILE_PREFIX_1
409 If defined, this macro supplies yet another prefix to try after the
410 standard prefixes. It is not searched when the @option{-b} option is
411 used, or when the compiler is built as a cross compiler.
413 @findex INIT_ENVIRONMENT
414 @item INIT_ENVIRONMENT
415 Define this macro as a C string constant if you wish to set environment
416 variables for programs called by the driver, such as the assembler and
417 loader. The driver passes the value of this macro to @code{putenv} to
418 initialize the necessary environment variables.
420 @findex LOCAL_INCLUDE_DIR
421 @item LOCAL_INCLUDE_DIR
422 Define this macro as a C string constant if you wish to override the
423 standard choice of @file{/usr/local/include} as the default prefix to
424 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
425 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
427 Cross compilers do not search either @file{/usr/local/include} or its
430 @findex MODIFY_TARGET_NAME
431 @item MODIFY_TARGET_NAME
432 Define this macro if you with to define command-line switches that modify the
435 For each switch, you can include a string to be appended to the first
436 part of the configuration name or a string to be deleted from the
437 configuration name, if present. The definition should be an initializer
438 for an array of structures. Each array element should have three
439 elements: the switch name (a string constant, including the initial
440 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
441 indicate whether the string should be inserted or deleted, and the string
442 to be inserted or deleted (a string constant).
444 For example, on a machine where @samp{64} at the end of the
445 configuration name denotes a 64-bit target and you want the @option{-32}
446 and @option{-64} switches to select between 32- and 64-bit targets, you would
450 #define MODIFY_TARGET_NAME \
451 @{ @{ "-32", DELETE, "64"@}, \
452 @{"-64", ADD, "64"@}@}
456 @findex SYSTEM_INCLUDE_DIR
457 @item SYSTEM_INCLUDE_DIR
458 Define this macro as a C string constant if you wish to specify a
459 system-specific directory to search for header files before the standard
460 directory. @code{SYSTEM_INCLUDE_DIR} comes before
461 @code{STANDARD_INCLUDE_DIR} in the search order.
463 Cross compilers do not use this macro and do not search the directory
466 @findex STANDARD_INCLUDE_DIR
467 @item STANDARD_INCLUDE_DIR
468 Define this macro as a C string constant if you wish to override the
469 standard choice of @file{/usr/include} as the default prefix to
470 try when searching for header files.
472 Cross compilers do not use this macro and do not search either
473 @file{/usr/include} or its replacement.
475 @findex STANDARD_INCLUDE_COMPONENT
476 @item STANDARD_INCLUDE_COMPONENT
477 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
478 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
479 If you do not define this macro, no component is used.
481 @findex INCLUDE_DEFAULTS
482 @item INCLUDE_DEFAULTS
483 Define this macro if you wish to override the entire default search path
484 for include files. For a native compiler, the default search path
485 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
486 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
487 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
488 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
489 and specify private search areas for GCC@. The directory
490 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
492 The definition should be an initializer for an array of structures.
493 Each array element should have four elements: the directory name (a
494 string constant), the component name (also a string constant), a flag
495 for C++-only directories,
496 and a flag showing that the includes in the directory don't need to be
497 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
498 the array with a null element.
500 The component name denotes what GNU package the include file is part of,
501 if any, in all upper-case letters. For example, it might be @samp{GCC}
502 or @samp{BINUTILS}. If the package is part of a vendor-supplied
503 operating system, code the component name as @samp{0}.
505 For example, here is the definition used for VAX/VMS:
508 #define INCLUDE_DEFAULTS \
510 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
511 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
512 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
519 Here is the order of prefixes tried for exec files:
523 Any prefixes specified by the user with @option{-B}.
526 The environment variable @code{GCC_EXEC_PREFIX}, if any.
529 The directories specified by the environment variable @code{COMPILER_PATH}.
532 The macro @code{STANDARD_EXEC_PREFIX}.
535 @file{/usr/lib/gcc/}.
538 The macro @code{MD_EXEC_PREFIX}, if any.
541 Here is the order of prefixes tried for startfiles:
545 Any prefixes specified by the user with @option{-B}.
548 The environment variable @code{GCC_EXEC_PREFIX}, if any.
551 The directories specified by the environment variable @code{LIBRARY_PATH}
552 (or port-specific name; native only, cross compilers do not use this).
555 The macro @code{STANDARD_EXEC_PREFIX}.
558 @file{/usr/lib/gcc/}.
561 The macro @code{MD_EXEC_PREFIX}, if any.
564 The macro @code{MD_STARTFILE_PREFIX}, if any.
567 The macro @code{STANDARD_STARTFILE_PREFIX}.
576 @node Run-time Target
577 @section Run-time Target Specification
578 @cindex run-time target specification
579 @cindex predefined macros
580 @cindex target specifications
582 @c prevent bad page break with this line
583 Here are run-time target specifications.
586 @findex TARGET_CPU_CPP_BUILTINS
587 @item TARGET_CPU_CPP_BUILTINS()
588 This function-like macro expands to a block of code that defines
589 built-in preprocessor macros and assertions for the target cpu, using
590 the functions @code{builtin_macro}, @code{builtin_macro_std} and
591 @code{builtin_assert} declared in @file{c-lex.h}. When the front end
592 calls this macro it provides a trailing semicolon, and since it has
593 finished command line option processing your code can use those
596 @code{builtin_assert} takes a string in the form you pass to the
597 command-line option @option{-A}, such as @code{cpu=mips}, and creates
598 the assertion. @code{builtin_macro} takes a string in the form
599 accepted by option @option{-D} and unconditionally defines the macro.
601 @code{builtin_macro_std} takes a string representing the name of an
602 object-like macro. If it doesn't lie in the user's namespace,
603 @code{builtin_macro_std} defines it unconditionally. Otherwise, it
604 defines a version with two leading underscores, and another version
605 with two leading and trailing underscores, and defines the original
606 only if an ISO standard was not requested on the command line. For
607 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
608 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
609 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
610 defines only @code{_ABI64}.
612 With @code{TARGET_OS_CPP_BUILTINS} this macro obsoletes the
613 @code{CPP_PREDEFINES} target macro.
615 @findex TARGET_OS_CPP_BUILTINS
616 @item TARGET_OS_CPP_BUILTINS()
617 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
618 and is used for the target operating system instead.
620 With @code{TARGET_CPU_CPP_BUILTINS} this macro obsoletes the
621 @code{CPP_PREDEFINES} target macro.
623 @findex CPP_PREDEFINES
625 Define this to be a string constant containing @option{-D} options to
626 define the predefined macros that identify this machine and system.
627 These macros will be predefined unless the @option{-ansi} option (or a
628 @option{-std} option for strict ISO C conformance) is specified.
630 In addition, a parallel set of macros are predefined, whose names are
631 made by appending @samp{__} at the beginning and at the end. These
632 @samp{__} macros are permitted by the ISO standard, so they are
633 predefined regardless of whether @option{-ansi} or a @option{-std} option
636 For example, on the Sun, one can use the following value:
639 "-Dmc68000 -Dsun -Dunix"
642 The result is to define the macros @code{__mc68000__}, @code{__sun__}
643 and @code{__unix__} unconditionally, and the macros @code{mc68000},
644 @code{sun} and @code{unix} provided @option{-ansi} is not specified.
646 @findex extern int target_flags
647 @item extern int target_flags;
648 This declaration should be present.
650 @cindex optional hardware or system features
651 @cindex features, optional, in system conventions
653 This series of macros is to allow compiler command arguments to
654 enable or disable the use of optional features of the target machine.
655 For example, one machine description serves both the 68000 and
656 the 68020; a command argument tells the compiler whether it should
657 use 68020-only instructions or not. This command argument works
658 by means of a macro @code{TARGET_68020} that tests a bit in
661 Define a macro @code{TARGET_@var{featurename}} for each such option.
662 Its definition should test a bit in @code{target_flags}. It is
663 recommended that a helper macro @code{TARGET_MASK_@var{featurename}}
664 is defined for each bit-value to test, and used in
665 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
669 #define TARGET_MASK_68020 1
670 #define TARGET_68020 (target_flags & TARGET_MASK_68020)
673 One place where these macros are used is in the condition-expressions
674 of instruction patterns. Note how @code{TARGET_68020} appears
675 frequently in the 68000 machine description file, @file{m68k.md}.
676 Another place they are used is in the definitions of the other
677 macros in the @file{@var{machine}.h} file.
679 @findex TARGET_SWITCHES
680 @item TARGET_SWITCHES
681 This macro defines names of command options to set and clear
682 bits in @code{target_flags}. Its definition is an initializer
683 with a subgrouping for each command option.
685 Each subgrouping contains a string constant, that defines the option
686 name, a number, which contains the bits to set in
687 @code{target_flags}, and a second string which is the description
688 displayed by @option{--help}. If the number is negative then the bits specified
689 by the number are cleared instead of being set. If the description
690 string is present but empty, then no help information will be displayed
691 for that option, but it will not count as an undocumented option. The
692 actual option name is made by appending @samp{-m} to the specified name.
693 Non-empty description strings should be marked with @code{N_(@dots{})} for
694 @command{xgettext}. Please do not mark empty strings because the empty
695 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
696 of the message catalog with meta information, not the empty string.
698 In addition to the description for @option{--help},
699 more detailed documentation for each option should be added to
702 One of the subgroupings should have a null string. The number in
703 this grouping is the default value for @code{target_flags}. Any
704 target options act starting with that value.
706 Here is an example which defines @option{-m68000} and @option{-m68020}
707 with opposite meanings, and picks the latter as the default:
710 #define TARGET_SWITCHES \
711 @{ @{ "68020", TARGET_MASK_68020, "" @}, \
712 @{ "68000", -TARGET_MASK_68020, \
713 N_("Compile for the 68000") @}, \
714 @{ "", TARGET_MASK_68020, "" @}@}
717 @findex TARGET_OPTIONS
719 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
720 options that have values. Its definition is an initializer with a
721 subgrouping for each command option.
723 Each subgrouping contains a string constant, that defines the fixed part
724 of the option name, the address of a variable, and a description string.
725 Non-empty description strings should be marked with @code{N_(@dots{})} for
726 @command{xgettext}. Please do not mark empty strings because the empty
727 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
728 of the message catalog with meta information, not the empty string.
730 The variable, type @code{char *}, is set to the variable part of the
731 given option if the fixed part matches. The actual option name is made
732 by appending @samp{-m} to the specified name. Again, each option should
733 also be documented in @file{invoke.texi}.
735 Here is an example which defines @option{-mshort-data-@var{number}}. If the
736 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
737 will be set to the string @code{"512"}.
740 extern char *m88k_short_data;
741 #define TARGET_OPTIONS \
742 @{ @{ "short-data-", &m88k_short_data, \
743 N_("Specify the size of the short data section") @} @}
746 @findex TARGET_VERSION
748 This macro is a C statement to print on @code{stderr} a string
749 describing the particular machine description choice. Every machine
750 description should define @code{TARGET_VERSION}. For example:
754 #define TARGET_VERSION \
755 fprintf (stderr, " (68k, Motorola syntax)");
757 #define TARGET_VERSION \
758 fprintf (stderr, " (68k, MIT syntax)");
762 @findex OVERRIDE_OPTIONS
763 @item OVERRIDE_OPTIONS
764 Sometimes certain combinations of command options do not make sense on
765 a particular target machine. You can define a macro
766 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
767 defined, is executed once just after all the command options have been
770 Don't use this macro to turn on various extra optimizations for
771 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
773 @findex OPTIMIZATION_OPTIONS
774 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
775 Some machines may desire to change what optimizations are performed for
776 various optimization levels. This macro, if defined, is executed once
777 just after the optimization level is determined and before the remainder
778 of the command options have been parsed. Values set in this macro are
779 used as the default values for the other command line options.
781 @var{level} is the optimization level specified; 2 if @option{-O2} is
782 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
784 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
786 You should not use this macro to change options that are not
787 machine-specific. These should uniformly selected by the same
788 optimization level on all supported machines. Use this macro to enable
789 machine-specific optimizations.
791 @strong{Do not examine @code{write_symbols} in
792 this macro!} The debugging options are not supposed to alter the
795 @findex CAN_DEBUG_WITHOUT_FP
796 @item CAN_DEBUG_WITHOUT_FP
797 Define this macro if debugging can be performed even without a frame
798 pointer. If this macro is defined, GCC will turn on the
799 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
802 @node Per-Function Data
803 @section Defining data structures for per-function information.
804 @cindex per-function data
805 @cindex data structures
807 If the target needs to store information on a per-function basis, GCC
808 provides a macro and a couple of variables to allow this. Note, just
809 using statics to store the information is a bad idea, since GCC supports
810 nested functions, so you can be halfway through encoding one function
811 when another one comes along.
813 GCC defines a data structure called @code{struct function} which
814 contains all of the data specific to an individual function. This
815 structure contains a field called @code{machine} whose type is
816 @code{struct machine_function *}, which can be used by targets to point
817 to their own specific data.
819 If a target needs per-function specific data it should define the type
820 @code{struct machine_function} and also the macro
821 @code{INIT_EXPANDERS}. This macro should be used to initialize some or
822 all of the function pointers @code{init_machine_status},
823 @code{free_machine_status} and @code{mark_machine_status}. These
824 pointers are explained below.
826 One typical use of per-function, target specific data is to create an
827 RTX to hold the register containing the function's return address. This
828 RTX can then be used to implement the @code{__builtin_return_address}
829 function, for level 0.
831 Note---earlier implementations of GCC used a single data area to hold
832 all of the per-function information. Thus when processing of a nested
833 function began the old per-function data had to be pushed onto a
834 stack, and when the processing was finished, it had to be popped off the
835 stack. GCC used to provide function pointers called
836 @code{save_machine_status} and @code{restore_machine_status} to handle
837 the saving and restoring of the target specific information. Since the
838 single data area approach is no longer used, these pointers are no
841 The macro and function pointers are described below.
844 @findex INIT_EXPANDERS
846 Macro called to initialize any target specific information. This macro
847 is called once per function, before generation of any RTL has begun.
848 The intention of this macro is to allow the initialization of the
849 function pointers below.
851 @findex init_machine_status
852 @item init_machine_status
853 This is a @code{void (*)(struct function *)} function pointer. If this
854 pointer is non-@code{NULL} it will be called once per function, before function
855 compilation starts, in order to allow the target to perform any target
856 specific initialization of the @code{struct function} structure. It is
857 intended that this would be used to initialize the @code{machine} of
860 @findex free_machine_status
861 @item free_machine_status
862 This is a @code{void (*)(struct function *)} function pointer. If this
863 pointer is non-@code{NULL} it will be called once per function, after the
864 function has been compiled, in order to allow any memory allocated
865 during the @code{init_machine_status} function call to be freed.
867 @findex mark_machine_status
868 @item mark_machine_status
869 This is a @code{void (*)(struct function *)} function pointer. If this
870 pointer is non-@code{NULL} it will be called once per function in order to mark
871 any data items in the @code{struct machine_function} structure which
872 need garbage collection.
877 @section Storage Layout
878 @cindex storage layout
880 Note that the definitions of the macros in this table which are sizes or
881 alignments measured in bits do not need to be constant. They can be C
882 expressions that refer to static variables, such as the @code{target_flags}.
883 @xref{Run-time Target}.
886 @findex BITS_BIG_ENDIAN
887 @item BITS_BIG_ENDIAN
888 Define this macro to have the value 1 if the most significant bit in a
889 byte has the lowest number; otherwise define it to have the value zero.
890 This means that bit-field instructions count from the most significant
891 bit. If the machine has no bit-field instructions, then this must still
892 be defined, but it doesn't matter which value it is defined to. This
893 macro need not be a constant.
895 This macro does not affect the way structure fields are packed into
896 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
898 @findex BYTES_BIG_ENDIAN
899 @item BYTES_BIG_ENDIAN
900 Define this macro to have the value 1 if the most significant byte in a
901 word has the lowest number. This macro need not be a constant.
903 @findex WORDS_BIG_ENDIAN
904 @item WORDS_BIG_ENDIAN
905 Define this macro to have the value 1 if, in a multiword object, the
906 most significant word has the lowest number. This applies to both
907 memory locations and registers; GCC fundamentally assumes that the
908 order of words in memory is the same as the order in registers. This
909 macro need not be a constant.
911 @findex LIBGCC2_WORDS_BIG_ENDIAN
912 @item LIBGCC2_WORDS_BIG_ENDIAN
913 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
914 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
915 used only when compiling @file{libgcc2.c}. Typically the value will be set
916 based on preprocessor defines.
918 @findex FLOAT_WORDS_BIG_ENDIAN
919 @item FLOAT_WORDS_BIG_ENDIAN
920 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
921 @code{TFmode} floating point numbers are stored in memory with the word
922 containing the sign bit at the lowest address; otherwise define it to
923 have the value 0. This macro need not be a constant.
925 You need not define this macro if the ordering is the same as for
928 @findex BITS_PER_UNIT
930 Define this macro to be the number of bits in an addressable storage
931 unit (byte). If you do not define this macro the default is 8.
933 @findex BITS_PER_WORD
935 Number of bits in a word. If you do not define this macro, the default
936 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
938 @findex MAX_BITS_PER_WORD
939 @item MAX_BITS_PER_WORD
940 Maximum number of bits in a word. If this is undefined, the default is
941 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
942 largest value that @code{BITS_PER_WORD} can have at run-time.
944 @findex UNITS_PER_WORD
946 Number of storage units in a word; normally 4.
948 @findex MIN_UNITS_PER_WORD
949 @item MIN_UNITS_PER_WORD
950 Minimum number of units in a word. If this is undefined, the default is
951 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
952 smallest value that @code{UNITS_PER_WORD} can have at run-time.
956 Width of a pointer, in bits. You must specify a value no wider than the
957 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
958 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
959 a value the default is @code{BITS_PER_WORD}.
961 @findex POINTERS_EXTEND_UNSIGNED
962 @item POINTERS_EXTEND_UNSIGNED
963 A C expression whose value is greater than zero if pointers that need to be
964 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
965 be zero-extended and zero if they are to be sign-extended. If the value
966 is less then zero then there must be an "ptr_extend" instruction that
967 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
969 You need not define this macro if the @code{POINTER_SIZE} is equal
970 to the width of @code{Pmode}.
973 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
974 A macro to update @var{m} and @var{unsignedp} when an object whose type
975 is @var{type} and which has the specified mode and signedness is to be
976 stored in a register. This macro is only called when @var{type} is a
979 On most RISC machines, which only have operations that operate on a full
980 register, define this macro to set @var{m} to @code{word_mode} if
981 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
982 cases, only integer modes should be widened because wider-precision
983 floating-point operations are usually more expensive than their narrower
986 For most machines, the macro definition does not change @var{unsignedp}.
987 However, some machines, have instructions that preferentially handle
988 either signed or unsigned quantities of certain modes. For example, on
989 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
990 sign-extend the result to 64 bits. On such machines, set
991 @var{unsignedp} according to which kind of extension is more efficient.
993 Do not define this macro if it would never modify @var{m}.
995 @findex PROMOTE_FUNCTION_ARGS
996 @item PROMOTE_FUNCTION_ARGS
997 Define this macro if the promotion described by @code{PROMOTE_MODE}
998 should also be done for outgoing function arguments.
1000 @findex PROMOTE_FUNCTION_RETURN
1001 @item PROMOTE_FUNCTION_RETURN
1002 Define this macro if the promotion described by @code{PROMOTE_MODE}
1003 should also be done for the return value of functions.
1005 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
1006 promotions done by @code{PROMOTE_MODE}.
1008 @findex PROMOTE_FOR_CALL_ONLY
1009 @item PROMOTE_FOR_CALL_ONLY
1010 Define this macro if the promotion described by @code{PROMOTE_MODE}
1011 should @emph{only} be performed for outgoing function arguments or
1012 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
1013 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
1015 @findex PARM_BOUNDARY
1017 Normal alignment required for function parameters on the stack, in
1018 bits. All stack parameters receive at least this much alignment
1019 regardless of data type. On most machines, this is the same as the
1022 @findex STACK_BOUNDARY
1023 @item STACK_BOUNDARY
1024 Define this macro to the minimum alignment enforced by hardware for the
1025 stack pointer on this machine. The definition is a C expression for the
1026 desired alignment (measured in bits). This value is used as a default
1027 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1028 this should be the same as @code{PARM_BOUNDARY}.
1030 @findex PREFERRED_STACK_BOUNDARY
1031 @item PREFERRED_STACK_BOUNDARY
1032 Define this macro if you wish to preserve a certain alignment for the
1033 stack pointer, greater than what the hardware enforces. The definition
1034 is a C expression for the desired alignment (measured in bits). This
1035 macro must evaluate to a value equal to or larger than
1036 @code{STACK_BOUNDARY}.
1038 @findex FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1039 @item FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1040 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1041 not guaranteed by the runtime and we should emit code to align the stack
1042 at the beginning of @code{main}.
1044 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1045 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1046 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1047 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1048 be momentarily unaligned while pushing arguments.
1050 @findex FUNCTION_BOUNDARY
1051 @item FUNCTION_BOUNDARY
1052 Alignment required for a function entry point, in bits.
1054 @findex BIGGEST_ALIGNMENT
1055 @item BIGGEST_ALIGNMENT
1056 Biggest alignment that any data type can require on this machine, in bits.
1058 @findex MINIMUM_ATOMIC_ALIGNMENT
1059 @item MINIMUM_ATOMIC_ALIGNMENT
1060 If defined, the smallest alignment, in bits, that can be given to an
1061 object that can be referenced in one operation, without disturbing any
1062 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1063 on machines that don't have byte or half-word store operations.
1065 @findex BIGGEST_FIELD_ALIGNMENT
1066 @item BIGGEST_FIELD_ALIGNMENT
1067 Biggest alignment that any structure or union field can require on this
1068 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1069 structure and union fields only, unless the field alignment has been set
1070 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1072 @findex ADJUST_FIELD_ALIGN
1073 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1074 An expression for the alignment of a structure field @var{field} if the
1075 alignment computed in the usual way is @var{computed}. GCC uses
1076 this value instead of the value in @code{BIGGEST_ALIGNMENT} or
1077 @code{BIGGEST_FIELD_ALIGNMENT}, if defined.
1079 @findex MAX_OFILE_ALIGNMENT
1080 @item MAX_OFILE_ALIGNMENT
1081 Biggest alignment supported by the object file format of this machine.
1082 Use this macro to limit the alignment which can be specified using the
1083 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1084 the default value is @code{BIGGEST_ALIGNMENT}.
1086 @findex DATA_ALIGNMENT
1087 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
1088 If defined, a C expression to compute the alignment for a variable in
1089 the static store. @var{type} is the data type, and @var{basic-align} is
1090 the alignment that the object would ordinarily have. The value of this
1091 macro is used instead of that alignment to align the object.
1093 If this macro is not defined, then @var{basic-align} is used.
1096 One use of this macro is to increase alignment of medium-size data to
1097 make it all fit in fewer cache lines. Another is to cause character
1098 arrays to be word-aligned so that @code{strcpy} calls that copy
1099 constants to character arrays can be done inline.
1101 @findex CONSTANT_ALIGNMENT
1102 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1103 If defined, a C expression to compute the alignment given to a constant
1104 that is being placed in memory. @var{constant} is the constant and
1105 @var{basic-align} is the alignment that the object would ordinarily
1106 have. The value of this macro is used instead of that alignment to
1109 If this macro is not defined, then @var{basic-align} is used.
1111 The typical use of this macro is to increase alignment for string
1112 constants to be word aligned so that @code{strcpy} calls that copy
1113 constants can be done inline.
1115 @findex LOCAL_ALIGNMENT
1116 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1117 If defined, a C expression to compute the alignment for a variable in
1118 the local store. @var{type} is the data type, and @var{basic-align} is
1119 the alignment that the object would ordinarily have. The value of this
1120 macro is used instead of that alignment to align the object.
1122 If this macro is not defined, then @var{basic-align} is used.
1124 One use of this macro is to increase alignment of medium-size data to
1125 make it all fit in fewer cache lines.
1127 @findex EMPTY_FIELD_BOUNDARY
1128 @item EMPTY_FIELD_BOUNDARY
1129 Alignment in bits to be given to a structure bit-field that follows an
1130 empty field such as @code{int : 0;}.
1132 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
1133 that results from an empty field.
1135 @findex STRUCTURE_SIZE_BOUNDARY
1136 @item STRUCTURE_SIZE_BOUNDARY
1137 Number of bits which any structure or union's size must be a multiple of.
1138 Each structure or union's size is rounded up to a multiple of this.
1140 If you do not define this macro, the default is the same as
1141 @code{BITS_PER_UNIT}.
1143 @findex STRICT_ALIGNMENT
1144 @item STRICT_ALIGNMENT
1145 Define this macro to be the value 1 if instructions will fail to work
1146 if given data not on the nominal alignment. If instructions will merely
1147 go slower in that case, define this macro as 0.
1149 @findex PCC_BITFIELD_TYPE_MATTERS
1150 @item PCC_BITFIELD_TYPE_MATTERS
1151 Define this if you wish to imitate the way many other C compilers handle
1152 alignment of bit-fields and the structures that contain them.
1154 The behavior is that the type written for a bit-field (@code{int},
1155 @code{short}, or other integer type) imposes an alignment for the
1156 entire structure, as if the structure really did contain an ordinary
1157 field of that type. In addition, the bit-field is placed within the
1158 structure so that it would fit within such a field, not crossing a
1161 Thus, on most machines, a bit-field whose type is written as @code{int}
1162 would not cross a four-byte boundary, and would force four-byte
1163 alignment for the whole structure. (The alignment used may not be four
1164 bytes; it is controlled by the other alignment parameters.)
1166 If the macro is defined, its definition should be a C expression;
1167 a nonzero value for the expression enables this behavior.
1169 Note that if this macro is not defined, or its value is zero, some
1170 bit-fields may cross more than one alignment boundary. The compiler can
1171 support such references if there are @samp{insv}, @samp{extv}, and
1172 @samp{extzv} insns that can directly reference memory.
1174 The other known way of making bit-fields work is to define
1175 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1176 Then every structure can be accessed with fullwords.
1178 Unless the machine has bit-field instructions or you define
1179 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1180 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1182 If your aim is to make GCC use the same conventions for laying out
1183 bit-fields as are used by another compiler, here is how to investigate
1184 what the other compiler does. Compile and run this program:
1203 printf ("Size of foo1 is %d\n",
1204 sizeof (struct foo1));
1205 printf ("Size of foo2 is %d\n",
1206 sizeof (struct foo2));
1211 If this prints 2 and 5, then the compiler's behavior is what you would
1212 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1214 @findex BITFIELD_NBYTES_LIMITED
1215 @item BITFIELD_NBYTES_LIMITED
1216 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1217 to aligning a bit-field within the structure.
1219 @findex MEMBER_TYPE_FORCES_BLK
1220 @item MEMBER_TYPE_FORCES_BLK (@var{field})
1221 Return 1 if a structure or array containing @var{field} should be accessed using
1224 Normally, this is not needed. See the file @file{c4x.h} for an example
1225 of how to use this macro to prevent a structure having a floating point
1226 field from being accessed in an integer mode.
1228 @findex ROUND_TYPE_SIZE
1229 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
1230 Define this macro as an expression for the overall size of a type
1231 (given by @var{type} as a tree node) when the size computed in the
1232 usual way is @var{computed} and the alignment is @var{specified}.
1234 The default is to round @var{computed} up to a multiple of @var{specified}.
1236 @findex ROUND_TYPE_SIZE_UNIT
1237 @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified})
1238 Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are
1239 specified in units (bytes). If you define @code{ROUND_TYPE_SIZE},
1240 you must also define this macro and they must be defined consistently
1243 @findex ROUND_TYPE_ALIGN
1244 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1245 Define this macro as an expression for the alignment of a type (given
1246 by @var{type} as a tree node) if the alignment computed in the usual
1247 way is @var{computed} and the alignment explicitly specified was
1250 The default is to use @var{specified} if it is larger; otherwise, use
1251 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1253 @findex MAX_FIXED_MODE_SIZE
1254 @item MAX_FIXED_MODE_SIZE
1255 An integer expression for the size in bits of the largest integer
1256 machine mode that should actually be used. All integer machine modes of
1257 this size or smaller can be used for structures and unions with the
1258 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1259 (DImode)} is assumed.
1261 @findex VECTOR_MODE_SUPPORTED_P
1262 @item VECTOR_MODE_SUPPORTED_P(@var{mode})
1263 Define this macro to be nonzero if the port is prepared to handle insns
1264 involving vector mode @var{mode}. At the very least, it must have move
1265 patterns for this mode.
1267 @findex STACK_SAVEAREA_MODE
1268 @item STACK_SAVEAREA_MODE (@var{save_level})
1269 If defined, an expression of type @code{enum machine_mode} that
1270 specifies the mode of the save area operand of a
1271 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1272 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1273 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1274 having its mode specified.
1276 You need not define this macro if it always returns @code{Pmode}. You
1277 would most commonly define this macro if the
1278 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1281 @findex STACK_SIZE_MODE
1282 @item STACK_SIZE_MODE
1283 If defined, an expression of type @code{enum machine_mode} that
1284 specifies the mode of the size increment operand of an
1285 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1287 You need not define this macro if it always returns @code{word_mode}.
1288 You would most commonly define this macro if the @code{allocate_stack}
1289 pattern needs to support both a 32- and a 64-bit mode.
1291 @findex CHECK_FLOAT_VALUE
1292 @item CHECK_FLOAT_VALUE (@var{mode}, @var{value}, @var{overflow})
1293 A C statement to validate the value @var{value} (of type
1294 @code{double}) for mode @var{mode}. This means that you check whether
1295 @var{value} fits within the possible range of values for mode
1296 @var{mode} on this target machine. The mode @var{mode} is always
1297 a mode of class @code{MODE_FLOAT}. @var{overflow} is nonzero if
1298 the value is already known to be out of range.
1300 If @var{value} is not valid or if @var{overflow} is nonzero, you should
1301 set @var{overflow} to 1 and then assign some valid value to @var{value}.
1302 Allowing an invalid value to go through the compiler can produce
1303 incorrect assembler code which may even cause Unix assemblers to crash.
1305 This macro need not be defined if there is no work for it to do.
1307 @findex TARGET_FLOAT_FORMAT
1308 @item TARGET_FLOAT_FORMAT
1309 A code distinguishing the floating point format of the target machine.
1310 There are five defined values:
1313 @findex IEEE_FLOAT_FORMAT
1314 @item IEEE_FLOAT_FORMAT
1315 This code indicates IEEE floating point. It is the default; there is no
1316 need to define this macro when the format is IEEE@.
1318 @findex VAX_FLOAT_FORMAT
1319 @item VAX_FLOAT_FORMAT
1320 This code indicates the ``D float'' format used on the VAX@.
1322 @findex IBM_FLOAT_FORMAT
1323 @item IBM_FLOAT_FORMAT
1324 This code indicates the format used on the IBM System/370.
1326 @findex C4X_FLOAT_FORMAT
1327 @item C4X_FLOAT_FORMAT
1328 This code indicates the format used on the TMS320C3x/C4x.
1330 @findex UNKNOWN_FLOAT_FORMAT
1331 @item UNKNOWN_FLOAT_FORMAT
1332 This code indicates any other format.
1335 The value of this macro is compared with @code{HOST_FLOAT_FORMAT}, which
1336 is defined by the @command{configure} script, to determine whether the
1337 target machine has the same format as the host machine. If any other
1338 formats are actually in use on supported machines, new codes should be
1341 The ordering of the component words of floating point values stored in
1342 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1344 @findex MODE_HAS_NANS
1345 @item MODE_HAS_NANS (@var{mode})
1346 When defined, this macro should be true if @var{mode} has a NaN
1347 representation. The compiler assumes that NaNs are not equal to
1348 anything (including themselves) and that addition, subtraction,
1349 multiplication and division all return NaNs when one operand is
1352 By default, this macro is true if @var{mode} is a floating-point
1353 mode and the target floating-point format is IEEE@.
1355 @findex MODE_HAS_INFINITIES
1356 @item MODE_HAS_INFINITIES (@var{mode})
1357 This macro should be true if @var{mode} can represent infinity. At
1358 present, the compiler uses this macro to decide whether @samp{x - x}
1359 is always defined. By default, the macro is true when @var{mode}
1360 is a floating-point mode and the target format is IEEE@.
1362 @findex MODE_HAS_SIGNED_ZEROS
1363 @item MODE_HAS_SIGNED_ZEROS (@var{mode})
1364 True if @var{mode} distinguishes between positive and negative zero.
1365 The rules are expected to follow the IEEE standard:
1369 @samp{x + x} has the same sign as @samp{x}.
1372 If the sum of two values with opposite sign is zero, the result is
1373 positive for all rounding modes expect towards @minus{}infinity, for
1374 which it is negative.
1377 The sign of a product or quotient is negative when exactly one
1378 of the operands is negative.
1381 The default definition is true if @var{mode} is a floating-point
1382 mode and the target format is IEEE@.
1384 @findex MODE_HAS_SIGN_DEPENDENT_ROUNDING
1385 @item MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1386 If defined, this macro should be true for @var{mode} if it has at
1387 least one rounding mode in which @samp{x} and @samp{-x} can be
1388 rounded to numbers of different magnitude. Two such modes are
1389 towards @minus{}infinity and towards +infinity.
1391 The default definition of this macro is true if @var{mode} is
1392 a floating-point mode and the target format is IEEE@.
1394 @findex ROUND_TOWARDS_ZERO
1395 @item ROUND_TOWARDS_ZERO
1396 If defined, this macro should be true if the prevailing rounding
1397 mode is towards zero. A true value has the following effects:
1401 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1404 @file{libgcc.a}'s floating-point emulator will round towards zero
1405 rather than towards nearest.
1408 The compiler's floating-point emulator will round towards zero after
1409 doing arithmetic, and when converting from the internal float format to
1413 The macro does not affect the parsing of string literals. When the
1414 primary rounding mode is towards zero, library functions like
1415 @code{strtod} might still round towards nearest, and the compiler's
1416 parser should behave like the target's @code{strtod} where possible.
1418 Not defining this macro is equivalent to returning zero.
1420 @findex LARGEST_EXPONENT_IS_NORMAL
1421 @item LARGEST_EXPONENT_IS_NORMAL (@var{size})
1422 This macro should only be defined when the target float format is
1423 described as IEEE@. It should return true if floats with @var{size}
1424 bits do not have a NaN or infinity representation, but use the largest
1425 exponent for normal numbers instead.
1427 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1428 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1429 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1430 floating-point arithmetic.
1432 The default definition of this macro returns false for all sizes.
1435 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1436 This target hook returns @code{true} if bit-fields in the given
1437 @var{record_type} are to be laid out following the rules of Microsoft
1438 Visual C/C++, namely: (i) a bit-field won't share the same storage
1439 unit with the previous bit-field if their underlying types have
1440 different sizes, and the bit-field will be aligned to the highest
1441 alignment of the underlying types of itself and of the previous
1442 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1443 the whole enclosing structure, even if it is unnamed; except that
1444 (iii) a zero-sized bit-field will be disregarded unless it follows
1445 another bit-field of non-zero size. If this hook returns @code{true},
1446 other macros that control bit-field layout are ignored.
1450 @section Layout of Source Language Data Types
1452 These macros define the sizes and other characteristics of the standard
1453 basic data types used in programs being compiled. Unlike the macros in
1454 the previous section, these apply to specific features of C and related
1455 languages, rather than to fundamental aspects of storage layout.
1458 @findex INT_TYPE_SIZE
1460 A C expression for the size in bits of the type @code{int} on the
1461 target machine. If you don't define this, the default is one word.
1463 @findex SHORT_TYPE_SIZE
1464 @item SHORT_TYPE_SIZE
1465 A C expression for the size in bits of the type @code{short} on the
1466 target machine. If you don't define this, the default is half a word.
1467 (If this would be less than one storage unit, it is rounded up to one
1470 @findex LONG_TYPE_SIZE
1471 @item LONG_TYPE_SIZE
1472 A C expression for the size in bits of the type @code{long} on the
1473 target machine. If you don't define this, the default is one word.
1475 @findex ADA_LONG_TYPE_SIZE
1476 @item ADA_LONG_TYPE_SIZE
1477 On some machines, the size used for the Ada equivalent of the type
1478 @code{long} by a native Ada compiler differs from that used by C. In
1479 that situation, define this macro to be a C expression to be used for
1480 the size of that type. If you don't define this, the default is the
1481 value of @code{LONG_TYPE_SIZE}.
1483 @findex MAX_LONG_TYPE_SIZE
1484 @item MAX_LONG_TYPE_SIZE
1485 Maximum number for the size in bits of the type @code{long} on the
1486 target machine. If this is undefined, the default is
1487 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1488 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1491 @findex LONG_LONG_TYPE_SIZE
1492 @item LONG_LONG_TYPE_SIZE
1493 A C expression for the size in bits of the type @code{long long} on the
1494 target machine. If you don't define this, the default is two
1495 words. If you want to support GNU Ada on your machine, the value of this
1496 macro must be at least 64.
1498 @findex CHAR_TYPE_SIZE
1499 @item CHAR_TYPE_SIZE
1500 A C expression for the size in bits of the type @code{char} on the
1501 target machine. If you don't define this, the default is
1502 @code{BITS_PER_UNIT}.
1504 @findex BOOL_TYPE_SIZE
1505 @item BOOL_TYPE_SIZE
1506 A C expression for the size in bits of the C++ type @code{bool} and
1507 C99 type @code{_Bool} on the target machine. If you don't define
1508 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1510 @findex FLOAT_TYPE_SIZE
1511 @item FLOAT_TYPE_SIZE
1512 A C expression for the size in bits of the type @code{float} on the
1513 target machine. If you don't define this, the default is one word.
