1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004 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 * PCH Target:: Validity checking for precompiled headers.
54 * C++ ABI:: Controlling C++ ABI changes.
55 * Misc:: Everything else.
58 @node Target Structure
59 @section The Global @code{targetm} Variable
61 @cindex target functions
63 @deftypevar {struct gcc_target} targetm
64 The target @file{.c} file must define the global @code{targetm} variable
65 which contains pointers to functions and data relating to the target
66 machine. The variable is declared in @file{target.h};
67 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
68 used to initialize the variable, and macros for the default initializers
69 for elements of the structure. The @file{.c} file should override those
70 macros for which the default definition is inappropriate. For example:
73 #include "target-def.h"
75 /* @r{Initialize the GCC target structure.} */
77 #undef TARGET_COMP_TYPE_ATTRIBUTES
78 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
80 struct gcc_target targetm = TARGET_INITIALIZER;
84 Where a macro should be defined in the @file{.c} file in this manner to
85 form part of the @code{targetm} structure, it is documented below as a
86 ``Target Hook'' with a prototype. Many macros will change in future
87 from being defined in the @file{.h} file to being part of the
88 @code{targetm} structure.
91 @section Controlling the Compilation Driver, @file{gcc}
93 @cindex controlling the compilation driver
95 @c prevent bad page break with this line
96 You can control the compilation driver.
98 @defmac 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
111 @defmac 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
125 @defmac 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.
140 @defmac 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{""}.
147 @defmac 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" @}
167 @defmac DRIVER_SELF_SPECS
168 A list of specs for the driver itself. It should be a suitable
169 initializer for an array of strings, with no surrounding braces.
171 The driver applies these specs to its own command line between loading
172 default @file{specs} files (but not command-line specified ones) and
173 choosing the multilib directory or running any subcommands. It
174 applies them in the order given, so each spec can depend on the
175 options added by earlier ones. It is also possible to remove options
176 using @samp{%<@var{option}} in the usual way.
178 This macro can be useful when a port has several interdependent target
179 options. It provides a way of standardizing the command line so
180 that the other specs are easier to write.
182 Do not define this macro if it does not need to do anything.
185 @defmac OPTION_DEFAULT_SPECS
186 A list of specs used to support configure-time default options (i.e.@:
187 @option{--with} options) in the driver. It should be a suitable initializer
188 for an array of structures, each containing two strings, without the
189 outermost pair of surrounding braces.
191 The first item in the pair is the name of the default. This must match
192 the code in @file{config.gcc} for the target. The second item is a spec
193 to apply if a default with this name was specified. The string
194 @samp{%(VALUE)} in the spec will be replaced by the value of the default
195 everywhere it occurs.
197 The driver will apply these specs to its own command line between loading
198 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
199 the same mechanism as @code{DRIVER_SELF_SPECS}.
201 Do not define this macro if it does not need to do anything.
205 A C string constant that tells the GCC driver program options to
206 pass to CPP@. It can also specify how to translate options you
207 give to GCC into options for GCC to pass to the CPP@.
209 Do not define this macro if it does not need to do anything.
212 @defmac CPLUSPLUS_CPP_SPEC
213 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
214 than C@. If you do not define this macro, then the value of
215 @code{CPP_SPEC} (if any) will be used instead.
219 A C string constant that tells the GCC driver program options to
220 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
222 It can also specify how to translate options you give to GCC into options
223 for GCC to pass to front ends.
225 Do not define this macro if it does not need to do anything.
229 A C string constant that tells the GCC driver program options to
230 pass to @code{cc1plus}. It can also specify how to translate options you
231 give to GCC into options for GCC to pass to the @code{cc1plus}.
233 Do not define this macro if it does not need to do anything.
234 Note that everything defined in CC1_SPEC is already passed to
235 @code{cc1plus} so there is no need to duplicate the contents of
236 CC1_SPEC in CC1PLUS_SPEC@.
240 A C string constant that tells the GCC driver program options to
241 pass to the assembler. It can also specify how to translate options
242 you give to GCC into options for GCC to pass to the assembler.
243 See the file @file{sun3.h} for an example of this.
245 Do not define this macro if it does not need to do anything.
248 @defmac ASM_FINAL_SPEC
249 A C string constant that tells the GCC driver program how to
250 run any programs which cleanup after the normal assembler.
251 Normally, this is not needed. See the file @file{mips.h} for
254 Do not define this macro if it does not need to do anything.
257 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
258 Define this macro, with no value, if the driver should give the assembler
259 an argument consisting of a single dash, @option{-}, to instruct it to
260 read from its standard input (which will be a pipe connected to the
261 output of the compiler proper). This argument is given after any
262 @option{-o} option specifying the name of the output file.
264 If you do not define this macro, the assembler is assumed to read its
265 standard input if given no non-option arguments. If your assembler
266 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
267 see @file{mips.h} for instance.
271 A C string constant that tells the GCC driver program options to
272 pass to the linker. It can also specify how to translate options you
273 give to GCC into options for GCC to pass to the linker.
275 Do not define this macro if it does not need to do anything.
279 Another C string constant used much like @code{LINK_SPEC}. The difference
280 between the two is that @code{LIB_SPEC} is used at the end of the
281 command given to the linker.
283 If this macro is not defined, a default is provided that
284 loads the standard C library from the usual place. See @file{gcc.c}.
288 Another C string constant that tells the GCC driver program
289 how and when to place a reference to @file{libgcc.a} into the
290 linker command line. This constant is placed both before and after
291 the value of @code{LIB_SPEC}.
293 If this macro is not defined, the GCC driver provides a default that
294 passes the string @option{-lgcc} to the linker.
297 @defmac REAL_LIBGCC_SPEC
298 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
299 @code{LIBGCC_SPEC} is not directly used by the driver program but is
300 instead modified to refer to different versions of @file{libgcc.a}
301 depending on the values of the command line flags @code{-static},
302 @code{-shared}, @code{-static-libgcc}, and @code{-shared-libgcc}. On
303 targets where these modifications are inappropriate, define
304 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
305 driver how to place a reference to @file{libgcc} on the link command
306 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
309 @defmac STARTFILE_SPEC
310 Another C string constant used much like @code{LINK_SPEC}. The
311 difference between the two is that @code{STARTFILE_SPEC} is used at
312 the very beginning of the command given to the linker.
314 If this macro is not defined, a default is provided that loads the
315 standard C startup file from the usual place. See @file{gcc.c}.
319 Another C string constant used much like @code{LINK_SPEC}. The
320 difference between the two is that @code{ENDFILE_SPEC} is used at
321 the very end of the command given to the linker.
323 Do not define this macro if it does not need to do anything.
326 @defmac THREAD_MODEL_SPEC
327 GCC @code{-v} will print the thread model GCC was configured to use.
328 However, this doesn't work on platforms that are multilibbed on thread
329 models, such as AIX 4.3. On such platforms, define
330 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
331 blanks that names one of the recognized thread models. @code{%*}, the
332 default value of this macro, will expand to the value of
333 @code{thread_file} set in @file{config.gcc}.
336 @defmac SYSROOT_SUFFIX_SPEC
337 Define this macro to add a suffix to the target sysroot when GCC is
338 configured with a sysroot. This will cause GCC to search for usr/lib,
339 et al, within sysroot+suffix.
342 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
343 Define this macro to add a headers_suffix to the target sysroot when
344 GCC is configured with a sysroot. This will cause GCC to pass the
345 updated sysroot+headers_suffix to CPP, causing it to search for
346 usr/include, et al, within sysroot+headers_suffix.
350 Define this macro to provide additional specifications to put in the
351 @file{specs} file that can be used in various specifications like
354 The definition should be an initializer for an array of structures,
355 containing a string constant, that defines the specification name, and a
356 string constant that provides the specification.
358 Do not define this macro if it does not need to do anything.
360 @code{EXTRA_SPECS} is useful when an architecture contains several
361 related targets, which have various @code{@dots{}_SPECS} which are similar
362 to each other, and the maintainer would like one central place to keep
365 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
366 define either @code{_CALL_SYSV} when the System V calling sequence is
367 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
370 The @file{config/rs6000/rs6000.h} target file defines:
373 #define EXTRA_SPECS \
374 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
376 #define CPP_SYS_DEFAULT ""
379 The @file{config/rs6000/sysv.h} target file defines:
383 "%@{posix: -D_POSIX_SOURCE @} \
384 %@{mcall-sysv: -D_CALL_SYSV @} \
385 %@{!mcall-sysv: %(cpp_sysv_default) @} \
386 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
388 #undef CPP_SYSV_DEFAULT
389 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
392 while the @file{config/rs6000/eabiaix.h} target file defines
393 @code{CPP_SYSV_DEFAULT} as:
396 #undef CPP_SYSV_DEFAULT
397 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
401 @defmac LINK_LIBGCC_SPECIAL
402 Define this macro if the driver program should find the library
403 @file{libgcc.a} itself and should not pass @option{-L} options to the
404 linker. If you do not define this macro, the driver program will pass
405 the argument @option{-lgcc} to tell the linker to do the search and will
406 pass @option{-L} options to it.
409 @defmac LINK_LIBGCC_SPECIAL_1
410 Define this macro if the driver program should find the library
411 @file{libgcc.a}. If you do not define this macro, the driver program will pass
412 the argument @option{-lgcc} to tell the linker to do the search.
413 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
414 not affect @option{-L} options.
417 @defmac LINK_GCC_C_SEQUENCE_SPEC
418 The sequence in which libgcc and libc are specified to the linker.
419 By default this is @code{%G %L %G}.
422 @defmac LINK_COMMAND_SPEC
423 A C string constant giving the complete command line need to execute the
424 linker. When you do this, you will need to update your port each time a
425 change is made to the link command line within @file{gcc.c}. Therefore,
426 define this macro only if you need to completely redefine the command
427 line for invoking the linker and there is no other way to accomplish
428 the effect you need. Overriding this macro may be avoidable by overriding
429 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
432 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
433 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
434 directories from linking commands. Do not give it a nonzero value if
435 removing duplicate search directories changes the linker's semantics.
438 @defmac MULTILIB_DEFAULTS
439 Define this macro as a C expression for the initializer of an array of
440 string to tell the driver program which options are defaults for this
441 target and thus do not need to be handled specially when using
442 @code{MULTILIB_OPTIONS}.
444 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
445 the target makefile fragment or if none of the options listed in
446 @code{MULTILIB_OPTIONS} are set by default.
447 @xref{Target Fragment}.
450 @defmac RELATIVE_PREFIX_NOT_LINKDIR
451 Define this macro to tell @command{gcc} that it should only translate
452 a @option{-B} prefix into a @option{-L} linker option if the prefix
453 indicates an absolute file name.
456 @defmac MD_EXEC_PREFIX
457 If defined, this macro is an additional prefix to try after
458 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
459 when the @option{-b} option is used, or the compiler is built as a cross
460 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
461 to the list of directories used to find the assembler in @file{configure.in}.
464 @defmac STANDARD_STARTFILE_PREFIX
465 Define this macro as a C string constant if you wish to override the
466 standard choice of @code{libdir} as the default prefix to
467 try when searching for startup files such as @file{crt0.o}.
468 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
469 is built as a cross compiler.
472 @defmac STANDARD_STARTFILE_PREFIX_1
473 Define this macro as a C string constant if you wish to override the
474 standard choice of @code{/lib} as a prefix to try after the default prefix
475 when searching for startup files such as @file{crt0.o}.
476 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
477 is built as a cross compiler.
480 @defmac STANDARD_STARTFILE_PREFIX_2
481 Define this macro as a C string constant if you wish to override the
482 standard choice of @code{/lib} as yet another prefix to try after the
483 default prefix when searching for startup files such as @file{crt0.o}.