1515 @findex DOUBLE_TYPE_SIZE
1516 @item DOUBLE_TYPE_SIZE
1517 A C expression for the size in bits of the type @code{double} on the
1518 target machine. If you don't define this, the default is two
1521 @findex LONG_DOUBLE_TYPE_SIZE
1522 @item LONG_DOUBLE_TYPE_SIZE
1523 A C expression for the size in bits of the type @code{long double} on
1524 the target machine. If you don't define this, the default is two
1527 @findex MAX_LONG_DOUBLE_TYPE_SIZE
1528 Maximum number for the size in bits of the type @code{long double} on the
1529 target machine. If this is undefined, the default is
1530 @code{LONG_DOUBLE_TYPE_SIZE}. Otherwise, it is the constant value that is
1531 the largest value that @code{LONG_DOUBLE_TYPE_SIZE} can have at run-time.
1532 This is used in @code{cpp}.
1534 @findex INTEL_EXTENDED_IEEE_FORMAT
1535 Define this macro to be 1 if the target machine uses 80-bit floating-point
1536 values with 128-bit size and alignment. This is used in @file{real.c}.
1538 @findex WIDEST_HARDWARE_FP_SIZE
1539 @item WIDEST_HARDWARE_FP_SIZE
1540 A C expression for the size in bits of the widest floating-point format
1541 supported by the hardware. If you define this macro, you must specify a
1542 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1543 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1546 @findex DEFAULT_SIGNED_CHAR
1547 @item DEFAULT_SIGNED_CHAR
1548 An expression whose value is 1 or 0, according to whether the type
1549 @code{char} should be signed or unsigned by default. The user can
1550 always override this default with the options @option{-fsigned-char}
1551 and @option{-funsigned-char}.
1553 @findex DEFAULT_SHORT_ENUMS
1554 @item DEFAULT_SHORT_ENUMS
1555 A C expression to determine whether to give an @code{enum} type
1556 only as many bytes as it takes to represent the range of possible values
1557 of that type. A nonzero value means to do that; a zero value means all
1558 @code{enum} types should be allocated like @code{int}.
1560 If you don't define the macro, the default is 0.
1564 A C expression for a string describing the name of the data type to use
1565 for size values. The typedef name @code{size_t} is defined using the
1566 contents of the string.
1568 The string can contain more than one keyword. If so, separate them with
1569 spaces, and write first any length keyword, then @code{unsigned} if
1570 appropriate, and finally @code{int}. The string must exactly match one
1571 of the data type names defined in the function
1572 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1573 omit @code{int} or change the order---that would cause the compiler to
1576 If you don't define this macro, the default is @code{"long unsigned
1579 @findex PTRDIFF_TYPE
1581 A C expression for a string describing the name of the data type to use
1582 for the result of subtracting two pointers. The typedef name
1583 @code{ptrdiff_t} is defined using the contents of the string. See
1584 @code{SIZE_TYPE} above for more information.
1586 If you don't define this macro, the default is @code{"long int"}.
1590 A C expression for a string describing the name of the data type to use
1591 for wide characters. The typedef name @code{wchar_t} is defined using
1592 the contents of the string. See @code{SIZE_TYPE} above for more
1595 If you don't define this macro, the default is @code{"int"}.
1597 @findex WCHAR_TYPE_SIZE
1598 @item WCHAR_TYPE_SIZE
1599 A C expression for the size in bits of the data type for wide
1600 characters. This is used in @code{cpp}, which cannot make use of
1603 @findex MAX_WCHAR_TYPE_SIZE
1604 @item MAX_WCHAR_TYPE_SIZE
1605 Maximum number for the size in bits of the data type for wide
1606 characters. If this is undefined, the default is
1607 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1608 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1611 @findex GCOV_TYPE_SIZE
1612 @item GCOV_TYPE_SIZE
1613 A C expression for the size in bits of the type used for gcov counters on the
1614 target machine. If you don't define this, the default is one
1615 @code{LONG_TYPE_SIZE} in case it is greater or equal to 64-bit and
1616 @code{LONG_LONG_TYPE_SIZE} otherwise. You may want to re-define the type to
1617 ensure atomicity for counters in multithreaded programs.
1621 A C expression for a string describing the name of the data type to
1622 use for wide characters passed to @code{printf} and returned from
1623 @code{getwc}. The typedef name @code{wint_t} is defined using the
1624 contents of the string. See @code{SIZE_TYPE} above for more
1627 If you don't define this macro, the default is @code{"unsigned int"}.
1631 A C expression for a string describing the name of the data type that
1632 can represent any value of any standard or extended signed integer type.
1633 The typedef name @code{intmax_t} is defined using the contents of the
1634 string. See @code{SIZE_TYPE} above for more information.
1636 If you don't define this macro, the default is the first of
1637 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1638 much precision as @code{long long int}.
1640 @findex UINTMAX_TYPE
1642 A C expression for a string describing the name of the data type that
1643 can represent any value of any standard or extended unsigned integer
1644 type. The typedef name @code{uintmax_t} is defined using the contents
1645 of the string. See @code{SIZE_TYPE} above for more information.
1647 If you don't define this macro, the default is the first of
1648 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1649 unsigned int"} that has as much precision as @code{long long unsigned
1652 @findex TARGET_PTRMEMFUNC_VBIT_LOCATION
1653 @item TARGET_PTRMEMFUNC_VBIT_LOCATION
1654 The C++ compiler represents a pointer-to-member-function with a struct
1661 ptrdiff_t vtable_index;
1668 The C++ compiler must use one bit to indicate whether the function that
1669 will be called through a pointer-to-member-function is virtual.
1670 Normally, we assume that the low-order bit of a function pointer must
1671 always be zero. Then, by ensuring that the vtable_index is odd, we can
1672 distinguish which variant of the union is in use. But, on some
1673 platforms function pointers can be odd, and so this doesn't work. In
1674 that case, we use the low-order bit of the @code{delta} field, and shift
1675 the remainder of the @code{delta} field to the left.
1677 GCC will automatically make the right selection about where to store
1678 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1679 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1680 set such that functions always start at even addresses, but the lowest
1681 bit of pointers to functions indicate whether the function at that
1682 address is in ARM or Thumb mode. If this is the case of your
1683 architecture, you should define this macro to
1684 @code{ptrmemfunc_vbit_in_delta}.
1686 In general, you should not have to define this macro. On architectures
1687 in which function addresses are always even, according to
1688 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1689 @code{ptrmemfunc_vbit_in_pfn}.
1691 @findex TARGET_VTABLE_USES_DESCRIPTORS
1692 @item TARGET_VTABLE_USES_DESCRIPTORS
1693 Normally, the C++ compiler uses function pointers in vtables. This
1694 macro allows the target to change to use ``function descriptors''
1695 instead. Function descriptors are found on targets for whom a
1696 function pointer is actually a small data structure. Normally the
1697 data structure consists of the actual code address plus a data
1698 pointer to which the function's data is relative.
1700 If vtables are used, the value of this macro should be the number
1701 of words that the function descriptor occupies.
1704 @node Escape Sequences
1705 @section Target Character Escape Sequences
1706 @cindex escape sequences
1708 By default, GCC assumes that the C character escape sequences take on
1709 their ASCII values for the target. If this is not correct, you must
1710 explicitly define all of the macros below.
1715 A C constant expression for the integer value for escape sequence
1720 A C constant expression for the integer value of the target escape
1721 character. As an extension, GCC evaluates the escape sequences
1722 @samp{\e} and @samp{\E} to this.
1726 @findex TARGET_NEWLINE
1729 @itemx TARGET_NEWLINE
1730 C constant expressions for the integer values for escape sequences
1731 @samp{\b}, @samp{\t} and @samp{\n}.
1739 C constant expressions for the integer values for escape sequences
1740 @samp{\v}, @samp{\f} and @samp{\r}.
1744 @section Register Usage
1745 @cindex register usage
1747 This section explains how to describe what registers the target machine
1748 has, and how (in general) they can be used.
1750 The description of which registers a specific instruction can use is
1751 done with register classes; see @ref{Register Classes}. For information
1752 on using registers to access a stack frame, see @ref{Frame Registers}.
1753 For passing values in registers, see @ref{Register Arguments}.
1754 For returning values in registers, see @ref{Scalar Return}.
1757 * Register Basics:: Number and kinds of registers.
1758 * Allocation Order:: Order in which registers are allocated.
1759 * Values in Registers:: What kinds of values each reg can hold.
1760 * Leaf Functions:: Renumbering registers for leaf functions.
1761 * Stack Registers:: Handling a register stack such as 80387.
1764 @node Register Basics
1765 @subsection Basic Characteristics of Registers
1767 @c prevent bad page break with this line
1768 Registers have various characteristics.
1771 @findex FIRST_PSEUDO_REGISTER
1772 @item FIRST_PSEUDO_REGISTER
1773 Number of hardware registers known to the compiler. They receive
1774 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1775 pseudo register's number really is assigned the number
1776 @code{FIRST_PSEUDO_REGISTER}.
1778 @item FIXED_REGISTERS
1779 @findex FIXED_REGISTERS
1780 @cindex fixed register
1781 An initializer that says which registers are used for fixed purposes
1782 all throughout the compiled code and are therefore not available for
1783 general allocation. These would include the stack pointer, the frame
1784 pointer (except on machines where that can be used as a general
1785 register when no frame pointer is needed), the program counter on
1786 machines where that is considered one of the addressable registers,
1787 and any other numbered register with a standard use.
1789 This information is expressed as a sequence of numbers, separated by
1790 commas and surrounded by braces. The @var{n}th number is 1 if
1791 register @var{n} is fixed, 0 otherwise.
1793 The table initialized from this macro, and the table initialized by
1794 the following one, may be overridden at run time either automatically,
1795 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1796 the user with the command options @option{-ffixed-@var{reg}},
1797 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1799 @findex CALL_USED_REGISTERS
1800 @item CALL_USED_REGISTERS
1801 @cindex call-used register
1802 @cindex call-clobbered register
1803 @cindex call-saved register
1804 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1805 clobbered (in general) by function calls as well as for fixed
1806 registers. This macro therefore identifies the registers that are not
1807 available for general allocation of values that must live across
1810 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1811 automatically saves it on function entry and restores it on function
1812 exit, if the register is used within the function.
1814 @findex CALL_REALLY_USED_REGISTERS
1815 @item CALL_REALLY_USED_REGISTERS
1816 @cindex call-used register
1817 @cindex call-clobbered register
1818 @cindex call-saved register
1819 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1820 that the entire set of @code{FIXED_REGISTERS} be included.
1821 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1822 This macro is optional. If not specified, it defaults to the value
1823 of @code{CALL_USED_REGISTERS}.
1825 @findex HARD_REGNO_CALL_PART_CLOBBERED
1826 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1827 @cindex call-used register
1828 @cindex call-clobbered register
1829 @cindex call-saved register
1830 A C expression that is nonzero if it is not permissible to store a
1831 value of mode @var{mode} in hard register number @var{regno} across a
1832 call without some part of it being clobbered. For most machines this
1833 macro need not be defined. It is only required for machines that do not
1834 preserve the entire contents of a register across a call.
1836 @findex CONDITIONAL_REGISTER_USAGE
1838 @findex call_used_regs
1839 @item CONDITIONAL_REGISTER_USAGE
1840 Zero or more C statements that may conditionally modify five variables
1841 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1842 @code{reg_names}, and @code{reg_class_contents}, to take into account
1843 any dependence of these register sets on target flags. The first three
1844 of these are of type @code{char []} (interpreted as Boolean vectors).
1845 @code{global_regs} is a @code{const char *[]}, and
1846 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1847 called, @code{fixed_regs}, @code{call_used_regs},
1848 @code{reg_class_contents}, and @code{reg_names} have been initialized
1849 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1850 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1851 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1852 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1853 command options have been applied.
1855 You need not define this macro if it has no work to do.
1857 @cindex disabling certain registers
1858 @cindex controlling register usage
1859 If the usage of an entire class of registers depends on the target
1860 flags, you may indicate this to GCC by using this macro to modify
1861 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1862 registers in the classes which should not be used by GCC@. Also define
1863 the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1864 is called with a letter for a class that shouldn't be used.
1866 (However, if this class is not included in @code{GENERAL_REGS} and all
1867 of the insn patterns whose constraints permit this class are
1868 controlled by target switches, then GCC will automatically avoid using
1869 these registers when the target switches are opposed to them.)
1871 @findex NON_SAVING_SETJMP
1872 @item NON_SAVING_SETJMP
1873 If this macro is defined and has a nonzero value, it means that
1874 @code{setjmp} and related functions fail to save the registers, or that
1875 @code{longjmp} fails to restore them. To compensate, the compiler
1876 avoids putting variables in registers in functions that use
1879 @findex INCOMING_REGNO
1880 @item INCOMING_REGNO (@var{out})
1881 Define this macro if the target machine has register windows. This C
1882 expression returns the register number as seen by the called function
1883 corresponding to the register number @var{out} as seen by the calling
1884 function. Return @var{out} if register number @var{out} is not an
1887 @findex OUTGOING_REGNO
1888 @item OUTGOING_REGNO (@var{in})
1889 Define this macro if the target machine has register windows. This C
1890 expression returns the register number as seen by the calling function
1891 corresponding to the register number @var{in} as seen by the called
1892 function. Return @var{in} if register number @var{in} is not an inbound
1896 @item LOCAL_REGNO (@var{regno})
1897 Define this macro if the target machine has register windows. This C
1898 expression returns true if the register is call-saved but is in the
1899 register window. Unlike most call-saved registers, such registers
1900 need not be explicitly restored on function exit or during non-local
1906 If the program counter has a register number, define this as that
1907 register number. Otherwise, do not define it.
1911 @node Allocation Order
1912 @subsection Order of Allocation of Registers
1913 @cindex order of register allocation
1914 @cindex register allocation order
1916 @c prevent bad page break with this line
1917 Registers are allocated in order.
1920 @findex REG_ALLOC_ORDER
1921 @item REG_ALLOC_ORDER
1922 If defined, an initializer for a vector of integers, containing the
1923 numbers of hard registers in the order in which GCC should prefer
1924 to use them (from most preferred to least).
1926 If this macro is not defined, registers are used lowest numbered first
1927 (all else being equal).
1929 One use of this macro is on machines where the highest numbered
1930 registers must always be saved and the save-multiple-registers
1931 instruction supports only sequences of consecutive registers. On such
1932 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1933 the highest numbered allocable register first.
1935 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1936 @item ORDER_REGS_FOR_LOCAL_ALLOC
1937 A C statement (sans semicolon) to choose the order in which to allocate
1938 hard registers for pseudo-registers local to a basic block.
1940 Store the desired register order in the array @code{reg_alloc_order}.
1941 Element 0 should be the register to allocate first; element 1, the next
1942 register; and so on.
1944 The macro body should not assume anything about the contents of
1945 @code{reg_alloc_order} before execution of the macro.
1947 On most machines, it is not necessary to define this macro.
1950 @node Values in Registers
1951 @subsection How Values Fit in Registers
1953 This section discusses the macros that describe which kinds of values
1954 (specifically, which machine modes) each register can hold, and how many
1955 consecutive registers are needed for a given mode.
1958 @findex HARD_REGNO_NREGS
1959 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1960 A C expression for the number of consecutive hard registers, starting
1961 at register number @var{regno}, required to hold a value of mode
1964 On a machine where all registers are exactly one word, a suitable
1965 definition of this macro is
1968 #define HARD_REGNO_NREGS(REGNO, MODE) \
1969 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1973 @findex HARD_REGNO_MODE_OK
1974 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1975 A C expression that is nonzero if it is permissible to store a value
1976 of mode @var{mode} in hard register number @var{regno} (or in several
1977 registers starting with that one). For a machine where all registers
1978 are equivalent, a suitable definition is
1981 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1984 You need not include code to check for the numbers of fixed registers,
1985 because the allocation mechanism considers them to be always occupied.
1987 @cindex register pairs
1988 On some machines, double-precision values must be kept in even/odd
1989 register pairs. You can implement that by defining this macro to reject
1990 odd register numbers for such modes.
1992 The minimum requirement for a mode to be OK in a register is that the
1993 @samp{mov@var{mode}} instruction pattern support moves between the
1994 register and other hard register in the same class and that moving a
1995 value into the register and back out not alter it.
1997 Since the same instruction used to move @code{word_mode} will work for
1998 all narrower integer modes, it is not necessary on any machine for
1999 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2000 you define patterns @samp{movhi}, etc., to take advantage of this. This
2001 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2002 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2005 Many machines have special registers for floating point arithmetic.
2006 Often people assume that floating point machine modes are allowed only
2007 in floating point registers. This is not true. Any registers that
2008 can hold integers can safely @emph{hold} a floating point machine
2009 mode, whether or not floating arithmetic can be done on it in those
2010 registers. Integer move instructions can be used to move the values.
2012 On some machines, though, the converse is true: fixed-point machine
2013 modes may not go in floating registers. This is true if the floating
2014 registers normalize any value stored in them, because storing a
2015 non-floating value there would garble it. In this case,
2016 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2017 floating registers. But if the floating registers do not automatically
2018 normalize, if you can store any bit pattern in one and retrieve it
2019 unchanged without a trap, then any machine mode may go in a floating
2020 register, so you can define this macro to say so.
2022 The primary significance of special floating registers is rather that
2023 they are the registers acceptable in floating point arithmetic
2024 instructions. However, this is of no concern to
2025 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2026 constraints for those instructions.
2028 On some machines, the floating registers are especially slow to access,
2029 so that it is better to store a value in a stack frame than in such a
2030 register if floating point arithmetic is not being done. As long as the
2031 floating registers are not in class @code{GENERAL_REGS}, they will not
2032 be used unless some pattern's constraint asks for one.
2034 @findex MODES_TIEABLE_P
2035 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2036 A C expression that is nonzero if a value of mode
2037 @var{mode1} is accessible in mode @var{mode2} without copying.
2039 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2040 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2041 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2042 should be nonzero. If they differ for any @var{r}, you should define
2043 this macro to return zero unless some other mechanism ensures the
2044 accessibility of the value in a narrower mode.
2046 You should define this macro to return nonzero in as many cases as
2047 possible since doing so will allow GCC to perform better register
2050 @findex AVOID_CCMODE_COPIES
2051 @item AVOID_CCMODE_COPIES
2052 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2053 registers. You should only define this macro if support for copying to/from
2054 @code{CCmode} is incomplete.
2057 @node Leaf Functions
2058 @subsection Handling Leaf Functions
2060 @cindex leaf functions
2061 @cindex functions, leaf
2062 On some machines, a leaf function (i.e., one which makes no calls) can run
2063 more efficiently if it does not make its own register window. Often this
2064 means it is required to receive its arguments in the registers where they
2065 are passed by the caller, instead of the registers where they would
2068 The special treatment for leaf functions generally applies only when
2069 other conditions are met; for example, often they may use only those
2070 registers for its own variables and temporaries. We use the term ``leaf
2071 function'' to mean a function that is suitable for this special
2072 handling, so that functions with no calls are not necessarily ``leaf
2075 GCC assigns register numbers before it knows whether the function is
2076 suitable for leaf function treatment. So it needs to renumber the
2077 registers in order to output a leaf function. The following macros
2081 @findex LEAF_REGISTERS
2082 @item LEAF_REGISTERS
2083 Name of a char vector, indexed by hard register number, which
2084 contains 1 for a register that is allowable in a candidate for leaf
2087 If leaf function treatment involves renumbering the registers, then the
2088 registers marked here should be the ones before renumbering---those that
2089 GCC would ordinarily allocate. The registers which will actually be
2090 used in the assembler code, after renumbering, should not be marked with 1
2093 Define this macro only if the target machine offers a way to optimize
2094 the treatment of leaf functions.
2096 @findex LEAF_REG_REMAP
2097 @item LEAF_REG_REMAP (@var{regno})
2098 A C expression whose value is the register number to which @var{regno}
2099 should be renumbered, when a function is treated as a leaf function.
2101 If @var{regno} is a register number which should not appear in a leaf
2102 function before renumbering, then the expression should yield @minus{}1, which
2103 will cause the compiler to abort.
2105 Define this macro only if the target machine offers a way to optimize the
2106 treatment of leaf functions, and registers need to be renumbered to do
2110 @findex current_function_is_leaf
2111 @findex current_function_uses_only_leaf_regs
2112 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2113 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2114 specially. They can test the C variable @code{current_function_is_leaf}
2115 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2116 set prior to local register allocation and is valid for the remaining
2117 compiler passes. They can also test the C variable
2118 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2119 functions which only use leaf registers.
2120 @code{current_function_uses_only_leaf_regs} is valid after reload and is
2121 only useful if @code{LEAF_REGISTERS} is defined.
2122 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2123 @c of the next paragraph?! --mew 2feb93
2125 @node Stack Registers
2126 @subsection Registers That Form a Stack
2128 There are special features to handle computers where some of the
2129 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
2130 Stack registers are normally written by pushing onto the stack, and are
2131 numbered relative to the top of the stack.
2133 Currently, GCC can only handle one group of stack-like registers, and
2134 they must be consecutively numbered.
2139 Define this if the machine has any stack-like registers.
2141 @findex FIRST_STACK_REG
2142 @item FIRST_STACK_REG
2143 The number of the first stack-like register. This one is the top
2146 @findex LAST_STACK_REG
2147 @item LAST_STACK_REG
2148 The number of the last stack-like register. This one is the bottom of
2152 @node Register Classes
2153 @section Register Classes
2154 @cindex register class definitions
2155 @cindex class definitions, register
2157 On many machines, the numbered registers are not all equivalent.
2158 For example, certain registers may not be allowed for indexed addressing;
2159 certain registers may not be allowed in some instructions. These machine
2160 restrictions are described to the compiler using @dfn{register classes}.
2162 You define a number of register classes, giving each one a name and saying
2163 which of the registers belong to it. Then you can specify register classes
2164 that are allowed as operands to particular instruction patterns.
2168 In general, each register will belong to several classes. In fact, one
2169 class must be named @code{ALL_REGS} and contain all the registers. Another
2170 class must be named @code{NO_REGS} and contain no registers. Often the
2171 union of two classes will be another class; however, this is not required.
2173 @findex GENERAL_REGS
2174 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2175 terribly special about the name, but the operand constraint letters
2176 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2177 the same as @code{ALL_REGS}, just define it as a macro which expands
2180 Order the classes so that if class @var{x} is contained in class @var{y}
2181 then @var{x} has a lower class number than @var{y}.
2183 The way classes other than @code{GENERAL_REGS} are specified in operand
2184 constraints is through machine-dependent operand constraint letters.
2185 You can define such letters to correspond to various classes, then use
2186 them in operand constraints.
2188 You should define a class for the union of two classes whenever some
2189 instruction allows both classes. For example, if an instruction allows
2190 either a floating point (coprocessor) register or a general register for a
2191 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2192 which includes both of them. Otherwise you will get suboptimal code.
2194 You must also specify certain redundant information about the register
2195 classes: for each class, which classes contain it and which ones are
2196 contained in it; for each pair of classes, the largest class contained
2199 When a value occupying several consecutive registers is expected in a
2200 certain class, all the registers used must belong to that class.
2201 Therefore, register classes cannot be used to enforce a requirement for
2202 a register pair to start with an even-numbered register. The way to
2203 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2205 Register classes used for input-operands of bitwise-and or shift
2206 instructions have a special requirement: each such class must have, for
2207 each fixed-point machine mode, a subclass whose registers can transfer that
2208 mode to or from memory. For example, on some machines, the operations for
2209 single-byte values (@code{QImode}) are limited to certain registers. When
2210 this is so, each register class that is used in a bitwise-and or shift
2211 instruction must have a subclass consisting of registers from which
2212 single-byte values can be loaded or stored. This is so that
2213 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2216 @findex enum reg_class
2217 @item enum reg_class
2218 An enumeral type that must be defined with all the register class names
2219 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
2220 must be the last register class, followed by one more enumeral value,
2221 @code{LIM_REG_CLASSES}, which is not a register class but rather
2222 tells how many classes there are.
2224 Each register class has a number, which is the value of casting
2225 the class name to type @code{int}. The number serves as an index
2226 in many of the tables described below.
2228 @findex N_REG_CLASSES
2230 The number of distinct register classes, defined as follows:
2233 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2236 @findex REG_CLASS_NAMES
2237 @item REG_CLASS_NAMES
2238 An initializer containing the names of the register classes as C string
2239 constants. These names are used in writing some of the debugging dumps.
2241 @findex REG_CLASS_CONTENTS
2242 @item REG_CLASS_CONTENTS
2243 An initializer containing the contents of the register classes, as integers
2244 which are bit masks. The @var{n}th integer specifies the contents of class
2245 @var{n}. The way the integer @var{mask} is interpreted is that
2246 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2248 When the machine has more than 32 registers, an integer does not suffice.
2249 Then the integers are replaced by sub-initializers, braced groupings containing
2250 several integers. Each sub-initializer must be suitable as an initializer
2251 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2252 In this situation, the first integer in each sub-initializer corresponds to
2253 registers 0 through 31, the second integer to registers 32 through 63, and
2256 @findex REGNO_REG_CLASS
2257 @item REGNO_REG_CLASS (@var{regno})
2258 A C expression whose value is a register class containing hard register
2259 @var{regno}. In general there is more than one such class; choose a class
2260 which is @dfn{minimal}, meaning that no smaller class also contains the
2263 @findex BASE_REG_CLASS
2264 @item BASE_REG_CLASS
2265 A macro whose definition is the name of the class to which a valid
2266 base register must belong. A base register is one used in an address
2267 which is the register value plus a displacement.
2269 @findex MODE_BASE_REG_CLASS
2270 @item MODE_BASE_REG_CLASS (@var{mode})
2271 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2272 the selection of a base register in a mode depenedent manner. If
2273 @var{mode} is VOIDmode then it should return the same value as
2274 @code{BASE_REG_CLASS}.
2276 @findex INDEX_REG_CLASS
2277 @item INDEX_REG_CLASS
2278 A macro whose definition is the name of the class to which a valid
2279 index register must belong. An index register is one used in an
2280 address where its value is either multiplied by a scale factor or
2281 added to another register (as well as added to a displacement).
2283 @findex REG_CLASS_FROM_LETTER
2284 @item REG_CLASS_FROM_LETTER (@var{char})
2285 A C expression which defines the machine-dependent operand constraint
2286 letters for register classes. If @var{char} is such a letter, the
2287 value should be the register class corresponding to it. Otherwise,
2288 the value should be @code{NO_REGS}. The register letter @samp{r},
2289 corresponding to class @code{GENERAL_REGS}, will not be passed
2290 to this macro; you do not need to handle it.
2292 @findex REGNO_OK_FOR_BASE_P
2293 @item REGNO_OK_FOR_BASE_P (@var{num})
2294 A C expression which is nonzero if register number @var{num} is
2295 suitable for use as a base register in operand addresses. It may be
2296 either a suitable hard register or a pseudo register that has been
2297 allocated such a hard register.
2299 @findex REGNO_MODE_OK_FOR_BASE_P
2300 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2301 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2302 that expression may examine the mode of the memory reference in
2303 @var{mode}. You should define this macro if the mode of the memory
2304 reference affects whether a register may be used as a base register. If
2305 you define this macro, the compiler will use it instead of
2306 @code{REGNO_OK_FOR_BASE_P}.
2308 @findex REGNO_OK_FOR_INDEX_P
2309 @item REGNO_OK_FOR_INDEX_P (@var{num})
2310 A C expression which is nonzero if register number @var{num} is
2311 suitable for use as an index register in operand addresses. It may be
2312 either a suitable hard register or a pseudo register that has been
2313 allocated such a hard register.
2315 The difference between an index register and a base register is that
2316 the index register may be scaled. If an address involves the sum of
2317 two registers, neither one of them scaled, then either one may be
2318 labeled the ``base'' and the other the ``index''; but whichever
2319 labeling is used must fit the machine's constraints of which registers
2320 may serve in each capacity. The compiler will try both labelings,
2321 looking for one that is valid, and will reload one or both registers
2322 only if neither labeling works.
2324 @findex PREFERRED_RELOAD_CLASS
2325 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2326 A C expression that places additional restrictions on the register class
2327 to use when it is necessary to copy value @var{x} into a register in class
2328 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2329 another, smaller class. On many machines, the following definition is
2333 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2336 Sometimes returning a more restrictive class makes better code. For
2337 example, on the 68000, when @var{x} is an integer constant that is in range
2338 for a @samp{moveq} instruction, the value of this macro is always
2339 @code{DATA_REGS} as long as @var{class} includes the data registers.
2340 Requiring a data register guarantees that a @samp{moveq} will be used.
2342 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2343 you can force @var{x} into a memory constant. This is useful on
2344 certain machines where immediate floating values cannot be loaded into
2345 certain kinds of registers.
2347 @findex PREFERRED_OUTPUT_RELOAD_CLASS
2348 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2349 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2350 input reloads. If you don't define this macro, the default is to use
2351 @var{class}, unchanged.
2353 @findex LIMIT_RELOAD_CLASS
2354 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2355 A C expression that places additional restrictions on the register class
2356 to use when it is necessary to be able to hold a value of mode
2357 @var{mode} in a reload register for which class @var{class} would
2360 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2361 there are certain modes that simply can't go in certain reload classes.
2363 The value is a register class; perhaps @var{class}, or perhaps another,
2366 Don't define this macro unless the target machine has limitations which
2367 require the macro to do something nontrivial.
2369 @findex SECONDARY_RELOAD_CLASS
2370 @findex SECONDARY_INPUT_RELOAD_CLASS
2371 @findex SECONDARY_OUTPUT_RELOAD_CLASS
2372 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2373 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2374 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2375 Many machines have some registers that cannot be copied directly to or
2376 from memory or even from other types of registers. An example is the
2377 @samp{MQ} register, which on most machines, can only be copied to or
2378 from general registers, but not memory. Some machines allow copying all
2379 registers to and from memory, but require a scratch register for stores
2380 to some memory locations (e.g., those with symbolic address on the RT,
2381 and those with certain symbolic address on the Sparc when compiling
2382 PIC)@. In some cases, both an intermediate and a scratch register are
2385 You should define these macros to indicate to the reload phase that it may
2386 need to allocate at least one register for a reload in addition to the
2387 register to contain the data. Specifically, if copying @var{x} to a
2388 register @var{class} in @var{mode} requires an intermediate register,
2389 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2390 largest register class all of whose registers can be used as
2391 intermediate registers or scratch registers.
2393 If copying a register @var{class} in @var{mode} to @var{x} requires an
2394 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2395 should be defined to return the largest register class required. If the
2396 requirements for input and output reloads are the same, the macro
2397 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2400 The values returned by these macros are often @code{GENERAL_REGS}.
2401 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2402 can be directly copied to or from a register of @var{class} in
2403 @var{mode} without requiring a scratch register. Do not define this
2404 macro if it would always return @code{NO_REGS}.
2406 If a scratch register is required (either with or without an
2407 intermediate register), you should define patterns for
2408 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2409 (@pxref{Standard Names}. These patterns, which will normally be
2410 implemented with a @code{define_expand}, should be similar to the
2411 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2414 Define constraints for the reload register and scratch register that
2415 contain a single register class. If the original reload register (whose
2416 class is @var{class}) can meet the constraint given in the pattern, the
2417 value returned by these macros is used for the class of the scratch
2418 register. Otherwise, two additional reload registers are required.
2419 Their classes are obtained from the constraints in the insn pattern.
2421 @var{x} might be a pseudo-register or a @code{subreg} of a
2422 pseudo-register, which could either be in a hard register or in memory.
2423 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2424 in memory and the hard register number if it is in a register.
2426 These macros should not be used in the case where a particular class of
2427 registers can only be copied to memory and not to another class of
2428 registers. In that case, secondary reload registers are not needed and
2429 would not be helpful. Instead, a stack location must be used to perform
2430 the copy and the @code{mov@var{m}} pattern should use memory as an
2431 intermediate storage. This case often occurs between floating-point and
2434 @findex SECONDARY_MEMORY_NEEDED
2435 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2436 Certain machines have the property that some registers cannot be copied
2437 to some other registers without using memory. Define this macro on
2438 those machines to be a C expression that is nonzero if objects of mode
2439 @var{m} in registers of @var{class1} can only be copied to registers of
2440 class @var{class2} by storing a register of @var{class1} into memory
2441 and loading that memory location into a register of @var{class2}.
2443 Do not define this macro if its value would always be zero.
2445 @findex SECONDARY_MEMORY_NEEDED_RTX
2446 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2447 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2448 allocates a stack slot for a memory location needed for register copies.
2449 If this macro is defined, the compiler instead uses the memory location
2450 defined by this macro.
2452 Do not define this macro if you do not define
2453 @code{SECONDARY_MEMORY_NEEDED}.
2455 @findex SECONDARY_MEMORY_NEEDED_MODE
2456 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2457 When the compiler needs a secondary memory location to copy between two
2458 registers of mode @var{mode}, it normally allocates sufficient memory to
2459 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2460 load operations in a mode that many bits wide and whose class is the
2461 same as that of @var{mode}.
2463 This is right thing to do on most machines because it ensures that all
2464 bits of the register are copied and prevents accesses to the registers
2465 in a narrower mode, which some machines prohibit for floating-point
2468 However, this default behavior is not correct on some machines, such as
2469 the DEC Alpha, that store short integers in floating-point registers
2470 differently than in integer registers. On those machines, the default
2471 widening will not work correctly and you must define this macro to
2472 suppress that widening in some cases. See the file @file{alpha.h} for
2475 Do not define this macro if you do not define
2476 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2477 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2479 @findex SMALL_REGISTER_CLASSES
2480 @item SMALL_REGISTER_CLASSES
2481 On some machines, it is risky to let hard registers live across arbitrary
2482 insns. Typically, these machines have instructions that require values
2483 to be in specific registers (like an accumulator), and reload will fail
2484 if the required hard register is used for another purpose across such an
2487 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2488 value on these machines. When this macro has a nonzero value, the
2489 compiler will try to minimize the lifetime of hard registers.
2491 It is always safe to define this macro with a nonzero value, but if you
2492 unnecessarily define it, you will reduce the amount of optimizations
2493 that can be performed in some cases. If you do not define this macro
2494 with a nonzero value when it is required, the compiler will run out of
2495 spill registers and print a fatal error message. For most machines, you
2496 should not define this macro at all.
2498 @findex CLASS_LIKELY_SPILLED_P
2499 @item CLASS_LIKELY_SPILLED_P (@var{class})
2500 A C expression whose value is nonzero if pseudos that have been assigned
2501 to registers of class @var{class} would likely be spilled because
2502 registers of @var{class} are needed for spill registers.
2504 The default value of this macro returns 1 if @var{class} has exactly one
2505 register and zero otherwise. On most machines, this default should be
2506 used. Only define this macro to some other expression if pseudos
2507 allocated by @file{local-alloc.c} end up in memory because their hard
2508 registers were needed for spill registers. If this macro returns nonzero
2509 for those classes, those pseudos will only be allocated by
2510 @file{global.c}, which knows how to reallocate the pseudo to another
2511 register. If there would not be another register available for
2512 reallocation, you should not change the definition of this macro since
2513 the only effect of such a definition would be to slow down register
2516 @findex CLASS_MAX_NREGS
2517 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2518 A C expression for the maximum number of consecutive registers
2519 of class @var{class} needed to hold a value of mode @var{mode}.
2521 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2522 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2523 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2524 @var{mode})} for all @var{regno} values in the class @var{class}.
2526 This macro helps control the handling of multiple-word values
2529 @item CLASS_CANNOT_CHANGE_MODE
2530 If defined, a C expression for a class that contains registers for
2531 which the compiler may not change modes arbitrarily.
2533 @item CLASS_CANNOT_CHANGE_MODE_P(@var{from}, @var{to})
2534 A C expression that is true if, for a register in
2535 @code{CLASS_CANNOT_CHANGE_MODE}, the requested mode punning is invalid.
2537 For the example, loading 32-bit integer or floating-point objects into
2538 floating-point registers on the Alpha extends them to 64-bits.
2539 Therefore loading a 64-bit object and then storing it as a 32-bit object
2540 does not store the low-order 32-bits, as would be the case for a normal
2541 register. Therefore, @file{alpha.h} defines @code{CLASS_CANNOT_CHANGE_MODE}
2542 as @code{FLOAT_REGS} and @code{CLASS_CANNOT_CHANGE_MODE_P} restricts
2543 mode changes to same-size modes.
2545 Compare this to IA-64, which extends floating-point values to 82-bits,
2546 and stores 64-bit integers in a different format than 64-bit doubles.
2547 Therefore @code{CLASS_CANNOT_CHANGE_MODE_P} is always true.
2550 Three other special macros describe which operands fit which constraint
2554 @findex CONST_OK_FOR_LETTER_P
2555 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2556 A C expression that defines the machine-dependent operand constraint
2557 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2558 particular ranges of integer values. If @var{c} is one of those
2559 letters, the expression should check that @var{value}, an integer, is in
2560 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2561 not one of those letters, the value should be 0 regardless of
2564 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2565 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2566 A C expression that defines the machine-dependent operand constraint
2567 letters that specify particular ranges of @code{const_double} values
2568 (@samp{G} or @samp{H}).
2570 If @var{c} is one of those letters, the expression should check that
2571 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2572 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2573 letters, the value should be 0 regardless of @var{value}.
2575 @code{const_double} is used for all floating-point constants and for
2576 @code{DImode} fixed-point constants. A given letter can accept either
2577 or both kinds of values. It can use @code{GET_MODE} to distinguish
2578 between these kinds.
2580 @findex EXTRA_CONSTRAINT
2581 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2582 A C expression that defines the optional machine-dependent constraint
2583 letters that can be used to segregate specific types of operands, usually
2584 memory references, for the target machine. Any letter that is not
2585 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER}
2586 may be used. Normally this macro will not be defined.
2588 If it is required for a particular target machine, it should return 1
2589 if @var{value} corresponds to the operand type represented by the
2590 constraint letter @var{c}. If @var{c} is not defined as an extra
2591 constraint, the value returned should be 0 regardless of @var{value}.
2593 For example, on the ROMP, load instructions cannot have their output
2594 in r0 if the memory reference contains a symbolic address. Constraint
2595 letter @samp{Q} is defined as representing a memory address that does
2596 @emph{not} contain a symbolic address. An alternative is specified with
2597 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2598 alternative specifies @samp{m} on the input and a register class that
2599 does not include r0 on the output.
2602 @node Stack and Calling
2603 @section Stack Layout and Calling Conventions
2604 @cindex calling conventions
2606 @c prevent bad page break with this line
2607 This describes the stack layout and calling conventions.
2611 * Exception Handling::
2616 * Register Arguments::
2618 * Aggregate Return::
2626 @subsection Basic Stack Layout
2627 @cindex stack frame layout
2628 @cindex frame layout
2630 @c prevent bad page break with this line
2631 Here is the basic stack layout.
2634 @findex STACK_GROWS_DOWNWARD
2635 @item STACK_GROWS_DOWNWARD
2636 Define this macro if pushing a word onto the stack moves the stack
2637 pointer to a smaller address.
2639 When we say, ``define this macro if @dots{},'' it means that the
2640 compiler checks this macro only with @code{#ifdef} so the precise
2641 definition used does not matter.
2643 @findex STACK_PUSH_CODE
2644 @item STACK_PUSH_CODE
2646 This macro defines the operation used when something is pushed
2647 on the stack. In RTL, a push operation will be
2648 @code{(set (mem (STACK_PUSH_CODE (reg sp))) ...)}
2650 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2651 and @code{POST_INC}. Which of these is correct depends on
2652 the stack direction and on whether the stack pointer points
2653 to the last item on the stack or whether it points to the
2654 space for the next item on the stack.
2656 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2657 defined, which is almost always right, and @code{PRE_INC} otherwise,
2658 which is often wrong.
2660 @findex FRAME_GROWS_DOWNWARD
2661 @item FRAME_GROWS_DOWNWARD
2662 Define this macro if the addresses of local variable slots are at negative
2663 offsets from the frame pointer.
2665 @findex ARGS_GROW_DOWNWARD
2666 @item ARGS_GROW_DOWNWARD
2667 Define this macro if successive arguments to a function occupy decreasing
2668 addresses on the stack.
2670 @findex STARTING_FRAME_OFFSET
2671 @item STARTING_FRAME_OFFSET
2672 Offset from the frame pointer to the first local variable slot to be allocated.
2674 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2675 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2676 Otherwise, it is found by adding the length of the first slot to the
2677 value @code{STARTING_FRAME_OFFSET}.
2678 @c i'm not sure if the above is still correct.. had to change it to get
2679 @c rid of an overfull. --mew 2feb93
2681 @findex STACK_POINTER_OFFSET
2682 @item STACK_POINTER_OFFSET
2683 Offset from the stack pointer register to the first location at which
2684 outgoing arguments are placed. If not specified, the default value of
2685 zero is used. This is the proper value for most machines.
2687 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2688 the first location at which outgoing arguments are placed.
2690 @findex FIRST_PARM_OFFSET
2691 @item FIRST_PARM_OFFSET (@var{fundecl})
2692 Offset from the argument pointer register to the first argument's
2693 address. On some machines it may depend on the data type of the
2696 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2697 the first argument's address.
2699 @findex STACK_DYNAMIC_OFFSET
2700 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2701 Offset from the stack pointer register to an item dynamically allocated
2702 on the stack, e.g., by @code{alloca}.
2704 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2705 length of the outgoing arguments. The default is correct for most
2706 machines. See @file{function.c} for details.
2708 @findex DYNAMIC_CHAIN_ADDRESS
2709 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2710 A C expression whose value is RTL representing the address in a stack
2711 frame where the pointer to the caller's frame is stored. Assume that
2712 @var{frameaddr} is an RTL expression for the address of the stack frame
2715 If you don't define this macro, the default is to return the value
2716 of @var{frameaddr}---that is, the stack frame address is also the
2717 address of the stack word that points to the previous frame.
2719 @findex SETUP_FRAME_ADDRESSES
2720 @item SETUP_FRAME_ADDRESSES
2721 If defined, a C expression that produces the machine-specific code to
2722 setup the stack so that arbitrary frames can be accessed. For example,
2723 on the Sparc, we must flush all of the register windows to the stack
2724 before we can access arbitrary stack frames. You will seldom need to
2727 @findex BUILTIN_SETJMP_FRAME_VALUE
2728 @item BUILTIN_SETJMP_FRAME_VALUE
2729 If defined, a C expression that contains an rtx that is used to store
2730 the address of the current frame into the built in @code{setjmp} buffer.
2731 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2732 machines. One reason you may need to define this macro is if
2733 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2735 @findex RETURN_ADDR_RTX
2736 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2737 A C expression whose value is RTL representing the value of the return
2738 address for the frame @var{count} steps up from the current frame, after
2739 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2740 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2741 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2743 The value of the expression must always be the correct address when
2744 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2745 determine the return address of other frames.
2747 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2748 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2749 Define this if the return address of a particular stack frame is accessed
2750 from the frame pointer of the previous stack frame.
2752 @findex INCOMING_RETURN_ADDR_RTX
2753 @item INCOMING_RETURN_ADDR_RTX
2754 A C expression whose value is RTL representing the location of the
2755 incoming return address at the beginning of any function, before the
2756 prologue. This RTL is either a @code{REG}, indicating that the return
2757 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2760 You only need to define this macro if you want to support call frame
2761 debugging information like that provided by DWARF 2.
2763 If this RTL is a @code{REG}, you should also define
2764 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2766 @findex INCOMING_FRAME_SP_OFFSET
2767 @item INCOMING_FRAME_SP_OFFSET
2768 A C expression whose value is an integer giving the offset, in bytes,
2769 from the value of the stack pointer register to the top of the stack
2770 frame at the beginning of any function, before the prologue. The top of
2771 the frame is defined to be the value of the stack pointer in the
2772 previous frame, just before the call instruction.
2774 You only need to define this macro if you want to support call frame
2775 debugging information like that provided by DWARF 2.
2777 @findex ARG_POINTER_CFA_OFFSET
2778 @item ARG_POINTER_CFA_OFFSET (@var{fundecl})
2779 A C expression whose value is an integer giving the offset, in bytes,
2780 from the argument pointer to the canonical frame address (cfa). The
2781 final value should coincide with that calculated by
2782 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2783 during virtual register instantiation.
2785 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2786 which is correct for most machines; in general, the arguments are found
2787 immediately before the stack frame. Note that this is not the case on
2788 some targets that save registers into the caller's frame, such as SPARC
2789 and rs6000, and so such targets need to define this macro.
2791 You only need to define this macro if the default is incorrect, and you
2792 want to support call frame debugging information like that provided by
2797 Define this macro if the stack size for the target is very small. This
2798 has the effect of disabling gcc's built-in @samp{alloca}, though
2799 @samp{__builtin_alloca} is not affected.
2802 @node Exception Handling
2803 @subsection Exception Handling Support
2804 @cindex exception handling
2807 @findex EH_RETURN_DATA_REGNO
2808 @item EH_RETURN_DATA_REGNO (@var{N})
2809 A C expression whose value is the @var{N}th register number used for
2810 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2811 @var{N} registers are usable.
2813 The exception handling library routines communicate with the exception
2814 handlers via a set of agreed upon registers. Ideally these registers
2815 should be call-clobbered; it is possible to use call-saved registers,
2816 but may negatively impact code size. The target must support at least
2817 2 data registers, but should define 4 if there are enough free registers.
2819 You must define this macro if you want to support call frame exception
2820 handling like that provided by DWARF 2.
2822 @findex EH_RETURN_STACKADJ_RTX
2823 @item EH_RETURN_STACKADJ_RTX
2824 A C expression whose value is RTL representing a location in which
2825 to store a stack adjustment to be applied before function return.
2826 This is used to unwind the stack to an exception handler's call frame.
2827 It will be assigned zero on code paths that return normally.
2829 Typically this is a call-clobbered hard register that is otherwise
2830 untouched by the epilogue, but could also be a stack slot.
2832 You must define this macro if you want to support call frame exception
2833 handling like that provided by DWARF 2.
2835 @findex EH_RETURN_HANDLER_RTX
2836 @item EH_RETURN_HANDLER_RTX
2837 A C expression whose value is RTL representing a location in which
2838 to store the address of an exception handler to which we should
2839 return. It will not be assigned on code paths that return normally.
2841 Typically this is the location in the call frame at which the normal
2842 return address is stored. For targets that return by popping an
2843 address off the stack, this might be a memory address just below
2844 the @emph{target} call frame rather than inside the current call
2845 frame. @code{EH_RETURN_STACKADJ_RTX} will have already been assigned,
2846 so it may be used to calculate the location of the target call frame.
2848 Some targets have more complex requirements than storing to an
2849 address calculable during initial code generation. In that case
2850 the @code{eh_return} instruction pattern should be used instead.
2852 If you want to support call frame exception handling, you must
2853 define either this macro or the @code{eh_return} instruction pattern.
2855 @findex ASM_PREFERRED_EH_DATA_FORMAT
2856 @item ASM_PREFERRED_EH_DATA_FORMAT(@var{code}, @var{global})
2857 This macro chooses the encoding of pointers embedded in the exception
2858 handling sections. If at all possible, this should be defined such
2859 that the exception handling section will not require dynamic relocations,
2860 and so may be read-only.
2862 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
2863 @var{global} is true if the symbol may be affected by dynamic relocations.
2864 The macro should return a combination of the @code{DW_EH_PE_*} defines
2865 as found in @file{dwarf2.h}.
2867 If this macro is not defined, pointers will not be encoded but
2868 represented directly.
2870 @findex ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX
2871 @item ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX(@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
2872 This macro allows the target to emit whatever special magic is required
2873 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
2874 Generic code takes care of pc-relative and indirect encodings; this must
2875 be defined if the target uses text-relative or data-relative encodings.
2877 This is a C statement that branches to @var{done} if the format was
2878 handled. @var{encoding} is the format chosen, @var{size} is the number
2879 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
2882 @findex MD_FALLBACK_FRAME_STATE_FOR
2883 @item MD_FALLBACK_FRAME_STATE_FOR(@var{context}, @var{fs}, @var{success})
2884 This macro allows the target to add cpu and operating system specific
2885 code to the call-frame unwinder for use when there is no unwind data
2886 available. The most common reason to implement this macro is to unwind
2887 through signal frames.
2889 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
2890 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2891 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
2892 for the address of the code being executed and @code{context->cfa} for
2893 the stack pointer value. If the frame can be decoded, the register save
2894 addresses should be updated in @var{fs} and the macro should branch to
2895 @var{success}. If the frame cannot be decoded, the macro should do
2899 @node Stack Checking
2900 @subsection Specifying How Stack Checking is Done
2902 GCC will check that stack references are within the boundaries of
2903 the stack, if the @option{-fstack-check} is specified, in one of three ways:
2907 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2908 will assume that you have arranged for stack checking to be done at
2909 appropriate places in the configuration files, e.g., in
2910 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
2914 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
2915 called @code{check_stack} in your @file{md} file, GCC will call that
2916 pattern with one argument which is the address to compare the stack
2917 value against. You must arrange for this pattern to report an error if
2918 the stack pointer is out of range.
2921 If neither of the above are true, GCC will generate code to periodically
2922 ``probe'' the stack pointer using the values of the macros defined below.
2925 Normally, you will use the default values of these macros, so GCC
2926 will use the third approach.
2929 @findex STACK_CHECK_BUILTIN
2930 @item STACK_CHECK_BUILTIN
2931 A nonzero value if stack checking is done by the configuration files in a
2932 machine-dependent manner. You should define this macro if stack checking
2933 is require by the ABI of your machine or if you would like to have to stack
2934 checking in some more efficient way than GCC's portable approach.
2935 The default value of this macro is zero.
2937 @findex STACK_CHECK_PROBE_INTERVAL
2938 @item STACK_CHECK_PROBE_INTERVAL
2939 An integer representing the interval at which GCC must generate stack
2940 probe instructions. You will normally define this macro to be no larger
2941 than the size of the ``guard pages'' at the end of a stack area. The
2942 default value of 4096 is suitable for most systems.
2944 @findex STACK_CHECK_PROBE_LOAD
2945 @item STACK_CHECK_PROBE_LOAD
2946 A integer which is nonzero if GCC should perform the stack probe
2947 as a load instruction and zero if GCC should use a store instruction.
2948 The default is zero, which is the most efficient choice on most systems.
2950 @findex STACK_CHECK_PROTECT
2951 @item STACK_CHECK_PROTECT
2952 The number of bytes of stack needed to recover from a stack overflow,
2953 for languages where such a recovery is supported. The default value of
2954 75 words should be adequate for most machines.
2956 @findex STACK_CHECK_MAX_FRAME_SIZE
2957 @item STACK_CHECK_MAX_FRAME_SIZE
2958 The maximum size of a stack frame, in bytes. GCC will generate probe
2959 instructions in non-leaf functions to ensure at least this many bytes of
2960 stack are available. If a stack frame is larger than this size, stack
2961 checking will not be reliable and GCC will issue a warning. The
2962 default is chosen so that GCC only generates one instruction on most
2963 systems. You should normally not change the default value of this macro.
2965 @findex STACK_CHECK_FIXED_FRAME_SIZE
2966 @item STACK_CHECK_FIXED_FRAME_SIZE
2967 GCC uses this value to generate the above warning message. It
2968 represents the amount of fixed frame used by a function, not including
2969 space for any callee-saved registers, temporaries and user variables.
2970 You need only specify an upper bound for this amount and will normally
2971 use the default of four words.
2973 @findex STACK_CHECK_MAX_VAR_SIZE
2974 @item STACK_CHECK_MAX_VAR_SIZE
2975 The maximum size, in bytes, of an object that GCC will place in the
2976 fixed area of the stack frame when the user specifies
2977 @option{-fstack-check}.
2978 GCC computed the default from the values of the above macros and you will
2979 normally not need to override that default.
2983 @node Frame Registers
2984 @subsection Registers That Address the Stack Frame
2986 @c prevent bad page break with this line
2987 This discusses registers that address the stack frame.
2990 @findex STACK_POINTER_REGNUM
2991 @item STACK_POINTER_REGNUM
2992 The register number of the stack pointer register, which must also be a
2993 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2994 the hardware determines which register this is.
2996 @findex FRAME_POINTER_REGNUM
2997 @item FRAME_POINTER_REGNUM
2998 The register number of the frame pointer register, which is used to
2999 access automatic variables in the stack frame. On some machines, the
3000 hardware determines which register this is. On other machines, you can
3001 choose any register you wish for this purpose.
3003 @findex HARD_FRAME_POINTER_REGNUM
3004 @item HARD_FRAME_POINTER_REGNUM
3005 On some machines the offset between the frame pointer and starting
3006 offset of the automatic variables is not known until after register
3007 allocation has been done (for example, because the saved registers are
3008 between these two locations). On those machines, define
3009 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3010 be used internally until the offset is known, and define
3011 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3012 used for the frame pointer.
3014 You should define this macro only in the very rare circumstances when it
3015 is not possible to calculate the offset between the frame pointer and
3016 the automatic variables until after register allocation has been
3017 completed. When this macro is defined, you must also indicate in your
3018 definition of @code{ELIMINABLE_REGS} how to eliminate
3019 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3020 or @code{STACK_POINTER_REGNUM}.
3022 Do not define this macro if it would be the same as
3023 @code{FRAME_POINTER_REGNUM}.
3025 @findex ARG_POINTER_REGNUM
3026 @item ARG_POINTER_REGNUM
3027 The register number of the arg pointer register, which is used to access
3028 the function's argument list. On some machines, this is the same as the
3029 frame pointer register. On some machines, the hardware determines which
3030 register this is. On other machines, you can choose any register you
3031 wish for this purpose. If this is not the same register as the frame
3032 pointer register, then you must mark it as a fixed register according to
3033 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3034 (@pxref{Elimination}).
3036 @findex RETURN_ADDRESS_POINTER_REGNUM
3037 @item RETURN_ADDRESS_POINTER_REGNUM
3038 The register number of the return address pointer register, which is used to
3039 access the current function's return address from the stack. On some
3040 machines, the return address is not at a fixed offset from the frame
3041 pointer or stack pointer or argument pointer. This register can be defined
3042 to point to the return address on the stack, and then be converted by
3043 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3045 Do not define this macro unless there is no other way to get the return
3046 address from the stack.
3048 @findex STATIC_CHAIN_REGNUM
3049 @findex STATIC_CHAIN_INCOMING_REGNUM
3050 @item STATIC_CHAIN_REGNUM
3051 @itemx STATIC_CHAIN_INCOMING_REGNUM
3052 Register numbers used for passing a function's static chain pointer. If
3053 register windows are used, the register number as seen by the called
3054 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3055 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3056 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3059 The static chain register need not be a fixed register.
3061 If the static chain is passed in memory, these macros should not be
3062 defined; instead, the next two macros should be defined.
3064 @findex STATIC_CHAIN
3065 @findex STATIC_CHAIN_INCOMING
3067 @itemx STATIC_CHAIN_INCOMING
3068 If the static chain is passed in memory, these macros provide rtx giving
3069 @code{mem} expressions that denote where they are stored.
3070 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3071 as seen by the calling and called functions, respectively. Often the former
3072 will be at an offset from the stack pointer and the latter at an offset from
3075 @findex stack_pointer_rtx
3076 @findex frame_pointer_rtx
3077 @findex arg_pointer_rtx
3078 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3079 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3080 macros and should be used to refer to those items.
3082 If the static chain is passed in a register, the two previous macros should
3085 @findex DWARF_FRAME_REGISTERS
3086 @item DWARF_FRAME_REGISTERS
3087 This macro specifies the maximum number of hard registers that can be
3088 saved in a call frame. This is used to size data structures used in
3089 DWARF2 exception handling.
3091 Prior to GCC 3.0, this macro was needed in order to establish a stable
3092 exception handling ABI in the face of adding new hard registers for ISA
3093 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3094 in the number of hard registers. Nevertheless, this macro can still be
3095 used to reduce the runtime memory requirements of the exception handling
3096 routines, which can be substantial if the ISA contains a lot of
3097 registers that are not call-saved.
3099 If this macro is not defined, it defaults to
3100 @code{FIRST_PSEUDO_REGISTER}.
3102 @findex PRE_GCC3_DWARF_FRAME_REGISTERS
3103 @item PRE_GCC3_DWARF_FRAME_REGISTERS
3105 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3106 for backward compatibility in pre GCC 3.0 compiled code.
3108 If this macro is not defined, it defaults to
3109 @code{DWARF_FRAME_REGISTERS}.
3114 @subsection Eliminating Frame Pointer and Arg Pointer
3116 @c prevent bad page break with this line
3117 This is about eliminating the frame pointer and arg pointer.
3120 @findex FRAME_POINTER_REQUIRED
3121 @item FRAME_POINTER_REQUIRED
3122 A C expression which is nonzero if a function must have and use a frame
3123 pointer. This expression is evaluated in the reload pass. If its value is
3124 nonzero the function will have a frame pointer.
3126 The expression can in principle examine the current function and decide
3127 according to the facts, but on most machines the constant 0 or the
3128 constant 1 suffices. Use 0 when the machine allows code to be generated
3129 with no frame pointer, and doing so saves some time or space. Use 1
3130 when there is no possible advantage to avoiding a frame pointer.
3132 In certain cases, the compiler does not know how to produce valid code
3133 without a frame pointer. The compiler recognizes those cases and
3134 automatically gives the function a frame pointer regardless of what
3135 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3138 In a function that does not require a frame pointer, the frame pointer
3139 register can be allocated for ordinary usage, unless you mark it as a
3140 fixed register. See @code{FIXED_REGISTERS} for more information.
3142 @findex INITIAL_FRAME_POINTER_OFFSET
3143 @findex get_frame_size
3144 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3145 A C statement to store in the variable @var{depth-var} the difference
3146 between the frame pointer and the stack pointer values immediately after
3147 the function prologue. The value would be computed from information
3148 such as the result of @code{get_frame_size ()} and the tables of
3149 registers @code{regs_ever_live} and @code{call_used_regs}.
3151 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3152 need not be defined. Otherwise, it must be defined even if
3153 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3154 case, you may set @var{depth-var} to anything.
3156 @findex ELIMINABLE_REGS
3157 @item ELIMINABLE_REGS
3158 If defined, this macro specifies a table of register pairs used to
3159 eliminate unneeded registers that point into the stack frame. If it is not
3160 defined, the only elimination attempted by the compiler is to replace
3161 references to the frame pointer with references to the stack pointer.
3163 The definition of this macro is a list of structure initializations, each
3164 of which specifies an original and replacement register.
3166 On some machines, the position of the argument pointer is not known until
3167 the compilation is completed. In such a case, a separate hard register
3168 must be used for the argument pointer. This register can be eliminated by
3169 replacing it with either the frame pointer or the argument pointer,
3170 depending on whether or not the frame pointer has been eliminated.
3172 In this case, you might specify:
3174 #define ELIMINABLE_REGS \
3175 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3176 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3177 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3180 Note that the elimination of the argument pointer with the stack pointer is
3181 specified first since that is the preferred elimination.
3183 @findex CAN_ELIMINATE
3184 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3185 A C expression that returns nonzero if the compiler is allowed to try
3186 to replace register number @var{from-reg} with register number
3187 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3188 is defined, and will usually be the constant 1, since most of the cases
3189 preventing register elimination are things that the compiler already
3192 @findex INITIAL_ELIMINATION_OFFSET
3193 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3194 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3195 specifies the initial difference between the specified pair of
3196 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3200 @node Stack Arguments
3201 @subsection Passing Function Arguments on the Stack
3202 @cindex arguments on stack
3203 @cindex stack arguments
3205 The macros in this section control how arguments are passed
3206 on the stack. See the following section for other macros that
3207 control passing certain arguments in registers.
3210 @findex PROMOTE_PROTOTYPES
3211 @item PROMOTE_PROTOTYPES
3212 A C expression whose value is nonzero if an argument declared in
3213 a prototype as an integral type smaller than @code{int} should
3214 actually be passed as an @code{int}. In addition to avoiding
3215 errors in certain cases of mismatch, it also makes for better
3216 code on certain machines. If the macro is not defined in target
3217 header files, it defaults to 0.
3221 A C expression. If nonzero, push insns will be used to pass
3223 If the target machine does not have a push instruction, set it to zero.
3224 That directs GCC to use an alternate strategy: to
3225 allocate the entire argument block and then store the arguments into
3226 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3227 On some machines, the definition
3229 @findex PUSH_ROUNDING
3230 @item PUSH_ROUNDING (@var{npushed})
3231 A C expression that is the number of bytes actually pushed onto the
3232 stack when an instruction attempts to push @var{npushed} bytes.
3234 On some machines, the definition
3237 #define PUSH_ROUNDING(BYTES) (BYTES)
3241 will suffice. But on other machines, instructions that appear
3242 to push one byte actually push two bytes in an attempt to maintain
3243 alignment. Then the definition should be
3246 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3249 @findex ACCUMULATE_OUTGOING_ARGS
3250 @findex current_function_outgoing_args_size
3251 @item ACCUMULATE_OUTGOING_ARGS
3252 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3253 will be computed and placed into the variable
3254 @code{current_function_outgoing_args_size}. No space will be pushed
3255 onto the stack for each call; instead, the function prologue should
3256 increase the stack frame size by this amount.
3258 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3261 @findex REG_PARM_STACK_SPACE
3262 @item REG_PARM_STACK_SPACE (@var{fndecl})
3263 Define this macro if functions should assume that stack space has been
3264 allocated for arguments even when their values are passed in
3267 The value of this macro is the size, in bytes, of the area reserved for
3268 arguments passed in registers for the function represented by @var{fndecl},
3269 which can be zero if GCC is calling a library function.
3271 This space can be allocated by the caller, or be a part of the
3272 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3274 @c above is overfull. not sure what to do. --mew 5feb93 did
3275 @c something, not sure if it looks good. --mew 10feb93
3277 @findex MAYBE_REG_PARM_STACK_SPACE
3278 @findex FINAL_REG_PARM_STACK_SPACE
3279 @item MAYBE_REG_PARM_STACK_SPACE
3280 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
3281 Define these macros in addition to the one above if functions might
3282 allocate stack space for arguments even when their values are passed
3283 in registers. These should be used when the stack space allocated
3284 for arguments in registers is not a simple constant independent of the
3285 function declaration.
3287 The value of the first macro is the size, in bytes, of the area that
3288 we should initially assume would be reserved for arguments passed in registers.
3290 The value of the second macro is the actual size, in bytes, of the area
3291 that will be reserved for arguments passed in registers. This takes two
3292 arguments: an integer representing the number of bytes of fixed sized
3293 arguments on the stack, and a tree representing the number of bytes of
3294 variable sized arguments on the stack.
3296 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
3297 called for libcall functions, the current function, or for a function
3298 being called when it is known that such stack space must be allocated.
3299 In each case this value can be easily computed.
3301 When deciding whether a called function needs such stack space, and how
3302 much space to reserve, GCC uses these two macros instead of
3303 @code{REG_PARM_STACK_SPACE}.
3305 @findex OUTGOING_REG_PARM_STACK_SPACE
3306 @item OUTGOING_REG_PARM_STACK_SPACE
3307 Define this if it is the responsibility of the caller to allocate the area
3308 reserved for arguments passed in registers.
3310 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3311 whether the space for these arguments counts in the value of
3312 @code{current_function_outgoing_args_size}.
3314 @findex STACK_PARMS_IN_REG_PARM_AREA
3315 @item STACK_PARMS_IN_REG_PARM_AREA
3316 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3317 stack parameters don't skip the area specified by it.
3318 @c i changed this, makes more sens and it should have taken care of the
3319 @c overfull.. not as specific, tho. --mew 5feb93
3321 Normally, when a parameter is not passed in registers, it is placed on the
3322 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3323 suppresses this behavior and causes the parameter to be passed on the
3324 stack in its natural location.
3326 @findex RETURN_POPS_ARGS
3327 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3328 A C expression that should indicate the number of bytes of its own
3329 arguments that a function pops on returning, or 0 if the
3330 function pops no arguments and the caller must therefore pop them all
3331 after the function returns.
3333 @var{fundecl} is a C variable whose value is a tree node that describes
3334 the function in question. Normally it is a node of type
3335 @code{FUNCTION_DECL} that describes the declaration of the function.
3336 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3338 @var{funtype} is a C variable whose value is a tree node that
3339 describes the function in question. Normally it is a node of type
3340 @code{FUNCTION_TYPE} that describes the data type of the function.
3341 From this it is possible to obtain the data types of the value and
3342 arguments (if known).
3344 When a call to a library function is being considered, @var{fundecl}
3345 will contain an identifier node for the library function. Thus, if
3346 you need to distinguish among various library functions, you can do so
3347 by their names. Note that ``library function'' in this context means
3348 a function used to perform arithmetic, whose name is known specially
3349 in the compiler and was not mentioned in the C code being compiled.
3351 @var{stack-size} is the number of bytes of arguments passed on the
3352 stack. If a variable number of bytes is passed, it is zero, and
3353 argument popping will always be the responsibility of the calling function.
3355 On the VAX, all functions always pop their arguments, so the definition
3356 of this macro is @var{stack-size}. On the 68000, using the standard
3357 calling convention, no functions pop their arguments, so the value of
3358 the macro is always 0 in this case. But an alternative calling
3359 convention is available in which functions that take a fixed number of
3360 arguments pop them but other functions (such as @code{printf}) pop
3361 nothing (the caller pops all). When this convention is in use,
3362 @var{funtype} is examined to determine whether a function takes a fixed
3363 number of arguments.
3365 @findex CALL_POPS_ARGS
3366 @item CALL_POPS_ARGS (@var{cum})
3367 A C expression that should indicate the number of bytes a call sequence
3368 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3369 when compiling a function call.
3371 @var{cum} is the variable in which all arguments to the called function
3372 have been accumulated.
3374 On certain architectures, such as the SH5, a call trampoline is used
3375 that pops certain registers off the stack, depending on the arguments
3376 that have been passed to the function. Since this is a property of the
3377 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3382 @node Register Arguments
3383 @subsection Passing Arguments in Registers
3384 @cindex arguments in registers
3385 @cindex registers arguments
3387 This section describes the macros which let you control how various
3388 types of arguments are passed in registers or how they are arranged in
3392 @findex FUNCTION_ARG
3393 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3394 A C expression that controls whether a function argument is passed
3395 in a register, and which register.
3397 The arguments are @var{cum}, which summarizes all the previous
3398 arguments; @var{mode}, the machine mode of the argument; @var{type},
3399 the data type of the argument as a tree node or 0 if that is not known
3400 (which happens for C support library functions); and @var{named},
3401 which is 1 for an ordinary argument and 0 for nameless arguments that
3402 correspond to @samp{@dots{}} in the called function's prototype.
3403 @var{type} can be an incomplete type if a syntax error has previously
3406 The value of the expression is usually either a @code{reg} RTX for the
3407 hard register in which to pass the argument, or zero to pass the
3408 argument on the stack.
3410 For machines like the VAX and 68000, where normally all arguments are
3411 pushed, zero suffices as a definition.
3413 The value of the expression can also be a @code{parallel} RTX@. This is
3414 used when an argument is passed in multiple locations. The mode of the
3415 of the @code{parallel} should be the mode of the entire argument. The
3416 @code{parallel} holds any number of @code{expr_list} pairs; each one
3417 describes where part of the argument is passed. In each
3418 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3419 register in which to pass this part of the argument, and the mode of the
3420 register RTX indicates how large this part of the argument is. The
3421 second operand of the @code{expr_list} is a @code{const_int} which gives
3422 the offset in bytes into the entire argument of where this part starts.
3423 As a special exception the first @code{expr_list} in the @code{parallel}
3424 RTX may have a first operand of zero. This indicates that the entire
3425 argument is also stored on the stack.
3427 The last time this macro is called, it is called with @code{MODE ==
3428 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3429 pattern as operands 2 and 3 respectively.
3431 @cindex @file{stdarg.h} and register arguments
3432 The usual way to make the ISO library @file{stdarg.h} work on a machine
3433 where some arguments are usually passed in registers, is to cause
3434 nameless arguments to be passed on the stack instead. This is done
3435 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3437 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3438 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3439 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3440 in the definition of this macro to determine if this argument is of a
3441 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3442 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3443 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3444 defined, the argument will be computed in the stack and then loaded into
3447 @findex MUST_PASS_IN_STACK
3448 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
3449 Define as a C expression that evaluates to nonzero if we do not know how
3450 to pass TYPE solely in registers. The file @file{expr.h} defines a
3451 definition that is usually appropriate, refer to @file{expr.h} for additional
3454 @findex FUNCTION_INCOMING_ARG
3455 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3456 Define this macro if the target machine has ``register windows'', so
3457 that the register in which a function sees an arguments is not
3458 necessarily the same as the one in which the caller passed the
3461 For such machines, @code{FUNCTION_ARG} computes the register in which
3462 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3463 be defined in a similar fashion to tell the function being called
3464 where the arguments will arrive.
3466 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3467 serves both purposes.
3469 @findex FUNCTION_ARG_PARTIAL_NREGS
3470 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3471 A C expression for the number of words, at the beginning of an
3472 argument, that must be put in registers. The value must be zero for
3473 arguments that are passed entirely in registers or that are entirely
3474 pushed on the stack.
3476 On some machines, certain arguments must be passed partially in
3477 registers and partially in memory. On these machines, typically the
3478 first @var{n} words of arguments are passed in registers, and the rest
3479 on the stack. If a multi-word argument (a @code{double} or a
3480 structure) crosses that boundary, its first few words must be passed
3481 in registers and the rest must be pushed. This macro tells the
3482 compiler when this occurs, and how many of the words should go in
3485 @code{FUNCTION_ARG} for these arguments should return the first
3486 register to be used by the caller for this argument; likewise
3487 @code{FUNCTION_INCOMING_ARG}, for the called function.
3489 @findex FUNCTION_ARG_PASS_BY_REFERENCE
3490 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3491 A C expression that indicates when an argument must be passed by reference.
3492 If nonzero for an argument, a copy of that argument is made in memory and a
3493 pointer to the argument is passed instead of the argument itself.
3494 The pointer is passed in whatever way is appropriate for passing a pointer
3497 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3498 definition of this macro might be
3500 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3501 (CUM, MODE, TYPE, NAMED) \
3502 MUST_PASS_IN_STACK (MODE, TYPE)
3504 @c this is *still* too long. --mew 5feb93
3506 @findex FUNCTION_ARG_CALLEE_COPIES
3507 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3508 If defined, a C expression that indicates when it is the called function's
3509 responsibility to make a copy of arguments passed by invisible reference.
3510 Normally, the caller makes a copy and passes the address of the copy to the
3511 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3512 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3513 ``live'' value. The called function must not modify this value. If it can be
3514 determined that the value won't be modified, it need not make a copy;
3515 otherwise a copy must be made.
3517 @findex FUNCTION_ARG_REG_LITTLE_ENDIAN
3518 @item FUNCTION_ARG_REG_LITTLE_ENDIAN
3519 If defined TRUE on a big-endian system then structure arguments passed
3520 (and returned) in registers are passed in a little-endian manner instead of
3521 the big-endian manner. On the HP-UX IA64 and PA64 platforms structures are
3522 aligned differently then integral values and setting this value to true will
3523 allow for the special handling of structure arguments and return values.
3525 @findex CUMULATIVE_ARGS
3526 @item CUMULATIVE_ARGS
3527 A C type for declaring a variable that is used as the first argument of
3528 @code{FUNCTION_ARG} and other related values. For some target machines,
3529 the type @code{int} suffices and can hold the number of bytes of
3532 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3533 arguments that have been passed on the stack. The compiler has other
3534 variables to keep track of that. For target machines on which all
3535 arguments are passed on the stack, there is no need to store anything in
3536 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3537 should not be empty, so use @code{int}.
3539 @findex INIT_CUMULATIVE_ARGS
3540 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
3541 A C statement (sans semicolon) for initializing the variable @var{cum}
3542 for the state at the beginning of the argument list. The variable has
3543 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
3544 for the data type of the function which will receive the args, or 0
3545 if the args are to a compiler support library function. The value of
3546 @var{indirect} is nonzero when processing an indirect call, for example
3547 a call through a function pointer. The value of @var{indirect} is zero
3548 for a call to an explicitly named function, a library function call, or when
3549 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3552 When processing a call to a compiler support library function,
3553 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3554 contains the name of the function, as a string. @var{libname} is 0 when
3555 an ordinary C function call is being processed. Thus, each time this
3556 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3557 never both of them at once.
3559 @findex INIT_CUMULATIVE_LIBCALL_ARGS
3560 @item INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3561 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3562 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3563 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3564 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3565 0)} is used instead.
3567 @findex INIT_CUMULATIVE_INCOMING_ARGS
3568 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3569 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3570 finding the arguments for the function being compiled. If this macro is
3571 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3573 The value passed for @var{libname} is always 0, since library routines
3574 with special calling conventions are never compiled with GCC@. The
3575 argument @var{libname} exists for symmetry with
3576 @code{INIT_CUMULATIVE_ARGS}.
3577 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3578 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3580 @findex FUNCTION_ARG_ADVANCE
3581 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3582 A C statement (sans semicolon) to update the summarizer variable
3583 @var{cum} to advance past an argument in the argument list. The
3584 values @var{mode}, @var{type} and @var{named} describe that argument.
3585 Once this is done, the variable @var{cum} is suitable for analyzing
3586 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3588 This macro need not do anything if the argument in question was passed
3589 on the stack. The compiler knows how to track the amount of stack space
3590 used for arguments without any special help.
3592 @findex FUNCTION_ARG_PADDING
3593 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3594 If defined, a C expression which determines whether, and in which direction,
3595 to pad out an argument with extra space. The value should be of type
3596 @code{enum direction}: either @code{upward} to pad above the argument,
3597 @code{downward} to pad below, or @code{none} to inhibit padding.
3599 The @emph{amount} of padding is always just enough to reach the next
3600 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3603 This macro has a default definition which is right for most systems.
3604 For little-endian machines, the default is to pad upward. For
3605 big-endian machines, the default is to pad downward for an argument of
3606 constant size shorter than an @code{int}, and upward otherwise.
3608 @findex PAD_VARARGS_DOWN
3609 @item PAD_VARARGS_DOWN
3610 If defined, a C expression which determines whether the default
3611 implementation of va_arg will attempt to pad down before reading the
3612 next argument, if that argument is smaller than its aligned space as
3613 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3614 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3616 @findex FUNCTION_ARG_BOUNDARY
3617 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3618 If defined, a C expression that gives the alignment boundary, in bits,
3619 of an argument with the specified mode and type. If it is not defined,
3620 @code{PARM_BOUNDARY} is used for all arguments.