484 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
485 is built as a cross compiler.
488 @defmac MD_STARTFILE_PREFIX
489 If defined, this macro supplies an additional prefix to try after the
490 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
491 @option{-b} option is used, or when the compiler is built as a cross
495 @defmac MD_STARTFILE_PREFIX_1
496 If defined, this macro supplies yet another prefix to try after the
497 standard prefixes. It is not searched when the @option{-b} option is
498 used, or when the compiler is built as a cross compiler.
501 @defmac INIT_ENVIRONMENT
502 Define this macro as a C string constant if you wish to set environment
503 variables for programs called by the driver, such as the assembler and
504 loader. The driver passes the value of this macro to @code{putenv} to
505 initialize the necessary environment variables.
508 @defmac LOCAL_INCLUDE_DIR
509 Define this macro as a C string constant if you wish to override the
510 standard choice of @file{/usr/local/include} as the default prefix to
511 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
512 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
514 Cross compilers do not search either @file{/usr/local/include} or its
518 @defmac MODIFY_TARGET_NAME
519 Define this macro if you wish to define command-line switches that
520 modify the default target name.
522 For each switch, you can include a string to be appended to the first
523 part of the configuration name or a string to be deleted from the
524 configuration name, if present. The definition should be an initializer
525 for an array of structures. Each array element should have three
526 elements: the switch name (a string constant, including the initial
527 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
528 indicate whether the string should be inserted or deleted, and the string
529 to be inserted or deleted (a string constant).
531 For example, on a machine where @samp{64} at the end of the
532 configuration name denotes a 64-bit target and you want the @option{-32}
533 and @option{-64} switches to select between 32- and 64-bit targets, you would
537 #define MODIFY_TARGET_NAME \
538 @{ @{ "-32", DELETE, "64"@}, \
539 @{"-64", ADD, "64"@}@}
543 @defmac SYSTEM_INCLUDE_DIR
544 Define this macro as a C string constant if you wish to specify a
545 system-specific directory to search for header files before the standard
546 directory. @code{SYSTEM_INCLUDE_DIR} comes before
547 @code{STANDARD_INCLUDE_DIR} in the search order.
549 Cross compilers do not use this macro and do not search the directory
553 @defmac STANDARD_INCLUDE_DIR
554 Define this macro as a C string constant if you wish to override the
555 standard choice of @file{/usr/include} as the default prefix to
556 try when searching for header files.
558 Cross compilers ignore this macro and do not search either
559 @file{/usr/include} or its replacement.
562 @defmac STANDARD_INCLUDE_COMPONENT
563 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
564 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
565 If you do not define this macro, no component is used.
568 @defmac INCLUDE_DEFAULTS
569 Define this macro if you wish to override the entire default search path
570 for include files. For a native compiler, the default search path
571 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
572 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
573 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
574 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
575 and specify private search areas for GCC@. The directory
576 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
578 The definition should be an initializer for an array of structures.
579 Each array element should have four elements: the directory name (a
580 string constant), the component name (also a string constant), a flag
581 for C++-only directories,
582 and a flag showing that the includes in the directory don't need to be
583 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
584 the array with a null element.
586 The component name denotes what GNU package the include file is part of,
587 if any, in all uppercase letters. For example, it might be @samp{GCC}
588 or @samp{BINUTILS}. If the package is part of a vendor-supplied
589 operating system, code the component name as @samp{0}.
591 For example, here is the definition used for VAX/VMS:
594 #define INCLUDE_DEFAULTS \
596 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
597 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
598 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
605 Here is the order of prefixes tried for exec files:
609 Any prefixes specified by the user with @option{-B}.
612 The environment variable @code{GCC_EXEC_PREFIX}, if any.
615 The directories specified by the environment variable @code{COMPILER_PATH}.
618 The macro @code{STANDARD_EXEC_PREFIX}.
621 @file{/usr/lib/gcc/}.
624 The macro @code{MD_EXEC_PREFIX}, if any.
627 Here is the order of prefixes tried for startfiles:
631 Any prefixes specified by the user with @option{-B}.
634 The environment variable @code{GCC_EXEC_PREFIX}, if any.
637 The directories specified by the environment variable @code{LIBRARY_PATH}
638 (or port-specific name; native only, cross compilers do not use this).
641 The macro @code{STANDARD_EXEC_PREFIX}.
644 @file{/usr/lib/gcc/}.
647 The macro @code{MD_EXEC_PREFIX}, if any.
650 The macro @code{MD_STARTFILE_PREFIX}, if any.
653 The macro @code{STANDARD_STARTFILE_PREFIX}.
662 @node Run-time Target
663 @section Run-time Target Specification
664 @cindex run-time target specification
665 @cindex predefined macros
666 @cindex target specifications
668 @c prevent bad page break with this line
669 Here are run-time target specifications.
671 @defmac TARGET_CPU_CPP_BUILTINS ()
672 This function-like macro expands to a block of code that defines
673 built-in preprocessor macros and assertions for the target cpu, using
674 the functions @code{builtin_define}, @code{builtin_define_std} and
675 @code{builtin_assert}. When the front end
676 calls this macro it provides a trailing semicolon, and since it has
677 finished command line option processing your code can use those
680 @code{builtin_assert} takes a string in the form you pass to the
681 command-line option @option{-A}, such as @code{cpu=mips}, and creates
682 the assertion. @code{builtin_define} takes a string in the form
683 accepted by option @option{-D} and unconditionally defines the macro.
685 @code{builtin_define_std} takes a string representing the name of an
686 object-like macro. If it doesn't lie in the user's namespace,
687 @code{builtin_define_std} defines it unconditionally. Otherwise, it
688 defines a version with two leading underscores, and another version
689 with two leading and trailing underscores, and defines the original
690 only if an ISO standard was not requested on the command line. For
691 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
692 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
693 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
694 defines only @code{_ABI64}.
696 You can also test for the C dialect being compiled. The variable
697 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
698 or @code{clk_objective_c}. Note that if we are preprocessing
699 assembler, this variable will be @code{clk_c} but the function-like
700 macro @code{preprocessing_asm_p()} will return true, so you might want
701 to check for that first. If you need to check for strict ANSI, the
702 variable @code{flag_iso} can be used. The function-like macro
703 @code{preprocessing_trad_p()} can be used to check for traditional
707 @defmac TARGET_OS_CPP_BUILTINS ()
708 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
709 and is used for the target operating system instead.
712 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
713 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
714 and is used for the target object format. @file{elfos.h} uses this
715 macro to define @code{__ELF__}, so you probably do not need to define
719 @deftypevar {extern int} target_flags
720 This declaration should be present.
723 @cindex optional hardware or system features
724 @cindex features, optional, in system conventions
726 @defmac TARGET_@var{featurename}
727 This series of macros is to allow compiler command arguments to
728 enable or disable the use of optional features of the target machine.
729 For example, one machine description serves both the 68000 and
730 the 68020; a command argument tells the compiler whether it should
731 use 68020-only instructions or not. This command argument works
732 by means of a macro @code{TARGET_68020} that tests a bit in
735 Define a macro @code{TARGET_@var{featurename}} for each such option.
736 Its definition should test a bit in @code{target_flags}. It is
737 recommended that a helper macro @code{MASK_@var{featurename}}
738 is defined for each bit-value to test, and used in
739 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
743 #define TARGET_MASK_68020 1
744 #define TARGET_68020 (target_flags & MASK_68020)
747 One place where these macros are used is in the condition-expressions
748 of instruction patterns. Note how @code{TARGET_68020} appears
749 frequently in the 68000 machine description file, @file{m68k.md}.
750 Another place they are used is in the definitions of the other
751 macros in the @file{@var{machine}.h} file.
754 @defmac TARGET_SWITCHES
755 This macro defines names of command options to set and clear
756 bits in @code{target_flags}. Its definition is an initializer
757 with a subgrouping for each command option.
759 Each subgrouping contains a string constant, that defines the option
760 name, a number, which contains the bits to set in
761 @code{target_flags}, and a second string which is the description
762 displayed by @option{--help}. If the number is negative then the bits specified
763 by the number are cleared instead of being set. If the description
764 string is present but empty, then no help information will be displayed
765 for that option, but it will not count as an undocumented option. The
766 actual option name is made by appending @samp{-m} to the specified name.
767 Non-empty description strings should be marked with @code{N_(@dots{})} for
768 @command{xgettext}. Please do not mark empty strings because the empty
769 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
770 of the message catalog with meta information, not the empty string.
772 In addition to the description for @option{--help},
773 more detailed documentation for each option should be added to
776 One of the subgroupings should have a null string. The number in
777 this grouping is the default value for @code{target_flags}. Any
778 target options act starting with that value.
780 Here is an example which defines @option{-m68000} and @option{-m68020}
781 with opposite meanings, and picks the latter as the default:
784 #define TARGET_SWITCHES \
785 @{ @{ "68020", MASK_68020, "" @}, \
786 @{ "68000", -MASK_68020, \
787 N_("Compile for the 68000") @}, \
788 @{ "", MASK_68020, "" @}, \
793 @defmac TARGET_OPTIONS
794 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
795 options that have values. Its definition is an initializer with a
796 subgrouping for each command option.
798 Each subgrouping contains a string constant, that defines the option
799 name, the address of a variable, a description string, and a value.
800 Non-empty description strings should be marked with @code{N_(@dots{})}
801 for @command{xgettext}. Please do not mark empty strings because the
802 empty string is reserved by GNU gettext. @code{gettext("")} returns the
803 header entry of the message catalog with meta information, not the empty
806 If the value listed in the table is @code{NULL}, then the variable, type
807 @code{char *}, is set to the variable part of the given option if the
808 fixed part matches. In other words, if the first part of the option
809 matches what's in the table, the variable will be set to point to the
810 rest of the option. This allows the user to specify a value for that
811 option. The actual option name is made by appending @samp{-m} to the
812 specified name. Again, each option should also be documented in
815 If the value listed in the table is non-@code{NULL}, then the option
816 must match the option in the table exactly (with @samp{-m}), and the
817 variable is set to point to the value listed in the table.
819 Here is an example which defines @option{-mshort-data-@var{number}}. If the
820 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
821 will be set to the string @code{"512"}.
824 extern char *m88k_short_data;
825 #define TARGET_OPTIONS \
826 @{ @{ "short-data-", &m88k_short_data, \
827 N_("Specify the size of the short data section"), 0 @} @}
830 Here is a variant of the above that allows the user to also specify
831 just @option{-mshort-data} where a default of @code{"64"} is used.
834 extern char *m88k_short_data;
835 #define TARGET_OPTIONS \
836 @{ @{ "short-data-", &m88k_short_data, \
837 N_("Specify the size of the short data section"), 0 @} \
838 @{ "short-data", &m88k_short_data, "", "64" @},
842 Here is an example which defines @option{-mno-alu}, @option{-malu1}, and
843 @option{-malu2} as a three-state switch, along with suitable macros for
844 checking the state of the option (documentation is elided for brevity).
848 char *chip_alu = ""; /* Specify default here. */
851 extern char *chip_alu;
852 #define TARGET_OPTIONS \
853 @{ @{ "no-alu", &chip_alu, "", "" @}, \
854 @{ "alu1", &chip_alu, "", "1" @}, \
855 @{ "alu2", &chip_alu, "", "2" @}, @}
856 #define TARGET_ALU (chip_alu[0] != '\0')
857 #define TARGET_ALU1 (chip_alu[0] == '1')
858 #define TARGET_ALU2 (chip_alu[0] == '2')
862 @defmac TARGET_VERSION
863 This macro is a C statement to print on @code{stderr} a string
864 describing the particular machine description choice. Every machine
865 description should define @code{TARGET_VERSION}. For example:
869 #define TARGET_VERSION \
870 fprintf (stderr, " (68k, Motorola syntax)");
872 #define TARGET_VERSION \
873 fprintf (stderr, " (68k, MIT syntax)");
878 @defmac OVERRIDE_OPTIONS
879 Sometimes certain combinations of command options do not make sense on
880 a particular target machine. You can define a macro
881 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
882 defined, is executed once just after all the command options have been
885 Don't use this macro to turn on various extra optimizations for
886 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
889 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
890 Some machines may desire to change what optimizations are performed for
891 various optimization levels. This macro, if defined, is executed once
892 just after the optimization level is determined and before the remainder
893 of the command options have been parsed. Values set in this macro are
894 used as the default values for the other command line options.
896 @var{level} is the optimization level specified; 2 if @option{-O2} is
897 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
899 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
901 You should not use this macro to change options that are not
902 machine-specific. These should uniformly selected by the same
903 optimization level on all supported machines. Use this macro to enable
904 machine-specific optimizations.
906 @strong{Do not examine @code{write_symbols} in
907 this macro!} The debugging options are not supposed to alter the
911 @defmac CAN_DEBUG_WITHOUT_FP
912 Define this macro if debugging can be performed even without a frame
913 pointer. If this macro is defined, GCC will turn on the
914 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
917 @node Per-Function Data
918 @section Defining data structures for per-function information.
919 @cindex per-function data
920 @cindex data structures
922 If the target needs to store information on a per-function basis, GCC
923 provides a macro and a couple of variables to allow this. Note, just
924 using statics to store the information is a bad idea, since GCC supports
925 nested functions, so you can be halfway through encoding one function
926 when another one comes along.
928 GCC defines a data structure called @code{struct function} which
929 contains all of the data specific to an individual function. This
930 structure contains a field called @code{machine} whose type is
931 @code{struct machine_function *}, which can be used by targets to point
932 to their own specific data.
934 If a target needs per-function specific data it should define the type
935 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
936 This macro should be used to initialize the function pointer
937 @code{init_machine_status}. This pointer is explained below.
939 One typical use of per-function, target specific data is to create an
940 RTX to hold the register containing the function's return address. This
941 RTX can then be used to implement the @code{__builtin_return_address}
942 function, for level 0.
944 Note---earlier implementations of GCC used a single data area to hold
945 all of the per-function information. Thus when processing of a nested
946 function began the old per-function data had to be pushed onto a
947 stack, and when the processing was finished, it had to be popped off the
948 stack. GCC used to provide function pointers called
949 @code{save_machine_status} and @code{restore_machine_status} to handle
950 the saving and restoring of the target specific information. Since the
951 single data area approach is no longer used, these pointers are no
954 @defmac INIT_EXPANDERS
955 Macro called to initialize any target specific information. This macro
956 is called once per function, before generation of any RTL has begun.
957 The intention of this macro is to allow the initialization of the
958 function pointer @code{init_machine_status}.
961 @deftypevar {void (*)(struct function *)} init_machine_status
962 If this function pointer is non-@code{NULL} it will be called once per
963 function, before function compilation starts, in order to allow the
964 target to perform any target specific initialization of the
965 @code{struct function} structure. It is intended that this would be
966 used to initialize the @code{machine} of that structure.
968 @code{struct machine_function} structures are expected to be freed by GC.
969 Generally, any memory that they reference must be allocated by using
970 @code{ggc_alloc}, including the structure itself.
974 @section Storage Layout
975 @cindex storage layout
977 Note that the definitions of the macros in this table which are sizes or
978 alignments measured in bits do not need to be constant. They can be C
979 expressions that refer to static variables, such as the @code{target_flags}.
980 @xref{Run-time Target}.
982 @defmac BITS_BIG_ENDIAN
983 Define this macro to have the value 1 if the most significant bit in a
984 byte has the lowest number; otherwise define it to have the value zero.
985 This means that bit-field instructions count from the most significant
986 bit. If the machine has no bit-field instructions, then this must still
987 be defined, but it doesn't matter which value it is defined to. This
988 macro need not be a constant.
990 This macro does not affect the way structure fields are packed into
991 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
994 @defmac BYTES_BIG_ENDIAN
995 Define this macro to have the value 1 if the most significant byte in a
996 word has the lowest number. This macro need not be a constant.
999 @defmac WORDS_BIG_ENDIAN
1000 Define this macro to have the value 1 if, in a multiword object, the
1001 most significant word has the lowest number. This applies to both
1002 memory locations and registers; GCC fundamentally assumes that the
1003 order of words in memory is the same as the order in registers. This
1004 macro need not be a constant.
1007 @defmac LIBGCC2_WORDS_BIG_ENDIAN
1008 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
1009 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
1010 used only when compiling @file{libgcc2.c}. Typically the value will be set
1011 based on preprocessor defines.
1014 @defmac FLOAT_WORDS_BIG_ENDIAN
1015 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
1016 @code{TFmode} floating point numbers are stored in memory with the word
1017 containing the sign bit at the lowest address; otherwise define it to
1018 have the value 0. This macro need not be a constant.
1020 You need not define this macro if the ordering is the same as for
1021 multi-word integers.
1024 @defmac BITS_PER_UNIT
1025 Define this macro to be the number of bits in an addressable storage
1026 unit (byte). If you do not define this macro the default is 8.
1029 @defmac BITS_PER_WORD
1030 Number of bits in a word. If you do not define this macro, the default
1031 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
1034 @defmac MAX_BITS_PER_WORD
1035 Maximum number of bits in a word. If this is undefined, the default is
1036 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
1037 largest value that @code{BITS_PER_WORD} can have at run-time.
1040 @defmac UNITS_PER_WORD
1041 Number of storage units in a word; normally 4.
1044 @defmac MIN_UNITS_PER_WORD
1045 Minimum number of units in a word. If this is undefined, the default is
1046 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
1047 smallest value that @code{UNITS_PER_WORD} can have at run-time.
1050 @defmac POINTER_SIZE
1051 Width of a pointer, in bits. You must specify a value no wider than the
1052 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1053 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1054 a value the default is @code{BITS_PER_WORD}.
1057 @defmac POINTERS_EXTEND_UNSIGNED
1058 A C expression whose value is greater than zero if pointers that need to be
1059 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
1060 be zero-extended and zero if they are to be sign-extended. If the value
1061 is less then zero then there must be an "ptr_extend" instruction that
1062 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
1064 You need not define this macro if the @code{POINTER_SIZE} is equal
1065 to the width of @code{Pmode}.
1068 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1069 A macro to update @var{m} and @var{unsignedp} when an object whose type
1070 is @var{type} and which has the specified mode and signedness is to be
1071 stored in a register. This macro is only called when @var{type} is a
1074 On most RISC machines, which only have operations that operate on a full
1075 register, define this macro to set @var{m} to @code{word_mode} if
1076 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1077 cases, only integer modes should be widened because wider-precision
1078 floating-point operations are usually more expensive than their narrower
1081 For most machines, the macro definition does not change @var{unsignedp}.
1082 However, some machines, have instructions that preferentially handle
1083 either signed or unsigned quantities of certain modes. For example, on
1084 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1085 sign-extend the result to 64 bits. On such machines, set
1086 @var{unsignedp} according to which kind of extension is more efficient.
1088 Do not define this macro if it would never modify @var{m}.
1091 @defmac PROMOTE_FUNCTION_MODE
1092 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
1093 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1094 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1096 The default is @code{PROMOTE_MODE}.
1099 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1100 This target hook should return @code{true} if the promotion described by
1101 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1105 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1106 This target hook should return @code{true} if the promotion described by
1107 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1110 If this target hook returns @code{true}, @code{FUNCTION_VALUE} must
1111 perform the same promotions done by @code{PROMOTE_FUNCTON_MODE}.
1114 @defmac PARM_BOUNDARY
1115 Normal alignment required for function parameters on the stack, in
1116 bits. All stack parameters receive at least this much alignment
1117 regardless of data type. On most machines, this is the same as the
1121 @defmac STACK_BOUNDARY
1122 Define this macro to the minimum alignment enforced by hardware for the
1123 stack pointer on this machine. The definition is a C expression for the
1124 desired alignment (measured in bits). This value is used as a default
1125 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1126 this should be the same as @code{PARM_BOUNDARY}.
1129 @defmac PREFERRED_STACK_BOUNDARY
1130 Define this macro if you wish to preserve a certain alignment for the
1131 stack pointer, greater than what the hardware enforces. The definition
1132 is a C expression for the desired alignment (measured in bits). This
1133 macro must evaluate to a value equal to or larger than
1134 @code{STACK_BOUNDARY}.
1137 @defmac FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1138 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1139 not guaranteed by the runtime and we should emit code to align the stack
1140 at the beginning of @code{main}.
1142 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1143 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1144 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1145 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1146 be momentarily unaligned while pushing arguments.
1149 @defmac FUNCTION_BOUNDARY
1150 Alignment required for a function entry point, in bits.
1153 @defmac BIGGEST_ALIGNMENT
1154 Biggest alignment that any data type can require on this machine, in bits.
1157 @defmac MINIMUM_ATOMIC_ALIGNMENT
1158 If defined, the smallest alignment, in bits, that can be given to an
1159 object that can be referenced in one operation, without disturbing any
1160 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1161 on machines that don't have byte or half-word store operations.
1164 @defmac BIGGEST_FIELD_ALIGNMENT
1165 Biggest alignment that any structure or union field can require on this
1166 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1167 structure and union fields only, unless the field alignment has been set
1168 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1171 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1172 An expression for the alignment of a structure field @var{field} if the
1173 alignment computed in the usual way (including applying of
1174 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1175 alignment) is @var{computed}. It overrides alignment only if the
1176 field alignment has not been set by the
1177 @code{__attribute__ ((aligned (@var{n})))} construct.
1180 @defmac MAX_OFILE_ALIGNMENT
1181 Biggest alignment supported by the object file format of this machine.
1182 Use this macro to limit the alignment which can be specified using the
1183 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1184 the default value is @code{BIGGEST_ALIGNMENT}.
1187 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1188 If defined, a C expression to compute the alignment for a variable in
1189 the static store. @var{type} is the data type, and @var{basic-align} is
1190 the alignment that the object would ordinarily have. The value of this
1191 macro is used instead of that alignment to align the object.
1193 If this macro is not defined, then @var{basic-align} is used.
1196 One use of this macro is to increase alignment of medium-size data to
1197 make it all fit in fewer cache lines. Another is to cause character
1198 arrays to be word-aligned so that @code{strcpy} calls that copy
1199 constants to character arrays can be done inline.
1202 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1203 If defined, a C expression to compute the alignment given to a constant
1204 that is being placed in memory. @var{constant} is the constant and
1205 @var{basic-align} is the alignment that the object would ordinarily
1206 have. The value of this macro is used instead of that alignment to
1209 If this macro is not defined, then @var{basic-align} is used.
1211 The typical use of this macro is to increase alignment for string
1212 constants to be word aligned so that @code{strcpy} calls that copy
1213 constants can be done inline.
1216 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1217 If defined, a C expression to compute the alignment for a variable in
1218 the local store. @var{type} is the data type, and @var{basic-align} is
1219 the alignment that the object would ordinarily have. The value of this
1220 macro is used instead of that alignment to align the object.
1222 If this macro is not defined, then @var{basic-align} is used.
1224 One use of this macro is to increase alignment of medium-size data to
1225 make it all fit in fewer cache lines.