3622 @findex FUNCTION_ARG_REGNO_P
3623 @item FUNCTION_ARG_REGNO_P (@var{regno})
3624 A C expression that is nonzero if @var{regno} is the number of a hard
3625 register in which function arguments are sometimes passed. This does
3626 @emph{not} include implicit arguments such as the static chain and
3627 the structure-value address. On many machines, no registers can be
3628 used for this purpose since all function arguments are pushed on the
3631 @findex LOAD_ARGS_REVERSED
3632 @item LOAD_ARGS_REVERSED
3633 If defined, the order in which arguments are loaded into their
3634 respective argument registers is reversed so that the last
3635 argument is loaded first. This macro only affects arguments
3636 passed in registers.
3641 @subsection How Scalar Function Values Are Returned
3642 @cindex return values in registers
3643 @cindex values, returned by functions
3644 @cindex scalars, returned as values
3646 This section discusses the macros that control returning scalars as
3647 values---values that can fit in registers.
3650 @findex FUNCTION_VALUE
3651 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3652 A C expression to create an RTX representing the place where a
3653 function returns a value of data type @var{valtype}. @var{valtype} is
3654 a tree node representing a data type. Write @code{TYPE_MODE
3655 (@var{valtype})} to get the machine mode used to represent that type.
3656 On many machines, only the mode is relevant. (Actually, on most
3657 machines, scalar values are returned in the same place regardless of
3660 The value of the expression is usually a @code{reg} RTX for the hard
3661 register where the return value is stored. The value can also be a
3662 @code{parallel} RTX, if the return value is in multiple places. See
3663 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3665 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3666 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3669 If the precise function being called is known, @var{func} is a tree
3670 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3671 pointer. This makes it possible to use a different value-returning
3672 convention for specific functions when all their calls are
3675 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3676 types, because these are returned in another way. See
3677 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3679 @findex FUNCTION_OUTGOING_VALUE
3680 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3681 Define this macro if the target machine has ``register windows''
3682 so that the register in which a function returns its value is not
3683 the same as the one in which the caller sees the value.
3685 For such machines, @code{FUNCTION_VALUE} computes the register in which
3686 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3687 defined in a similar fashion to tell the function where to put the
3690 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3691 @code{FUNCTION_VALUE} serves both purposes.
3693 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3694 aggregate data types, because these are returned in another way. See
3695 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3697 @findex LIBCALL_VALUE
3698 @item LIBCALL_VALUE (@var{mode})
3699 A C expression to create an RTX representing the place where a library
3700 function returns a value of mode @var{mode}. If the precise function
3701 being called is known, @var{func} is a tree node
3702 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3703 pointer. This makes it possible to use a different value-returning
3704 convention for specific functions when all their calls are
3707 Note that ``library function'' in this context means a compiler
3708 support routine, used to perform arithmetic, whose name is known
3709 specially by the compiler and was not mentioned in the C code being
3712 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3713 data types, because none of the library functions returns such types.
3715 @findex FUNCTION_VALUE_REGNO_P
3716 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3717 A C expression that is nonzero if @var{regno} is the number of a hard
3718 register in which the values of called function may come back.
3720 A register whose use for returning values is limited to serving as the
3721 second of a pair (for a value of type @code{double}, say) need not be
3722 recognized by this macro. So for most machines, this definition
3726 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3729 If the machine has register windows, so that the caller and the called
3730 function use different registers for the return value, this macro
3731 should recognize only the caller's register numbers.
3733 @findex APPLY_RESULT_SIZE
3734 @item APPLY_RESULT_SIZE
3735 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3736 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3737 saving and restoring an arbitrary return value.
3740 @node Aggregate Return
3741 @subsection How Large Values Are Returned
3742 @cindex aggregates as return values
3743 @cindex large return values
3744 @cindex returning aggregate values
3745 @cindex structure value address
3747 When a function value's mode is @code{BLKmode} (and in some other
3748 cases), the value is not returned according to @code{FUNCTION_VALUE}
3749 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3750 block of memory in which the value should be stored. This address
3751 is called the @dfn{structure value address}.
3753 This section describes how to control returning structure values in
3757 @findex RETURN_IN_MEMORY
3758 @item RETURN_IN_MEMORY (@var{type})
3759 A C expression which can inhibit the returning of certain function
3760 values in registers, based on the type of value. A nonzero value says
3761 to return the function value in memory, just as large structures are
3762 always returned. Here @var{type} will be a C expression of type
3763 @code{tree}, representing the data type of the value.
3765 Note that values of mode @code{BLKmode} must be explicitly handled
3766 by this macro. Also, the option @option{-fpcc-struct-return}
3767 takes effect regardless of this macro. On most systems, it is
3768 possible to leave the macro undefined; this causes a default
3769 definition to be used, whose value is the constant 1 for @code{BLKmode}
3770 values, and 0 otherwise.
3772 Do not use this macro to indicate that structures and unions should always
3773 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3776 @findex DEFAULT_PCC_STRUCT_RETURN
3777 @item DEFAULT_PCC_STRUCT_RETURN
3778 Define this macro to be 1 if all structure and union return values must be
3779 in memory. Since this results in slower code, this should be defined
3780 only if needed for compatibility with other compilers or with an ABI@.
3781 If you define this macro to be 0, then the conventions used for structure
3782 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3784 If not defined, this defaults to the value 1.
3786 @findex STRUCT_VALUE_REGNUM
3787 @item STRUCT_VALUE_REGNUM
3788 If the structure value address is passed in a register, then
3789 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3791 @findex STRUCT_VALUE
3793 If the structure value address is not passed in a register, define
3794 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3795 where the address is passed. If it returns 0, the address is passed as
3796 an ``invisible'' first argument.
3798 @findex STRUCT_VALUE_INCOMING_REGNUM
3799 @item STRUCT_VALUE_INCOMING_REGNUM
3800 On some architectures the place where the structure value address
3801 is found by the called function is not the same place that the
3802 caller put it. This can be due to register windows, or it could
3803 be because the function prologue moves it to a different place.
3805 If the incoming location of the structure value address is in a
3806 register, define this macro as the register number.
3808 @findex STRUCT_VALUE_INCOMING
3809 @item STRUCT_VALUE_INCOMING
3810 If the incoming location is not a register, then you should define
3811 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3812 called function should find the value. If it should find the value on
3813 the stack, define this to create a @code{mem} which refers to the frame
3814 pointer. A definition of 0 means that the address is passed as an
3815 ``invisible'' first argument.
3817 @findex PCC_STATIC_STRUCT_RETURN
3818 @item PCC_STATIC_STRUCT_RETURN
3819 Define this macro if the usual system convention on the target machine
3820 for returning structures and unions is for the called function to return
3821 the address of a static variable containing the value.
3823 Do not define this if the usual system convention is for the caller to
3824 pass an address to the subroutine.
3826 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3827 nothing when you use @option{-freg-struct-return} mode.
3831 @subsection Caller-Saves Register Allocation
3833 If you enable it, GCC can save registers around function calls. This
3834 makes it possible to use call-clobbered registers to hold variables that
3835 must live across calls.
3838 @findex DEFAULT_CALLER_SAVES
3839 @item DEFAULT_CALLER_SAVES
3840 Define this macro if function calls on the target machine do not preserve
3841 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3842 for all registers. When defined, this macro enables @option{-fcaller-saves}
3843 by default for all optimization levels. It has no effect for optimization
3844 levels 2 and higher, where @option{-fcaller-saves} is the default.
3846 @findex CALLER_SAVE_PROFITABLE
3847 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3848 A C expression to determine whether it is worthwhile to consider placing
3849 a pseudo-register in a call-clobbered hard register and saving and
3850 restoring it around each function call. The expression should be 1 when
3851 this is worth doing, and 0 otherwise.
3853 If you don't define this macro, a default is used which is good on most
3854 machines: @code{4 * @var{calls} < @var{refs}}.
3856 @findex HARD_REGNO_CALLER_SAVE_MODE
3857 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3858 A C expression specifying which mode is required for saving @var{nregs}
3859 of a pseudo-register in call-clobbered hard register @var{regno}. If
3860 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3861 returned. For most machines this macro need not be defined since GCC
3862 will select the smallest suitable mode.
3865 @node Function Entry
3866 @subsection Function Entry and Exit
3867 @cindex function entry and exit
3871 This section describes the macros that output function entry
3872 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3874 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
3875 If defined, a function that outputs the assembler code for entry to a
3876 function. The prologue is responsible for setting up the stack frame,
3877 initializing the frame pointer register, saving registers that must be
3878 saved, and allocating @var{size} additional bytes of storage for the
3879 local variables. @var{size} is an integer. @var{file} is a stdio
3880 stream to which the assembler code should be output.
3882 The label for the beginning of the function need not be output by this
3883 macro. That has already been done when the macro is run.
3885 @findex regs_ever_live
3886 To determine which registers to save, the macro can refer to the array
3887 @code{regs_ever_live}: element @var{r} is nonzero if hard register
3888 @var{r} is used anywhere within the function. This implies the function
3889 prologue should save register @var{r}, provided it is not one of the
3890 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
3891 @code{regs_ever_live}.)
3893 On machines that have ``register windows'', the function entry code does
3894 not save on the stack the registers that are in the windows, even if
3895 they are supposed to be preserved by function calls; instead it takes
3896 appropriate steps to ``push'' the register stack, if any non-call-used
3897 registers are used in the function.
3899 @findex frame_pointer_needed
3900 On machines where functions may or may not have frame-pointers, the
3901 function entry code must vary accordingly; it must set up the frame
3902 pointer if one is wanted, and not otherwise. To determine whether a
3903 frame pointer is in wanted, the macro can refer to the variable
3904 @code{frame_pointer_needed}. The variable's value will be 1 at run
3905 time in a function that needs a frame pointer. @xref{Elimination}.
3907 The function entry code is responsible for allocating any stack space
3908 required for the function. This stack space consists of the regions
3909 listed below. In most cases, these regions are allocated in the
3910 order listed, with the last listed region closest to the top of the
3911 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
3912 the highest address if it is not defined). You can use a different order
3913 for a machine if doing so is more convenient or required for
3914 compatibility reasons. Except in cases where required by standard
3915 or by a debugger, there is no reason why the stack layout used by GCC
3916 need agree with that used by other compilers for a machine.
3919 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
3920 If defined, a function that outputs assembler code at the end of a
3921 prologue. This should be used when the function prologue is being
3922 emitted as RTL, and you have some extra assembler that needs to be
3923 emitted. @xref{prologue instruction pattern}.
3926 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
3927 If defined, a function that outputs assembler code at the start of an
3928 epilogue. This should be used when the function epilogue is being
3929 emitted as RTL, and you have some extra assembler that needs to be
3930 emitted. @xref{epilogue instruction pattern}.
3933 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
3934 If defined, a function that outputs the assembler code for exit from a
3935 function. The epilogue is responsible for restoring the saved
3936 registers and stack pointer to their values when the function was
3937 called, and returning control to the caller. This macro takes the
3938 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
3939 registers to restore are determined from @code{regs_ever_live} and
3940 @code{CALL_USED_REGISTERS} in the same way.
3942 On some machines, there is a single instruction that does all the work
3943 of returning from the function. On these machines, give that
3944 instruction the name @samp{return} and do not define the macro
3945 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
3947 Do not define a pattern named @samp{return} if you want the
3948 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
3949 switches to control whether return instructions or epilogues are used,
3950 define a @samp{return} pattern with a validity condition that tests the
3951 target switches appropriately. If the @samp{return} pattern's validity
3952 condition is false, epilogues will be used.
3954 On machines where functions may or may not have frame-pointers, the
3955 function exit code must vary accordingly. Sometimes the code for these
3956 two cases is completely different. To determine whether a frame pointer
3957 is wanted, the macro can refer to the variable
3958 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
3959 a function that needs a frame pointer.
3961 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
3962 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
3963 The C variable @code{current_function_is_leaf} is nonzero for such a
3964 function. @xref{Leaf Functions}.
3966 On some machines, some functions pop their arguments on exit while
3967 others leave that for the caller to do. For example, the 68020 when
3968 given @option{-mrtd} pops arguments in functions that take a fixed
3969 number of arguments.
3971 @findex current_function_pops_args
3972 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
3973 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
3974 needs to know what was decided. The variable that is called
3975 @code{current_function_pops_args} is the number of bytes of its
3976 arguments that a function should pop. @xref{Scalar Return}.
3977 @c what is the "its arguments" in the above sentence referring to, pray
3978 @c tell? --mew 5feb93
3985 @findex current_function_pretend_args_size
3986 A region of @code{current_function_pretend_args_size} bytes of
3987 uninitialized space just underneath the first argument arriving on the
3988 stack. (This may not be at the very start of the allocated stack region
3989 if the calling sequence has pushed anything else since pushing the stack
3990 arguments. But usually, on such machines, nothing else has been pushed
3991 yet, because the function prologue itself does all the pushing.) This
3992 region is used on machines where an argument may be passed partly in
3993 registers and partly in memory, and, in some cases to support the
3994 features in @code{<varargs.h>} and @code{<stdarg.h>}.
3997 An area of memory used to save certain registers used by the function.
3998 The size of this area, which may also include space for such things as
3999 the return address and pointers to previous stack frames, is
4000 machine-specific and usually depends on which registers have been used
4001 in the function. Machines with register windows often do not require
4005 A region of at least @var{size} bytes, possibly rounded up to an allocation
4006 boundary, to contain the local variables of the function. On some machines,
4007 this region and the save area may occur in the opposite order, with the
4008 save area closer to the top of the stack.
4011 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4012 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4013 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4014 argument lists of the function. @xref{Stack Arguments}.
4017 Normally, it is necessary for the macros
4018 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4019 @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially.
4020 The C variable @code{current_function_is_leaf} is nonzero for such a
4023 @findex EXIT_IGNORE_STACK
4024 @item EXIT_IGNORE_STACK
4025 Define this macro as a C expression that is nonzero if the return
4026 instruction or the function epilogue ignores the value of the stack
4027 pointer; in other words, if it is safe to delete an instruction to
4028 adjust the stack pointer before a return from the function.
4030 Note that this macro's value is relevant only for functions for which
4031 frame pointers are maintained. It is never safe to delete a final
4032 stack adjustment in a function that has no frame pointer, and the
4033 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4035 @findex EPILOGUE_USES
4036 @item EPILOGUE_USES (@var{regno})
4037 Define this macro as a C expression that is nonzero for registers that are
4038 used by the epilogue or the @samp{return} pattern. The stack and frame
4039 pointer registers are already be assumed to be used as needed.
4042 @item EH_USES (@var{regno})
4043 Define this macro as a C expression that is nonzero for registers that are
4044 used by the exception handling mechanism, and so should be considered live
4045 on entry to an exception edge.
4047 @findex DELAY_SLOTS_FOR_EPILOGUE
4048 @item DELAY_SLOTS_FOR_EPILOGUE
4049 Define this macro if the function epilogue contains delay slots to which
4050 instructions from the rest of the function can be ``moved''. The
4051 definition should be a C expression whose value is an integer
4052 representing the number of delay slots there.
4054 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
4055 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4056 A C expression that returns 1 if @var{insn} can be placed in delay
4057 slot number @var{n} of the epilogue.
4059 The argument @var{n} is an integer which identifies the delay slot now
4060 being considered (since different slots may have different rules of
4061 eligibility). It is never negative and is always less than the number
4062 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4063 If you reject a particular insn for a given delay slot, in principle, it
4064 may be reconsidered for a subsequent delay slot. Also, other insns may
4065 (at least in principle) be considered for the so far unfilled delay
4068 @findex current_function_epilogue_delay_list
4069 @findex final_scan_insn
4070 The insns accepted to fill the epilogue delay slots are put in an RTL
4071 list made with @code{insn_list} objects, stored in the variable
4072 @code{current_function_epilogue_delay_list}. The insn for the first
4073 delay slot comes first in the list. Your definition of the macro
4074 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4075 outputting the insns in this list, usually by calling
4076 @code{final_scan_insn}.
4078 You need not define this macro if you did not define
4079 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4081 @findex ASM_OUTPUT_MI_THUNK
4082 @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function})
4083 A C compound statement that outputs the assembler code for a thunk
4084 function, used to implement C++ virtual function calls with multiple
4085 inheritance. The thunk acts as a wrapper around a virtual function,
4086 adjusting the implicit object parameter before handing control off to
4089 First, emit code to add the integer @var{delta} to the location that
4090 contains the incoming first argument. Assume that this argument
4091 contains a pointer, and is the one used to pass the @code{this} pointer
4092 in C++. This is the incoming argument @emph{before} the function prologue,
4093 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4094 all other incoming arguments.
4096 After the addition, emit code to jump to @var{function}, which is a
4097 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4098 not touch the return address. Hence returning from @var{FUNCTION} will
4099 return to whoever called the current @samp{thunk}.
4101 The effect must be as if @var{function} had been called directly with
4102 the adjusted first argument. This macro is responsible for emitting all
4103 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4104 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4106 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4107 have already been extracted from it.) It might possibly be useful on
4108 some targets, but probably not.
4110 If you do not define this macro, the target-independent code in the C++
4111 front end will generate a less efficient heavyweight thunk that calls
4112 @var{function} instead of jumping to it. The generic approach does
4113 not support varargs.
4117 @subsection Generating Code for Profiling
4118 @cindex profiling, code generation
4120 These macros will help you generate code for profiling.
4123 @findex FUNCTION_PROFILER
4124 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
4125 A C statement or compound statement to output to @var{file} some
4126 assembler code to call the profiling subroutine @code{mcount}.
4129 The details of how @code{mcount} expects to be called are determined by
4130 your operating system environment, not by GCC@. To figure them out,
4131 compile a small program for profiling using the system's installed C
4132 compiler and look at the assembler code that results.
4134 Older implementations of @code{mcount} expect the address of a counter
4135 variable to be loaded into some register. The name of this variable is
4136 @samp{LP} followed by the number @var{labelno}, so you would generate
4137 the name using @samp{LP%d} in a @code{fprintf}.
4139 @findex PROFILE_HOOK
4141 A C statement or compound statement to output to @var{file} some assembly
4142 code to call the profiling subroutine @code{mcount} even the target does
4143 not support profiling.
4145 @findex NO_PROFILE_COUNTERS
4146 @item NO_PROFILE_COUNTERS
4147 Define this macro if the @code{mcount} subroutine on your system does
4148 not need a counter variable allocated for each function. This is true
4149 for almost all modern implementations. If you define this macro, you
4150 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4152 @findex PROFILE_BEFORE_PROLOGUE
4153 @item PROFILE_BEFORE_PROLOGUE
4154 Define this macro if the code for function profiling should come before
4155 the function prologue. Normally, the profiling code comes after.
4158 @findex TARGET_ALLOWS_PROFILING_WITHOUT_FRAME_POINTER
4159 @item TARGET_ALLOWS_PROFILING_WITHOUT_FRAME_POINTER
4160 On some targets, it is impossible to use profiling when the frame
4161 pointer has been omitted. For example, on x86 GNU/Linux systems,
4162 the @code{mcount} routine provided by the GNU C Library finds the
4163 address of the routine that called the routine that called @code{mcount}
4164 by looking in the immediate caller's stack frame. If the immediate
4165 caller has no frame pointer, this lookup will fail.
4167 By default, GCC assumes that the target does allow profiling when the
4168 frame pointer is omitted. This macro should be defined to a C
4169 expression that evaluates to @code{false} if the target does not allow
4170 profiling when the frame pointer is omitted.
4175 @subsection Permitting tail calls
4179 @findex FUNCTION_OK_FOR_SIBCALL
4180 @item FUNCTION_OK_FOR_SIBCALL (@var{decl})
4181 A C expression that evaluates to true if it is ok to perform a sibling
4182 call to @var{decl} from the current function.
4184 It is not uncommon for limitations of calling conventions to prevent
4185 tail calls to functions outside the current unit of translation, or
4186 during PIC compilation. Use this macro to enforce these restrictions,
4187 as the @code{sibcall} md pattern can not fail, or fall over to a
4192 @section Implementing the Varargs Macros
4193 @cindex varargs implementation
4195 GCC comes with an implementation of @code{<varargs.h>} and
4196 @code{<stdarg.h>} that work without change on machines that pass arguments
4197 on the stack. Other machines require their own implementations of
4198 varargs, and the two machine independent header files must have
4199 conditionals to include it.
4201 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4202 the calling convention for @code{va_start}. The traditional
4203 implementation takes just one argument, which is the variable in which
4204 to store the argument pointer. The ISO implementation of
4205 @code{va_start} takes an additional second argument. The user is
4206 supposed to write the last named argument of the function here.
4208 However, @code{va_start} should not use this argument. The way to find
4209 the end of the named arguments is with the built-in functions described
4213 @findex __builtin_saveregs
4214 @item __builtin_saveregs ()
4215 Use this built-in function to save the argument registers in memory so
4216 that the varargs mechanism can access them. Both ISO and traditional
4217 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4218 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
4220 On some machines, @code{__builtin_saveregs} is open-coded under the
4221 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
4222 it calls a routine written in assembler language, found in
4225 Code generated for the call to @code{__builtin_saveregs} appears at the
4226 beginning of the function, as opposed to where the call to
4227 @code{__builtin_saveregs} is written, regardless of what the code is.
4228 This is because the registers must be saved before the function starts
4229 to use them for its own purposes.
4230 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4233 @findex __builtin_args_info
4234 @item __builtin_args_info (@var{category})
4235 Use this built-in function to find the first anonymous arguments in
4238 In general, a machine may have several categories of registers used for
4239 arguments, each for a particular category of data types. (For example,
4240 on some machines, floating-point registers are used for floating-point
4241 arguments while other arguments are passed in the general registers.)
4242 To make non-varargs functions use the proper calling convention, you
4243 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4244 registers in each category have been used so far
4246 @code{__builtin_args_info} accesses the same data structure of type
4247 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4248 with it, with @var{category} specifying which word to access. Thus, the
4249 value indicates the first unused register in a given category.
4251 Normally, you would use @code{__builtin_args_info} in the implementation
4252 of @code{va_start}, accessing each category just once and storing the
4253 value in the @code{va_list} object. This is because @code{va_list} will
4254 have to update the values, and there is no way to alter the
4255 values accessed by @code{__builtin_args_info}.
4257 @findex __builtin_next_arg
4258 @item __builtin_next_arg (@var{lastarg})
4259 This is the equivalent of @code{__builtin_args_info}, for stack
4260 arguments. It returns the address of the first anonymous stack
4261 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4262 returns the address of the location above the first anonymous stack
4263 argument. Use it in @code{va_start} to initialize the pointer for
4264 fetching arguments from the stack. Also use it in @code{va_start} to
4265 verify that the second parameter @var{lastarg} is the last named argument
4266 of the current function.
4268 @findex __builtin_classify_type
4269 @item __builtin_classify_type (@var{object})
4270 Since each machine has its own conventions for which data types are
4271 passed in which kind of register, your implementation of @code{va_arg}
4272 has to embody these conventions. The easiest way to categorize the
4273 specified data type is to use @code{__builtin_classify_type} together
4274 with @code{sizeof} and @code{__alignof__}.
4276 @code{__builtin_classify_type} ignores the value of @var{object},
4277 considering only its data type. It returns an integer describing what
4278 kind of type that is---integer, floating, pointer, structure, and so on.
4280 The file @file{typeclass.h} defines an enumeration that you can use to
4281 interpret the values of @code{__builtin_classify_type}.
4284 These machine description macros help implement varargs:
4287 @findex EXPAND_BUILTIN_SAVEREGS
4288 @item EXPAND_BUILTIN_SAVEREGS ()
4289 If defined, is a C expression that produces the machine-specific code
4290 for a call to @code{__builtin_saveregs}. This code will be moved to the
4291 very beginning of the function, before any parameter access are made.
4292 The return value of this function should be an RTX that contains the
4293 value to use as the return of @code{__builtin_saveregs}.
4295 @findex SETUP_INCOMING_VARARGS
4296 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
4297 This macro offers an alternative to using @code{__builtin_saveregs} and
4298 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
4299 anonymous register arguments into the stack so that all the arguments
4300 appear to have been passed consecutively on the stack. Once this is
4301 done, you can use the standard implementation of varargs that works for
4302 machines that pass all their arguments on the stack.
4304 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
4305 structure, containing the values that are obtained after processing the
4306 named arguments. The arguments @var{mode} and @var{type} describe the
4307 last named argument---its machine mode and its data type as a tree node.
4309 The macro implementation should do two things: first, push onto the
4310 stack all the argument registers @emph{not} used for the named
4311 arguments, and second, store the size of the data thus pushed into the
4312 @code{int}-valued variable whose name is supplied as the argument
4313 @var{pretend_args_size}. The value that you store here will serve as
4314 additional offset for setting up the stack frame.
4316 Because you must generate code to push the anonymous arguments at
4317 compile time without knowing their data types,
4318 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
4319 a single category of argument register and use it uniformly for all data
4322 If the argument @var{second_time} is nonzero, it means that the
4323 arguments of the function are being analyzed for the second time. This
4324 happens for an inline function, which is not actually compiled until the
4325 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
4326 not generate any instructions in this case.
4328 @findex STRICT_ARGUMENT_NAMING
4329 @item STRICT_ARGUMENT_NAMING
4330 Define this macro to be a nonzero value if the location where a function
4331 argument is passed depends on whether or not it is a named argument.
4333 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
4334 is set for varargs and stdarg functions. If this macro returns a
4335 nonzero value, the @var{named} argument is always true for named
4336 arguments, and false for unnamed arguments. If it returns a value of
4337 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
4338 are treated as named. Otherwise, all named arguments except the last
4339 are treated as named.
4341 You need not define this macro if it always returns zero.
4343 @findex PRETEND_OUTGOING_VARARGS_NAMED
4344 @item PRETEND_OUTGOING_VARARGS_NAMED
4345 If you need to conditionally change ABIs so that one works with
4346 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
4347 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
4348 defined, then define this macro to return nonzero if
4349 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
4350 Otherwise, you should not define this macro.
4354 @section Trampolines for Nested Functions
4355 @cindex trampolines for nested functions
4356 @cindex nested functions, trampolines for
4358 A @dfn{trampoline} is a small piece of code that is created at run time
4359 when the address of a nested function is taken. It normally resides on
4360 the stack, in the stack frame of the containing function. These macros
4361 tell GCC how to generate code to allocate and initialize a
4364 The instructions in the trampoline must do two things: load a constant
4365 address into the static chain register, and jump to the real address of
4366 the nested function. On CISC machines such as the m68k, this requires
4367 two instructions, a move immediate and a jump. Then the two addresses
4368 exist in the trampoline as word-long immediate operands. On RISC
4369 machines, it is often necessary to load each address into a register in
4370 two parts. Then pieces of each address form separate immediate
4373 The code generated to initialize the trampoline must store the variable
4374 parts---the static chain value and the function address---into the
4375 immediate operands of the instructions. On a CISC machine, this is
4376 simply a matter of copying each address to a memory reference at the
4377 proper offset from the start of the trampoline. On a RISC machine, it
4378 may be necessary to take out pieces of the address and store them
4382 @findex TRAMPOLINE_TEMPLATE
4383 @item TRAMPOLINE_TEMPLATE (@var{file})
4384 A C statement to output, on the stream @var{file}, assembler code for a
4385 block of data that contains the constant parts of a trampoline. This
4386 code should not include a label---the label is taken care of
4389 If you do not define this macro, it means no template is needed
4390 for the target. Do not define this macro on systems where the block move
4391 code to copy the trampoline into place would be larger than the code
4392 to generate it on the spot.
4394 @findex TRAMPOLINE_SECTION
4395 @item TRAMPOLINE_SECTION
4396 The name of a subroutine to switch to the section in which the
4397 trampoline template is to be placed (@pxref{Sections}). The default is
4398 a value of @samp{readonly_data_section}, which places the trampoline in
4399 the section containing read-only data.
4401 @findex TRAMPOLINE_SIZE
4402 @item TRAMPOLINE_SIZE
4403 A C expression for the size in bytes of the trampoline, as an integer.
4405 @findex TRAMPOLINE_ALIGNMENT
4406 @item TRAMPOLINE_ALIGNMENT
4407 Alignment required for trampolines, in bits.
4409 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4410 is used for aligning trampolines.
4412 @findex INITIALIZE_TRAMPOLINE
4413 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4414 A C statement to initialize the variable parts of a trampoline.
4415 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4416 an RTX for the address of the nested function; @var{static_chain} is an
4417 RTX for the static chain value that should be passed to the function
4420 @findex TRAMPOLINE_ADJUST_ADDRESS
4421 @item TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4422 A C statement that should perform any machine-specific adjustment in
4423 the address of the trampoline. Its argument contains the address that
4424 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4425 used for a function call should be different from the address in which
4426 the template was stored, the different address should be assigned to
4427 @var{addr}. If this macro is not defined, @var{addr} will be used for
4430 @findex ALLOCATE_TRAMPOLINE
4431 @item ALLOCATE_TRAMPOLINE (@var{fp})
4432 A C expression to allocate run-time space for a trampoline. The
4433 expression value should be an RTX representing a memory reference to the
4434 space for the trampoline.
4436 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4437 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4438 If this macro is not defined, by default the trampoline is allocated as
4439 a stack slot. This default is right for most machines. The exceptions
4440 are machines where it is impossible to execute instructions in the stack
4441 area. On such machines, you may have to implement a separate stack,
4442 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4443 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4445 @var{fp} points to a data structure, a @code{struct function}, which
4446 describes the compilation status of the immediate containing function of
4447 the function which the trampoline is for. Normally (when
4448 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
4449 trampoline is in the stack frame of this containing function. Other
4450 allocation strategies probably must do something analogous with this
4454 Implementing trampolines is difficult on many machines because they have
4455 separate instruction and data caches. Writing into a stack location
4456 fails to clear the memory in the instruction cache, so when the program
4457 jumps to that location, it executes the old contents.
4459 Here are two possible solutions. One is to clear the relevant parts of
4460 the instruction cache whenever a trampoline is set up. The other is to
4461 make all trampolines identical, by having them jump to a standard
4462 subroutine. The former technique makes trampoline execution faster; the
4463 latter makes initialization faster.
4465 To clear the instruction cache when a trampoline is initialized, define
4466 the following macros which describe the shape of the cache.
4469 @findex INSN_CACHE_SIZE
4470 @item INSN_CACHE_SIZE
4471 The total size in bytes of the cache.
4473 @findex INSN_CACHE_LINE_WIDTH
4474 @item INSN_CACHE_LINE_WIDTH
4475 The length in bytes of each cache line. The cache is divided into cache
4476 lines which are disjoint slots, each holding a contiguous chunk of data
4477 fetched from memory. Each time data is brought into the cache, an
4478 entire line is read at once. The data loaded into a cache line is
4479 always aligned on a boundary equal to the line size.
4481 @findex INSN_CACHE_DEPTH
4482 @item INSN_CACHE_DEPTH
4483 The number of alternative cache lines that can hold any particular memory
4487 Alternatively, if the machine has system calls or instructions to clear
4488 the instruction cache directly, you can define the following macro.
4491 @findex CLEAR_INSN_CACHE
4492 @item CLEAR_INSN_CACHE (@var{beg}, @var{end})
4493 If defined, expands to a C expression clearing the @emph{instruction
4494 cache} in the specified interval. If it is not defined, and the macro
4495 @code{INSN_CACHE_SIZE} is defined, some generic code is generated to clear the
4496 cache. The definition of this macro would typically be a series of
4497 @code{asm} statements. Both @var{beg} and @var{end} are both pointer
4501 To use a standard subroutine, define the following macro. In addition,
4502 you must make sure that the instructions in a trampoline fill an entire
4503 cache line with identical instructions, or else ensure that the
4504 beginning of the trampoline code is always aligned at the same point in
4505 its cache line. Look in @file{m68k.h} as a guide.
4508 @findex TRANSFER_FROM_TRAMPOLINE
4509 @item TRANSFER_FROM_TRAMPOLINE
4510 Define this macro if trampolines need a special subroutine to do their
4511 work. The macro should expand to a series of @code{asm} statements
4512 which will be compiled with GCC@. They go in a library function named
4513 @code{__transfer_from_trampoline}.
4515 If you need to avoid executing the ordinary prologue code of a compiled
4516 C function when you jump to the subroutine, you can do so by placing a
4517 special label of your own in the assembler code. Use one @code{asm}
4518 statement to generate an assembler label, and another to make the label
4519 global. Then trampolines can use that label to jump directly to your
4520 special assembler code.
4524 @section Implicit Calls to Library Routines
4525 @cindex library subroutine names
4526 @cindex @file{libgcc.a}
4528 @c prevent bad page break with this line
4529 Here is an explanation of implicit calls to library routines.
4532 @findex MULSI3_LIBCALL
4533 @item MULSI3_LIBCALL
4534 A C string constant giving the name of the function to call for
4535 multiplication of one signed full-word by another. If you do not
4536 define this macro, the default name is used, which is @code{__mulsi3},
4537 a function defined in @file{libgcc.a}.
4539 @findex DIVSI3_LIBCALL
4540 @item DIVSI3_LIBCALL
4541 A C string constant giving the name of the function to call for
4542 division of one signed full-word by another. If you do not define
4543 this macro, the default name is used, which is @code{__divsi3}, a
4544 function defined in @file{libgcc.a}.
4546 @findex UDIVSI3_LIBCALL
4547 @item UDIVSI3_LIBCALL
4548 A C string constant giving the name of the function to call for
4549 division of one unsigned full-word by another. If you do not define
4550 this macro, the default name is used, which is @code{__udivsi3}, a
4551 function defined in @file{libgcc.a}.
4553 @findex MODSI3_LIBCALL
4554 @item MODSI3_LIBCALL
4555 A C string constant giving the name of the function to call for the
4556 remainder in division of one signed full-word by another. If you do
4557 not define this macro, the default name is used, which is
4558 @code{__modsi3}, a function defined in @file{libgcc.a}.
4560 @findex UMODSI3_LIBCALL
4561 @item UMODSI3_LIBCALL
4562 A C string constant giving the name of the function to call for the
4563 remainder in division of one unsigned full-word by another. If you do
4564 not define this macro, the default name is used, which is
4565 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4567 @findex MULDI3_LIBCALL
4568 @item MULDI3_LIBCALL
4569 A C string constant giving the name of the function to call for
4570 multiplication of one signed double-word by another. If you do not
4571 define this macro, the default name is used, which is @code{__muldi3},
4572 a function defined in @file{libgcc.a}.
4574 @findex DIVDI3_LIBCALL
4575 @item DIVDI3_LIBCALL
4576 A C string constant giving the name of the function to call for
4577 division of one signed double-word by another. If you do not define
4578 this macro, the default name is used, which is @code{__divdi3}, a
4579 function defined in @file{libgcc.a}.
4581 @findex UDIVDI3_LIBCALL
4582 @item UDIVDI3_LIBCALL
4583 A C string constant giving the name of the function to call for
4584 division of one unsigned full-word by another. If you do not define
4585 this macro, the default name is used, which is @code{__udivdi3}, a
4586 function defined in @file{libgcc.a}.
4588 @findex MODDI3_LIBCALL
4589 @item MODDI3_LIBCALL
4590 A C string constant giving the name of the function to call for the
4591 remainder in division of one signed double-word by another. If you do
4592 not define this macro, the default name is used, which is
4593 @code{__moddi3}, a function defined in @file{libgcc.a}.
4595 @findex UMODDI3_LIBCALL
4596 @item UMODDI3_LIBCALL
4597 A C string constant giving the name of the function to call for the
4598 remainder in division of one unsigned full-word by another. If you do
4599 not define this macro, the default name is used, which is
4600 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4602 @findex INIT_TARGET_OPTABS
4603 @item INIT_TARGET_OPTABS
4604 Define this macro as a C statement that declares additional library
4605 routines renames existing ones. @code{init_optabs} calls this macro after
4606 initializing all the normal library routines.
4608 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4609 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4610 Define this macro as a C statement that returns nonzero if a call to
4611 the floating point comparison library function will return a boolean
4612 value that indicates the result of the comparison. It should return
4613 zero if one of gcc's own libgcc functions is called.
4615 Most ports don't need to define this macro.
4618 @cindex @code{EDOM}, implicit usage
4620 The value of @code{EDOM} on the target machine, as a C integer constant
4621 expression. If you don't define this macro, GCC does not attempt to
4622 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4623 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4626 If you do not define @code{TARGET_EDOM}, then compiled code reports
4627 domain errors by calling the library function and letting it report the
4628 error. If mathematical functions on your system use @code{matherr} when
4629 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4630 that @code{matherr} is used normally.
4632 @findex GEN_ERRNO_RTX
4633 @cindex @code{errno}, implicit usage
4635 Define this macro as a C expression to create an rtl expression that
4636 refers to the global ``variable'' @code{errno}. (On certain systems,
4637 @code{errno} may not actually be a variable.) If you don't define this
4638 macro, a reasonable default is used.
4640 @findex TARGET_MEM_FUNCTIONS
4641 @cindex @code{bcopy}, implicit usage
4642 @cindex @code{memcpy}, implicit usage
4643 @cindex @code{memmove}, implicit usage
4644 @cindex @code{bzero}, implicit usage
4645 @cindex @code{memset}, implicit usage
4646 @item TARGET_MEM_FUNCTIONS
4647 Define this macro if GCC should generate calls to the ISO C
4648 (and System V) library functions @code{memcpy}, @code{memmove} and
4649 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4651 @findex LIBGCC_NEEDS_DOUBLE
4652 @item LIBGCC_NEEDS_DOUBLE
4653 Define this macro if @code{float} arguments cannot be passed to library
4654 routines (so they must be converted to @code{double}). This macro
4655 affects both how library calls are generated and how the library
4656 routines in @file{libgcc.a} accept their arguments. It is useful on
4657 machines where floating and fixed point arguments are passed
4658 differently, such as the i860.