1228 @defmac EMPTY_FIELD_BOUNDARY
1229 Alignment in bits to be given to a structure bit-field that follows an
1230 empty field such as @code{int : 0;}.
1232 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1235 @defmac STRUCTURE_SIZE_BOUNDARY
1236 Number of bits which any structure or union's size must be a multiple of.
1237 Each structure or union's size is rounded up to a multiple of this.
1239 If you do not define this macro, the default is the same as
1240 @code{BITS_PER_UNIT}.
1243 @defmac STRICT_ALIGNMENT
1244 Define this macro to be the value 1 if instructions will fail to work
1245 if given data not on the nominal alignment. If instructions will merely
1246 go slower in that case, define this macro as 0.
1249 @defmac PCC_BITFIELD_TYPE_MATTERS
1250 Define this if you wish to imitate the way many other C compilers handle
1251 alignment of bit-fields and the structures that contain them.
1253 The behavior is that the type written for a named bit-field (@code{int},
1254 @code{short}, or other integer type) imposes an alignment for the entire
1255 structure, as if the structure really did contain an ordinary field of
1256 that type. In addition, the bit-field is placed within the structure so
1257 that it would fit within such a field, not crossing a boundary for it.
1259 Thus, on most machines, a named bit-field whose type is written as
1260 @code{int} would not cross a four-byte boundary, and would force
1261 four-byte alignment for the whole structure. (The alignment used may
1262 not be four bytes; it is controlled by the other alignment parameters.)
1264 An unnamed bit-field will not affect the alignment of the containing
1267 If the macro is defined, its definition should be a C expression;
1268 a nonzero value for the expression enables this behavior.
1270 Note that if this macro is not defined, or its value is zero, some
1271 bit-fields may cross more than one alignment boundary. The compiler can
1272 support such references if there are @samp{insv}, @samp{extv}, and
1273 @samp{extzv} insns that can directly reference memory.
1275 The other known way of making bit-fields work is to define
1276 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1277 Then every structure can be accessed with fullwords.
1279 Unless the machine has bit-field instructions or you define
1280 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1281 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1283 If your aim is to make GCC use the same conventions for laying out
1284 bit-fields as are used by another compiler, here is how to investigate
1285 what the other compiler does. Compile and run this program:
1304 printf ("Size of foo1 is %d\n",
1305 sizeof (struct foo1));
1306 printf ("Size of foo2 is %d\n",
1307 sizeof (struct foo2));
1312 If this prints 2 and 5, then the compiler's behavior is what you would
1313 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1316 @defmac BITFIELD_NBYTES_LIMITED
1317 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1318 to aligning a bit-field within the structure.
1321 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1322 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1323 whether unnamed bitfields affect the alignment of the containing
1324 structure. The hook should return true if the structure should inherit
1325 the alignment requirements of an unnamed bitfield's type.
1328 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1329 Return 1 if a structure or array containing @var{field} should be accessed using
1332 If @var{field} is the only field in the structure, @var{mode} is its
1333 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1334 case where structures of one field would require the structure's mode to
1335 retain the field's mode.
1337 Normally, this is not needed. See the file @file{c4x.h} for an example
1338 of how to use this macro to prevent a structure having a floating point
1339 field from being accessed in an integer mode.
1342 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1343 Define this macro as an expression for the alignment of a type (given
1344 by @var{type} as a tree node) if the alignment computed in the usual
1345 way is @var{computed} and the alignment explicitly specified was
1348 The default is to use @var{specified} if it is larger; otherwise, use
1349 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1352 @defmac MAX_FIXED_MODE_SIZE
1353 An integer expression for the size in bits of the largest integer
1354 machine mode that should actually be used. All integer machine modes of
1355 this size or smaller can be used for structures and unions with the
1356 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1357 (DImode)} is assumed.
1360 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1361 If defined, an expression of type @code{enum machine_mode} that
1362 specifies the mode of the save area operand of a
1363 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1364 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1365 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1366 having its mode specified.
1368 You need not define this macro if it always returns @code{Pmode}. You
1369 would most commonly define this macro if the
1370 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1374 @defmac STACK_SIZE_MODE
1375 If defined, an expression of type @code{enum machine_mode} that
1376 specifies the mode of the size increment operand of an
1377 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1379 You need not define this macro if it always returns @code{word_mode}.
1380 You would most commonly define this macro if the @code{allocate_stack}
1381 pattern needs to support both a 32- and a 64-bit mode.
1384 @defmac TARGET_FLOAT_FORMAT
1385 A code distinguishing the floating point format of the target machine.
1386 There are four defined values:
1389 @item IEEE_FLOAT_FORMAT
1390 This code indicates IEEE floating point. It is the default; there is no
1391 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1393 @item VAX_FLOAT_FORMAT
1394 This code indicates the ``F float'' (for @code{float}) and ``D float''
1395 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1397 @item IBM_FLOAT_FORMAT
1398 This code indicates the format used on the IBM System/370.
1400 @item C4X_FLOAT_FORMAT
1401 This code indicates the format used on the TMS320C3x/C4x.
1404 If your target uses a floating point format other than these, you must
1405 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1406 it to @file{real.c}.
1408 The ordering of the component words of floating point values stored in
1409 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1412 @defmac MODE_HAS_NANS (@var{mode})
1413 When defined, this macro should be true if @var{mode} has a NaN
1414 representation. The compiler assumes that NaNs are not equal to
1415 anything (including themselves) and that addition, subtraction,
1416 multiplication and division all return NaNs when one operand is
1419 By default, this macro is true if @var{mode} is a floating-point
1420 mode and the target floating-point format is IEEE@.
1423 @defmac MODE_HAS_INFINITIES (@var{mode})
1424 This macro should be true if @var{mode} can represent infinity. At
1425 present, the compiler uses this macro to decide whether @samp{x - x}
1426 is always defined. By default, the macro is true when @var{mode}
1427 is a floating-point mode and the target format is IEEE@.
1430 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1431 True if @var{mode} distinguishes between positive and negative zero.
1432 The rules are expected to follow the IEEE standard:
1436 @samp{x + x} has the same sign as @samp{x}.
1439 If the sum of two values with opposite sign is zero, the result is
1440 positive for all rounding modes expect towards @minus{}infinity, for
1441 which it is negative.
1444 The sign of a product or quotient is negative when exactly one
1445 of the operands is negative.
1448 The default definition is true if @var{mode} is a floating-point
1449 mode and the target format is IEEE@.
1452 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1453 If defined, this macro should be true for @var{mode} if it has at
1454 least one rounding mode in which @samp{x} and @samp{-x} can be
1455 rounded to numbers of different magnitude. Two such modes are
1456 towards @minus{}infinity and towards +infinity.
1458 The default definition of this macro is true if @var{mode} is
1459 a floating-point mode and the target format is IEEE@.
1462 @defmac ROUND_TOWARDS_ZERO
1463 If defined, this macro should be true if the prevailing rounding
1464 mode is towards zero. A true value has the following effects:
1468 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1471 @file{libgcc.a}'s floating-point emulator will round towards zero
1472 rather than towards nearest.
1475 The compiler's floating-point emulator will round towards zero after
1476 doing arithmetic, and when converting from the internal float format to
1480 The macro does not affect the parsing of string literals. When the
1481 primary rounding mode is towards zero, library functions like
1482 @code{strtod} might still round towards nearest, and the compiler's
1483 parser should behave like the target's @code{strtod} where possible.
1485 Not defining this macro is equivalent to returning zero.
1488 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1489 This macro should return true if floats with @var{size}
1490 bits do not have a NaN or infinity representation, but use the largest
1491 exponent for normal numbers instead.
1493 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1494 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1495 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1496 floating-point arithmetic.
1498 The default definition of this macro returns false for all sizes.
1501 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1502 This target hook should return @code{true} a vector is opaque. That
1503 is, if no cast is needed when copying a vector value of type
1504 @var{type} into another vector lvalue of the same size. Vector opaque
1505 types cannot be initialized. The default is that there are no such
1509 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1510 This target hook returns @code{true} if bit-fields in the given
1511 @var{record_type} are to be laid out following the rules of Microsoft
1512 Visual C/C++, namely: (i) a bit-field won't share the same storage
1513 unit with the previous bit-field if their underlying types have
1514 different sizes, and the bit-field will be aligned to the highest
1515 alignment of the underlying types of itself and of the previous
1516 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1517 the whole enclosing structure, even if it is unnamed; except that
1518 (iii) a zero-sized bit-field will be disregarded unless it follows
1519 another bit-field of nonzero size. If this hook returns @code{true},
1520 other macros that control bit-field layout are ignored.
1522 When a bit-field is inserted into a packed record, the whole size
1523 of the underlying type is used by one or more same-size adjacent
1524 bit-fields (that is, if its long:3, 32 bits is used in the record,
1525 and any additional adjacent long bit-fields are packed into the same
1526 chunk of 32 bits. However, if the size changes, a new field of that
1527 size is allocated). In an unpacked record, this is the same as using
1528 alignment, but not equivalent when packing.
1530 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1531 the latter will take precedence. If @samp{__attribute__((packed))} is
1532 used on a single field when MS bit-fields are in use, it will take
1533 precedence for that field, but the alignment of the rest of the structure
1534 may affect its placement.
1537 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_FUNDAMENTAL_TYPE (tree @var{type})
1538 If your target defines any fundamental types, define this hook to
1539 return the appropriate encoding for these types as part of a C++
1540 mangled name. The @var{type} argument is the tree structure
1541 representing the type to be mangled. The hook may be applied to trees
1542 which are not target-specific fundamental types; it should return
1543 @code{NULL} for all such types, as well as arguments it does not
1544 recognize. If the return value is not @code{NULL}, it must point to
1545 a statically-allocated string constant.
1547 Target-specific fundamental types might be new fundamental types or
1548 qualified versions of ordinary fundamental types. Encode new
1549 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1550 is the name used for the type in source code, and @var{n} is the
1551 length of @var{name} in decimal. Encode qualified versions of
1552 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1553 @var{name} is the name used for the type qualifier in source code,
1554 @var{n} is the length of @var{name} as above, and @var{code} is the
1555 code used to represent the unqualified version of this type. (See
1556 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1557 codes.) In both cases the spaces are for clarity; do not include any
1558 spaces in your string.
1560 The default version of this hook always returns @code{NULL}, which is
1561 appropriate for a target that does not define any new fundamental
1566 @section Layout of Source Language Data Types
1568 These macros define the sizes and other characteristics of the standard
1569 basic data types used in programs being compiled. Unlike the macros in
1570 the previous section, these apply to specific features of C and related
1571 languages, rather than to fundamental aspects of storage layout.
1573 @defmac INT_TYPE_SIZE
1574 A C expression for the size in bits of the type @code{int} on the
1575 target machine. If you don't define this, the default is one word.
1578 @defmac SHORT_TYPE_SIZE
1579 A C expression for the size in bits of the type @code{short} on the
1580 target machine. If you don't define this, the default is half a word.
1581 (If this would be less than one storage unit, it is rounded up to one
1585 @defmac LONG_TYPE_SIZE
1586 A C expression for the size in bits of the type @code{long} on the
1587 target machine. If you don't define this, the default is one word.
1590 @defmac ADA_LONG_TYPE_SIZE
1591 On some machines, the size used for the Ada equivalent of the type
1592 @code{long} by a native Ada compiler differs from that used by C. In
1593 that situation, define this macro to be a C expression to be used for
1594 the size of that type. If you don't define this, the default is the
1595 value of @code{LONG_TYPE_SIZE}.
1598 @defmac LONG_LONG_TYPE_SIZE
1599 A C expression for the size in bits of the type @code{long long} on the
1600 target machine. If you don't define this, the default is two
1601 words. If you want to support GNU Ada on your machine, the value of this
1602 macro must be at least 64.
1605 @defmac CHAR_TYPE_SIZE
1606 A C expression for the size in bits of the type @code{char} on the
1607 target machine. If you don't define this, the default is
1608 @code{BITS_PER_UNIT}.
1611 @defmac BOOL_TYPE_SIZE
1612 A C expression for the size in bits of the C++ type @code{bool} and
1613 C99 type @code{_Bool} on the target machine. If you don't define
1614 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1617 @defmac FLOAT_TYPE_SIZE
1618 A C expression for the size in bits of the type @code{float} on the
1619 target machine. If you don't define this, the default is one word.
1622 @defmac DOUBLE_TYPE_SIZE
1623 A C expression for the size in bits of the type @code{double} on the
1624 target machine. If you don't define this, the default is two
1628 @defmac LONG_DOUBLE_TYPE_SIZE
1629 A C expression for the size in bits of the type @code{long double} on
1630 the target machine. If you don't define this, the default is two
1634 @defmac TARGET_FLT_EVAL_METHOD
1635 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1636 assuming, if applicable, that the floating-point control word is in its
1637 default state. If you do not define this macro the value of
1638 @code{FLT_EVAL_METHOD} will be zero.
1641 @defmac WIDEST_HARDWARE_FP_SIZE
1642 A C expression for the size in bits of the widest floating-point format
1643 supported by the hardware. If you define this macro, you must specify a
1644 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1645 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1649 @defmac DEFAULT_SIGNED_CHAR
1650 An expression whose value is 1 or 0, according to whether the type
1651 @code{char} should be signed or unsigned by default. The user can
1652 always override this default with the options @option{-fsigned-char}
1653 and @option{-funsigned-char}.
1656 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1657 This target hook should return true if the compiler should give an
1658 @code{enum} type only as many bytes as it takes to represent the range
1659 of possible values of that type. It should return false if all
1660 @code{enum} types should be allocated like @code{int}.
1662 The default is to return false.
1666 A C expression for a string describing the name of the data type to use
1667 for size values. The typedef name @code{size_t} is defined using the
1668 contents of the string.
1670 The string can contain more than one keyword. If so, separate them with
1671 spaces, and write first any length keyword, then @code{unsigned} if
1672 appropriate, and finally @code{int}. The string must exactly match one
1673 of the data type names defined in the function
1674 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1675 omit @code{int} or change the order---that would cause the compiler to
1678 If you don't define this macro, the default is @code{"long unsigned
1682 @defmac PTRDIFF_TYPE
1683 A C expression for a string describing the name of the data type to use
1684 for the result of subtracting two pointers. The typedef name
1685 @code{ptrdiff_t} is defined using the contents of the string. See
1686 @code{SIZE_TYPE} above for more information.
1688 If you don't define this macro, the default is @code{"long int"}.
1692 A C expression for a string describing the name of the data type to use
1693 for wide characters. The typedef name @code{wchar_t} is defined using
1694 the contents of the string. See @code{SIZE_TYPE} above for more
1697 If you don't define this macro, the default is @code{"int"}.
1700 @defmac WCHAR_TYPE_SIZE
1701 A C expression for the size in bits of the data type for wide
1702 characters. This is used in @code{cpp}, which cannot make use of
1707 A C expression for a string describing the name of the data type to
1708 use for wide characters passed to @code{printf} and returned from
1709 @code{getwc}. The typedef name @code{wint_t} is defined using the
1710 contents of the string. See @code{SIZE_TYPE} above for more
1713 If you don't define this macro, the default is @code{"unsigned int"}.
1717 A C expression for a string describing the name of the data type that
1718 can represent any value of any standard or extended signed integer type.
1719 The typedef name @code{intmax_t} is defined using the contents of the
1720 string. See @code{SIZE_TYPE} above for more information.
1722 If you don't define this macro, the default is the first of
1723 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1724 much precision as @code{long long int}.
1727 @defmac UINTMAX_TYPE
1728 A C expression for a string describing the name of the data type that
1729 can represent any value of any standard or extended unsigned integer
1730 type. The typedef name @code{uintmax_t} is defined using the contents
1731 of the string. See @code{SIZE_TYPE} above for more information.
1733 If you don't define this macro, the default is the first of
1734 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1735 unsigned int"} that has as much precision as @code{long long unsigned
1739 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1740 The C++ compiler represents a pointer-to-member-function with a struct
1747 ptrdiff_t vtable_index;
1754 The C++ compiler must use one bit to indicate whether the function that
1755 will be called through a pointer-to-member-function is virtual.
1756 Normally, we assume that the low-order bit of a function pointer must
1757 always be zero. Then, by ensuring that the vtable_index is odd, we can
1758 distinguish which variant of the union is in use. But, on some
1759 platforms function pointers can be odd, and so this doesn't work. In
1760 that case, we use the low-order bit of the @code{delta} field, and shift
1761 the remainder of the @code{delta} field to the left.
1763 GCC will automatically make the right selection about where to store
1764 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1765 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1766 set such that functions always start at even addresses, but the lowest
1767 bit of pointers to functions indicate whether the function at that
1768 address is in ARM or Thumb mode. If this is the case of your
1769 architecture, you should define this macro to
1770 @code{ptrmemfunc_vbit_in_delta}.
1772 In general, you should not have to define this macro. On architectures
1773 in which function addresses are always even, according to
1774 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1775 @code{ptrmemfunc_vbit_in_pfn}.
1778 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1779 Normally, the C++ compiler uses function pointers in vtables. This
1780 macro allows the target to change to use ``function descriptors''
1781 instead. Function descriptors are found on targets for whom a
1782 function pointer is actually a small data structure. Normally the
1783 data structure consists of the actual code address plus a data
1784 pointer to which the function's data is relative.
1786 If vtables are used, the value of this macro should be the number
1787 of words that the function descriptor occupies.
1790 @defmac TARGET_VTABLE_ENTRY_ALIGN
1791 By default, the vtable entries are void pointers, the so the alignment
1792 is the same as pointer alignment. The value of this macro specifies
1793 the alignment of the vtable entry in bits. It should be defined only
1794 when special alignment is necessary. */
1797 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1798 There are a few non-descriptor entries in the vtable at offsets below
1799 zero. If these entries must be padded (say, to preserve the alignment
1800 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1801 of words in each data entry.
1804 @node Escape Sequences
1805 @section Target Character Escape Sequences
1806 @cindex escape sequences
1808 By default, GCC assumes that the C character escape sequences and other
1809 characters take on their ASCII values for the target. If this is not
1810 correct, you must explicitly define all of the macros below. All of
1811 them must evaluate to constants; they are used in @code{case}
1817 @findex TARGET_DIGIT0
1820 @findex TARGET_NEWLINE
1823 @multitable {@code{TARGET_NEWLINE}} {Escape} {ASCII character}
1824 @item Macro @tab Escape @tab ASCII character
1825 @item @code{TARGET_BELL} @tab @kbd{\a} @tab @code{07}, @code{BEL}
1826 @item @code{TARGET_BS} @tab @kbd{\b} @tab @code{08}, @code{BS}
1827 @item @code{TARGET_CR} @tab @kbd{\r} @tab @code{0D}, @code{CR}
1828 @item @code{TARGET_DIGIT0} @tab @kbd{0} @tab @code{30}, @code{ZERO}
1829 @item @code{TARGET_ESC} @tab @kbd{\e}, @kbd{\E} @tab @code{1B}, @code{ESC}
1830 @item @code{TARGET_FF} @tab @kbd{\f} @tab @code{0C}, @code{FF}
1831 @item @code{TARGET_NEWLINE} @tab @kbd{\n} @tab @code{0A}, @code{LF}
1832 @item @code{TARGET_TAB} @tab @kbd{\t} @tab @code{09}, @code{HT}
1833 @item @code{TARGET_VT} @tab @kbd{\v} @tab @code{0B}, @code{VT}
1837 Note that the @kbd{\e} and @kbd{\E} escapes are GNU extensions, not
1838 part of the C standard.
1841 @section Register Usage
1842 @cindex register usage
1844 This section explains how to describe what registers the target machine
1845 has, and how (in general) they can be used.
1847 The description of which registers a specific instruction can use is
1848 done with register classes; see @ref{Register Classes}. For information
1849 on using registers to access a stack frame, see @ref{Frame Registers}.
1850 For passing values in registers, see @ref{Register Arguments}.
1851 For returning values in registers, see @ref{Scalar Return}.
1854 * Register Basics:: Number and kinds of registers.
1855 * Allocation Order:: Order in which registers are allocated.
1856 * Values in Registers:: What kinds of values each reg can hold.
1857 * Leaf Functions:: Renumbering registers for leaf functions.
1858 * Stack Registers:: Handling a register stack such as 80387.
1861 @node Register Basics
1862 @subsection Basic Characteristics of Registers
1864 @c prevent bad page break with this line
1865 Registers have various characteristics.
1867 @defmac FIRST_PSEUDO_REGISTER
1868 Number of hardware registers known to the compiler. They receive
1869 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1870 pseudo register's number really is assigned the number
1871 @code{FIRST_PSEUDO_REGISTER}.
1874 @defmac FIXED_REGISTERS
1875 @cindex fixed register
1876 An initializer that says which registers are used for fixed purposes
1877 all throughout the compiled code and are therefore not available for
1878 general allocation. These would include the stack pointer, the frame
1879 pointer (except on machines where that can be used as a general
1880 register when no frame pointer is needed), the program counter on
1881 machines where that is considered one of the addressable registers,
1882 and any other numbered register with a standard use.
1884 This information is expressed as a sequence of numbers, separated by
1885 commas and surrounded by braces. The @var{n}th number is 1 if
1886 register @var{n} is fixed, 0 otherwise.
1888 The table initialized from this macro, and the table initialized by
1889 the following one, may be overridden at run time either automatically,
1890 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1891 the user with the command options @option{-ffixed-@var{reg}},
1892 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1895 @defmac CALL_USED_REGISTERS
1896 @cindex call-used register
1897 @cindex call-clobbered register
1898 @cindex call-saved register
1899 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1900 clobbered (in general) by function calls as well as for fixed
1901 registers. This macro therefore identifies the registers that are not
1902 available for general allocation of values that must live across
1905 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1906 automatically saves it on function entry and restores it on function
1907 exit, if the register is used within the function.
1910 @defmac CALL_REALLY_USED_REGISTERS
1911 @cindex call-used register
1912 @cindex call-clobbered register
1913 @cindex call-saved register
1914 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1915 that the entire set of @code{FIXED_REGISTERS} be included.
1916 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1917 This macro is optional. If not specified, it defaults to the value
1918 of @code{CALL_USED_REGISTERS}.
1921 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1922 @cindex call-used register
1923 @cindex call-clobbered register
1924 @cindex call-saved register
1925 A C expression that is nonzero if it is not permissible to store a
1926 value of mode @var{mode} in hard register number @var{regno} across a
1927 call without some part of it being clobbered. For most machines this
1928 macro need not be defined. It is only required for machines that do not
1929 preserve the entire contents of a register across a call.
1933 @findex call_used_regs
1936 @findex reg_class_contents
1937 @defmac CONDITIONAL_REGISTER_USAGE
1938 Zero or more C statements that may conditionally modify five variables
1939 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1940 @code{reg_names}, and @code{reg_class_contents}, to take into account
1941 any dependence of these register sets on target flags. The first three
1942 of these are of type @code{char []} (interpreted as Boolean vectors).
1943 @code{global_regs} is a @code{const char *[]}, and
1944 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1945 called, @code{fixed_regs}, @code{call_used_regs},
1946 @code{reg_class_contents}, and @code{reg_names} have been initialized
1947 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1948 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1949 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1950 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1951 command options have been applied.