4660 @findex NEXT_OBJC_RUNTIME
4661 @item NEXT_OBJC_RUNTIME
4662 Define this macro to generate code for Objective-C message sending using
4663 the calling convention of the NeXT system. This calling convention
4664 involves passing the object, the selector and the method arguments all
4665 at once to the method-lookup library function.
4667 The default calling convention passes just the object and the selector
4668 to the lookup function, which returns a pointer to the method.
4671 @node Addressing Modes
4672 @section Addressing Modes
4673 @cindex addressing modes
4675 @c prevent bad page break with this line
4676 This is about addressing modes.
4679 @findex HAVE_PRE_INCREMENT
4680 @findex HAVE_PRE_DECREMENT
4681 @findex HAVE_POST_INCREMENT
4682 @findex HAVE_POST_DECREMENT
4683 @item HAVE_PRE_INCREMENT
4684 @itemx HAVE_PRE_DECREMENT
4685 @itemx HAVE_POST_INCREMENT
4686 @itemx HAVE_POST_DECREMENT
4687 A C expression that is nonzero if the machine supports pre-increment,
4688 pre-decrement, post-increment, or post-decrement addressing respectively.
4690 @findex HAVE_POST_MODIFY_DISP
4691 @findex HAVE_PRE_MODIFY_DISP
4692 @item HAVE_PRE_MODIFY_DISP
4693 @itemx HAVE_POST_MODIFY_DISP
4694 A C expression that is nonzero if the machine supports pre- or
4695 post-address side-effect generation involving constants other than
4696 the size of the memory operand.
4698 @findex HAVE_POST_MODIFY_REG
4699 @findex HAVE_PRE_MODIFY_REG
4700 @item HAVE_PRE_MODIFY_REG
4701 @itemx HAVE_POST_MODIFY_REG
4702 A C expression that is nonzero if the machine supports pre- or
4703 post-address side-effect generation involving a register displacement.
4705 @findex CONSTANT_ADDRESS_P
4706 @item CONSTANT_ADDRESS_P (@var{x})
4707 A C expression that is 1 if the RTX @var{x} is a constant which
4708 is a valid address. On most machines, this can be defined as
4709 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4710 in which constant addresses are supported.
4713 @code{CONSTANT_P} accepts integer-values expressions whose values are
4714 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4715 @code{high} expressions and @code{const} arithmetic expressions, in
4716 addition to @code{const_int} and @code{const_double} expressions.
4718 @findex MAX_REGS_PER_ADDRESS
4719 @item MAX_REGS_PER_ADDRESS
4720 A number, the maximum number of registers that can appear in a valid
4721 memory address. Note that it is up to you to specify a value equal to
4722 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4725 @findex GO_IF_LEGITIMATE_ADDRESS
4726 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4727 A C compound statement with a conditional @code{goto @var{label};}
4728 executed if @var{x} (an RTX) is a legitimate memory address on the
4729 target machine for a memory operand of mode @var{mode}.
4731 It usually pays to define several simpler macros to serve as
4732 subroutines for this one. Otherwise it may be too complicated to
4735 This macro must exist in two variants: a strict variant and a
4736 non-strict one. The strict variant is used in the reload pass. It
4737 must be defined so that any pseudo-register that has not been
4738 allocated a hard register is considered a memory reference. In
4739 contexts where some kind of register is required, a pseudo-register
4740 with no hard register must be rejected.
4742 The non-strict variant is used in other passes. It must be defined to
4743 accept all pseudo-registers in every context where some kind of
4744 register is required.
4746 @findex REG_OK_STRICT
4747 Compiler source files that want to use the strict variant of this
4748 macro define the macro @code{REG_OK_STRICT}. You should use an
4749 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4750 in that case and the non-strict variant otherwise.
4752 Subroutines to check for acceptable registers for various purposes (one
4753 for base registers, one for index registers, and so on) are typically
4754 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4755 Then only these subroutine macros need have two variants; the higher
4756 levels of macros may be the same whether strict or not.
4758 Normally, constant addresses which are the sum of a @code{symbol_ref}
4759 and an integer are stored inside a @code{const} RTX to mark them as
4760 constant. Therefore, there is no need to recognize such sums
4761 specifically as legitimate addresses. Normally you would simply
4762 recognize any @code{const} as legitimate.
4764 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4765 sums that are not marked with @code{const}. It assumes that a naked
4766 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4767 naked constant sums as illegitimate addresses, so that none of them will
4768 be given to @code{PRINT_OPERAND_ADDRESS}.
4770 @cindex @code{ENCODE_SECTION_INFO} and address validation
4771 On some machines, whether a symbolic address is legitimate depends on
4772 the section that the address refers to. On these machines, define the
4773 macro @code{ENCODE_SECTION_INFO} to store the information into the
4774 @code{symbol_ref}, and then check for it here. When you see a
4775 @code{const}, you will have to look inside it to find the
4776 @code{symbol_ref} in order to determine the section. @xref{Assembler
4779 @findex saveable_obstack
4780 The best way to modify the name string is by adding text to the
4781 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4782 the new name in @code{saveable_obstack}. You will have to modify
4783 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4784 output the name accordingly, and define @code{STRIP_NAME_ENCODING} to
4785 access the original name string.
4787 You can check the information stored here into the @code{symbol_ref} in
4788 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4789 @code{PRINT_OPERAND_ADDRESS}.
4791 @findex REG_OK_FOR_BASE_P
4792 @item REG_OK_FOR_BASE_P (@var{x})
4793 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4794 RTX) is valid for use as a base register. For hard registers, it
4795 should always accept those which the hardware permits and reject the
4796 others. Whether the macro accepts or rejects pseudo registers must be
4797 controlled by @code{REG_OK_STRICT} as described above. This usually
4798 requires two variant definitions, of which @code{REG_OK_STRICT}
4799 controls the one actually used.
4801 @findex REG_MODE_OK_FOR_BASE_P
4802 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4803 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4804 that expression may examine the mode of the memory reference in
4805 @var{mode}. You should define this macro if the mode of the memory
4806 reference affects whether a register may be used as a base register. If
4807 you define this macro, the compiler will use it instead of
4808 @code{REG_OK_FOR_BASE_P}.
4810 @findex REG_OK_FOR_INDEX_P
4811 @item REG_OK_FOR_INDEX_P (@var{x})
4812 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4813 RTX) is valid for use as an index register.
4815 The difference between an index register and a base register is that
4816 the index register may be scaled. If an address involves the sum of
4817 two registers, neither one of them scaled, then either one may be
4818 labeled the ``base'' and the other the ``index''; but whichever
4819 labeling is used must fit the machine's constraints of which registers
4820 may serve in each capacity. The compiler will try both labelings,
4821 looking for one that is valid, and will reload one or both registers
4822 only if neither labeling works.
4824 @findex FIND_BASE_TERM
4825 @item FIND_BASE_TERM (@var{x})
4826 A C expression to determine the base term of address @var{x}.
4827 This macro is used in only one place: `find_base_term' in alias.c.
4829 It is always safe for this macro to not be defined. It exists so
4830 that alias analysis can understand machine-dependent addresses.
4832 The typical use of this macro is to handle addresses containing
4833 a label_ref or symbol_ref within an UNSPEC@.
4835 @findex LEGITIMIZE_ADDRESS
4836 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4837 A C compound statement that attempts to replace @var{x} with a valid
4838 memory address for an operand of mode @var{mode}. @var{win} will be a
4839 C statement label elsewhere in the code; the macro definition may use
4842 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4846 to avoid further processing if the address has become legitimate.
4848 @findex break_out_memory_refs
4849 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4850 and @var{oldx} will be the operand that was given to that function to produce
4853 The code generated by this macro should not alter the substructure of
4854 @var{x}. If it transforms @var{x} into a more legitimate form, it
4855 should assign @var{x} (which will always be a C variable) a new value.
4857 It is not necessary for this macro to come up with a legitimate
4858 address. The compiler has standard ways of doing so in all cases. In
4859 fact, it is safe for this macro to do nothing. But often a
4860 machine-dependent strategy can generate better code.
4862 @findex LEGITIMIZE_RELOAD_ADDRESS
4863 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4864 A C compound statement that attempts to replace @var{x}, which is an address
4865 that needs reloading, with a valid memory address for an operand of mode
4866 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4867 It is not necessary to define this macro, but it might be useful for
4868 performance reasons.
4870 For example, on the i386, it is sometimes possible to use a single
4871 reload register instead of two by reloading a sum of two pseudo
4872 registers into a register. On the other hand, for number of RISC
4873 processors offsets are limited so that often an intermediate address
4874 needs to be generated in order to address a stack slot. By defining
4875 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4876 generated for adjacent some stack slots can be made identical, and thus
4879 @emph{Note}: This macro should be used with caution. It is necessary
4880 to know something of how reload works in order to effectively use this,
4881 and it is quite easy to produce macros that build in too much knowledge
4882 of reload internals.
4884 @emph{Note}: This macro must be able to reload an address created by a
4885 previous invocation of this macro. If it fails to handle such addresses
4886 then the compiler may generate incorrect code or abort.
4889 The macro definition should use @code{push_reload} to indicate parts that
4890 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4891 suitable to be passed unaltered to @code{push_reload}.
4893 The code generated by this macro must not alter the substructure of
4894 @var{x}. If it transforms @var{x} into a more legitimate form, it
4895 should assign @var{x} (which will always be a C variable) a new value.
4896 This also applies to parts that you change indirectly by calling
4899 @findex strict_memory_address_p
4900 The macro definition may use @code{strict_memory_address_p} to test if
4901 the address has become legitimate.
4904 If you want to change only a part of @var{x}, one standard way of doing
4905 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4906 single level of rtl. Thus, if the part to be changed is not at the
4907 top level, you'll need to replace first the top level.
4908 It is not necessary for this macro to come up with a legitimate
4909 address; but often a machine-dependent strategy can generate better code.
4911 @findex GO_IF_MODE_DEPENDENT_ADDRESS
4912 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4913 A C statement or compound statement with a conditional @code{goto
4914 @var{label};} executed if memory address @var{x} (an RTX) can have
4915 different meanings depending on the machine mode of the memory
4916 reference it is used for or if the address is valid for some modes
4919 Autoincrement and autodecrement addresses typically have mode-dependent
4920 effects because the amount of the increment or decrement is the size
4921 of the operand being addressed. Some machines have other mode-dependent
4922 addresses. Many RISC machines have no mode-dependent addresses.
4924 You may assume that @var{addr} is a valid address for the machine.
4926 @findex LEGITIMATE_CONSTANT_P
4927 @item LEGITIMATE_CONSTANT_P (@var{x})
4928 A C expression that is nonzero if @var{x} is a legitimate constant for
4929 an immediate operand on the target machine. You can assume that
4930 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4931 @samp{1} is a suitable definition for this macro on machines where
4932 anything @code{CONSTANT_P} is valid.
4935 @node Condition Code
4936 @section Condition Code Status
4937 @cindex condition code status
4939 @c prevent bad page break with this line
4940 This describes the condition code status.
4943 The file @file{conditions.h} defines a variable @code{cc_status} to
4944 describe how the condition code was computed (in case the interpretation of
4945 the condition code depends on the instruction that it was set by). This
4946 variable contains the RTL expressions on which the condition code is
4947 currently based, and several standard flags.
4949 Sometimes additional machine-specific flags must be defined in the machine
4950 description header file. It can also add additional machine-specific
4951 information by defining @code{CC_STATUS_MDEP}.
4954 @findex CC_STATUS_MDEP
4955 @item CC_STATUS_MDEP
4956 C code for a data type which is used for declaring the @code{mdep}
4957 component of @code{cc_status}. It defaults to @code{int}.
4959 This macro is not used on machines that do not use @code{cc0}.
4961 @findex CC_STATUS_MDEP_INIT
4962 @item CC_STATUS_MDEP_INIT
4963 A C expression to initialize the @code{mdep} field to ``empty''.
4964 The default definition does nothing, since most machines don't use
4965 the field anyway. If you want to use the field, you should probably
4966 define this macro to initialize it.
4968 This macro is not used on machines that do not use @code{cc0}.
4970 @findex NOTICE_UPDATE_CC
4971 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4972 A C compound statement to set the components of @code{cc_status}
4973 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4974 this macro's responsibility to recognize insns that set the condition
4975 code as a byproduct of other activity as well as those that explicitly
4978 This macro is not used on machines that do not use @code{cc0}.
4980 If there are insns that do not set the condition code but do alter
4981 other machine registers, this macro must check to see whether they
4982 invalidate the expressions that the condition code is recorded as
4983 reflecting. For example, on the 68000, insns that store in address
4984 registers do not set the condition code, which means that usually
4985 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4986 insns. But suppose that the previous insn set the condition code
4987 based on location @samp{a4@@(102)} and the current insn stores a new
4988 value in @samp{a4}. Although the condition code is not changed by
4989 this, it will no longer be true that it reflects the contents of
4990 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4991 @code{cc_status} in this case to say that nothing is known about the
4992 condition code value.
4994 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4995 with the results of peephole optimization: insns whose patterns are
4996 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4997 constants which are just the operands. The RTL structure of these
4998 insns is not sufficient to indicate what the insns actually do. What
4999 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5000 @code{CC_STATUS_INIT}.
5002 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5003 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5004 @samp{cc}. This avoids having detailed information about patterns in
5005 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5007 @findex EXTRA_CC_MODES
5008 @item EXTRA_CC_MODES
5009 A list of additional modes for condition code values in registers
5010 (@pxref{Jump Patterns}). This macro should expand to a sequence of
5011 calls of the macro @code{CC} separated by white space. @code{CC} takes
5012 two arguments. The first is the enumeration name of the mode, which
5013 should begin with @samp{CC} and end with @samp{mode}. The second is a C
5014 string giving the printable name of the mode; it should be the same as
5015 the first argument, but with the trailing @samp{mode} removed.
5017 You should only define this macro if additional modes are required.
5019 A sample definition of @code{EXTRA_CC_MODES} is:
5021 #define EXTRA_CC_MODES \
5022 CC(CC_NOOVmode, "CC_NOOV") \
5023 CC(CCFPmode, "CCFP") \
5024 CC(CCFPEmode, "CCFPE")
5027 @findex SELECT_CC_MODE
5028 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5029 Returns a mode from class @code{MODE_CC} to be used when comparison
5030 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5031 example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
5032 @pxref{Jump Patterns} for a description of the reason for this
5036 #define SELECT_CC_MODE(OP,X,Y) \
5037 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5038 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5039 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5040 || GET_CODE (X) == NEG) \
5041 ? CC_NOOVmode : CCmode))
5044 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
5046 @findex CANONICALIZE_COMPARISON
5047 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5048 On some machines not all possible comparisons are defined, but you can
5049 convert an invalid comparison into a valid one. For example, the Alpha
5050 does not have a @code{GT} comparison, but you can use an @code{LT}
5051 comparison instead and swap the order of the operands.
5053 On such machines, define this macro to be a C statement to do any
5054 required conversions. @var{code} is the initial comparison code
5055 and @var{op0} and @var{op1} are the left and right operands of the
5056 comparison, respectively. You should modify @var{code}, @var{op0}, and
5057 @var{op1} as required.
5059 GCC will not assume that the comparison resulting from this macro is
5060 valid but will see if the resulting insn matches a pattern in the
5063 You need not define this macro if it would never change the comparison
5066 @findex REVERSIBLE_CC_MODE
5067 @item REVERSIBLE_CC_MODE (@var{mode})
5068 A C expression whose value is one if it is always safe to reverse a
5069 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5070 can ever return @var{mode} for a floating-point inequality comparison,
5071 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5073 You need not define this macro if it would always returns zero or if the
5074 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5075 For example, here is the definition used on the Sparc, where floating-point
5076 inequality comparisons are always given @code{CCFPEmode}:
5079 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5082 @findex REVERSE_CONDITION (@var{code}, @var{mode})
5083 A C expression whose value is reversed condition code of the @var{code} for
5084 comparison done in CC_MODE @var{mode}. The macro is used only in case
5085 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5086 machine has some non-standard way how to reverse certain conditionals. For
5087 instance in case all floating point conditions are non-trapping, compiler may
5088 freely convert unordered compares to ordered one. Then definition may look
5092 #define REVERSE_CONDITION(CODE, MODE) \
5093 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5094 : reverse_condition_maybe_unordered (CODE))
5097 @findex REVERSE_CONDEXEC_PREDICATES_P
5098 @item REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
5099 A C expression that returns true if the conditional execution predicate
5100 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
5101 return 0 if the target has conditional execution predicates that cannot be
5102 reversed safely. If no expansion is specified, this macro is defined as
5106 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5107 ((x) == reverse_condition (y))
5113 @section Describing Relative Costs of Operations
5114 @cindex costs of instructions
5115 @cindex relative costs
5116 @cindex speed of instructions
5118 These macros let you describe the relative speed of various operations
5119 on the target machine.
5123 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
5124 A part of a C @code{switch} statement that describes the relative costs
5125 of constant RTL expressions. It must contain @code{case} labels for
5126 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
5127 @code{label_ref} and @code{const_double}. Each case must ultimately
5128 reach a @code{return} statement to return the relative cost of the use
5129 of that kind of constant value in an expression. The cost may depend on
5130 the precise value of the constant, which is available for examination in
5131 @var{x}, and the rtx code of the expression in which it is contained,
5132 found in @var{outer_code}.
5134 @var{code} is the expression code---redundant, since it can be
5135 obtained with @code{GET_CODE (@var{x})}.
5138 @findex COSTS_N_INSNS
5139 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
5140 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
5141 This can be used, for example, to indicate how costly a multiply
5142 instruction is. In writing this macro, you can use the construct
5143 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5144 instructions. @var{outer_code} is the code of the expression in which
5145 @var{x} is contained.
5147 This macro is optional; do not define it if the default cost assumptions
5148 are adequate for the target machine.
5150 @findex DEFAULT_RTX_COSTS
5151 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
5152 This macro, if defined, is called for any case not handled by the
5153 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
5154 to put case labels into the macro, but the code, or any functions it
5155 calls, must assume that the RTL in @var{x} could be of any type that has
5156 not already been handled. The arguments are the same as for
5157 @code{RTX_COSTS}, and the macro should execute a return statement giving
5158 the cost of any RTL expressions that it can handle. The default cost
5159 calculation is used for any RTL for which this macro does not return a
5162 This macro is optional; do not define it if the default cost assumptions
5163 are adequate for the target machine.
5165 @findex ADDRESS_COST
5166 @item ADDRESS_COST (@var{address})
5167 An expression giving the cost of an addressing mode that contains
5168 @var{address}. If not defined, the cost is computed from
5169 the @var{address} expression and the @code{CONST_COSTS} values.
5171 For most CISC machines, the default cost is a good approximation of the
5172 true cost of the addressing mode. However, on RISC machines, all
5173 instructions normally have the same length and execution time. Hence
5174 all addresses will have equal costs.
5176 In cases where more than one form of an address is known, the form with
5177 the lowest cost will be used. If multiple forms have the same, lowest,
5178 cost, the one that is the most complex will be used.
5180 For example, suppose an address that is equal to the sum of a register
5181 and a constant is used twice in the same basic block. When this macro
5182 is not defined, the address will be computed in a register and memory
5183 references will be indirect through that register. On machines where
5184 the cost of the addressing mode containing the sum is no higher than
5185 that of a simple indirect reference, this will produce an additional
5186 instruction and possibly require an additional register. Proper
5187 specification of this macro eliminates this overhead for such machines.
5189 Similar use of this macro is made in strength reduction of loops.
5191 @var{address} need not be valid as an address. In such a case, the cost
5192 is not relevant and can be any value; invalid addresses need not be
5193 assigned a different cost.
5195 On machines where an address involving more than one register is as
5196 cheap as an address computation involving only one register, defining
5197 @code{ADDRESS_COST} to reflect this can cause two registers to be live
5198 over a region of code where only one would have been if
5199 @code{ADDRESS_COST} were not defined in that manner. This effect should
5200 be considered in the definition of this macro. Equivalent costs should
5201 probably only be given to addresses with different numbers of registers
5202 on machines with lots of registers.
5204 This macro will normally either not be defined or be defined as a
5207 @findex REGISTER_MOVE_COST
5208 @item REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5209 A C expression for the cost of moving data of mode @var{mode} from a
5210 register in class @var{from} to one in class @var{to}. The classes are
5211 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5212 value of 2 is the default; other values are interpreted relative to
5215 It is not required that the cost always equal 2 when @var{from} is the
5216 same as @var{to}; on some machines it is expensive to move between
5217 registers if they are not general registers.
5219 If reload sees an insn consisting of a single @code{set} between two
5220 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5221 classes returns a value of 2, reload does not check to ensure that the
5222 constraints of the insn are met. Setting a cost of other than 2 will
5223 allow reload to verify that the constraints are met. You should do this
5224 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5226 @findex MEMORY_MOVE_COST
5227 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5228 A C expression for the cost of moving data of mode @var{mode} between a
5229 register of class @var{class} and memory; @var{in} is zero if the value
5230 is to be written to memory, nonzero if it is to be read in. This cost
5231 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5232 registers and memory is more expensive than between two registers, you
5233 should define this macro to express the relative cost.
5235 If you do not define this macro, GCC uses a default cost of 4 plus
5236 the cost of copying via a secondary reload register, if one is
5237 needed. If your machine requires a secondary reload register to copy
5238 between memory and a register of @var{class} but the reload mechanism is
5239 more complex than copying via an intermediate, define this macro to
5240 reflect the actual cost of the move.
5242 GCC defines the function @code{memory_move_secondary_cost} if
5243 secondary reloads are needed. It computes the costs due to copying via
5244 a secondary register. If your machine copies from memory using a
5245 secondary register in the conventional way but the default base value of
5246 4 is not correct for your machine, define this macro to add some other
5247 value to the result of that function. The arguments to that function
5248 are the same as to this macro.
5252 A C expression for the cost of a branch instruction. A value of 1 is
5253 the default; other values are interpreted relative to that.
5256 Here are additional macros which do not specify precise relative costs,
5257 but only that certain actions are more expensive than GCC would
5261 @findex SLOW_BYTE_ACCESS
5262 @item SLOW_BYTE_ACCESS
5263 Define this macro as a C expression which is nonzero if accessing less
5264 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5265 faster than accessing a word of memory, i.e., if such access
5266 require more than one instruction or if there is no difference in cost
5267 between byte and (aligned) word loads.
5269 When this macro is not defined, the compiler will access a field by
5270 finding the smallest containing object; when it is defined, a fullword
5271 load will be used if alignment permits. Unless bytes accesses are
5272 faster than word accesses, using word accesses is preferable since it
5273 may eliminate subsequent memory access if subsequent accesses occur to
5274 other fields in the same word of the structure, but to different bytes.
5276 @findex SLOW_UNALIGNED_ACCESS
5277 @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5278 Define this macro to be the value 1 if memory accesses described by the
5279 @var{mode} and @var{alignment} parameters have a cost many times greater
5280 than aligned accesses, for example if they are emulated in a trap
5283 When this macro is nonzero, the compiler will act as if
5284 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5285 moves. This can cause significantly more instructions to be produced.
5286 Therefore, do not set this macro nonzero if unaligned accesses only add a
5287 cycle or two to the time for a memory access.
5289 If the value of this macro is always zero, it need not be defined. If
5290 this macro is defined, it should produce a nonzero value when
5291 @code{STRICT_ALIGNMENT} is nonzero.
5293 @findex DONT_REDUCE_ADDR
5294 @item DONT_REDUCE_ADDR
5295 Define this macro to inhibit strength reduction of memory addresses.
5296 (On some machines, such strength reduction seems to do harm rather
5301 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5302 which a sequence of insns should be generated instead of a
5303 string move insn or a library call. Increasing the value will always
5304 make code faster, but eventually incurs high cost in increased code size.
5306 Note that on machines where the corresponding move insn is a
5307 @code{define_expand} that emits a sequence of insns, this macro counts
5308 the number of such sequences.
5310 If you don't define this, a reasonable default is used.
5312 @findex MOVE_BY_PIECES_P
5313 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5314 A C expression used to determine whether @code{move_by_pieces} will be used to
5315 copy a chunk of memory, or whether some other block move mechanism
5316 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5317 than @code{MOVE_RATIO}.
5319 @findex MOVE_MAX_PIECES
5320 @item MOVE_MAX_PIECES
5321 A C expression used by @code{move_by_pieces} to determine the largest unit
5322 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5324 @findex USE_LOAD_POST_INCREMENT
5325 @item USE_LOAD_POST_INCREMENT (@var{mode})
5326 A C expression used to determine whether a load postincrement is a good
5327 thing to use for a given mode. Defaults to the value of
5328 @code{HAVE_POST_INCREMENT}.
5330 @findex USE_LOAD_POST_DECREMENT
5331 @item USE_LOAD_POST_DECREMENT (@var{mode})
5332 A C expression used to determine whether a load postdecrement is a good
5333 thing to use for a given mode. Defaults to the value of
5334 @code{HAVE_POST_DECREMENT}.
5336 @findex USE_LOAD_PRE_INCREMENT
5337 @item USE_LOAD_PRE_INCREMENT (@var{mode})
5338 A C expression used to determine whether a load preincrement is a good
5339 thing to use for a given mode. Defaults to the value of
5340 @code{HAVE_PRE_INCREMENT}.
5342 @findex USE_LOAD_PRE_DECREMENT
5343 @item USE_LOAD_PRE_DECREMENT (@var{mode})
5344 A C expression used to determine whether a load predecrement is a good
5345 thing to use for a given mode. Defaults to the value of
5346 @code{HAVE_PRE_DECREMENT}.
5348 @findex USE_STORE_POST_INCREMENT
5349 @item USE_STORE_POST_INCREMENT (@var{mode})
5350 A C expression used to determine whether a store postincrement is a good
5351 thing to use for a given mode. Defaults to the value of
5352 @code{HAVE_POST_INCREMENT}.
5354 @findex USE_STORE_POST_DECREMENT
5355 @item USE_STORE_POST_DECREMENT (@var{mode})
5356 A C expression used to determine whether a store postdecrement is a good
5357 thing to use for a given mode. Defaults to the value of
5358 @code{HAVE_POST_DECREMENT}.
5360 @findex USE_STORE_PRE_INCREMENT
5361 @item USE_STORE_PRE_INCREMENT (@var{mode})
5362 This macro is used to determine whether a store preincrement is a good
5363 thing to use for a given mode. Defaults to the value of
5364 @code{HAVE_PRE_INCREMENT}.
5366 @findex USE_STORE_PRE_DECREMENT
5367 @item USE_STORE_PRE_DECREMENT (@var{mode})
5368 This macro is used to determine whether a store predecrement is a good
5369 thing to use for a given mode. Defaults to the value of
5370 @code{HAVE_PRE_DECREMENT}.
5372 @findex NO_FUNCTION_CSE
5373 @item NO_FUNCTION_CSE
5374 Define this macro if it is as good or better to call a constant
5375 function address than to call an address kept in a register.
5377 @findex NO_RECURSIVE_FUNCTION_CSE
5378 @item NO_RECURSIVE_FUNCTION_CSE
5379 Define this macro if it is as good or better for a function to call
5380 itself with an explicit address than to call an address kept in a
5385 @section Adjusting the Instruction Scheduler
5387 The instruction scheduler may need a fair amount of machine-specific
5388 adjustment in order to produce good code. GCC provides several target
5389 hooks for this purpose. It is usually enough to define just a few of
5390 them: try the first ones in this list first.
5392 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5393 This hook returns the maximum number of instructions that can ever
5394 issue at the same time on the target machine. The default is one.
5395 Although the insn scheduler can define itself the possibility of issue
5396 an insn on the same cycle, the value can serve as an additional
5397 constraint to issue insns on the same simulated processor cycle (see
5398 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5399 This value must be constant over the entire compilation. If you need
5400 it to vary depending on what the instructions are, you must use
5401 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5403 You could use the value of macro @samp{MAX_DFA_ISSUE_RATE} to return
5404 the value of the hook @samp{TARGET_SCHED_ISSUE_RATE} for the automaton
5405 based pipeline interface.
5408 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5409 This hook is executed by the scheduler after it has scheduled an insn
5410 from the ready list. It should return the number of insns which can
5411 still be issued in the current cycle. Normally this is
5412 @samp{@w{@var{more} - 1}}. You should define this hook if some insns
5413 take more machine resources than others, so that fewer insns can follow
5414 them in the same cycle. @var{file} is either a null pointer, or a stdio
5415 stream to write any debug output to. @var{verbose} is the verbose level
5416 provided by @option{-fsched-verbose-@var{n}}. @var{insn} is the
5417 instruction that was scheduled.
5420 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5421 This function corrects the value of @var{cost} based on the
5422 relationship between @var{insn} and @var{dep_insn} through the
5423 dependence @var{link}. It should return the new value. The default
5424 is to make no adjustment to @var{cost}. This can be used for example
5425 to specify to the scheduler using the traditional pipeline description
5426 that an output- or anti-dependence does not incur the same cost as a
5427 data-dependence. If the scheduler using the automaton based pipeline
5428 description, the cost of anti-dependence is zero and the cost of
5429 output-dependence is maximum of one and the difference of latency
5430 times of the first and the second insns. If these values are not
5431 acceptable, you could use the hook to modify them too. See also
5432 @pxref{Automaton pipeline description}.
5435 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5436 This hook adjusts the integer scheduling priority @var{priority} of
5437 @var{insn}. It should return the new priority. Reduce the priority to
5438 execute @var{insn} earlier, increase the priority to execute @var{insn}
5439 later. Do not define this hook if you do not need to adjust the
5440 scheduling priorities of insns.
5443 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5444 This hook is executed by the scheduler after it has scheduled the ready
5445 list, to allow the machine description to reorder it (for example to
5446 combine two small instructions together on @samp{VLIW} machines).
5447 @var{file} is either a null pointer, or a stdio stream to write any
5448 debug output to. @var{verbose} is the verbose level provided by
5449 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5450 list of instructions that are ready to be scheduled. @var{n_readyp} is
5451 a pointer to the number of elements in the ready list. The scheduler
5452 reads the ready list in reverse order, starting with
5453 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5454 is the timer tick of the scheduler. You may modify the ready list and
5455 the number of ready insns. The return value is the number of insns that
5456 can issue this cycle; normally this is just @code{issue_rate}. See also
5457 @samp{TARGET_SCHED_REORDER2}.
5460 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5461 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5462 function is called whenever the scheduler starts a new cycle. This one
5463 is called once per iteration over a cycle, immediately after
5464 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5465 return the number of insns to be scheduled in the same cycle. Defining
5466 this hook can be useful if there are frequent situations where
5467 scheduling one insn causes other insns to become ready in the same
5468 cycle. These other insns can then be taken into account properly.
5471 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5472 This hook is executed by the scheduler at the beginning of each block of
5473 instructions that are to be scheduled. @var{file} is either a null
5474 pointer, or a stdio stream to write any debug output to. @var{verbose}
5475 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5476 @var{max_ready} is the maximum number of insns in the current scheduling
5477 region that can be live at the same time. This can be used to allocate
5478 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5481 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5482 This hook is executed by the scheduler at the end of each block of
5483 instructions that are to be scheduled. It can be used to perform
5484 cleanup of any actions done by the other scheduling hooks. @var{file}
5485 is either a null pointer, or a stdio stream to write any debug output
5486 to. @var{verbose} is the verbose level provided by
5487 @option{-fsched-verbose-@var{n}}.
5490 @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void)
5491 This hook is called many times during insn scheduling. If the hook
5492 returns nonzero, the automaton based pipeline description is used for
5493 insn scheduling. Otherwise the traditional pipeline description is
5494 used. The default is usage of the traditional pipeline description.
5496 You should also remember that to simplify the insn scheduler sources
5497 an empty traditional pipeline description interface is generated even
5498 if there is no a traditional pipeline description in the @file{.md}
5499 file. The same is true for the automaton based pipeline description.
5500 That means that you should be accurate in defining the hook.
5503 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5504 The hook returns an RTL insn. The automaton state used in the
5505 pipeline hazard recognizer is changed as if the insn were scheduled
5506 when the new simulated processor cycle starts. Usage of the hook may
5507 simplify the automaton pipeline description for some @acronym{VLIW}
5508 processors. If the hook is defined, it is used only for the automaton
5509 based pipeline description. The default is not to change the state
5510 when the new simulated processor cycle starts.
5513 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5514 The hook can be used to initialize data used by the previous hook.
5517 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5518 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5519 to changed the state as if the insn were scheduled when the new
5520 simulated processor cycle finishes.
5523 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5524 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5525 used to initialize data used by the previous hook.
5528 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5529 This hook controls better choosing an insn from the ready insn queue
5530 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5531 chooses the first insn from the queue. If the hook returns a positive
5532 value, an additional scheduler code tries all permutations of
5533 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5534 subsequent ready insns to choose an insn whose issue will result in
5535 maximal number of issued insns on the same cycle. For the
5536 @acronym{VLIW} processor, the code could actually solve the problem of
5537 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5538 rules of @acronym{VLIW} packing are described in the automaton.
5540 This code also could be used for superscalar @acronym{RISC}
5541 processors. Let us consider a superscalar @acronym{RISC} processor
5542 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5543 @var{B}, some insns can be executed only in pipelines @var{B} or
5544 @var{C}, and one insn can be executed in pipeline @var{B}. The
5545 processor may issue the 1st insn into @var{A} and the 2nd one into
5546 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5547 until the next cycle. If the scheduler issues the 3rd insn the first,
5548 the processor could issue all 3 insns per cycle.
5550 Actually this code demonstrates advantages of the automaton based
5551 pipeline hazard recognizer. We try quickly and easy many insn
5552 schedules to choose the best one.
5554 The default is no multipass scheduling.
5557 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void)
5558 The @acronym{DFA}-based scheduler could take the insertion of nop
5559 operations for better insn scheduling into account. It can be done
5560 only if the multi-pass insn scheduling works (see hook
5561 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}).
5563 Let us consider a @acronym{VLIW} processor insn with 3 slots. Each
5564 insn can be placed only in one of the three slots. We have 3 ready
5565 insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be
5566 placed only in the 1st slot, @var{B} can be placed only in the 3rd
5567 slot. We described the automaton which does not permit empty slot
5568 gaps between insns (usually such description is simpler). Without
5569 this code the scheduler would place each insn in 3 separate
5570 @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd
5571 slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is
5572 the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook
5573 @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or
5574 create the nop insns.
5576 You should remember that the scheduler does not insert the nop insns.
5577 It is not wise because of the following optimizations. The scheduler
5578 only considers such possibility to improve the result schedule. The
5579 nop insns should be inserted lately, e.g. on the final phase.
5582 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index})
5583 This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert
5584 nop operations for better insn scheduling when @acronym{DFA}-based
5585 scheduler makes multipass insn scheduling (see also description of
5586 hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook
5587 returns a nop insn with given @var{index}. The indexes start with
5588 zero. The hook should return @code{NULL} if there are no more nop
5589 insns with indexes greater than given index.
5592 Macros in the following table are generated by the program
5593 @file{genattr} and can be useful for writing the hooks.
5596 @findex TRADITIONAL_PIPELINE_INTERFACE
5597 @item TRADITIONAL_PIPELINE_INTERFACE
5598 The macro definition is generated if there is a traditional pipeline
5599 description in @file{.md} file. You should also remember that to
5600 simplify the insn scheduler sources an empty traditional pipeline
5601 description interface is generated even if there is no a traditional
5602 pipeline description in the @file{.md} file. The macro can be used to
5603 distinguish the two types of the traditional interface.
5605 @findex DFA_PIPELINE_INTERFACE
5606 @item DFA_PIPELINE_INTERFACE
5607 The macro definition is generated if there is an automaton pipeline
5608 description in @file{.md} file. You should also remember that to
5609 simplify the insn scheduler sources an empty automaton pipeline
5610 description interface is generated even if there is no an automaton
5611 pipeline description in the @file{.md} file. The macro can be used to
5612 distinguish the two types of the automaton interface.
5614 @findex MAX_DFA_ISSUE_RATE
5615 @item MAX_DFA_ISSUE_RATE
5616 The macro definition is generated in the automaton based pipeline
5617 description interface. Its value is calculated from the automaton
5618 based pipeline description and is equal to maximal number of all insns
5619 described in constructions @samp{define_insn_reservation} which can be
5620 issued on the same processor cycle.
5625 @section Dividing the Output into Sections (Texts, Data, @dots{})
5626 @c the above section title is WAY too long. maybe cut the part between
5627 @c the (...)? --mew 10feb93
5629 An object file is divided into sections containing different types of
5630 data. In the most common case, there are three sections: the @dfn{text
5631 section}, which holds instructions and read-only data; the @dfn{data
5632 section}, which holds initialized writable data; and the @dfn{bss
5633 section}, which holds uninitialized data. Some systems have other kinds
5636 The compiler must tell the assembler when to switch sections. These
5637 macros control what commands to output to tell the assembler this. You
5638 can also define additional sections.
5641 @findex TEXT_SECTION_ASM_OP
5642 @item TEXT_SECTION_ASM_OP
5643 A C expression whose value is a string, including spacing, containing the
5644 assembler operation that should precede instructions and read-only data.
5645 Normally @code{"\t.text"} is right.
5647 @findex TEXT_SECTION
5649 A C statement that switches to the default section containing instructions.
5650 Normally this is not needed, as simply defining @code{TEXT_SECTION_ASM_OP}
5651 is enough. The MIPS port uses this to sort all functions after all data
5654 @findex HOT_TEXT_SECTION_NAME
5655 @item HOT_TEXT_SECTION_NAME
5656 If defined, a C string constant for the name of the section containing most
5657 frequently executed functions of the program. If not defined, GCC will provide
5658 a default definition if the target supports named sections.
5660 @findex UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5661 @item UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5662 If defined, a C string constant for the name of the section containing unlikely
5663 executed functions in the program.
5665 @findex DATA_SECTION_ASM_OP
5666 @item DATA_SECTION_ASM_OP
5667 A C expression whose value is a string, including spacing, containing the
5668 assembler operation to identify the following data as writable initialized
5669 data. Normally @code{"\t.data"} is right.
5671 @findex READONLY_DATA_SECTION_ASM_OP
5672 @item READONLY_DATA_SECTION_ASM_OP
5673 A C expression whose value is a string, including spacing, containing the
5674 assembler operation to identify the following data as read-only initialized
5677 @findex READONLY_DATA_SECTION
5678 @item READONLY_DATA_SECTION
5679 A macro naming a function to call to switch to the proper section for
5680 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5681 if defined, else fall back to @code{text_section}.
5683 The most common definition will be @code{data_section}, if the target
5684 does not have a special read-only data section, and does not put data
5685 in the text section.
5687 @findex SHARED_SECTION_ASM_OP
5688 @item SHARED_SECTION_ASM_OP
5689 If defined, a C expression whose value is a string, including spacing,
5690 containing the assembler operation to identify the following data as
5691 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5693 @findex BSS_SECTION_ASM_OP
5694 @item BSS_SECTION_ASM_OP
5695 If defined, a C expression whose value is a string, including spacing,
5696 containing the assembler operation to identify the following data as
5697 uninitialized global data. If not defined, and neither
5698 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5699 uninitialized global data will be output in the data section if
5700 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5703 @findex SHARED_BSS_SECTION_ASM_OP
5704 @item SHARED_BSS_SECTION_ASM_OP
5705 If defined, a C expression whose value is a string, including spacing,
5706 containing the assembler operation to identify the following data as
5707 uninitialized global shared data. If not defined, and
5708 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5710 @findex INIT_SECTION_ASM_OP
5711 @item INIT_SECTION_ASM_OP
5712 If defined, a C expression whose value is a string, including spacing,
5713 containing the assembler operation to identify the following data as
5714 initialization code. If not defined, GCC will assume such a section does
5717 @findex FINI_SECTION_ASM_OP
5718 @item FINI_SECTION_ASM_OP
5719 If defined, a C expression whose value is a string, including spacing,
5720 containing the assembler operation to identify the following data as
5721 finalization code. If not defined, GCC will assume such a section does
5724 @findex CRT_CALL_STATIC_FUNCTION
5725 @item CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5726 If defined, an ASM statement that switches to a different section
5727 via @var{section_op}, calls @var{function}, and switches back to
5728 the text section. This is used in @file{crtstuff.c} if
5729 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5730 to initialization and finalization functions from the init and fini
5731 sections. By default, this macro uses a simple function call. Some
5732 ports need hand-crafted assembly code to avoid dependencies on
5733 registers initialized in the function prologue or to ensure that
5734 constant pools don't end up too far way in the text section.
5736 @findex FORCE_CODE_SECTION_ALIGN
5737 @item FORCE_CODE_SECTION_ALIGN
5738 If defined, an ASM statement that aligns a code section to some
5739 arbitrary boundary. This is used to force all fragments of the
5740 @code{.init} and @code{.fini} sections to have to same alignment
5741 and thus prevent the linker from having to add any padding.
5743 @findex EXTRA_SECTIONS
5746 @item EXTRA_SECTIONS
5747 A list of names for sections other than the standard two, which are
5748 @code{in_text} and @code{in_data}. You need not define this macro
5749 on a system with no other sections (that GCC needs to use).
5751 @findex EXTRA_SECTION_FUNCTIONS
5752 @findex text_section
5753 @findex data_section
5754 @item EXTRA_SECTION_FUNCTIONS
5755 One or more functions to be defined in @file{varasm.c}. These
5756 functions should do jobs analogous to those of @code{text_section} and
5757 @code{data_section}, for your additional sections. Do not define this
5758 macro if you do not define @code{EXTRA_SECTIONS}.
5760 @findex SELECT_RTX_SECTION
5761 @item SELECT_RTX_SECTION (@var{mode}, @var{rtx}, @var{align})
5762 A C statement or statements to switch to the appropriate section for
5763 output of @var{rtx} in mode @var{mode}. You can assume that @var{rtx}
5764 is some kind of constant in RTL@. The argument @var{mode} is redundant
5765 except in the case of a @code{const_int} rtx. Select the section by
5766 calling @code{text_section} or one of the alternatives for other
5767 sections. @var{align} is the constant alignment in bits.
5769 Do not define this macro if you put all constants in the read-only
5772 @findex JUMP_TABLES_IN_TEXT_SECTION
5773 @item JUMP_TABLES_IN_TEXT_SECTION
5774 Define this macro to be an expression with a nonzero value if jump
5775 tables (for @code{tablejump} insns) should be output in the text
5776 section, along with the assembler instructions. Otherwise, the
5777 readonly data section is used.
5779 This macro is irrelevant if there is no separate readonly data section.
5781 @findex ENCODE_SECTION_INFO
5782 @item ENCODE_SECTION_INFO (@var{decl}, @var{new_decl_p})
5783 Define this macro if references to a symbol or a constant must be
5784 treated differently depending on something about the variable or
5785 function named by the symbol (such as what section it is in).
5787 The macro definition, if any, is executed under two circumstances. One
5788 is immediately after the rtl for @var{decl} that represents a variable
5789 or a function has been created and stored in @code{DECL_RTL(@var{decl})}.
5790 The value of the rtl will be a @code{mem} whose address is a @code{symbol_ref}.
5791 The other is immediately after the rtl for @var{decl} that represents a
5792 constant has been created and stored in @code{TREE_CST_RTL (@var{decl})}.
5793 The macro is called once for each distinct constant in a source file.
5795 The @var{new_decl_p} argument will be true if this is the first time that
5796 @code{ENCODE_SECTION_INFO} has been invoked on this decl. It will
5797 be false for subsequent invocations, which will happen for duplicate
5798 declarations. Whether or not anything must be done for the duplicate
5799 declaration depends on whether @code{ENCODE_SECTION_INFO} examines
5800 @code{DECL_ATTRIBUTES}.
5802 @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO}
5803 The usual thing for this macro to do is to record a flag in the
5804 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
5805 modified name string in the @code{symbol_ref} (if one bit is not
5806 enough information).
5808 @findex STRIP_NAME_ENCODING
5809 @item STRIP_NAME_ENCODING (@var{var}, @var{sym_name})
5810 Decode @var{sym_name} and store the real name part in @var{var}, sans
5811 the characters that encode section info. Define this macro if
5812 @code{ENCODE_SECTION_INFO} alters the symbol's name string.
5815 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5816 Switches to the appropriate section for output of @var{exp}. You can
5817 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5818 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5819 requires link-time relocations. Bit 0 is set when variable contains
5820 local relocations only, while bit 1 is set for global relocations.
5821 Select the section by calling @code{data_section} or one of the
5822 alternatives for other sections. @var{align} is the constant alignment
5825 The default version of this function takes care of putting read-only
5826 variables in @code{readonly_data_section}.
5829 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5830 Build up a unique section name, expressed as a @code{STRING_CST} node,
5831 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5832 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5833 the initial value of @var{exp} requires link-time relocations.
5835 The default version of this function appends the symbol name to the
5836 ELF section name that would normally be used for the symbol. For
5837 example, the function @code{foo} would be placed in @code{.text.foo}.
5838 Whatever the actual target object format, this is often good enough.
5842 @section Position Independent Code
5843 @cindex position independent code
5846 This section describes macros that help implement generation of position
5847 independent code. Simply defining these macros is not enough to
5848 generate valid PIC; you must also add support to the macros
5849 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5850 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5851 @samp{movsi} to do something appropriate when the source operand
5852 contains a symbolic address. You may also need to alter the handling of
5853 switch statements so that they use relative addresses.
5854 @c i rearranged the order of the macros above to try to force one of
5855 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5858 @findex PIC_OFFSET_TABLE_REGNUM
5859 @item PIC_OFFSET_TABLE_REGNUM
5860 The register number of the register used to address a table of static
5861 data addresses in memory. In some cases this register is defined by a
5862 processor's ``application binary interface'' (ABI)@. When this macro
5863 is defined, RTL is generated for this register once, as with the stack
5864 pointer and frame pointer registers. If this macro is not defined, it
5865 is up to the machine-dependent files to allocate such a register (if
5866 necessary). Note that this register must be fixed when in use (e.g.@:
5867 when @code{flag_pic} is true).
5869 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5870 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5871 Define this macro if the register defined by
5872 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
5873 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5875 @findex FINALIZE_PIC
5877 By generating position-independent code, when two different programs (A
5878 and B) share a common library (libC.a), the text of the library can be
5879 shared whether or not the library is linked at the same address for both
5880 programs. In some of these environments, position-independent code
5881 requires not only the use of different addressing modes, but also
5882 special code to enable the use of these addressing modes.
5884 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
5885 codes once the function is being compiled into assembly code, but not
5886 before. (It is not done before, because in the case of compiling an
5887 inline function, it would lead to multiple PIC prologues being
5888 included in functions which used inline functions and were compiled to
5891 @findex LEGITIMATE_PIC_OPERAND_P
5892 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
5893 A C expression that is nonzero if @var{x} is a legitimate immediate
5894 operand on the target machine when generating position independent code.
5895 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5896 check this. You can also assume @var{flag_pic} is true, so you need not
5897 check it either. You need not define this macro if all constants
5898 (including @code{SYMBOL_REF}) can be immediate operands when generating
5899 position independent code.
5902 @node Assembler Format
5903 @section Defining the Output Assembler Language
5905 This section describes macros whose principal purpose is to describe how
5906 to write instructions in assembler language---rather than what the
5910 * File Framework:: Structural information for the assembler file.
5911 * Data Output:: Output of constants (numbers, strings, addresses).
5912 * Uninitialized Data:: Output of uninitialized variables.
5913 * Label Output:: Output and generation of labels.
5914 * Initialization:: General principles of initialization
5915 and termination routines.
5916 * Macros for Initialization::
5917 Specific macros that control the handling of
5918 initialization and termination routines.
5919 * Instruction Output:: Output of actual instructions.
5920 * Dispatch Tables:: Output of jump tables.
5921 * Exception Region Output:: Output of exception region code.
5922 * Alignment Output:: Pseudo ops for alignment and skipping data.
5925 @node File Framework
5926 @subsection The Overall Framework of an Assembler File
5927 @cindex assembler format
5928 @cindex output of assembler code
5930 @c prevent bad page break with this line
5931 This describes the overall framework of an assembler file.
5934 @findex ASM_FILE_START
5935 @item ASM_FILE_START (@var{stream})
5936 A C expression which outputs to the stdio stream @var{stream}
5937 some appropriate text to go at the start of an assembler file.
5939 Normally this macro is defined to output a line containing
5940 @samp{#NO_APP}, which is a comment that has no effect on most
5941 assemblers but tells the GNU assembler that it can save time by not
5942 checking for certain assembler constructs.
5944 On systems that use SDB, it is necessary to output certain commands;
5945 see @file{attasm.h}.
5947 @findex ASM_FILE_END
5948 @item ASM_FILE_END (@var{stream})
5949 A C expression which outputs to the stdio stream @var{stream}
5950 some appropriate text to go at the end of an assembler file.
5952 If this macro is not defined, the default is to output nothing
5953 special at the end of the file. Most systems don't require any
5956 On systems that use SDB, it is necessary to output certain commands;
5957 see @file{attasm.h}.
5959 @findex ASM_COMMENT_START
5960 @item ASM_COMMENT_START
5961 A C string constant describing how to begin a comment in the target
5962 assembler language. The compiler assumes that the comment will end at
5963 the end of the line.
5967 A C string constant for text to be output before each @code{asm}
5968 statement or group of consecutive ones. Normally this is
5969 @code{"#APP"}, which is a comment that has no effect on most
5970 assemblers but tells the GNU assembler that it must check the lines
5971 that follow for all valid assembler constructs.
5975 A C string constant for text to be output after each @code{asm}
5976 statement or group of consecutive ones. Normally this is
5977 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5978 time-saving assumptions that are valid for ordinary compiler output.
5980 @findex ASM_OUTPUT_SOURCE_FILENAME
5981 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5982 A C statement to output COFF information or DWARF debugging information
5983 which indicates that filename @var{name} is the current source file to
5984 the stdio stream @var{stream}.
5986 This macro need not be defined if the standard form of output
5987 for the file format in use is appropriate.
5989 @findex OUTPUT_QUOTED_STRING
5990 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5991 A C statement to output the string @var{string} to the stdio stream
5992 @var{stream}. If you do not call the function @code{output_quoted_string}
5993 in your config files, GCC will only call it to output filenames to
5994 the assembler source. So you can use it to canonicalize the format
5995 of the filename using this macro.
5997 @findex ASM_OUTPUT_SOURCE_LINE
5998 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
5999 A C statement to output DBX or SDB debugging information before code
6000 for line number @var{line} of the current source file to the
6001 stdio stream @var{stream}.
6003 This macro need not be defined if the standard form of debugging
6004 information for the debugger in use is appropriate.
6006 @findex ASM_OUTPUT_IDENT
6007 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6008 A C statement to output something to the assembler file to handle a
6009 @samp{#ident} directive containing the text @var{string}. If this
6010 macro is not defined, nothing is output for a @samp{#ident} directive.
6012 @findex OBJC_PROLOGUE
6014 A C statement to output any assembler statements which are required to
6015 precede any Objective-C object definitions or message sending. The
6016 statement is executed only when compiling an Objective-C program.
6019 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6020 Output assembly directives to switch to section @var{name}. The section
6021 should have attributes as specified by @var{flags}, which is a bit mask
6022 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6023 is nonzero, it contains an alignment in bytes to be used for the section,
6024 otherwise some target default should be used. Only targets that must
6025 specify an alignment within the section directive need pay attention to
6026 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6029 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6030 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6033 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6034 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6035 based on a variable or function decl, a section name, and whether or not the
6036 declaration's initializer may contain runtime relocations. @var{decl} may be
6037 null, in which case read-write data should be assumed.
6039 The default version if this function handles choosing code vs data,
6040 read-only vs read-write data, and @code{flag_pic}. You should only
6041 need to override this if your target has special flags that might be
6042 set via @code{__attribute__}.
6047 @subsection Output of Data
6050 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6051 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6052 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6053 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6054 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6055 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6056 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6057 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6058 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6059 These hooks specify assembly directives for creating certain kinds
6060 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6061 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6062 aligned two-byte object, and so on. Any of the hooks may be
6063 @code{NULL}, indicating that no suitable directive is available.
6065 The compiler will print these strings at the start of a new line,
6066 followed immediately by the object's initial value. In most cases,
6067 the string should contain a tab, a pseudo-op, and then another tab.
6070 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6071 The @code{assemble_integer} function uses this hook to output an
6072 integer object. @var{x} is the object's value, @var{size} is its size
6073 in bytes and @var{aligned_p} indicates whether it is aligned. The
6074 function should return @code{true} if it was able to output the
6075 object. If it returns false, @code{assemble_integer} will try to
6076 split the object into smaller parts.
6078 The default implementation of this hook will use the
6079 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6080 when the relevant string is @code{NULL}.
6084 @findex OUTPUT_ADDR_CONST_EXTRA
6085 @item OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6086 A C statement to recognize @var{rtx} patterns that
6087 @code{output_addr_const} can't deal with, and output assembly code to
6088 @var{stream} corresponding to the pattern @var{x}. This may be used to
6089 allow machine-dependent @code{UNSPEC}s to appear within constants.
6091 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6092 @code{goto fail}, so that a standard error message is printed. If it
6093 prints an error message itself, by calling, for example,
6094 @code{output_operand_lossage}, it may just complete normally.
6096 @findex ASM_OUTPUT_ASCII
6097 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6098 A C statement to output to the stdio stream @var{stream} an assembler
6099 instruction to assemble a string constant containing the @var{len}
6100 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6101 @code{char *} and @var{len} a C expression of type @code{int}.
6103 If the assembler has a @code{.ascii} pseudo-op as found in the
6104 Berkeley Unix assembler, do not define the macro
6105 @code{ASM_OUTPUT_ASCII}.
6107 @findex ASM_OUTPUT_FDESC
6108 @item ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6109 A C statement to output word @var{n} of a function descriptor for
6110 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6111 is defined, and is otherwise unused.
6113 @findex CONSTANT_POOL_BEFORE_FUNCTION
6114 @item CONSTANT_POOL_BEFORE_FUNCTION
6115 You may define this macro as a C expression. You should define the
6116 expression to have a nonzero value if GCC should output the constant
6117 pool for a function before the code for the function, or a zero value if
6118 GCC should output the constant pool after the function. If you do
6119 not define this macro, the usual case, GCC will output the constant
6120 pool before the function.
6122 @findex ASM_OUTPUT_POOL_PROLOGUE
6123 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6124 A C statement to output assembler commands to define the start of the
6125 constant pool for a function. @var{funname} is a string giving
6126 the name of the function. Should the return type of the function
6127 be required, it can be obtained via @var{fundecl}. @var{size}
6128 is the size, in bytes, of the constant pool that will be written
6129 immediately after this call.
6131 If no constant-pool prefix is required, the usual case, this macro need
6134 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
6135 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6136 A C statement (with or without semicolon) to output a constant in the
6137 constant pool, if it needs special treatment. (This macro need not do
6138 anything for RTL expressions that can be output normally.)
6140 The argument @var{file} is the standard I/O stream to output the
6141 assembler code on. @var{x} is the RTL expression for the constant to
6142 output, and @var{mode} is the machine mode (in case @var{x} is a
6143 @samp{const_int}). @var{align} is the required alignment for the value
6144 @var{x}; you should output an assembler directive to force this much
6147 The argument @var{labelno} is a number to use in an internal label for
6148 the address of this pool entry. The definition of this macro is
6149 responsible for outputting the label definition at the proper place.
6150 Here is how to do this:
6153 ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
6156 When you output a pool entry specially, you should end with a
6157 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6158 entry from being output a second time in the usual manner.
6160 You need not define this macro if it would do nothing.
6162 @findex CONSTANT_AFTER_FUNCTION_P
6163 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
6164 Define this macro as a C expression which is nonzero if the constant
6165 @var{exp}, of type @code{tree}, should be output after the code for a
6166 function. The compiler will normally output all constants before the
6167 function; you need not define this macro if this is OK@.
6169 @findex ASM_OUTPUT_POOL_EPILOGUE
6170 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6171 A C statement to output assembler commands to at the end of the constant
6172 pool for a function. @var{funname} is a string giving the name of the
6173 function. Should the return type of the function be required, you can
6174 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6175 constant pool that GCC wrote immediately before this call.
6177 If no constant-pool epilogue is required, the usual case, you need not
6180 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
6181 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6182 Define this macro as a C expression which is nonzero if @var{C} is
6183 used as a logical line separator by the assembler.
6185 If you do not define this macro, the default is that only
6186 the character @samp{;} is treated as a logical line separator.
6189 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6190 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6191 These target hooks are C string constants, describing the syntax in the
6192 assembler for grouping arithmetic expressions. If not overridden, they
6193 default to normal parentheses, which is correct for most assemblers.
6196 These macros are provided by @file{real.h} for writing the definitions
6197 of @code{ASM_OUTPUT_DOUBLE} and the like:
6200 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6201 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6202 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6203 @findex REAL_VALUE_TO_TARGET_SINGLE
6204 @findex REAL_VALUE_TO_TARGET_DOUBLE
6205 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
6206 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6207 floating point representation, and store its bit pattern in the variable
6208 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6209 be a simple @code{long int}. For the others, it should be an array of
6210 @code{long int}. The number of elements in this array is determined by
6211 the size of the desired target floating point data type: 32 bits of it
6212 go in each @code{long int} array element. Each array element holds 32
6213 bits of the result, even if @code{long int} is wider than 32 bits on the
6216 The array element values are designed so that you can print them out
6217 using @code{fprintf} in the order they should appear in the target
6220 @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
6221 @findex REAL_VALUE_TO_DECIMAL
6222 This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
6223 decimal number and stores it as a string into @var{string}.
6224 You must pass, as @var{string}, the address of a long enough block
6225 of space to hold the result.
6227 The argument @var{format} is a @code{printf}-specification that serves
6228 as a suggestion for how to format the output string.
6231 @node Uninitialized Data
6232 @subsection Output of Uninitialized Variables
6234 Each of the macros in this section is used to do the whole job of
6235 outputting a single uninitialized variable.
6238 @findex ASM_OUTPUT_COMMON
6239 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6240 A C statement (sans semicolon) to output to the stdio stream
6241 @var{stream} the assembler definition of a common-label named
6242 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6243 is the size rounded up to whatever alignment the caller wants.
6245 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6246 output the name itself; before and after that, output the additional
6247 assembler syntax for defining the name, and a newline.
6249 This macro controls how the assembler definitions of uninitialized
6250 common global variables are output.
6252 @findex ASM_OUTPUT_ALIGNED_COMMON
6253 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6254 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6255 separate, explicit argument. If you define this macro, it is used in
6256 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6257 handling the required alignment of the variable. The alignment is specified
6258 as the number of bits.
6260 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
6261 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6262 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6263 variable to be output, if there is one, or @code{NULL_TREE} if there
6264 is no corresponding variable. If you define this macro, GCC will use it
6265 in place of both @code{ASM_OUTPUT_COMMON} and
6266 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6267 the variable's decl in order to chose what to output.
6269 @findex ASM_OUTPUT_SHARED_COMMON
6270 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6271 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6272 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6275 @findex ASM_OUTPUT_BSS
6276 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6277 A C statement (sans semicolon) to output to the stdio stream
6278 @var{stream} the assembler definition of uninitialized global @var{decl} named
6279 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6280 is the size rounded up to whatever alignment the caller wants.
6282 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6283 defining this macro. If unable, use the expression
6284 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6285 before and after that, output the additional assembler syntax for defining
6286 the name, and a newline.
6288 This macro controls how the assembler definitions of uninitialized global
6289 variables are output. This macro exists to properly support languages like
6290 C++ which do not have @code{common} data. However, this macro currently
6291 is not defined for all targets. If this macro and
6292 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6293 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6294 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6296 @findex ASM_OUTPUT_ALIGNED_BSS
6297 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6298 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6299 separate, explicit argument. If you define this macro, it is used in
6300 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6301 handling the required alignment of the variable. The alignment is specified
6302 as the number of bits.
6304 Try to use function @code{asm_output_aligned_bss} defined in file
6305 @file{varasm.c} when defining this macro.
6307 @findex ASM_OUTPUT_SHARED_BSS
6308 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6309 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6310 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6313 @findex ASM_OUTPUT_LOCAL
6314 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6315 A C statement (sans semicolon) to output to the stdio stream
6316 @var{stream} the assembler definition of a local-common-label named
6317 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6318 is the size rounded up to whatever alignment the caller wants.
6320 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6321 output the name itself; before and after that, output the additional
6322 assembler syntax for defining the name, and a newline.
6324 This macro controls how the assembler definitions of uninitialized
6325 static variables are output.
6327 @findex ASM_OUTPUT_ALIGNED_LOCAL
6328 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6329 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6330 separate, explicit argument. If you define this macro, it is used in
6331 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6332 handling the required alignment of the variable. The alignment is specified
6333 as the number of bits.
6335 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
6336 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6337 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6338 variable to be output, if there is one, or @code{NULL_TREE} if there
6339 is no corresponding variable. If you define this macro, GCC will use it
6340 in place of both @code{ASM_OUTPUT_DECL} and
6341 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6342 the variable's decl in order to chose what to output.
6344 @findex ASM_OUTPUT_SHARED_LOCAL
6345 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6346 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6347 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6352 @subsection Output and Generation of Labels
6354 @c prevent bad page break with this line
6355 This is about outputting labels.
6358 @findex ASM_OUTPUT_LABEL
6359 @findex assemble_name
6360 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6361 A C statement (sans semicolon) to output to the stdio stream
6362 @var{stream} the assembler definition of a label named @var{name}.
6363 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6364 output the name itself; before and after that, output the additional
6365 assembler syntax for defining the name, and a newline.
6367 @findex ASM_DECLARE_FUNCTION_NAME
6368 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6369 A C statement (sans semicolon) to output to the stdio stream
6370 @var{stream} any text necessary for declaring the name @var{name} of a
6371 function which is being defined. This macro is responsible for
6372 outputting the label definition (perhaps using
6373 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6374 @code{FUNCTION_DECL} tree node representing the function.
6376 If this macro is not defined, then the function name is defined in the
6377 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6379 @findex ASM_DECLARE_FUNCTION_SIZE
6380 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6381 A C statement (sans semicolon) to output to the stdio stream
6382 @var{stream} any text necessary for declaring the size of a function
6383 which is being defined. The argument @var{name} is the name of the
6384 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6385 representing the function.
6387 If this macro is not defined, then the function size is not defined.
6389 @findex ASM_DECLARE_OBJECT_NAME
6390 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6391 A C statement (sans semicolon) to output to the stdio stream
6392 @var{stream} any text necessary for declaring the name @var{name} of an
6393 initialized variable which is being defined. This macro must output the
6394 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6395 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6397 If this macro is not defined, then the variable name is defined in the
6398 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6400 @findex ASM_DECLARE_REGISTER_GLOBAL
6401 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6402 A C statement (sans semicolon) to output to the stdio stream
6403 @var{stream} any text necessary for claiming a register @var{regno}
6404 for a global variable @var{decl} with name @var{name}.
6406 If you don't define this macro, that is equivalent to defining it to do
6409 @findex ASM_FINISH_DECLARE_OBJECT
6410 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6411 A C statement (sans semicolon) to finish up declaring a variable name
6412 once the compiler has processed its initializer fully and thus has had a
6413 chance to determine the size of an array when controlled by an
6414 initializer. This is used on systems where it's necessary to declare
6415 something about the size of the object.
6417 If you don't define this macro, that is equivalent to defining it to do
6420 @findex ASM_GLOBALIZE_LABEL
6421 @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
6422 A C statement (sans semicolon) to output to the stdio stream
6423 @var{stream} some commands that will make the label @var{name} global;
6424 that is, available for reference from other files. Use the expression
6425 @code{assemble_name (@var{stream}, @var{name})} to output the name
6426 itself; before and after that, output the additional assembler syntax
6427 for making that name global, and a newline.
6429 @findex ASM_WEAKEN_LABEL
6430 @item ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6431 A C statement (sans semicolon) to output to the stdio stream
6432 @var{stream} some commands that will make the label @var{name} weak;
6433 that is, available for reference from other files but only used if
6434 no other definition is available. Use the expression
6435 @code{assemble_name (@var{stream}, @var{name})} to output the name
6436 itself; before and after that, output the additional assembler syntax
6437 for making that name weak, and a newline.
6439 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6440 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6443 @findex ASM_WEAKEN_DECL
6444 @item ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6445 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6446 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6447 or variable decl. If @var{value} is not @code{NULL}, this C statement
6448 should output to the stdio stream @var{stream} assembler code which
6449 defines (equates) the weak symbol @var{name} to have the value
6450 @var{value}. If @var{value} is @code{NULL}, it should output commands
6451 to make @var{name} weak.
6453 @findex SUPPORTS_WEAK
6455 A C expression which evaluates to true if the target supports weak symbols.
6457 If you don't define this macro, @file{defaults.h} provides a default
6458 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6459 is defined, the default definition is @samp{1}; otherwise, it is
6460 @samp{0}. Define this macro if you want to control weak symbol support
6461 with a compiler flag such as @option{-melf}.
6463 @findex MAKE_DECL_ONE_ONLY (@var{decl})
6464 @item MAKE_DECL_ONE_ONLY
6465 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6466 public symbol such that extra copies in multiple translation units will
6467 be discarded by the linker. Define this macro if your object file
6468 format provides support for this concept, such as the @samp{COMDAT}
6469 section flags in the Microsoft Windows PE/COFF format, and this support
6470 requires changes to @var{decl}, such as putting it in a separate section.
6472 @findex SUPPORTS_ONE_ONLY
6473 @item SUPPORTS_ONE_ONLY
6474 A C expression which evaluates to true if the target supports one-only
6477 If you don't define this macro, @file{varasm.c} provides a default
6478 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6479 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6480 you want to control one-only symbol support with a compiler flag, or if
6481 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6482 be emitted as one-only.
6484 @findex ASM_OUTPUT_EXTERNAL
6485 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6486 A C statement (sans semicolon) to output to the stdio stream
6487 @var{stream} any text necessary for declaring the name of an external
6488 symbol named @var{name} which is referenced in this compilation but
6489 not defined. The value of @var{decl} is the tree node for the
6492 This macro need not be defined if it does not need to output anything.
6493 The GNU assembler and most Unix assemblers don't require anything.
6495 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
6496 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
6497 A C statement (sans semicolon) to output on @var{stream} an assembler
6498 pseudo-op to declare a library function name external. The name of the
6499 library function is given by @var{symref}, which has type @code{rtx} and
6500 is a @code{symbol_ref}.
6502 This macro need not be defined if it does not need to output anything.
6503 The GNU assembler and most Unix assemblers don't require anything.
6505 @findex ASM_OUTPUT_LABELREF
6506 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6507 A C statement (sans semicolon) to output to the stdio stream
6508 @var{stream} a reference in assembler syntax to a label named
6509 @var{name}. This should add @samp{_} to the front of the name, if that
6510 is customary on your operating system, as it is in most Berkeley Unix
6511 systems. This macro is used in @code{assemble_name}.
6513 @findex ASM_OUTPUT_SYMBOL_REF
6514 @item ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6515 A C statement (sans semicolon) to output a reference to
6516 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6517 will be used to output the name of the symbol. This macro may be used
6518 to modify the way a symbol is referenced depending on information
6519 encoded by @code{ENCODE_SECTION_INFO}.
6521 @findex ASM_OUTPUT_LABEL_REF
6522 @item ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6523 A C statement (sans semicolon) to output a reference to @var{buf}, the
6524 result of ASM_GENERATE_INTERNAL_LABEL. If not defined,
6525 @code{assemble_name} will be used to output the name of the symbol.
6526 This macro is not used by @code{output_asm_label}, or the @code{%l}
6527 specifier that calls it; the intention is that this macro should be set
6528 when it is necessary to output a label differently when its address
6531 @findex ASM_OUTPUT_INTERNAL_LABEL
6532 @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
6533 A C statement to output to the stdio stream @var{stream} a label whose
6534 name is made from the string @var{prefix} and the number @var{num}.
6536 It is absolutely essential that these labels be distinct from the labels
6537 used for user-level functions and variables. Otherwise, certain programs
6538 will have name conflicts with internal labels.
6540 It is desirable to exclude internal labels from the symbol table of the
6541 object file. Most assemblers have a naming convention for labels that
6542 should be excluded; on many systems, the letter @samp{L} at the
6543 beginning of a label has this effect. You should find out what
6544 convention your system uses, and follow it.
6546 The usual definition of this macro is as follows:
6549 fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
6552 @findex ASM_OUTPUT_DEBUG_LABEL
6553 @item ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6554 A C statement to output to the stdio stream @var{stream} a debug info
6555 label whose name is made from the string @var{prefix} and the number
6556 @var{num}. This is useful for VLIW targets, where debug info labels
6557 may need to be treated differently than branch target labels. On some
6558 systems, branch target labels must be at the beginning of instruction
6559 bundles, but debug info labels can occur in the middle of instruction
6562 If this macro is not defined, then @code{ASM_OUTPUT_INTERNAL_LABEL} will be
6565 @findex ASM_OUTPUT_ALTERNATE_LABEL_NAME
6566 @item ASM_OUTPUT_ALTERNATE_LABEL_NAME (@var{stream}, @var{string})
6567 A C statement to output to the stdio stream @var{stream} the string
6570 The default definition of this macro is as follows:
6573 fprintf (@var{stream}, "%s:\n", LABEL_ALTERNATE_NAME (INSN))
6576 @findex ASM_GENERATE_INTERNAL_LABEL
6577 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6578 A C statement to store into the string @var{string} a label whose name
6579 is made from the string @var{prefix} and the number @var{num}.
6581 This string, when output subsequently by @code{assemble_name}, should
6582 produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
6583 with the same @var{prefix} and @var{num}.
6585 If the string begins with @samp{*}, then @code{assemble_name} will
6586 output the rest of the string unchanged. It is often convenient for
6587 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6588 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6589 to output the string, and may change it. (Of course,
6590 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6591 you should know what it does on your machine.)
6593 @findex ASM_FORMAT_PRIVATE_NAME
6594 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6595 A C expression to assign to @var{outvar} (which is a variable of type
6596 @code{char *}) a newly allocated string made from the string
6597 @var{name} and the number @var{number}, with some suitable punctuation
6598 added. Use @code{alloca} to get space for the string.
6600 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6601 produce an assembler label for an internal static variable whose name is
6602 @var{name}. Therefore, the string must be such as to result in valid
6603 assembler code. The argument @var{number} is different each time this
6604 macro is executed; it prevents conflicts between similarly-named
6605 internal static variables in different scopes.
6607 Ideally this string should not be a valid C identifier, to prevent any
6608 conflict with the user's own symbols. Most assemblers allow periods
6609 or percent signs in assembler symbols; putting at least one of these
6610 between the name and the number will suffice.
6612 @findex ASM_OUTPUT_DEF
6613 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6614 A C statement to output to the stdio stream @var{stream} assembler code
6615 which defines (equates) the symbol @var{name} to have the value @var{value}.
6618 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6619 correct for most systems.
6621 @findex ASM_OUTPUT_DEF_FROM_DECLS
6622 @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6623 A C statement to output to the stdio stream @var{stream} assembler code
6624 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6625 to have the value of the tree node @var{decl_of_value}. This macro will
6626 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6627 the tree nodes are available.
6629 @findex ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL
6630 @item ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL (@var{stream}, @var{symbol}, @var{high}, @var{low})
6631 A C statement to output to the stdio stream @var{stream} assembler code
6632 which defines (equates) the symbol @var{symbol} to have a value equal to
6633 the difference of the two symbols @var{high} and @var{low},
6634 i.e.@: @var{high} minus @var{low}. GCC guarantees that the symbols @var{high}
6635 and @var{low} are already known by the assembler so that the difference
6636 resolves into a constant.
6639 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6640 correct for most systems.
6642 @findex ASM_OUTPUT_WEAK_ALIAS
6643 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6644 A C statement to output to the stdio stream @var{stream} assembler code
6645 which defines (equates) the weak symbol @var{name} to have the value
6646 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6647 an undefined weak symbol.
6649 Define this macro if the target only supports weak aliases; define
6650 @code{ASM_OUTPUT_DEF} instead if possible.
6652 @findex OBJC_GEN_METHOD_LABEL
6653 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6654 Define this macro to override the default assembler names used for
6655 Objective-C methods.
6657 The default name is a unique method number followed by the name of the
6658 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6659 the category is also included in the assembler name (e.g.@:
6662 These names are safe on most systems, but make debugging difficult since
6663 the method's selector is not present in the name. Therefore, particular
6664 systems define other ways of computing names.
6666 @var{buf} is an expression of type @code{char *} which gives you a
6667 buffer in which to store the name; its length is as long as
6668 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6669 50 characters extra.
6671 The argument @var{is_inst} specifies whether the method is an instance
6672 method or a class method; @var{class_name} is the name of the class;
6673 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6674 in a category); and @var{sel_name} is the name of the selector.
6676 On systems where the assembler can handle quoted names, you can use this
6677 macro to provide more human-readable names.
6679 @findex ASM_DECLARE_CLASS_REFERENCE
6680 @item ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
6681 A C statement (sans semicolon) to output to the stdio stream
6682 @var{stream} commands to declare that the label @var{name} is an
6683 Objective-C class reference. This is only needed for targets whose
6684 linkers have special support for NeXT-style runtimes.
6686 @findex ASM_DECLARE_UNRESOLVED_REFERENCE
6687 @item ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6688 A C statement (sans semicolon) to output to the stdio stream
6689 @var{stream} commands to declare that the label @var{name} is an
6690 unresolved Objective-C class reference. This is only needed for targets
6691 whose linkers have special support for NeXT-style runtimes.
6694 @node Initialization
6695 @subsection How Initialization Functions Are Handled
6696 @cindex initialization routines
6697 @cindex termination routines
6698 @cindex constructors, output of
6699 @cindex destructors, output of
6701 The compiled code for certain languages includes @dfn{constructors}
6702 (also called @dfn{initialization routines})---functions to initialize
6703 data in the program when the program is started. These functions need
6704 to be called before the program is ``started''---that is to say, before
6705 @code{main} is called.
6707 Compiling some languages generates @dfn{destructors} (also called
6708 @dfn{termination routines}) that should be called when the program
6711 To make the initialization and termination functions work, the compiler
6712 must output something in the assembler code to cause those functions to
6713 be called at the appropriate time. When you port the compiler to a new
6714 system, you need to specify how to do this.
6716 There are two major ways that GCC currently supports the execution of
6717 initialization and termination functions. Each way has two variants.
6718 Much of the structure is common to all four variations.
6720 @findex __CTOR_LIST__
6721 @findex __DTOR_LIST__
6722 The linker must build two lists of these functions---a list of
6723 initialization functions, called @code{__CTOR_LIST__}, and a list of
6724 termination functions, called @code{__DTOR_LIST__}.