1953 You need not define this macro if it has no work to do.
1955 @cindex disabling certain registers
1956 @cindex controlling register usage
1957 If the usage of an entire class of registers depends on the target
1958 flags, you may indicate this to GCC by using this macro to modify
1959 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1960 registers in the classes which should not be used by GCC@. Also define
1961 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1962 to return @code{NO_REGS} if it
1963 is called with a letter for a class that shouldn't be used.
1965 (However, if this class is not included in @code{GENERAL_REGS} and all
1966 of the insn patterns whose constraints permit this class are
1967 controlled by target switches, then GCC will automatically avoid using
1968 these registers when the target switches are opposed to them.)
1971 @defmac NON_SAVING_SETJMP
1972 If this macro is defined and has a nonzero value, it means that
1973 @code{setjmp} and related functions fail to save the registers, or that
1974 @code{longjmp} fails to restore them. To compensate, the compiler
1975 avoids putting variables in registers in functions that use
1979 @defmac INCOMING_REGNO (@var{out})
1980 Define this macro if the target machine has register windows. This C
1981 expression returns the register number as seen by the called function
1982 corresponding to the register number @var{out} as seen by the calling
1983 function. Return @var{out} if register number @var{out} is not an
1987 @defmac OUTGOING_REGNO (@var{in})
1988 Define this macro if the target machine has register windows. This C
1989 expression returns the register number as seen by the calling function
1990 corresponding to the register number @var{in} as seen by the called
1991 function. Return @var{in} if register number @var{in} is not an inbound
1995 @defmac LOCAL_REGNO (@var{regno})
1996 Define this macro if the target machine has register windows. This C
1997 expression returns true if the register is call-saved but is in the
1998 register window. Unlike most call-saved registers, such registers
1999 need not be explicitly restored on function exit or during non-local
2004 If the program counter has a register number, define this as that
2005 register number. Otherwise, do not define it.
2008 @node Allocation Order
2009 @subsection Order of Allocation of Registers
2010 @cindex order of register allocation
2011 @cindex register allocation order
2013 @c prevent bad page break with this line
2014 Registers are allocated in order.
2016 @defmac REG_ALLOC_ORDER
2017 If defined, an initializer for a vector of integers, containing the
2018 numbers of hard registers in the order in which GCC should prefer
2019 to use them (from most preferred to least).
2021 If this macro is not defined, registers are used lowest numbered first
2022 (all else being equal).
2024 One use of this macro is on machines where the highest numbered
2025 registers must always be saved and the save-multiple-registers
2026 instruction supports only sequences of consecutive registers. On such
2027 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2028 the highest numbered allocable register first.
2031 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
2032 A C statement (sans semicolon) to choose the order in which to allocate
2033 hard registers for pseudo-registers local to a basic block.
2035 Store the desired register order in the array @code{reg_alloc_order}.
2036 Element 0 should be the register to allocate first; element 1, the next
2037 register; and so on.
2039 The macro body should not assume anything about the contents of
2040 @code{reg_alloc_order} before execution of the macro.
2042 On most machines, it is not necessary to define this macro.
2045 @node Values in Registers
2046 @subsection How Values Fit in Registers
2048 This section discusses the macros that describe which kinds of values
2049 (specifically, which machine modes) each register can hold, and how many
2050 consecutive registers are needed for a given mode.
2052 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2053 A C expression for the number of consecutive hard registers, starting
2054 at register number @var{regno}, required to hold a value of mode
2057 On a machine where all registers are exactly one word, a suitable
2058 definition of this macro is
2061 #define HARD_REGNO_NREGS(REGNO, MODE) \
2062 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2067 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2068 Define this macro if the natural size of registers that hold values
2069 of mode @var{mode} is not the word size. It is a C expression that
2070 should give the natural size in bytes for the specified mode. It is
2071 used by the register allocator to try to optimize its results. This
2072 happens for example on SPARC 64-bit where the natural size of
2073 floating-point registers is still 32-bit.
2076 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2077 A C expression that is nonzero if it is permissible to store a value
2078 of mode @var{mode} in hard register number @var{regno} (or in several
2079 registers starting with that one). For a machine where all registers
2080 are equivalent, a suitable definition is
2083 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2086 You need not include code to check for the numbers of fixed registers,
2087 because the allocation mechanism considers them to be always occupied.
2089 @cindex register pairs
2090 On some machines, double-precision values must be kept in even/odd
2091 register pairs. You can implement that by defining this macro to reject
2092 odd register numbers for such modes.
2094 The minimum requirement for a mode to be OK in a register is that the
2095 @samp{mov@var{mode}} instruction pattern support moves between the
2096 register and other hard register in the same class and that moving a
2097 value into the register and back out not alter it.
2099 Since the same instruction used to move @code{word_mode} will work for
2100 all narrower integer modes, it is not necessary on any machine for
2101 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2102 you define patterns @samp{movhi}, etc., to take advantage of this. This
2103 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2104 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2107 Many machines have special registers for floating point arithmetic.
2108 Often people assume that floating point machine modes are allowed only
2109 in floating point registers. This is not true. Any registers that
2110 can hold integers can safely @emph{hold} a floating point machine
2111 mode, whether or not floating arithmetic can be done on it in those
2112 registers. Integer move instructions can be used to move the values.
2114 On some machines, though, the converse is true: fixed-point machine
2115 modes may not go in floating registers. This is true if the floating
2116 registers normalize any value stored in them, because storing a
2117 non-floating value there would garble it. In this case,
2118 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2119 floating registers. But if the floating registers do not automatically
2120 normalize, if you can store any bit pattern in one and retrieve it
2121 unchanged without a trap, then any machine mode may go in a floating
2122 register, so you can define this macro to say so.
2124 The primary significance of special floating registers is rather that
2125 they are the registers acceptable in floating point arithmetic
2126 instructions. However, this is of no concern to
2127 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2128 constraints for those instructions.
2130 On some machines, the floating registers are especially slow to access,
2131 so that it is better to store a value in a stack frame than in such a
2132 register if floating point arithmetic is not being done. As long as the
2133 floating registers are not in class @code{GENERAL_REGS}, they will not
2134 be used unless some pattern's constraint asks for one.
2137 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2138 A C expression that is nonzero if it is OK to rename a hard register
2139 @var{from} to another hard register @var{to}.
2141 One common use of this macro is to prevent renaming of a register to
2142 another register that is not saved by a prologue in an interrupt
2145 The default is always nonzero.
2148 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2149 A C expression that is nonzero if a value of mode
2150 @var{mode1} is accessible in mode @var{mode2} without copying.
2152 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2153 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2154 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2155 should be nonzero. If they differ for any @var{r}, you should define
2156 this macro to return zero unless some other mechanism ensures the
2157 accessibility of the value in a narrower mode.
2159 You should define this macro to return nonzero in as many cases as
2160 possible since doing so will allow GCC to perform better register
2164 @defmac AVOID_CCMODE_COPIES
2165 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2166 registers. You should only define this macro if support for copying to/from
2167 @code{CCmode} is incomplete.
2170 @node Leaf Functions
2171 @subsection Handling Leaf Functions
2173 @cindex leaf functions
2174 @cindex functions, leaf
2175 On some machines, a leaf function (i.e., one which makes no calls) can run
2176 more efficiently if it does not make its own register window. Often this
2177 means it is required to receive its arguments in the registers where they
2178 are passed by the caller, instead of the registers where they would
2181 The special treatment for leaf functions generally applies only when
2182 other conditions are met; for example, often they may use only those
2183 registers for its own variables and temporaries. We use the term ``leaf
2184 function'' to mean a function that is suitable for this special
2185 handling, so that functions with no calls are not necessarily ``leaf
2188 GCC assigns register numbers before it knows whether the function is
2189 suitable for leaf function treatment. So it needs to renumber the
2190 registers in order to output a leaf function. The following macros
2193 @defmac LEAF_REGISTERS
2194 Name of a char vector, indexed by hard register number, which
2195 contains 1 for a register that is allowable in a candidate for leaf
2198 If leaf function treatment involves renumbering the registers, then the
2199 registers marked here should be the ones before renumbering---those that
2200 GCC would ordinarily allocate. The registers which will actually be
2201 used in the assembler code, after renumbering, should not be marked with 1
2204 Define this macro only if the target machine offers a way to optimize
2205 the treatment of leaf functions.
2208 @defmac LEAF_REG_REMAP (@var{regno})
2209 A C expression whose value is the register number to which @var{regno}
2210 should be renumbered, when a function is treated as a leaf function.
2212 If @var{regno} is a register number which should not appear in a leaf
2213 function before renumbering, then the expression should yield @minus{}1, which
2214 will cause the compiler to abort.
2216 Define this macro only if the target machine offers a way to optimize the
2217 treatment of leaf functions, and registers need to be renumbered to do
2221 @findex current_function_is_leaf
2222 @findex current_function_uses_only_leaf_regs
2223 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2224 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2225 specially. They can test the C variable @code{current_function_is_leaf}
2226 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2227 set prior to local register allocation and is valid for the remaining
2228 compiler passes. They can also test the C variable
2229 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2230 functions which only use leaf registers.
2231 @code{current_function_uses_only_leaf_regs} is valid after all passes
2232 that modify the instructions have been run and is only useful if
2233 @code{LEAF_REGISTERS} is defined.
2234 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2235 @c of the next paragraph?! --mew 2feb93
2237 @node Stack Registers
2238 @subsection Registers That Form a Stack
2240 There are special features to handle computers where some of the
2241 ``registers'' form a stack. Stack registers are normally written by
2242 pushing onto the stack, and are numbered relative to the top of the
2245 Currently, GCC can only handle one group of stack-like registers, and
2246 they must be consecutively numbered. Furthermore, the existing
2247 support for stack-like registers is specific to the 80387 floating
2248 point coprocessor. If you have a new architecture that uses
2249 stack-like registers, you will need to do substantial work on
2250 @file{reg-stack.c} and write your machine description to cooperate
2251 with it, as well as defining these macros.
2254 Define this if the machine has any stack-like registers.
2257 @defmac FIRST_STACK_REG
2258 The number of the first stack-like register. This one is the top
2262 @defmac LAST_STACK_REG
2263 The number of the last stack-like register. This one is the bottom of
2267 @node Register Classes
2268 @section Register Classes
2269 @cindex register class definitions
2270 @cindex class definitions, register
2272 On many machines, the numbered registers are not all equivalent.
2273 For example, certain registers may not be allowed for indexed addressing;
2274 certain registers may not be allowed in some instructions. These machine
2275 restrictions are described to the compiler using @dfn{register classes}.
2277 You define a number of register classes, giving each one a name and saying
2278 which of the registers belong to it. Then you can specify register classes
2279 that are allowed as operands to particular instruction patterns.
2283 In general, each register will belong to several classes. In fact, one
2284 class must be named @code{ALL_REGS} and contain all the registers. Another
2285 class must be named @code{NO_REGS} and contain no registers. Often the
2286 union of two classes will be another class; however, this is not required.
2288 @findex GENERAL_REGS
2289 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2290 terribly special about the name, but the operand constraint letters
2291 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2292 the same as @code{ALL_REGS}, just define it as a macro which expands
2295 Order the classes so that if class @var{x} is contained in class @var{y}
2296 then @var{x} has a lower class number than @var{y}.
2298 The way classes other than @code{GENERAL_REGS} are specified in operand
2299 constraints is through machine-dependent operand constraint letters.
2300 You can define such letters to correspond to various classes, then use
2301 them in operand constraints.