6726 Each list always begins with an ignored function pointer (which may hold
6727 0, @minus{}1, or a count of the function pointers after it, depending on
6728 the environment). This is followed by a series of zero or more function
6729 pointers to constructors (or destructors), followed by a function
6730 pointer containing zero.
6732 Depending on the operating system and its executable file format, either
6733 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6734 time and exit time. Constructors are called in reverse order of the
6735 list; destructors in forward order.
6737 The best way to handle static constructors works only for object file
6738 formats which provide arbitrarily-named sections. A section is set
6739 aside for a list of constructors, and another for a list of destructors.
6740 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6741 object file that defines an initialization function also puts a word in
6742 the constructor section to point to that function. The linker
6743 accumulates all these words into one contiguous @samp{.ctors} section.
6744 Termination functions are handled similarly.
6746 This method will be chosen as the default by @file{target-def.h} if
6747 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
6748 support arbitrary sections, but does support special designated
6749 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
6750 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
6752 When arbitrary sections are available, there are two variants, depending
6753 upon how the code in @file{crtstuff.c} is called. On systems that
6754 support a @dfn{.init} section which is executed at program startup,
6755 parts of @file{crtstuff.c} are compiled into that section. The
6756 program is linked by the @code{gcc} driver like this:
6759 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
6762 The prologue of a function (@code{__init}) appears in the @code{.init}
6763 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
6764 for the function @code{__fini} in the @dfn{.fini} section. Normally these
6765 files are provided by the operating system or by the GNU C library, but
6766 are provided by GCC for a few targets.
6768 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
6769 compiled from @file{crtstuff.c}. They contain, among other things, code
6770 fragments within the @code{.init} and @code{.fini} sections that branch
6771 to routines in the @code{.text} section. The linker will pull all parts
6772 of a section together, which results in a complete @code{__init} function
6773 that invokes the routines we need at startup.
6775 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
6778 If no init section is available, when GCC compiles any function called
6779 @code{main} (or more accurately, any function designated as a program
6780 entry point by the language front end calling @code{expand_main_function}),
6781 it inserts a procedure call to @code{__main} as the first executable code
6782 after the function prologue. The @code{__main} function is defined
6783 in @file{libgcc2.c} and runs the global constructors.
6785 In file formats that don't support arbitrary sections, there are again
6786 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
6787 and an `a.out' format must be used. In this case,
6788 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
6789 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
6790 and with the address of the void function containing the initialization
6791 code as its value. The GNU linker recognizes this as a request to add
6792 the value to a @dfn{set}; the values are accumulated, and are eventually
6793 placed in the executable as a vector in the format described above, with
6794 a leading (ignored) count and a trailing zero element.
6795 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
6796 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
6797 the compilation of @code{main} to call @code{__main} as above, starting
6798 the initialization process.
6800 The last variant uses neither arbitrary sections nor the GNU linker.
6801 This is preferable when you want to do dynamic linking and when using
6802 file formats which the GNU linker does not support, such as `ECOFF'@. In
6803 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
6804 termination functions are recognized simply by their names. This requires
6805 an extra program in the linkage step, called @command{collect2}. This program
6806 pretends to be the linker, for use with GCC; it does its job by running
6807 the ordinary linker, but also arranges to include the vectors of
6808 initialization and termination functions. These functions are called
6809 via @code{__main} as described above. In order to use this method,
6810 @code{use_collect2} must be defined in the target in @file{config.gcc}.
6813 The following section describes the specific macros that control and
6814 customize the handling of initialization and termination functions.
6817 @node Macros for Initialization
6818 @subsection Macros Controlling Initialization Routines
6820 Here are the macros that control how the compiler handles initialization
6821 and termination functions:
6824 @findex INIT_SECTION_ASM_OP
6825 @item INIT_SECTION_ASM_OP
6826 If defined, a C string constant, including spacing, for the assembler
6827 operation to identify the following data as initialization code. If not
6828 defined, GCC will assume such a section does not exist. When you are
6829 using special sections for initialization and termination functions, this
6830 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6831 run the initialization functions.
6833 @item HAS_INIT_SECTION
6834 @findex HAS_INIT_SECTION
6835 If defined, @code{main} will not call @code{__main} as described above.
6836 This macro should be defined for systems that control start-up code
6837 on a symbol-by-symbol basis, such as OSF/1, and should not
6838 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
6840 @item LD_INIT_SWITCH
6841 @findex LD_INIT_SWITCH
6842 If defined, a C string constant for a switch that tells the linker that
6843 the following symbol is an initialization routine.
6845 @item LD_FINI_SWITCH
6846 @findex LD_FINI_SWITCH
6847 If defined, a C string constant for a switch that tells the linker that
6848 the following symbol is a finalization routine.
6850 @item COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
6851 If defined, a C statement that will write a function that can be
6852 automatically called when a shared library is loaded. The function
6853 should call @var{func}, which takes no arguments. If not defined, and
6854 the object format requires an explicit initialization function, then a
6855 function called @code{_GLOBAL__DI} will be generated.
6857 This function and the following one are used by collect2 when linking a
6858 shared library that needs constructors or destructors, or has DWARF2
6859 exception tables embedded in the code.
6861 @item COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
6862 If defined, a C statement that will write a function that can be
6863 automatically called when a shared library is unloaded. The function
6864 should call @var{func}, which takes no arguments. If not defined, and
6865 the object format requires an explicit finalization function, then a
6866 function called @code{_GLOBAL__DD} will be generated.
6869 @findex INVOKE__main
6870 If defined, @code{main} will call @code{__main} despite the presence of
6871 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6872 where the init section is not actually run automatically, but is still
6873 useful for collecting the lists of constructors and destructors.
6875 @item SUPPORTS_INIT_PRIORITY
6876 @findex SUPPORTS_INIT_PRIORITY
6877 If nonzero, the C++ @code{init_priority} attribute is supported and the
6878 compiler should emit instructions to control the order of initialization
6879 of objects. If zero, the compiler will issue an error message upon
6880 encountering an @code{init_priority} attribute.
6883 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
6884 This value is true if the target supports some ``native'' method of
6885 collecting constructors and destructors to be run at startup and exit.
6886 It is false if we must use @command{collect2}.
6889 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
6890 If defined, a function that outputs assembler code to arrange to call
6891 the function referenced by @var{symbol} at initialization time.
6893 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
6894 no arguments and with no return value. If the target supports initialization
6895 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
6896 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
6898 If this macro is not defined by the target, a suitable default will
6899 be chosen if (1) the target supports arbitrary section names, (2) the
6900 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
6904 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
6905 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
6906 functions rather than initialization functions.
6909 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
6910 generated for the generated object file will have static linkage.
6912 If your system uses @command{collect2} as the means of processing
6913 constructors, then that program normally uses @command{nm} to scan
6914 an object file for constructor functions to be called.
6916 On certain kinds of systems, you can define these macros to make
6917 @command{collect2} work faster (and, in some cases, make it work at all):
6920 @findex OBJECT_FORMAT_COFF
6921 @item OBJECT_FORMAT_COFF
6922 Define this macro if the system uses COFF (Common Object File Format)
6923 object files, so that @command{collect2} can assume this format and scan
6924 object files directly for dynamic constructor/destructor functions.
6926 @findex OBJECT_FORMAT_ROSE
6927 @item OBJECT_FORMAT_ROSE
6928 Define this macro if the system uses ROSE format object files, so that
6929 @command{collect2} can assume this format and scan object files directly
6930 for dynamic constructor/destructor functions.
6932 These macros are effective only in a native compiler; @command{collect2} as
6933 part of a cross compiler always uses @command{nm} for the target machine.
6935 @findex REAL_NM_FILE_NAME
6936 @item REAL_NM_FILE_NAME
6937 Define this macro as a C string constant containing the file name to use
6938 to execute @command{nm}. The default is to search the path normally for
6941 If your system supports shared libraries and has a program to list the
6942 dynamic dependencies of a given library or executable, you can define
6943 these macros to enable support for running initialization and
6944 termination functions in shared libraries:
6948 Define this macro to a C string constant containing the name of the program
6949 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
6951 @findex PARSE_LDD_OUTPUT
6952 @item PARSE_LDD_OUTPUT (@var{ptr})
6953 Define this macro to be C code that extracts filenames from the output
6954 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
6955 of type @code{char *} that points to the beginning of a line of output
6956 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6957 code must advance @var{ptr} to the beginning of the filename on that
6958 line. Otherwise, it must set @var{ptr} to @code{NULL}.
6961 @node Instruction Output
6962 @subsection Output of Assembler Instructions
6964 @c prevent bad page break with this line
6965 This describes assembler instruction output.
6968 @findex REGISTER_NAMES
6969 @item REGISTER_NAMES
6970 A C initializer containing the assembler's names for the machine
6971 registers, each one as a C string constant. This is what translates
6972 register numbers in the compiler into assembler language.
6974 @findex ADDITIONAL_REGISTER_NAMES
6975 @item ADDITIONAL_REGISTER_NAMES
6976 If defined, a C initializer for an array of structures containing a name
6977 and a register number. This macro defines additional names for hard
6978 registers, thus allowing the @code{asm} option in declarations to refer
6979 to registers using alternate names.
6981 @findex ASM_OUTPUT_OPCODE
6982 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6983 Define this macro if you are using an unusual assembler that
6984 requires different names for the machine instructions.
6986 The definition is a C statement or statements which output an
6987 assembler instruction opcode to the stdio stream @var{stream}. The
6988 macro-operand @var{ptr} is a variable of type @code{char *} which
6989 points to the opcode name in its ``internal'' form---the form that is
6990 written in the machine description. The definition should output the
6991 opcode name to @var{stream}, performing any translation you desire, and
6992 increment the variable @var{ptr} to point at the end of the opcode
6993 so that it will not be output twice.
6995 In fact, your macro definition may process less than the entire opcode
6996 name, or more than the opcode name; but if you want to process text
6997 that includes @samp{%}-sequences to substitute operands, you must take
6998 care of the substitution yourself. Just be sure to increment
6999 @var{ptr} over whatever text should not be output normally.
7001 @findex recog_data.operand
7002 If you need to look at the operand values, they can be found as the
7003 elements of @code{recog_data.operand}.
7005 If the macro definition does nothing, the instruction is output
7008 @findex FINAL_PRESCAN_INSN
7009 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7010 If defined, a C statement to be executed just prior to the output of
7011 assembler code for @var{insn}, to modify the extracted operands so
7012 they will be output differently.
7014 Here the argument @var{opvec} is the vector containing the operands
7015 extracted from @var{insn}, and @var{noperands} is the number of
7016 elements of the vector which contain meaningful data for this insn.
7017 The contents of this vector are what will be used to convert the insn
7018 template into assembler code, so you can change the assembler output
7019 by changing the contents of the vector.
7021 This macro is useful when various assembler syntaxes share a single
7022 file of instruction patterns; by defining this macro differently, you
7023 can cause a large class of instructions to be output differently (such
7024 as with rearranged operands). Naturally, variations in assembler
7025 syntax affecting individual insn patterns ought to be handled by
7026 writing conditional output routines in those patterns.
7028 If this macro is not defined, it is equivalent to a null statement.
7030 @findex FINAL_PRESCAN_LABEL
7031 @item FINAL_PRESCAN_LABEL
7032 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
7033 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
7034 @var{noperands} will be zero.
7036 @findex PRINT_OPERAND
7037 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7038 A C compound statement to output to stdio stream @var{stream} the
7039 assembler syntax for an instruction operand @var{x}. @var{x} is an
7042 @var{code} is a value that can be used to specify one of several ways
7043 of printing the operand. It is used when identical operands must be
7044 printed differently depending on the context. @var{code} comes from
7045 the @samp{%} specification that was used to request printing of the
7046 operand. If the specification was just @samp{%@var{digit}} then
7047 @var{code} is 0; if the specification was @samp{%@var{ltr}
7048 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7051 If @var{x} is a register, this macro should print the register's name.
7052 The names can be found in an array @code{reg_names} whose type is
7053 @code{char *[]}. @code{reg_names} is initialized from
7054 @code{REGISTER_NAMES}.
7056 When the machine description has a specification @samp{%@var{punct}}
7057 (a @samp{%} followed by a punctuation character), this macro is called
7058 with a null pointer for @var{x} and the punctuation character for
7061 @findex PRINT_OPERAND_PUNCT_VALID_P
7062 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7063 A C expression which evaluates to true if @var{code} is a valid
7064 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7065 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7066 punctuation characters (except for the standard one, @samp{%}) are used
7069 @findex PRINT_OPERAND_ADDRESS
7070 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7071 A C compound statement to output to stdio stream @var{stream} the
7072 assembler syntax for an instruction operand that is a memory reference
7073 whose address is @var{x}. @var{x} is an RTL expression.
7075 @cindex @code{ENCODE_SECTION_INFO} usage
7076 On some machines, the syntax for a symbolic address depends on the
7077 section that the address refers to. On these machines, define the macro
7078 @code{ENCODE_SECTION_INFO} to store the information into the
7079 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
7081 @findex DBR_OUTPUT_SEQEND
7082 @findex dbr_sequence_length
7083 @item DBR_OUTPUT_SEQEND(@var{file})
7084 A C statement, to be executed after all slot-filler instructions have
7085 been output. If necessary, call @code{dbr_sequence_length} to
7086 determine the number of slots filled in a sequence (zero if not
7087 currently outputting a sequence), to decide how many no-ops to output,
7090 Don't define this macro if it has nothing to do, but it is helpful in
7091 reading assembly output if the extent of the delay sequence is made
7092 explicit (e.g.@: with white space).
7094 @findex final_sequence
7095 Note that output routines for instructions with delay slots must be
7096 prepared to deal with not being output as part of a sequence
7097 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7098 found.) The variable @code{final_sequence} is null when not
7099 processing a sequence, otherwise it contains the @code{sequence} rtx
7102 @findex REGISTER_PREFIX
7103 @findex LOCAL_LABEL_PREFIX
7104 @findex USER_LABEL_PREFIX
7105 @findex IMMEDIATE_PREFIX
7107 @item REGISTER_PREFIX
7108 @itemx LOCAL_LABEL_PREFIX
7109 @itemx USER_LABEL_PREFIX
7110 @itemx IMMEDIATE_PREFIX
7111 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7112 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7113 @file{final.c}). These are useful when a single @file{md} file must
7114 support multiple assembler formats. In that case, the various @file{tm.h}
7115 files can define these macros differently.
7117 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
7118 @findex ASM_FPRINTF_EXTENSIONS
7119 If defined this macro should expand to a series of @code{case}
7120 statements which will be parsed inside the @code{switch} statement of
7121 the @code{asm_fprintf} function. This allows targets to define extra
7122 printf formats which may useful when generating their assembler
7123 statements. Note that upper case letters are reserved for future
7124 generic extensions to asm_fprintf, and so are not available to target
7125 specific code. The output file is given by the parameter @var{file}.
7126 The varargs input pointer is @var{argptr} and the rest of the format
7127 string, starting the character after the one that is being switched
7128 upon, is pointed to by @var{format}.
7130 @findex ASSEMBLER_DIALECT
7131 @item ASSEMBLER_DIALECT
7132 If your target supports multiple dialects of assembler language (such as
7133 different opcodes), define this macro as a C expression that gives the
7134 numeric index of the assembler language dialect to use, with zero as the
7137 If this macro is defined, you may use constructs of the form
7139 @samp{@{option0|option1|option2@dots{}@}}
7142 in the output templates of patterns (@pxref{Output Template}) or in the
7143 first argument of @code{asm_fprintf}. This construct outputs
7144 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7145 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7146 within these strings retain their usual meaning. If there are fewer
7147 alternatives within the braces than the value of
7148 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7150 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7151 @samp{@}} do not have any special meaning when used in templates or
7152 operands to @code{asm_fprintf}.
7154 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7155 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7156 the variations in assembler language syntax with that mechanism. Define
7157 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7158 if the syntax variant are larger and involve such things as different
7159 opcodes or operand order.
7161 @findex ASM_OUTPUT_REG_PUSH
7162 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7163 A C expression to output to @var{stream} some assembler code
7164 which will push hard register number @var{regno} onto the stack.
7165 The code need not be optimal, since this macro is used only when
7168 @findex ASM_OUTPUT_REG_POP
7169 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7170 A C expression to output to @var{stream} some assembler code
7171 which will pop hard register number @var{regno} off of the stack.
7172 The code need not be optimal, since this macro is used only when
7176 @node Dispatch Tables
7177 @subsection Output of Dispatch Tables
7179 @c prevent bad page break with this line
7180 This concerns dispatch tables.
7183 @cindex dispatch table
7184 @findex ASM_OUTPUT_ADDR_DIFF_ELT
7185 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7186 A C statement to output to the stdio stream @var{stream} an assembler
7187 pseudo-instruction to generate a difference between two labels.
7188 @var{value} and @var{rel} are the numbers of two internal labels. The
7189 definitions of these labels are output using
7190 @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
7191 way here. For example,
7194 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7195 @var{value}, @var{rel})
7198 You must provide this macro on machines where the addresses in a
7199 dispatch table are relative to the table's own address. If defined, GCC
7200 will also use this macro on all machines when producing PIC@.
7201 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7202 mode and flags can be read.
7204 @findex ASM_OUTPUT_ADDR_VEC_ELT
7205 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7206 This macro should be provided on machines where the addresses
7207 in a dispatch table are absolute.
7209 The definition should be a C statement to output to the stdio stream
7210 @var{stream} an assembler pseudo-instruction to generate a reference to
7211 a label. @var{value} is the number of an internal label whose
7212 definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
7216 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7219 @findex ASM_OUTPUT_CASE_LABEL
7220 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7221 Define this if the label before a jump-table needs to be output
7222 specially. The first three arguments are the same as for
7223 @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
7224 jump-table which follows (a @code{jump_insn} containing an
7225 @code{addr_vec} or @code{addr_diff_vec}).
7227 This feature is used on system V to output a @code{swbeg} statement
7230 If this macro is not defined, these labels are output with
7231 @code{ASM_OUTPUT_INTERNAL_LABEL}.
7233 @findex ASM_OUTPUT_CASE_END
7234 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7235 Define this if something special must be output at the end of a
7236 jump-table. The definition should be a C statement to be executed
7237 after the assembler code for the table is written. It should write
7238 the appropriate code to stdio stream @var{stream}. The argument
7239 @var{table} is the jump-table insn, and @var{num} is the label-number
7240 of the preceding label.
7242 If this macro is not defined, nothing special is output at the end of
7246 @node Exception Region Output
7247 @subsection Assembler Commands for Exception Regions
7249 @c prevent bad page break with this line
7251 This describes commands marking the start and the end of an exception
7255 @findex EH_FRAME_SECTION_NAME
7256 @item EH_FRAME_SECTION_NAME
7257 If defined, a C string constant for the name of the section containing
7258 exception handling frame unwind information. If not defined, GCC will
7259 provide a default definition if the target supports named sections.
7260 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7262 You should define this symbol if your target supports DWARF 2 frame
7263 unwind information and the default definition does not work.
7265 @findex EH_FRAME_IN_DATA_SECTION
7266 @item EH_FRAME_IN_DATA_SECTION
7267 If defined, DWARF 2 frame unwind information will be placed in the
7268 data section even though the target supports named sections. This
7269 might be necessary, for instance, if the system linker does garbage
7270 collection and sections cannot be marked as not to be collected.
7272 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7275 @findex MASK_RETURN_ADDR
7276 @item MASK_RETURN_ADDR
7277 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7278 that it does not contain any extraneous set bits in it.
7280 @findex DWARF2_UNWIND_INFO
7281 @item DWARF2_UNWIND_INFO
7282 Define this macro to 0 if your target supports DWARF 2 frame unwind
7283 information, but it does not yet work with exception handling.
7284 Otherwise, if your target supports this information (if it defines
7285 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7286 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7289 If this macro is defined to 1, the DWARF 2 unwinder will be the default
7290 exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by
7293 If this macro is defined to anything, the DWARF 2 unwinder will be used
7294 instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case.
7296 @findex DWARF_CIE_DATA_ALIGNMENT
7297 @item DWARF_CIE_DATA_ALIGNMENT
7298 This macro need only be defined if the target might save registers in the
7299 function prologue at an offset to the stack pointer that is not aligned to
7300 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7301 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7302 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7303 the target supports DWARF 2 frame unwind information.
7307 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7308 If defined, a function that switches to the section in which the main
7309 exception table is to be placed (@pxref{Sections}). The default is a
7310 function that switches to a section named @code{.gcc_except_table} on
7311 machines that support named sections via
7312 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7313 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7314 @code{readonly_data_section}.
7317 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7318 If defined, a function that switches to the section in which the DWARF 2
7319 frame unwind information to be placed (@pxref{Sections}). The default
7320 is a function that outputs a standard GAS section directive, if
7321 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7322 directive followed by a synthetic label.
7325 @node Alignment Output
7326 @subsection Assembler Commands for Alignment
7328 @c prevent bad page break with this line
7329 This describes commands for alignment.
7333 @item JUMP_ALIGN (@var{label})
7334 The alignment (log base 2) to put in front of @var{label}, which is
7335 a common destination of jumps and has no fallthru incoming edge.
7337 This macro need not be defined if you don't want any special alignment
7338 to be done at such a time. Most machine descriptions do not currently
7341 Unless it's necessary to inspect the @var{label} parameter, it is better
7342 to set the variable @var{align_jumps} in the target's
7343 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7344 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7346 @findex LABEL_ALIGN_AFTER_BARRIER
7347 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
7348 The alignment (log base 2) to put in front of @var{label}, which follows
7351 This macro need not be defined if you don't want any special alignment
7352 to be done at such a time. Most machine descriptions do not currently
7355 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7356 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7357 The maximum number of bytes to skip when applying
7358 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7359 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7362 @item LOOP_ALIGN (@var{label})
7363 The alignment (log base 2) to put in front of @var{label}, which follows
7364 a @code{NOTE_INSN_LOOP_BEG} note.
7366 This macro need not be defined if you don't want any special alignment
7367 to be done at such a time. Most machine descriptions do not currently
7370 Unless it's necessary to inspect the @var{label} parameter, it is better
7371 to set the variable @code{align_loops} in the target's
7372 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7373 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7375 @findex LOOP_ALIGN_MAX_SKIP
7376 @item LOOP_ALIGN_MAX_SKIP
7377 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7378 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7381 @item LABEL_ALIGN (@var{label})
7382 The alignment (log base 2) to put in front of @var{label}.
7383 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7384 the maximum of the specified values is used.
7386 Unless it's necessary to inspect the @var{label} parameter, it is better
7387 to set the variable @code{align_labels} in the target's
7388 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7389 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7391 @findex LABEL_ALIGN_MAX_SKIP
7392 @item LABEL_ALIGN_MAX_SKIP
7393 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7394 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7396 @findex ASM_OUTPUT_SKIP
7397 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7398 A C statement to output to the stdio stream @var{stream} an assembler
7399 instruction to advance the location counter by @var{nbytes} bytes.
7400 Those bytes should be zero when loaded. @var{nbytes} will be a C
7401 expression of type @code{int}.
7403 @findex ASM_NO_SKIP_IN_TEXT
7404 @item ASM_NO_SKIP_IN_TEXT
7405 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7406 text section because it fails to put zeros in the bytes that are skipped.
7407 This is true on many Unix systems, where the pseudo--op to skip bytes
7408 produces no-op instructions rather than zeros when used in the text
7411 @findex ASM_OUTPUT_ALIGN
7412 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7413 A C statement to output to the stdio stream @var{stream} an assembler
7414 command to advance the location counter to a multiple of 2 to the
7415 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7417 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
7418 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7419 A C statement to output to the stdio stream @var{stream} an assembler
7420 command to advance the location counter to a multiple of 2 to the
7421 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7422 satisfy the alignment request. @var{power} and @var{max_skip} will be
7423 a C expression of type @code{int}.
7427 @node Debugging Info
7428 @section Controlling Debugging Information Format
7430 @c prevent bad page break with this line
7431 This describes how to specify debugging information.
7434 * All Debuggers:: Macros that affect all debugging formats uniformly.
7435 * DBX Options:: Macros enabling specific options in DBX format.
7436 * DBX Hooks:: Hook macros for varying DBX format.
7437 * File Names and DBX:: Macros controlling output of file names in DBX format.
7438 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7439 * VMS Debug:: Macros for VMS debug format.
7443 @subsection Macros Affecting All Debugging Formats
7445 @c prevent bad page break with this line
7446 These macros affect all debugging formats.
7449 @findex DBX_REGISTER_NUMBER
7450 @item DBX_REGISTER_NUMBER (@var{regno})
7451 A C expression that returns the DBX register number for the compiler
7452 register number @var{regno}. In the default macro provided, the value
7453 of this expression will be @var{regno} itself. But sometimes there are
7454 some registers that the compiler knows about and DBX does not, or vice
7455 versa. In such cases, some register may need to have one number in the
7456 compiler and another for DBX@.
7458 If two registers have consecutive numbers inside GCC, and they can be
7459 used as a pair to hold a multiword value, then they @emph{must} have
7460 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7461 Otherwise, debuggers will be unable to access such a pair, because they
7462 expect register pairs to be consecutive in their own numbering scheme.
7464 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7465 does not preserve register pairs, then what you must do instead is
7466 redefine the actual register numbering scheme.
7468 @findex DEBUGGER_AUTO_OFFSET
7469 @item DEBUGGER_AUTO_OFFSET (@var{x})
7470 A C expression that returns the integer offset value for an automatic
7471 variable having address @var{x} (an RTL expression). The default
7472 computation assumes that @var{x} is based on the frame-pointer and
7473 gives the offset from the frame-pointer. This is required for targets
7474 that produce debugging output for DBX or COFF-style debugging output
7475 for SDB and allow the frame-pointer to be eliminated when the
7476 @option{-g} options is used.
7478 @findex DEBUGGER_ARG_OFFSET
7479 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7480 A C expression that returns the integer offset value for an argument
7481 having address @var{x} (an RTL expression). The nominal offset is
7484 @findex PREFERRED_DEBUGGING_TYPE
7485 @item PREFERRED_DEBUGGING_TYPE
7486 A C expression that returns the type of debugging output GCC should
7487 produce when the user specifies just @option{-g}. Define
7488 this if you have arranged for GCC to support more than one format of
7489 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7490 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7491 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7493 When the user specifies @option{-ggdb}, GCC normally also uses the
7494 value of this macro to select the debugging output format, but with two
7495 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7496 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7497 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7498 defined, GCC uses @code{DBX_DEBUG}.
7500 The value of this macro only affects the default debugging output; the
7501 user can always get a specific type of output by using @option{-gstabs},
7502 @option{-gcoff}, @option{-gdwarf-1}, @option{-gdwarf-2}, @option{-gxcoff},
7507 @subsection Specific Options for DBX Output
7509 @c prevent bad page break with this line
7510 These are specific options for DBX output.
7513 @findex DBX_DEBUGGING_INFO
7514 @item DBX_DEBUGGING_INFO
7515 Define this macro if GCC should produce debugging output for DBX
7516 in response to the @option{-g} option.
7518 @findex XCOFF_DEBUGGING_INFO
7519 @item XCOFF_DEBUGGING_INFO
7520 Define this macro if GCC should produce XCOFF format debugging output
7521 in response to the @option{-g} option. This is a variant of DBX format.
7523 @findex DEFAULT_GDB_EXTENSIONS
7524 @item DEFAULT_GDB_EXTENSIONS
7525 Define this macro to control whether GCC should by default generate
7526 GDB's extended version of DBX debugging information (assuming DBX-format
7527 debugging information is enabled at all). If you don't define the
7528 macro, the default is 1: always generate the extended information
7529 if there is any occasion to.
7531 @findex DEBUG_SYMS_TEXT
7532 @item DEBUG_SYMS_TEXT
7533 Define this macro if all @code{.stabs} commands should be output while
7534 in the text section.
7536 @findex ASM_STABS_OP
7538 A C string constant, including spacing, naming the assembler pseudo op to
7539 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7540 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7541 applies only to DBX debugging information format.
7543 @findex ASM_STABD_OP
7545 A C string constant, including spacing, naming the assembler pseudo op to
7546 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7547 value is the current location. If you don't define this macro,
7548 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7551 @findex ASM_STABN_OP
7553 A C string constant, including spacing, naming the assembler pseudo op to
7554 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7555 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7556 macro applies only to DBX debugging information format.
7558 @findex DBX_NO_XREFS
7560 Define this macro if DBX on your system does not support the construct
7561 @samp{xs@var{tagname}}. On some systems, this construct is used to
7562 describe a forward reference to a structure named @var{tagname}.
7563 On other systems, this construct is not supported at all.
7565 @findex DBX_CONTIN_LENGTH
7566 @item DBX_CONTIN_LENGTH
7567 A symbol name in DBX-format debugging information is normally
7568 continued (split into two separate @code{.stabs} directives) when it
7569 exceeds a certain length (by default, 80 characters). On some
7570 operating systems, DBX requires this splitting; on others, splitting
7571 must not be done. You can inhibit splitting by defining this macro
7572 with the value zero. You can override the default splitting-length by
7573 defining this macro as an expression for the length you desire.
7575 @findex DBX_CONTIN_CHAR
7576 @item DBX_CONTIN_CHAR
7577 Normally continuation is indicated by adding a @samp{\} character to
7578 the end of a @code{.stabs} string when a continuation follows. To use
7579 a different character instead, define this macro as a character
7580 constant for the character you want to use. Do not define this macro
7581 if backslash is correct for your system.
7583 @findex DBX_STATIC_STAB_DATA_SECTION
7584 @item DBX_STATIC_STAB_DATA_SECTION
7585 Define this macro if it is necessary to go to the data section before
7586 outputting the @samp{.stabs} pseudo-op for a non-global static
7589 @findex DBX_TYPE_DECL_STABS_CODE
7590 @item DBX_TYPE_DECL_STABS_CODE
7591 The value to use in the ``code'' field of the @code{.stabs} directive
7592 for a typedef. The default is @code{N_LSYM}.
7594 @findex DBX_STATIC_CONST_VAR_CODE
7595 @item DBX_STATIC_CONST_VAR_CODE
7596 The value to use in the ``code'' field of the @code{.stabs} directive
7597 for a static variable located in the text section. DBX format does not
7598 provide any ``right'' way to do this. The default is @code{N_FUN}.
7600 @findex DBX_REGPARM_STABS_CODE
7601 @item DBX_REGPARM_STABS_CODE
7602 The value to use in the ``code'' field of the @code{.stabs} directive
7603 for a parameter passed in registers. DBX format does not provide any
7604 ``right'' way to do this. The default is @code{N_RSYM}.
7606 @findex DBX_REGPARM_STABS_LETTER
7607 @item DBX_REGPARM_STABS_LETTER
7608 The letter to use in DBX symbol data to identify a symbol as a parameter
7609 passed in registers. DBX format does not customarily provide any way to
7610 do this. The default is @code{'P'}.
7612 @findex DBX_MEMPARM_STABS_LETTER
7613 @item DBX_MEMPARM_STABS_LETTER
7614 The letter to use in DBX symbol data to identify a symbol as a stack
7615 parameter. The default is @code{'p'}.
7617 @findex DBX_FUNCTION_FIRST
7618 @item DBX_FUNCTION_FIRST
7619 Define this macro if the DBX information for a function and its
7620 arguments should precede the assembler code for the function. Normally,
7621 in DBX format, the debugging information entirely follows the assembler
7624 @findex DBX_LBRAC_FIRST
7625 @item DBX_LBRAC_FIRST
7626 Define this macro if the @code{N_LBRAC} symbol for a block should
7627 precede the debugging information for variables and functions defined in
7628 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
7631 @findex DBX_BLOCKS_FUNCTION_RELATIVE
7632 @item DBX_BLOCKS_FUNCTION_RELATIVE
7633 Define this macro if the value of a symbol describing the scope of a
7634 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7635 of the enclosing function. Normally, GCC uses an absolute address.
7637 @findex DBX_USE_BINCL
7639 Define this macro if GCC should generate @code{N_BINCL} and
7640 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7641 macro also directs GCC to output a type number as a pair of a file
7642 number and a type number within the file. Normally, GCC does not
7643 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7644 number for a type number.
7648 @subsection Open-Ended Hooks for DBX Format
7650 @c prevent bad page break with this line
7651 These are hooks for DBX format.
7654 @findex DBX_OUTPUT_LBRAC
7655 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7656 Define this macro to say how to output to @var{stream} the debugging
7657 information for the start of a scope level for variable names. The
7658 argument @var{name} is the name of an assembler symbol (for use with
7659 @code{assemble_name}) whose value is the address where the scope begins.
7661 @findex DBX_OUTPUT_RBRAC
7662 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7663 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7665 @findex DBX_OUTPUT_ENUM
7666 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
7667 Define this macro if the target machine requires special handling to
7668 output an enumeration type. The definition should be a C statement
7669 (sans semicolon) to output the appropriate information to @var{stream}
7670 for the type @var{type}.
7672 @findex DBX_OUTPUT_FUNCTION_END
7673 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7674 Define this macro if the target machine requires special output at the
7675 end of the debugging information for a function. The definition should
7676 be a C statement (sans semicolon) to output the appropriate information
7677 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7680 @findex DBX_OUTPUT_STANDARD_TYPES
7681 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7682 Define this macro if you need to control the order of output of the
7683 standard data types at the beginning of compilation. The argument
7684 @var{syms} is a @code{tree} which is a chain of all the predefined
7685 global symbols, including names of data types.
7687 Normally, DBX output starts with definitions of the types for integers
7688 and characters, followed by all the other predefined types of the
7689 particular language in no particular order.
7691 On some machines, it is necessary to output different particular types
7692 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
7693 those symbols in the necessary order. Any predefined types that you
7694 don't explicitly output will be output afterward in no particular order.
7696 Be careful not to define this macro so that it works only for C@. There
7697 are no global variables to access most of the built-in types, because
7698 another language may have another set of types. The way to output a
7699 particular type is to look through @var{syms} to see if you can find it.
7705 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7706 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
7708 dbxout_symbol (decl);
7714 This does nothing if the expected type does not exist.
7716 See the function @code{init_decl_processing} in @file{c-decl.c} to find
7717 the names to use for all the built-in C types.
7719 Here is another way of finding a particular type:
7721 @c this is still overfull. --mew 10feb93
7725 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7726 if (TREE_CODE (decl) == TYPE_DECL
7727 && (TREE_CODE (TREE_TYPE (decl))
7729 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
7730 && TYPE_UNSIGNED (TREE_TYPE (decl)))
7732 /* @r{This must be @code{unsigned short}.} */
7733 dbxout_symbol (decl);
7739 @findex NO_DBX_FUNCTION_END
7740 @item NO_DBX_FUNCTION_END
7741 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7742 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
7743 On those machines, define this macro to turn this feature off without
7744 disturbing the rest of the gdb extensions.
7748 @node File Names and DBX
7749 @subsection File Names in DBX Format
7751 @c prevent bad page break with this line
7752 This describes file names in DBX format.
7755 @findex DBX_WORKING_DIRECTORY
7756 @item DBX_WORKING_DIRECTORY
7757 Define this if DBX wants to have the current directory recorded in each
7760 Note that the working directory is always recorded if GDB extensions are
7763 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
7764 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7765 A C statement to output DBX debugging information to the stdio stream
7766 @var{stream} which indicates that file @var{name} is the main source
7767 file---the file specified as the input file for compilation.
7768 This macro is called only once, at the beginning of compilation.
7770 This macro need not be defined if the standard form of output
7771 for DBX debugging information is appropriate.
7773 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
7774 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
7775 A C statement to output DBX debugging information to the stdio stream
7776 @var{stream} which indicates that the current directory during
7777 compilation is named @var{name}.
7779 This macro need not be defined if the standard form of output
7780 for DBX debugging information is appropriate.
7782 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
7783 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7784 A C statement to output DBX debugging information at the end of
7785 compilation of the main source file @var{name}.
7787 If you don't define this macro, nothing special is output at the end
7788 of compilation, which is correct for most machines.
7790 @findex DBX_OUTPUT_SOURCE_FILENAME
7791 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7792 A C statement to output DBX debugging information to the stdio stream
7793 @var{stream} which indicates that file @var{name} is the current source
7794 file. This output is generated each time input shifts to a different
7795 source file as a result of @samp{#include}, the end of an included file,
7796 or a @samp{#line} command.
7798 This macro need not be defined if the standard form of output
7799 for DBX debugging information is appropriate.
7804 @subsection Macros for SDB and DWARF Output
7806 @c prevent bad page break with this line
7807 Here are macros for SDB and DWARF output.
7810 @findex SDB_DEBUGGING_INFO
7811 @item SDB_DEBUGGING_INFO
7812 Define this macro if GCC should produce COFF-style debugging output
7813 for SDB in response to the @option{-g} option.
7815 @findex DWARF_DEBUGGING_INFO
7816 @item DWARF_DEBUGGING_INFO
7817 Define this macro if GCC should produce dwarf format debugging output
7818 in response to the @option{-g} option.
7820 @findex DWARF2_DEBUGGING_INFO
7821 @item DWARF2_DEBUGGING_INFO
7822 Define this macro if GCC should produce dwarf version 2 format
7823 debugging output in response to the @option{-g} option.
7825 To support optional call frame debugging information, you must also
7826 define @code{INCOMING_RETURN_ADDR_RTX} and either set
7827 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
7828 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
7829 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
7831 @findex DWARF2_FRAME_INFO
7832 @item DWARF2_FRAME_INFO
7833 Define this macro to a nonzero value if GCC should always output
7834 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
7835 (@pxref{Exception Region Output} is nonzero, GCC will output this
7836 information not matter how you define @code{DWARF2_FRAME_INFO}.