2303 You should define a class for the union of two classes whenever some
2304 instruction allows both classes. For example, if an instruction allows
2305 either a floating point (coprocessor) register or a general register for a
2306 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2307 which includes both of them. Otherwise you will get suboptimal code.
2309 You must also specify certain redundant information about the register
2310 classes: for each class, which classes contain it and which ones are
2311 contained in it; for each pair of classes, the largest class contained
2314 When a value occupying several consecutive registers is expected in a
2315 certain class, all the registers used must belong to that class.
2316 Therefore, register classes cannot be used to enforce a requirement for
2317 a register pair to start with an even-numbered register. The way to
2318 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2320 Register classes used for input-operands of bitwise-and or shift
2321 instructions have a special requirement: each such class must have, for
2322 each fixed-point machine mode, a subclass whose registers can transfer that
2323 mode to or from memory. For example, on some machines, the operations for
2324 single-byte values (@code{QImode}) are limited to certain registers. When
2325 this is so, each register class that is used in a bitwise-and or shift
2326 instruction must have a subclass consisting of registers from which
2327 single-byte values can be loaded or stored. This is so that
2328 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2330 @deftp {Data type} {enum reg_class}
2331 An enumerated type that must be defined with all the register class names
2332 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2333 must be the last register class, followed by one more enumerated value,
2334 @code{LIM_REG_CLASSES}, which is not a register class but rather
2335 tells how many classes there are.
2337 Each register class has a number, which is the value of casting
2338 the class name to type @code{int}. The number serves as an index
2339 in many of the tables described below.
2342 @defmac N_REG_CLASSES
2343 The number of distinct register classes, defined as follows:
2346 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2350 @defmac REG_CLASS_NAMES
2351 An initializer containing the names of the register classes as C string
2352 constants. These names are used in writing some of the debugging dumps.
2355 @defmac REG_CLASS_CONTENTS
2356 An initializer containing the contents of the register classes, as integers
2357 which are bit masks. The @var{n}th integer specifies the contents of class
2358 @var{n}. The way the integer @var{mask} is interpreted is that
2359 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2361 When the machine has more than 32 registers, an integer does not suffice.
2362 Then the integers are replaced by sub-initializers, braced groupings containing
2363 several integers. Each sub-initializer must be suitable as an initializer
2364 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2365 In this situation, the first integer in each sub-initializer corresponds to
2366 registers 0 through 31, the second integer to registers 32 through 63, and
2370 @defmac REGNO_REG_CLASS (@var{regno})
2371 A C expression whose value is a register class containing hard register
2372 @var{regno}. In general there is more than one such class; choose a class
2373 which is @dfn{minimal}, meaning that no smaller class also contains the
2377 @defmac BASE_REG_CLASS
2378 A macro whose definition is the name of the class to which a valid
2379 base register must belong. A base register is one used in an address
2380 which is the register value plus a displacement.
2383 @defmac MODE_BASE_REG_CLASS (@var{mode})
2384 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2385 the selection of a base register in a mode dependent manner. If
2386 @var{mode} is VOIDmode then it should return the same value as
2387 @code{BASE_REG_CLASS}.
2390 @defmac INDEX_REG_CLASS
2391 A macro whose definition is the name of the class to which a valid
2392 index register must belong. An index register is one used in an
2393 address where its value is either multiplied by a scale factor or
2394 added to another register (as well as added to a displacement).
2397 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2398 For the constraint at the start of @var{str}, which starts with the letter
2399 @var{c}, return the length. This allows you to have register class /
2400 constant / extra constraints that are longer than a single letter;
2401 you don't need to define this macro if you can do with single-letter
2402 constraints only. The definition of this macro should use
2403 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2404 to handle specially.
2405 There are some sanity checks in genoutput.c that check the constraint lengths
2406 for the md file, so you can also use this macro to help you while you are
2407 transitioning from a byzantine single-letter-constraint scheme: when you
2408 return a negative length for a constraint you want to re-use, genoutput
2409 will complain about every instance where it is used in the md file.
2412 @defmac REG_CLASS_FROM_LETTER (@var{char})
2413 A C expression which defines the machine-dependent operand constraint
2414 letters for register classes. If @var{char} is such a letter, the
2415 value should be the register class corresponding to it. Otherwise,
2416 the value should be @code{NO_REGS}. The register letter @samp{r},
2417 corresponding to class @code{GENERAL_REGS}, will not be passed
2418 to this macro; you do not need to handle it.
2421 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2422 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2423 passed in @var{str}, so that you can use suffixes to distinguish between
2427 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2428 A C expression which is nonzero if register number @var{num} is
2429 suitable for use as a base register in operand addresses. It may be
2430 either a suitable hard register or a pseudo register that has been
2431 allocated such a hard register.
2434 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2435 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2436 that expression may examine the mode of the memory reference in
2437 @var{mode}. You should define this macro if the mode of the memory
2438 reference affects whether a register may be used as a base register. If
2439 you define this macro, the compiler will use it instead of
2440 @code{REGNO_OK_FOR_BASE_P}.
2443 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2444 A C expression which is nonzero if register number @var{num} is
2445 suitable for use as an index register in operand addresses. It may be
2446 either a suitable hard register or a pseudo register that has been
2447 allocated such a hard register.
2449 The difference between an index register and a base register is that
2450 the index register may be scaled. If an address involves the sum of
2451 two registers, neither one of them scaled, then either one may be
2452 labeled the ``base'' and the other the ``index''; but whichever
2453 labeling is used must fit the machine's constraints of which registers
2454 may serve in each capacity. The compiler will try both labelings,
2455 looking for one that is valid, and will reload one or both registers
2456 only if neither labeling works.
2459 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2460 A C expression that places additional restrictions on the register class
2461 to use when it is necessary to copy value @var{x} into a register in class
2462 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2463 another, smaller class. On many machines, the following definition is
2467 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2470 Sometimes returning a more restrictive class makes better code. For
2471 example, on the 68000, when @var{x} is an integer constant that is in range
2472 for a @samp{moveq} instruction, the value of this macro is always
2473 @code{DATA_REGS} as long as @var{class} includes the data registers.
2474 Requiring a data register guarantees that a @samp{moveq} will be used.
2476 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2477 @var{class} is if @var{x} is a legitimate constant which cannot be
2478 loaded into some register class. By returning @code{NO_REGS} you can
2479 force @var{x} into a memory location. For example, rs6000 can load
2480 immediate values into general-purpose registers, but does not have an
2481 instruction for loading an immediate value into a floating-point
2482 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2483 @var{x} is a floating-point constant. If the constant can't be loaded
2484 into any kind of register, code generation will be better if
2485 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2486 of using @code{PREFERRED_RELOAD_CLASS}.
2489 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2490 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2491 input reloads. If you don't define this macro, the default is to use
2492 @var{class}, unchanged.
2495 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2496 A C expression that places additional restrictions on the register class
2497 to use when it is necessary to be able to hold a value of mode
2498 @var{mode} in a reload register for which class @var{class} would
2501 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2502 there are certain modes that simply can't go in certain reload classes.
2504 The value is a register class; perhaps @var{class}, or perhaps another,
2507 Don't define this macro unless the target machine has limitations which
2508 require the macro to do something nontrivial.
2511 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2512 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2513 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2514 Many machines have some registers that cannot be copied directly to or
2515 from memory or even from other types of registers. An example is the
2516 @samp{MQ} register, which on most machines, can only be copied to or
2517 from general registers, but not memory. Some machines allow copying all
2518 registers to and from memory, but require a scratch register for stores
2519 to some memory locations (e.g., those with symbolic address on the RT,
2520 and those with certain symbolic address on the SPARC when compiling
2521 PIC)@. In some cases, both an intermediate and a scratch register are
2524 You should define these macros to indicate to the reload phase that it may
2525 need to allocate at least one register for a reload in addition to the
2526 register to contain the data. Specifically, if copying @var{x} to a
2527 register @var{class} in @var{mode} requires an intermediate register,
2528 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2529 largest register class all of whose registers can be used as
2530 intermediate registers or scratch registers.
2532 If copying a register @var{class} in @var{mode} to @var{x} requires an
2533 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2534 should be defined to return the largest register class required. If the
2535 requirements for input and output reloads are the same, the macro
2536 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2539 The values returned by these macros are often @code{GENERAL_REGS}.
2540 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2541 can be directly copied to or from a register of @var{class} in
2542 @var{mode} without requiring a scratch register. Do not define this
2543 macro if it would always return @code{NO_REGS}.
2545 If a scratch register is required (either with or without an
2546 intermediate register), you should define patterns for
2547 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2548 (@pxref{Standard Names}. These patterns, which will normally be
2549 implemented with a @code{define_expand}, should be similar to the
2550 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2553 Define constraints for the reload register and scratch register that
2554 contain a single register class. If the original reload register (whose
2555 class is @var{class}) can meet the constraint given in the pattern, the
2556 value returned by these macros is used for the class of the scratch
2557 register. Otherwise, two additional reload registers are required.
2558 Their classes are obtained from the constraints in the insn pattern.
2560 @var{x} might be a pseudo-register or a @code{subreg} of a
2561 pseudo-register, which could either be in a hard register or in memory.
2562 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2563 in memory and the hard register number if it is in a register.
2565 These macros should not be used in the case where a particular class of
2566 registers can only be copied to memory and not to another class of
2567 registers. In that case, secondary reload registers are not needed and
2568 would not be helpful. Instead, a stack location must be used to perform
2569 the copy and the @code{mov@var{m}} pattern should use memory as an
2570 intermediate storage. This case often occurs between floating-point and
2574 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2575 Certain machines have the property that some registers cannot be copied
2576 to some other registers without using memory. Define this macro on
2577 those machines to be a C expression that is nonzero if objects of mode
2578 @var{m} in registers of @var{class1} can only be copied to registers of
2579 class @var{class2} by storing a register of @var{class1} into memory
2580 and loading that memory location into a register of @var{class2}.
2582 Do not define this macro if its value would always be zero.
2585 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2586 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2587 allocates a stack slot for a memory location needed for register copies.
2588 If this macro is defined, the compiler instead uses the memory location
2589 defined by this macro.
2591 Do not define this macro if you do not define
2592 @code{SECONDARY_MEMORY_NEEDED}.
2595 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2596 When the compiler needs a secondary memory location to copy between two
2597 registers of mode @var{mode}, it normally allocates sufficient memory to
2598 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2599 load operations in a mode that many bits wide and whose class is the
2600 same as that of @var{mode}.
2602 This is right thing to do on most machines because it ensures that all
2603 bits of the register are copied and prevents accesses to the registers
2604 in a narrower mode, which some machines prohibit for floating-point
2607 However, this default behavior is not correct on some machines, such as
2608 the DEC Alpha, that store short integers in floating-point registers
2609 differently than in integer registers. On those machines, the default
2610 widening will not work correctly and you must define this macro to
2611 suppress that widening in some cases. See the file @file{alpha.h} for
2614 Do not define this macro if you do not define
2615 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2616 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2619 @defmac SMALL_REGISTER_CLASSES
2620 On some machines, it is risky to let hard registers live across arbitrary
2621 insns. Typically, these machines have instructions that require values
2622 to be in specific registers (like an accumulator), and reload will fail
2623 if the required hard register is used for another purpose across such an
2626 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2627 value on these machines. When this macro has a nonzero value, the
2628 compiler will try to minimize the lifetime of hard registers.
2630 It is always safe to define this macro with a nonzero value, but if you
2631 unnecessarily define it, you will reduce the amount of optimizations
2632 that can be performed in some cases. If you do not define this macro
2633 with a nonzero value when it is required, the compiler will run out of
2634 spill registers and print a fatal error message. For most machines, you
2635 should not define this macro at all.