7838 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
7839 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
7840 Define this macro if the linker does not work with Dwarf version 2.
7841 Normally, if the user specifies only @option{-ggdb} GCC will use Dwarf
7842 version 2 if available; this macro disables this. See the description
7843 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
7845 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
7846 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
7847 By default, the Dwarf 2 debugging information generator will generate a
7848 label to mark the beginning of the text section. If it is better simply
7849 to use the name of the text section itself, rather than an explicit label,
7850 to indicate the beginning of the text section, define this macro to zero.
7852 @findex DWARF2_ASM_LINE_DEBUG_INFO
7853 @item DWARF2_ASM_LINE_DEBUG_INFO
7854 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
7855 line debug info sections. This will result in much more compact line number
7856 tables, and hence is desirable if it works.
7858 @findex PUT_SDB_@dots{}
7859 @item PUT_SDB_@dots{}
7860 Define these macros to override the assembler syntax for the special
7861 SDB assembler directives. See @file{sdbout.c} for a list of these
7862 macros and their arguments. If the standard syntax is used, you need
7863 not define them yourself.
7867 Some assemblers do not support a semicolon as a delimiter, even between
7868 SDB assembler directives. In that case, define this macro to be the
7869 delimiter to use (usually @samp{\n}). It is not necessary to define
7870 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
7873 @findex SDB_GENERATE_FAKE
7874 @item SDB_GENERATE_FAKE
7875 Define this macro to override the usual method of constructing a dummy
7876 name for anonymous structure and union types. See @file{sdbout.c} for
7879 @findex SDB_ALLOW_UNKNOWN_REFERENCES
7880 @item SDB_ALLOW_UNKNOWN_REFERENCES
7881 Define this macro to allow references to unknown structure,
7882 union, or enumeration tags to be emitted. Standard COFF does not
7883 allow handling of unknown references, MIPS ECOFF has support for
7886 @findex SDB_ALLOW_FORWARD_REFERENCES
7887 @item SDB_ALLOW_FORWARD_REFERENCES
7888 Define this macro to allow references to structure, union, or
7889 enumeration tags that have not yet been seen to be handled. Some
7890 assemblers choke if forward tags are used, while some require it.
7895 @subsection Macros for VMS Debug Format
7897 @c prevent bad page break with this line
7898 Here are macros for VMS debug format.
7901 @findex VMS_DEBUGGING_INFO
7902 @item VMS_DEBUGGING_INFO
7903 Define this macro if GCC should produce debugging output for VMS
7904 in response to the @option{-g} option. The default behavior for VMS
7905 is to generate minimal debug info for a traceback in the absence of
7906 @option{-g} unless explicitly overridden with @option{-g0}. This
7907 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
7908 @code{OVERRIDE_OPTIONS}.
7911 @node Floating Point
7912 @section Cross Compilation and Floating Point
7913 @cindex cross compilation and floating point
7914 @cindex floating point and cross compilation
7916 While all modern machines use twos-complement representation for integers,
7917 there are a variety of representations for floating point numbers. This
7918 means that in a cross-compiler the representation of floating point numbers
7919 in the compiled program may be different from that used in the machine
7920 doing the compilation.
7922 Because different representation systems may offer different amounts of
7923 range and precision, all floating point constants must be represented in
7924 the target machine's format. Therefore, the cross compiler cannot
7925 safely use the host machine's floating point arithmetic; it must emulate
7926 the target's arithmetic. To ensure consistency, GCC always uses
7927 emulation to work with floating point values, even when the host and
7928 target floating point formats are identical.
7930 The following macros are provided by @file{real.h} for the compiler to
7931 use. All parts of the compiler which generate or optimize
7932 floating-point calculations must use these macros. They may evaluate
7933 their operands more than once, so operands must not have side effects.
7935 @defmac REAL_VALUE_TYPE
7936 The C data type to be used to hold a floating point value in the target
7937 machine's format. Typically this is a @code{struct} containing an
7938 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
7942 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
7943 Compares for equality the two values, @var{x} and @var{y}. If the target
7944 floating point format supports negative zeroes and/or NaNs,
7945 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
7946 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
7949 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
7950 Tests whether @var{x} is less than @var{y}.
7954 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_LDEXP (REAL_VALUE_TYPE @var{x}, int @var{scale})
7955 Multiplies @var{x} by 2 raised to the power @var{scale}.
7958 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
7959 Truncates @var{x} to a signed integer, rounding toward zero.
7962 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
7963 Truncates @var{x} to an unsigned integer, rounding toward zero. If
7964 @var{x} is negative, returns zero.
7967 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_RNDZINT (REAL_VALUE_TYPE @var{x})
7968 Rounds the target-machine floating point value @var{x} towards zero to an
7969 integer value, but leaves it represented as a floating point number.
7972 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_UNSIGNED_RNDZINT (REAL_VALUE_TYPE @var{x})
7973 Rounds the target-machine floating point value @var{x} towards zero to an
7974 unsigned integer value, but leaves it represented as a floating point
7975 number. If @var{x} is negative, returns (positive) zero.
7978 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
7979 Converts @var{string} into a floating point number in the target machine's
7980 representation for mode @var{mode}. This routine can handle both
7981 decimal and hexadecimal floating point constants, using the syntax
7982 defined by the C language for both.
7985 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
7986 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
7989 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
7990 Determines whether @var{x} represents infinity (positive or negative).
7993 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
7994 Determines whether @var{x} represents a ``NaN'' (not-a-number).
7997 @deftypefn Macro void REAL_ARITHMETIC (REAL_VALUE_TYPE @var{output}, enum tree_code @var{code}, REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
7998 Calculates an arithmetic operation on the two floating point values
7999 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8002 The operation to be performed is specified by @var{code}. Only the
8003 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8004 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8006 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8007 target's floating point format cannot represent infinity, it will call
8008 @code{abort}. Callers should check for this situation first, using
8009 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8012 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8013 Returns the negative of the floating point value @var{x}.
8016 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8017 Returns the absolute value of @var{x}.
8020 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8021 Truncates the floating point value @var{x} to fit in @var{mode}. The
8022 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8023 appropriate bit pattern to be output asa floating constant whose
8024 precision accords with mode @var{mode}.
8027 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8028 Converts a floating point value @var{x} into a double-precision integer
8029 which is then stored into @var{low} and @var{high}. If the value is not
8030 integral, it is truncated.
8033 @deftypefn Macro void REAL_VALUE_FROM_INT (REAL_VALUE_TYPE @var{x}, HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, enum machine_mode @var{mode})
8034 @findex REAL_VALUE_FROM_INT
8035 Converts a double-precision integer found in @var{low} and @var{high},
8036 into a floating point value which is then stored into @var{x}. The
8037 value is truncated to fit in mode @var{mode}.
8040 @node Mode Switching
8041 @section Mode Switching Instructions
8042 @cindex mode switching
8043 The following macros control mode switching optimizations:
8046 @findex OPTIMIZE_MODE_SWITCHING
8047 @item OPTIMIZE_MODE_SWITCHING (@var{entity})
8048 Define this macro if the port needs extra instructions inserted for mode
8049 switching in an optimizing compilation.
8051 For an example, the SH4 can perform both single and double precision
8052 floating point operations, but to perform a single precision operation,
8053 the FPSCR PR bit has to be cleared, while for a double precision
8054 operation, this bit has to be set. Changing the PR bit requires a general
8055 purpose register as a scratch register, hence these FPSCR sets have to
8056 be inserted before reload, i.e.@: you can't put this into instruction emitting
8057 or @code{MACHINE_DEPENDENT_REORG}.
8059 You can have multiple entities that are mode-switched, and select at run time
8060 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8061 return nonzero for any @var{entity} that needs mode-switching.
8062 If you define this macro, you also have to define
8063 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8064 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8065 @code{NORMAL_MODE} is optional.
8067 @findex NUM_MODES_FOR_MODE_SWITCHING
8068 @item NUM_MODES_FOR_MODE_SWITCHING
8069 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8070 initializer for an array of integers. Each initializer element
8071 N refers to an entity that needs mode switching, and specifies the number
8072 of different modes that might need to be set for this entity.
8073 The position of the initializer in the initializer - starting counting at
8074 zero - determines the integer that is used to refer to the mode-switched
8076 In macros that take mode arguments / yield a mode result, modes are
8077 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8078 switch is needed / supplied.
8081 @item MODE_NEEDED (@var{entity}, @var{insn})
8082 @var{entity} is an integer specifying a mode-switched entity. If
8083 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8084 return an integer value not larger than the corresponding element in
8085 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8086 be switched into prior to the execution of @var{insn}.
8089 @item NORMAL_MODE (@var{entity})
8090 If this macro is defined, it is evaluated for every @var{entity} that needs
8091 mode switching. It should evaluate to an integer, which is a mode that
8092 @var{entity} is assumed to be switched to at function entry and exit.
8094 @findex MODE_PRIORITY_TO_MODE
8095 @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8096 This macro specifies the order in which modes for @var{entity} are processed.
8097 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8098 lowest. The value of the macro should be an integer designating a mode
8099 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8100 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8101 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8103 @findex EMIT_MODE_SET
8104 @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8105 Generate one or more insns to set @var{entity} to @var{mode}.
8106 @var{hard_reg_live} is the set of hard registers live at the point where
8107 the insn(s) are to be inserted.
8110 @node Target Attributes
8111 @section Defining target-specific uses of @code{__attribute__}
8112 @cindex target attributes
8113 @cindex machine attributes
8114 @cindex attributes, target-specific
8116 Target-specific attributes may be defined for functions, data and types.
8117 These are described using the following target hooks; they also need to
8118 be documented in @file{extend.texi}.
8120 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8121 If defined, this target hook points to an array of @samp{struct
8122 attribute_spec} (defined in @file{tree.h}) specifying the machine
8123 specific attributes for this target and some of the restrictions on the
8124 entities to which these attributes are applied and the arguments they
8128 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8129 If defined, this target hook is a function which returns zero if the attributes on
8130 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8131 and two if they are nearly compatible (which causes a warning to be
8132 generated). If this is not defined, machine-specific attributes are
8133 supposed always to be compatible.
8136 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8137 If defined, this target hook is a function which assigns default attributes to
8138 newly defined @var{type}.
8141 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8142 Define this target hook if the merging of type attributes needs special
8143 handling. If defined, the result is a list of the combined
8144 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8145 that @code{comptypes} has already been called and returned 1. This
8146 function may call @code{merge_attributes} to handle machine-independent
8150 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8151 Define this target hook if the merging of decl attributes needs special
8152 handling. If defined, the result is a list of the combined
8153 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8154 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8155 when this is needed are when one attribute overrides another, or when an
8156 attribute is nullified by a subsequent definition. This function may
8157 call @code{merge_attributes} to handle machine-independent merging.
8159 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8160 If the only target-specific handling you require is @samp{dllimport} for
8161 Windows targets, you should define the macro
8162 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function
8163 called @code{merge_dllimport_decl_attributes} which can then be defined
8164 as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done
8165 in @file{i386/cygwin.h} and @file{i386/i386.c}, for example.
8168 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8169 Define this target hook if you want to be able to add attributes to a decl
8170 when it is being created. This is normally useful for back ends which
8171 wish to implement a pragma by using the attributes which correspond to
8172 the pragma's effect. The @var{node} argument is the decl which is being
8173 created. The @var{attr_ptr} argument is a pointer to the attribute list
8174 for this decl. The list itself should not be modified, since it may be
8175 shared with other decls, but attributes may be chained on the head of
8176 the list and @code{*@var{attr_ptr}} modified to point to the new
8177 attributes, or a copy of the list may be made if further changes are
8181 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8183 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8184 into the current function, despite its having target-specific
8185 attributes, @code{false} otherwise. By default, if a function has a
8186 target specific attribute attached to it, it will not be inlined.
8189 @node MIPS Coprocessors
8190 @section Defining coprocessor specifics for MIPS targets.
8191 @cindex MIPS coprocessor-definition macros
8193 The MIPS specification allows MIPS implementations to have as many as 4
8194 coprocessors, each with as many as 32 private registers. gcc supports
8195 accessing these registers and transferring values between the registers
8196 and memory using asm-ized variables. For example:
8199 register unsigned int cp0count asm ("c0r1");
8205 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8206 names may be added as described below, or the default names may be
8207 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8209 Coprocessor registers are assumed to be epilogue-used; sets to them will
8210 be preserved even if it does not appear that the register is used again
8211 later in the function.
8213 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8214 the FPU. One accesses COP1 registers through standard mips
8215 floating-point support; they are not included in this mechanism.
8217 There is one macro used in defining the MIPS coprocessor interface which
8218 you may want to override in subtargets; it is described below.
8222 @item ALL_COP_ADDITIONAL_REGISTER_NAMES
8223 @findex ALL_COP_ADDITIONAL_REGISTER_NAMES
8224 A comma-separated list (with leading comma) of pairs describing the
8225 alternate names of coprocessor registers. The format of each entry should be
8227 @{ @var{alternatename}, @var{register_number}@}
8234 @section Miscellaneous Parameters
8235 @cindex parameters, miscellaneous
8237 @c prevent bad page break with this line
8238 Here are several miscellaneous parameters.
8241 @item PREDICATE_CODES
8242 @findex PREDICATE_CODES
8243 Define this if you have defined special-purpose predicates in the file
8244 @file{@var{machine}.c}. This macro is called within an initializer of an
8245 array of structures. The first field in the structure is the name of a
8246 predicate and the second field is an array of rtl codes. For each
8247 predicate, list all rtl codes that can be in expressions matched by the
8248 predicate. The list should have a trailing comma. Here is an example
8249 of two entries in the list for a typical RISC machine:
8252 #define PREDICATE_CODES \
8253 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8254 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8257 Defining this macro does not affect the generated code (however,
8258 incorrect definitions that omit an rtl code that may be matched by the
8259 predicate can cause the compiler to malfunction). Instead, it allows
8260 the table built by @file{genrecog} to be more compact and efficient,
8261 thus speeding up the compiler. The most important predicates to include
8262 in the list specified by this macro are those used in the most insn
8265 For each predicate function named in @code{PREDICATE_CODES}, a
8266 declaration will be generated in @file{insn-codes.h}.
8268 @item SPECIAL_MODE_PREDICATES
8269 @findex SPECIAL_MODE_PREDICATES
8270 Define this if you have special predicates that know special things
8271 about modes. Genrecog will warn about certain forms of
8272 @code{match_operand} without a mode; if the operand predicate is
8273 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8276 Here is an example from the IA-32 port (@code{ext_register_operand}
8277 specially checks for @code{HImode} or @code{SImode} in preparation
8278 for a byte extraction from @code{%ah} etc.).
8281 #define SPECIAL_MODE_PREDICATES \
8282 "ext_register_operand",
8285 @findex CASE_VECTOR_MODE
8286 @item CASE_VECTOR_MODE
8287 An alias for a machine mode name. This is the machine mode that
8288 elements of a jump-table should have.
8290 @findex CASE_VECTOR_SHORTEN_MODE
8291 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8292 Optional: return the preferred mode for an @code{addr_diff_vec}
8293 when the minimum and maximum offset are known. If you define this,
8294 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8295 To make this work, you also have to define INSN_ALIGN and
8296 make the alignment for @code{addr_diff_vec} explicit.
8297 The @var{body} argument is provided so that the offset_unsigned and scale
8298 flags can be updated.
8300 @findex CASE_VECTOR_PC_RELATIVE
8301 @item CASE_VECTOR_PC_RELATIVE
8302 Define this macro to be a C expression to indicate when jump-tables
8303 should contain relative addresses. If jump-tables never contain
8304 relative addresses, then you need not define this macro.
8306 @findex CASE_DROPS_THROUGH
8307 @item CASE_DROPS_THROUGH
8308 Define this if control falls through a @code{case} insn when the index
8309 value is out of range. This means the specified default-label is
8310 actually ignored by the @code{case} insn proper.
8312 @findex CASE_VALUES_THRESHOLD
8313 @item CASE_VALUES_THRESHOLD
8314 Define this to be the smallest number of different values for which it
8315 is best to use a jump-table instead of a tree of conditional branches.
8316 The default is four for machines with a @code{casesi} instruction and
8317 five otherwise. This is best for most machines.
8319 @findex WORD_REGISTER_OPERATIONS
8320 @item WORD_REGISTER_OPERATIONS
8321 Define this macro if operations between registers with integral mode
8322 smaller than a word are always performed on the entire register.
8323 Most RISC machines have this property and most CISC machines do not.
8325 @findex LOAD_EXTEND_OP
8326 @item LOAD_EXTEND_OP (@var{mode})
8327 Define this macro to be a C expression indicating when insns that read
8328 memory in @var{mode}, an integral mode narrower than a word, set the
8329 bits outside of @var{mode} to be either the sign-extension or the
8330 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8331 of @var{mode} for which the
8332 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8333 @code{NIL} for other modes.
8335 This macro is not called with @var{mode} non-integral or with a width
8336 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8337 value in this case. Do not define this macro if it would always return
8338 @code{NIL}. On machines where this macro is defined, you will normally
8339 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8341 @findex SHORT_IMMEDIATES_SIGN_EXTEND
8342 @item SHORT_IMMEDIATES_SIGN_EXTEND
8343 Define this macro if loading short immediate values into registers sign
8346 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
8347 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
8348 Define this macro if the same instructions that convert a floating
8349 point number to a signed fixed point number also convert validly to an
8354 The maximum number of bytes that a single instruction can move quickly
8355 between memory and registers or between two memory locations.
8357 @findex MAX_MOVE_MAX
8359 The maximum number of bytes that a single instruction can move quickly
8360 between memory and registers or between two memory locations. If this
8361 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8362 constant value that is the largest value that @code{MOVE_MAX} can have
8365 @findex SHIFT_COUNT_TRUNCATED
8366 @item SHIFT_COUNT_TRUNCATED
8367 A C expression that is nonzero if on this machine the number of bits
8368 actually used for the count of a shift operation is equal to the number
8369 of bits needed to represent the size of the object being shifted. When
8370 this macro is nonzero, the compiler will assume that it is safe to omit
8371 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8372 truncates the count of a shift operation. On machines that have
8373 instructions that act on bit-fields at variable positions, which may
8374 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8375 also enables deletion of truncations of the values that serve as
8376 arguments to bit-field instructions.
8378 If both types of instructions truncate the count (for shifts) and
8379 position (for bit-field operations), or if no variable-position bit-field
8380 instructions exist, you should define this macro.
8382 However, on some machines, such as the 80386 and the 680x0, truncation
8383 only applies to shift operations and not the (real or pretended)
8384 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8385 such machines. Instead, add patterns to the @file{md} file that include
8386 the implied truncation of the shift instructions.
8388 You need not define this macro if it would always have the value of zero.
8390 @findex TRULY_NOOP_TRUNCATION
8391 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8392 A C expression which is nonzero if on this machine it is safe to
8393 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8394 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8395 operating on it as if it had only @var{outprec} bits.
8397 On many machines, this expression can be 1.
8399 @c rearranged this, removed the phrase "it is reported that". this was
8400 @c to fix an overfull hbox. --mew 10feb93
8401 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8402 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8403 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8404 such cases may improve things.
8406 @findex STORE_FLAG_VALUE
8407 @item STORE_FLAG_VALUE
8408 A C expression describing the value returned by a comparison operator
8409 with an integral mode and stored by a store-flag instruction
8410 (@samp{s@var{cond}}) when the condition is true. This description must
8411 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8412 comparison operators whose results have a @code{MODE_INT} mode.
8414 A value of 1 or @minus{}1 means that the instruction implementing the
8415 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8416 and 0 when the comparison is false. Otherwise, the value indicates
8417 which bits of the result are guaranteed to be 1 when the comparison is
8418 true. This value is interpreted in the mode of the comparison
8419 operation, which is given by the mode of the first operand in the
8420 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8421 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8424 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8425 generate code that depends only on the specified bits. It can also
8426 replace comparison operators with equivalent operations if they cause
8427 the required bits to be set, even if the remaining bits are undefined.
8428 For example, on a machine whose comparison operators return an
8429 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8430 @samp{0x80000000}, saying that just the sign bit is relevant, the
8434 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8441 (ashift:SI @var{x} (const_int @var{n}))
8445 where @var{n} is the appropriate shift count to move the bit being
8446 tested into the sign bit.
8448 There is no way to describe a machine that always sets the low-order bit
8449 for a true value, but does not guarantee the value of any other bits,
8450 but we do not know of any machine that has such an instruction. If you
8451 are trying to port GCC to such a machine, include an instruction to
8452 perform a logical-and of the result with 1 in the pattern for the
8453 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8455 Often, a machine will have multiple instructions that obtain a value
8456 from a comparison (or the condition codes). Here are rules to guide the
8457 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8462 Use the shortest sequence that yields a valid definition for
8463 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8464 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8465 comparison operators to do so because there may be opportunities to
8466 combine the normalization with other operations.
8469 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8470 slightly preferred on machines with expensive jumps and 1 preferred on
8474 As a second choice, choose a value of @samp{0x80000001} if instructions
8475 exist that set both the sign and low-order bits but do not define the
8479 Otherwise, use a value of @samp{0x80000000}.
8482 Many machines can produce both the value chosen for
8483 @code{STORE_FLAG_VALUE} and its negation in the same number of
8484 instructions. On those machines, you should also define a pattern for
8485 those cases, e.g., one matching
8488 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8491 Some machines can also perform @code{and} or @code{plus} operations on
8492 condition code values with less instructions than the corresponding
8493 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8494 machines, define the appropriate patterns. Use the names @code{incscc}
8495 and @code{decscc}, respectively, for the patterns which perform
8496 @code{plus} or @code{minus} operations on condition code values. See
8497 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8498 find such instruction sequences on other machines.
8500 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8503 @findex FLOAT_STORE_FLAG_VALUE
8504 @item FLOAT_STORE_FLAG_VALUE (@var{mode})
8505 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8506 returned when comparison operators with floating-point results are true.
8507 Define this macro on machine that have comparison operations that return
8508 floating-point values. If there are no such operations, do not define
8513 An alias for the machine mode for pointers. On most machines, define
8514 this to be the integer mode corresponding to the width of a hardware
8515 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8516 On some machines you must define this to be one of the partial integer
8517 modes, such as @code{PSImode}.
8519 The width of @code{Pmode} must be at least as large as the value of
8520 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8521 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8524 @findex FUNCTION_MODE
8526 An alias for the machine mode used for memory references to functions
8527 being called, in @code{call} RTL expressions. On most machines this
8528 should be @code{QImode}.
8530 @findex INTEGRATE_THRESHOLD
8531 @item INTEGRATE_THRESHOLD (@var{decl})
8532 A C expression for the maximum number of instructions above which the
8533 function @var{decl} should not be inlined. @var{decl} is a
8534 @code{FUNCTION_DECL} node.
8536 The default definition of this macro is 64 plus 8 times the number of
8537 arguments that the function accepts. Some people think a larger
8538 threshold should be used on RISC machines.
8540 @findex STDC_0_IN_SYSTEM_HEADERS
8541 @item STDC_0_IN_SYSTEM_HEADERS
8542 In normal operation, the preprocessor expands @code{__STDC__} to the
8543 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8544 hosts, like Solaris, the system compiler uses a different convention,
8545 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8546 strict conformance to the C Standard.
8548 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8549 convention when processing system header files, but when processing user
8550 files @code{__STDC__} will always expand to 1.
8552 @findex SCCS_DIRECTIVE
8553 @item SCCS_DIRECTIVE
8554 Define this if the preprocessor should ignore @code{#sccs} directives
8555 and print no error message.
8557 @findex NO_IMPLICIT_EXTERN_C
8558 @item NO_IMPLICIT_EXTERN_C
8559 Define this macro if the system header files support C++ as well as C@.
8560 This macro inhibits the usual method of using system header files in
8561 C++, which is to pretend that the file's contents are enclosed in
8562 @samp{extern "C" @{@dots{}@}}.
8564 @findex HANDLE_PRAGMA
8565 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
8566 This macro is no longer supported. You must use
8567 @code{REGISTER_TARGET_PRAGMAS} instead.
8569 @findex REGISTER_TARGET_PRAGMAS
8572 @item REGISTER_TARGET_PRAGMAS (@var{pfile})
8573 Define this macro if you want to implement any target-specific pragmas.
8574 If defined, it is a C expression which makes a series of calls to
8575 @code{cpp_register_pragma} for each pragma, with @var{pfile} passed as
8576 the first argument to to these functions. The macro may also do any
8577 setup required for the pragmas.
8579 The primary reason to define this macro is to provide compatibility with
8580 other compilers for the same target. In general, we discourage
8581 definition of target-specific pragmas for GCC@.
8583 If the pragma can be implemented by attributes then you should consider
8584 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8586 Preprocessor macros that appear on pragma lines are not expanded. All
8587 @samp{#pragma} directives that do not match any registered pragma are
8588 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8590 @deftypefun void cpp_register_pragma (cpp_reader *@var{pfile}, const char *@var{space}, const char *@var{name}, void (*@var{callback}) (cpp_reader *))
8592 Each call to @code{cpp_register_pragma} establishes one pragma. The
8593 @var{callback} routine will be called when the preprocessor encounters a
8597 #pragma [@var{space}] @var{name} @dots{}
8600 @var{space} is the case-sensitive namespace of the pragma, or
8601 @code{NULL} to put the pragma in the global namespace. The callback
8602 routine receives @var{pfile} as its first argument, which can be passed
8603 on to cpplib's functions if necessary. You can lex tokens after the
8604 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8605 callback will be silently ignored. The end of the line is indicated by
8606 a token of type @code{CPP_EOF}.
8608 For an example use of this routine, see @file{c4x.h} and the callback
8609 routines defined in @file{c4x-c.c}.
8611 Note that the use of @code{c_lex} is specific to the C and C++
8612 compilers. It will not work in the Java or Fortran compilers, or any
8613 other language compilers for that matter. Thus if @code{c_lex} is going
8614 to be called from target-specific code, it must only be done so when
8615 building the C and C++ compilers. This can be done by defining the
8616 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8617 target entry in the @file{config.gcc} file. These variables should name
8618 the target-specific, language-specific object file which contains the
8619 code that uses @code{c_lex}. Note it will also be necessary to add a
8620 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8621 how to build this object file.
8624 @findex HANDLE_SYSV_PRAGMA
8627 @item HANDLE_SYSV_PRAGMA
8628 Define this macro (to a value of 1) if you want the System V style
8629 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8630 [=<value>]} to be supported by gcc.
8632 The pack pragma specifies the maximum alignment (in bytes) of fields
8633 within a structure, in much the same way as the @samp{__aligned__} and
8634 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8635 the behavior to the default.
8637 The weak pragma only works if @code{SUPPORTS_WEAK} and
8638 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8639 of specifically named weak labels, optionally with a value.
8641 @findex HANDLE_PRAGMA_PACK_PUSH_POP
8644 @item HANDLE_PRAGMA_PACK_PUSH_POP
8645 Define this macro (to a value of 1) if you want to support the Win32
8646 style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma
8647 pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment
8648 (in bytes) of fields within a structure, in much the same way as the
8649 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8650 pack value of zero resets the behavior to the default. Successive
8651 invocations of this pragma cause the previous values to be stacked, so
8652 that invocations of @samp{#pragma pack(pop)} will return to the previous
8655 @findex DOLLARS_IN_IDENTIFIERS
8656 @item DOLLARS_IN_IDENTIFIERS
8657 Define this macro to control use of the character @samp{$} in identifier
8658 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
8659 1 is the default; there is no need to define this macro in that case.
8660 This macro controls the compiler proper; it does not affect the preprocessor.
8662 @findex NO_DOLLAR_IN_LABEL
8663 @item NO_DOLLAR_IN_LABEL
8664 Define this macro if the assembler does not accept the character
8665 @samp{$} in label names. By default constructors and destructors in
8666 G++ have @samp{$} in the identifiers. If this macro is defined,
8667 @samp{.} is used instead.
8669 @findex NO_DOT_IN_LABEL
8670 @item NO_DOT_IN_LABEL
8671 Define this macro if the assembler does not accept the character
8672 @samp{.} in label names. By default constructors and destructors in G++
8673 have names that use @samp{.}. If this macro is defined, these names
8674 are rewritten to avoid @samp{.}.
8676 @findex DEFAULT_MAIN_RETURN
8677 @item DEFAULT_MAIN_RETURN
8678 Define this macro if the target system expects every program's @code{main}
8679 function to return a standard ``success'' value by default (if no other
8680 value is explicitly returned).
8682 The definition should be a C statement (sans semicolon) to generate the
8683 appropriate rtl instructions. It is used only when compiling the end of
8688 Define this if the target system lacks the function @code{atexit}
8689 from the ISO C standard. If this macro is defined, a default definition
8690 will be provided to support C++. If @code{ON_EXIT} is not defined,
8691 a default @code{exit} function will also be provided.
8695 Define this macro if the target has another way to implement atexit
8696 functionality without replacing @code{exit}. For instance, SunOS 4 has
8697 a similar @code{on_exit} library function.
8699 The definition should be a functional macro which can be used just like
8700 the @code{atexit} function.
8704 Define this if your @code{exit} function needs to do something
8705 besides calling an external function @code{_cleanup} before
8706 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
8707 only needed if @code{NEED_ATEXIT} is defined and @code{ON_EXIT} is not
8710 @findex INSN_SETS_ARE_DELAYED
8711 @item INSN_SETS_ARE_DELAYED (@var{insn})
8712 Define this macro as a C expression that is nonzero if it is safe for the
8713 delay slot scheduler to place instructions in the delay slot of @var{insn},
8714 even if they appear to use a resource set or clobbered in @var{insn}.
8715 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8716 every @code{call_insn} has this behavior. On machines where some @code{insn}
8717 or @code{jump_insn} is really a function call and hence has this behavior,
8718 you should define this macro.
8720 You need not define this macro if it would always return zero.
8722 @findex INSN_REFERENCES_ARE_DELAYED
8723 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
8724 Define this macro as a C expression that is nonzero if it is safe for the
8725 delay slot scheduler to place instructions in the delay slot of @var{insn},
8726 even if they appear to set or clobber a resource referenced in @var{insn}.
8727 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8728 some @code{insn} or @code{jump_insn} is really a function call and its operands
8729 are registers whose use is actually in the subroutine it calls, you should
8730 define this macro. Doing so allows the delay slot scheduler to move
8731 instructions which copy arguments into the argument registers into the delay
8734 You need not define this macro if it would always return zero.
8736 @findex MACHINE_DEPENDENT_REORG
8737 @item MACHINE_DEPENDENT_REORG (@var{insn})
8738 In rare cases, correct code generation requires extra machine
8739 dependent processing between the second jump optimization pass and
8740 delayed branch scheduling. On those machines, define this macro as a C
8741 statement to act on the code starting at @var{insn}.
8743 @findex MULTIPLE_SYMBOL_SPACES
8744 @item MULTIPLE_SYMBOL_SPACES
8745 Define this macro if in some cases global symbols from one translation
8746 unit may not be bound to undefined symbols in another translation unit
8747 without user intervention. For instance, under Microsoft Windows
8748 symbols must be explicitly imported from shared libraries (DLLs).
8750 @findex MD_ASM_CLOBBERS
8751 @item MD_ASM_CLOBBERS (@var{clobbers})
8752 A C statement that adds to @var{clobbers} @code{STRING_CST} trees for
8753 any hard regs the port wishes to automatically clobber for all asms.
8755 @findex MAX_INTEGER_COMPUTATION_MODE
8756 @item MAX_INTEGER_COMPUTATION_MODE
8757 Define this to the largest integer machine mode which can be used for
8758 operations other than load, store and copy operations.
8760 You need only define this macro if the target holds values larger than
8761 @code{word_mode} in general purpose registers. Most targets should not define
8764 @findex MATH_LIBRARY
8766 Define this macro as a C string constant for the linker argument to link
8767 in the system math library, or @samp{""} if the target does not have a
8768 separate math library.
8770 You need only define this macro if the default of @samp{"-lm"} is wrong.
8772 @findex LIBRARY_PATH_ENV
8773 @item LIBRARY_PATH_ENV
8774 Define this macro as a C string constant for the environment variable that
8775 specifies where the linker should look for libraries.
8777 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
8780 @findex TARGET_HAS_F_SETLKW
8781 @item TARGET_HAS_F_SETLKW
8782 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
8783 Note that this functionality is part of POSIX@.
8784 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
8785 to use file locking when exiting a program, which avoids race conditions
8786 if the program has forked.
8788 @findex MAX_CONDITIONAL_EXECUTE
8789 @item MAX_CONDITIONAL_EXECUTE
8791 A C expression for the maximum number of instructions to execute via
8792 conditional execution instructions instead of a branch. A value of
8793 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8794 1 if it does use cc0.
8796 @findex IFCVT_MODIFY_TESTS
8797 @item IFCVT_MODIFY_TESTS
8798 A C expression to modify the tests in @code{TRUE_EXPR}, and
8799 @code{FALSE_EXPR} for use in converting insns in @code{TEST_BB},
8800 @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB} basic blocks to
8801 conditional execution. Set either @code{TRUE_EXPR} or @code{FALSE_EXPR}
8802 to a null pointer if the tests cannot be converted.
8804 @findex IFCVT_MODIFY_INSN
8805 @item IFCVT_MODIFY_INSN
8806 A C expression to modify the @code{PATTERN} of an @code{INSN} that is to
8807 be converted to conditional execution format.
8809 @findex IFCVT_MODIFY_FINAL
8810 @item IFCVT_MODIFY_FINAL
8811 A C expression to perform any final machine dependent modifications in
8812 converting code to conditional execution in the basic blocks
8813 @code{TEST_BB}, @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB}.
8815 @findex IFCVT_MODIFY_CANCEL
8816 @item IFCVT_MODIFY_CANCEL
8817 A C expression to cancel any machine dependent modifications in
8818 converting code to conditional execution in the basic blocks
8819 @code{TEST_BB}, @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB}.
8822 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
8823 Define this hook if you have any machine-specific built-in functions
8824 that need to be defined. It should be a function that performs the
8827 Machine specific built-in functions can be useful to expand special machine
8828 instructions that would otherwise not normally be generated because
8829 they have no equivalent in the source language (for example, SIMD vector
8830 instructions or prefetch instructions).
8832 To create a built-in function, call the function @code{builtin_function}
8833 which is defined by the language front end. You can use any type nodes set
8834 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
8835 only language front ends that use those two functions will call
8836 @samp{TARGET_INIT_BUILTINS}.
8839 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
8841 Expand a call to a machine specific built-in function that was set up by
8842 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
8843 function call; the result should go to @var{target} if that is
8844 convenient, and have mode @var{mode} if that is convenient.
8845 @var{subtarget} may be used as the target for computing one of
8846 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
8847 ignored. This function should return the result of the call to the
8852 @findex MD_CAN_REDIRECT_BRANCH
8853 @item MD_CAN_REDIRECT_BRANCH(@var{branch1}, @var{branch2})
8855 Take a branch insn in @var{branch1} and another in @var{branch2}.
8856 Return true if redirecting @var{branch1} to the destination of
8857 @var{branch2} is possible.
8859 On some targets, branches may have a limited range. Optimizing the
8860 filling of delay slots can result in branches being redirected, and this
8861 may in turn cause a branch offset to overflow.
8863 @findex ALLOCATE_INITIAL_VALUE
8864 @item ALLOCATE_INITIAL_VALUE(@var{hard_reg})
8866 When the initial value of a hard register has been copied in a pseudo
8867 register, it is often not necessary to actually allocate another register
8868 to this pseudo register, because the original hard register or a stack slot
8869 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
8870 defined, is called at the start of register allocation once for each
8871 hard register that had its initial value copied by using
8872 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
8873 Possible values are @code{NULL_RTX}, if you don't want
8874 to do any special allocation, a @code{REG} rtx---that would typically be
8875 the hard register itself, if it is known not to be clobbered---or a
8877 If you are returning a @code{MEM}, this is only a hint for the allocator;
8878 it might decide to use another register anyways.
8879 You may use @code{current_function_leaf_function} in the definition of the
8880 macro, functions that use @code{REG_N_SETS}, to determine if the hard
8881 register in question will not be clobbered.
8883 @findex TARGET_OBJECT_SUFFIX
8884 @item TARGET_OBJECT_SUFFIX
8885 Define this macro to be a C string representing the suffix for object
8886 files on your target machine. If you do not define this macro, GCC will
8887 use @samp{.o} as the suffix for object files.
8889 @findex TARGET_EXECUTABLE_SUFFIX
8890 @item TARGET_EXECUTABLE_SUFFIX
8891 Define this macro to be a C string representing the suffix to be
8892 automatically added to executable files on your target machine. If you
8893 do not define this macro, GCC will use the null string as the suffix for
8896 @findex COLLECT_EXPORT_LIST
8897 @item COLLECT_EXPORT_LIST
8898 If defined, @code{collect2} will scan the individual object files
8899 specified on its command line and create an export list for the linker.
8900 Define this macro for systems like AIX, where the linker discards
8901 object files that are not referenced from @code{main} and uses export
8906 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
8907 This target hook returns @code{true} past the point in which new jump
8908 instructions could be created. On machines that require a register for
8909 every jump such as the SHmedia ISA of SH5, this point would typically be
8910 reload, so this target hook should be defined to a function such as:
8914 cannot_modify_jumps_past_reload_p ()
8916 return (reload_completed || reload_in_progress);