2638 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2639 A C expression whose value is nonzero if pseudos that have been assigned
2640 to registers of class @var{class} would likely be spilled because
2641 registers of @var{class} are needed for spill registers.
2643 The default value of this macro returns 1 if @var{class} has exactly one
2644 register and zero otherwise. On most machines, this default should be
2645 used. Only define this macro to some other expression if pseudos
2646 allocated by @file{local-alloc.c} end up in memory because their hard
2647 registers were needed for spill registers. If this macro returns nonzero
2648 for those classes, those pseudos will only be allocated by
2649 @file{global.c}, which knows how to reallocate the pseudo to another
2650 register. If there would not be another register available for
2651 reallocation, you should not change the definition of this macro since
2652 the only effect of such a definition would be to slow down register
2656 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2657 A C expression for the maximum number of consecutive registers
2658 of class @var{class} needed to hold a value of mode @var{mode}.
2660 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2661 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2662 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2663 @var{mode})} for all @var{regno} values in the class @var{class}.
2665 This macro helps control the handling of multiple-word values
2669 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2670 If defined, a C expression that returns nonzero for a @var{class} for which
2671 a change from mode @var{from} to mode @var{to} is invalid.
2673 For the example, loading 32-bit integer or floating-point objects into
2674 floating-point registers on the Alpha extends them to 64 bits.
2675 Therefore loading a 64-bit object and then storing it as a 32-bit object
2676 does not store the low-order 32 bits, as would be the case for a normal
2677 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2681 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2682 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2683 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2687 Three other special macros describe which operands fit which constraint
2690 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2691 A C expression that defines the machine-dependent operand constraint
2692 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2693 particular ranges of integer values. If @var{c} is one of those
2694 letters, the expression should check that @var{value}, an integer, is in
2695 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2696 not one of those letters, the value should be 0 regardless of
2700 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2701 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2702 string passed in @var{str}, so that you can use suffixes to distinguish
2703 between different variants.
2706 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2707 A C expression that defines the machine-dependent operand constraint
2708 letters that specify particular ranges of @code{const_double} values
2709 (@samp{G} or @samp{H}).
2711 If @var{c} is one of those letters, the expression should check that
2712 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2713 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2714 letters, the value should be 0 regardless of @var{value}.
2716 @code{const_double} is used for all floating-point constants and for
2717 @code{DImode} fixed-point constants. A given letter can accept either
2718 or both kinds of values. It can use @code{GET_MODE} to distinguish
2719 between these kinds.
2722 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2723 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2724 string passed in @var{str}, so that you can use suffixes to distinguish
2725 between different variants.
2728 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2729 A C expression that defines the optional machine-dependent constraint
2730 letters that can be used to segregate specific types of operands, usually
2731 memory references, for the target machine. Any letter that is not
2732 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2733 @code{REG_CLASS_FROM_CONSTRAINT}
2734 may be used. Normally this macro will not be defined.
2736 If it is required for a particular target machine, it should return 1
2737 if @var{value} corresponds to the operand type represented by the
2738 constraint letter @var{c}. If @var{c} is not defined as an extra
2739 constraint, the value returned should be 0 regardless of @var{value}.
2741 For example, on the ROMP, load instructions cannot have their output
2742 in r0 if the memory reference contains a symbolic address. Constraint
2743 letter @samp{Q} is defined as representing a memory address that does
2744 @emph{not} contain a symbolic address. An alternative is specified with
2745 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2746 alternative specifies @samp{m} on the input and a register class that
2747 does not include r0 on the output.
2750 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2751 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2752 in @var{str}, so that you can use suffixes to distinguish between different
2756 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2757 A C expression that defines the optional machine-dependent constraint
2758 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2759 be treated like memory constraints by the reload pass.
2761 It should return 1 if the operand type represented by the constraint
2762 at the start of @var{str}, the first letter of which is the letter @var{c},
2763 comprises a subset of all memory references including
2764 all those whose address is simply a base register. This allows the reload
2765 pass to reload an operand, if it does not directly correspond to the operand
2766 type of @var{c}, by copying its address into a base register.
2768 For example, on the S/390, some instructions do not accept arbitrary
2769 memory references, but only those that do not make use of an index
2770 register. The constraint letter @samp{Q} is defined via
2771 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2772 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2773 a @samp{Q} constraint can handle any memory operand, because the
2774 reload pass knows it can be reloaded by copying the memory address
2775 into a base register if required. This is analogous to the way
2776 a @samp{o} constraint can handle any memory operand.
2779 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2780 A C expression that defines the optional machine-dependent constraint
2781 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2782 @code{EXTRA_CONSTRAINT_STR}, that should
2783 be treated like address constraints by the reload pass.
2785 It should return 1 if the operand type represented by the constraint
2786 at the start of @var{str}, which starts with the letter @var{c}, comprises
2787 a subset of all memory addresses including
2788 all those that consist of just a base register. This allows the reload
2789 pass to reload an operand, if it does not directly correspond to the operand
2790 type of @var{str}, by copying it into a base register.
2792 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2793 be used with the @code{address_operand} predicate. It is treated
2794 analogously to the @samp{p} constraint.
2797 @node Stack and Calling
2798 @section Stack Layout and Calling Conventions
2799 @cindex calling conventions
2801 @c prevent bad page break with this line
2802 This describes the stack layout and calling conventions.
2806 * Exception Handling::
2811 * Register Arguments::
2813 * Aggregate Return::
2821 @subsection Basic Stack Layout
2822 @cindex stack frame layout
2823 @cindex frame layout
2825 @c prevent bad page break with this line
2826 Here is the basic stack layout.
2828 @defmac STACK_GROWS_DOWNWARD
2829 Define this macro if pushing a word onto the stack moves the stack
2830 pointer to a smaller address.
2832 When we say, ``define this macro if @dots{},'' it means that the
2833 compiler checks this macro only with @code{#ifdef} so the precise
2834 definition used does not matter.
2837 @defmac STACK_PUSH_CODE
2838 This macro defines the operation used when something is pushed
2839 on the stack. In RTL, a push operation will be
2840 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2842 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2843 and @code{POST_INC}. Which of these is correct depends on
2844 the stack direction and on whether the stack pointer points
2845 to the last item on the stack or whether it points to the
2846 space for the next item on the stack.
2848 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2849 defined, which is almost always right, and @code{PRE_INC} otherwise,
2850 which is often wrong.
2853 @defmac FRAME_GROWS_DOWNWARD
2854 Define this macro if the addresses of local variable slots are at negative
2855 offsets from the frame pointer.
2858 @defmac ARGS_GROW_DOWNWARD
2859 Define this macro if successive arguments to a function occupy decreasing
2860 addresses on the stack.
2863 @defmac STARTING_FRAME_OFFSET
2864 Offset from the frame pointer to the first local variable slot to be allocated.
2866 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2867 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2868 Otherwise, it is found by adding the length of the first slot to the
2869 value @code{STARTING_FRAME_OFFSET}.
2870 @c i'm not sure if the above is still correct.. had to change it to get
2871 @c rid of an overfull. --mew 2feb93
2874 @defmac STACK_ALIGNMENT_NEEDED
2875 Define to zero to disable final alignment of the stack during reload.
2876 The nonzero default for this macro is suitable for most ports.
2878 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2879 is a register save block following the local block that doesn't require
2880 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2881 stack alignment and do it in the backend.
2884 @defmac STACK_POINTER_OFFSET
2885 Offset from the stack pointer register to the first location at which
2886 outgoing arguments are placed. If not specified, the default value of
2887 zero is used. This is the proper value for most machines.
2889 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2890 the first location at which outgoing arguments are placed.
2893 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2894 Offset from the argument pointer register to the first argument's
2895 address. On some machines it may depend on the data type of the
2898 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2899 the first argument's address.
2902 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2903 Offset from the stack pointer register to an item dynamically allocated
2904 on the stack, e.g., by @code{alloca}.
2906 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2907 length of the outgoing arguments. The default is correct for most
2908 machines. See @file{function.c} for details.
2911 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2912 A C expression whose value is RTL representing the address in a stack
2913 frame where the pointer to the caller's frame is stored. Assume that
2914 @var{frameaddr} is an RTL expression for the address of the stack frame
2917 If you don't define this macro, the default is to return the value
2918 of @var{frameaddr}---that is, the stack frame address is also the
2919 address of the stack word that points to the previous frame.
2922 @defmac SETUP_FRAME_ADDRESSES
2923 If defined, a C expression that produces the machine-specific code to
2924 setup the stack so that arbitrary frames can be accessed. For example,
2925 on the SPARC, we must flush all of the register windows to the stack
2926 before we can access arbitrary stack frames. You will seldom need to
2930 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
2931 This target hook should return an rtx that is used to store
2932 the address of the current frame into the built in @code{setjmp} buffer.
2933 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2934 machines. One reason you may need to define this target hook is if
2935 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2938 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2939 A C expression whose value is RTL representing the value of the return
2940 address for the frame @var{count} steps up from the current frame, after
2941 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2942 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2943 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2945 The value of the expression must always be the correct address when
2946 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2947 determine the return address of other frames.
2950 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2951 Define this if the return address of a particular stack frame is accessed
2952 from the frame pointer of the previous stack frame.
2955 @defmac INCOMING_RETURN_ADDR_RTX
2956 A C expression whose value is RTL representing the location of the
2957 incoming return address at the beginning of any function, before the
2958 prologue. This RTL is either a @code{REG}, indicating that the return
2959 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2962 You only need to define this macro if you want to support call frame
2963 debugging information like that provided by DWARF 2.
2965 If this RTL is a @code{REG}, you should also define
2966 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2969 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2970 A C expression whose value is an integer giving a DWARF 2 column
2971 number that may be used as an alternate return column. This should
2972 be defined only if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2973 general register, but an alternate column needs to be used for
2977 @defmac INCOMING_FRAME_SP_OFFSET
2978 A C expression whose value is an integer giving the offset, in bytes,
2979 from the value of the stack pointer register to the top of the stack
2980 frame at the beginning of any function, before the prologue. The top of
2981 the frame is defined to be the value of the stack pointer in the
2982 previous frame, just before the call instruction.
2984 You only need to define this macro if you want to support call frame
2985 debugging information like that provided by DWARF 2.
2988 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2989 A C expression whose value is an integer giving the offset, in bytes,
2990 from the argument pointer to the canonical frame address (cfa). The
2991 final value should coincide with that calculated by
2992 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2993 during virtual register instantiation.
2995 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2996 which is correct for most machines; in general, the arguments are found
2997 immediately before the stack frame. Note that this is not the case on
2998 some targets that save registers into the caller's frame, such as SPARC
2999 and rs6000, and so such targets need to define this macro.
3001 You only need to define this macro if the default is incorrect, and you
3002 want to support call frame debugging information like that provided by
3006 @node Exception Handling
3007 @subsection Exception Handling Support
3008 @cindex exception handling
3010 @defmac EH_RETURN_DATA_REGNO (@var{N})
3011 A C expression whose value is the @var{N}th register number used for
3012 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3013 @var{N} registers are usable.
3015 The exception handling library routines communicate with the exception
3016 handlers via a set of agreed upon registers. Ideally these registers
3017 should be call-clobbered; it is possible to use call-saved registers,
3018 but may negatively impact code size. The target must support at least
3019 2 data registers, but should define 4 if there are enough free registers.
3021 You must define this macro if you want to support call frame exception
3022 handling like that provided by DWARF 2.
3025 @defmac EH_RETURN_STACKADJ_RTX
3026 A C expression whose value is RTL representing a location in which
3027 to store a stack adjustment to be applied before function return.
3028 This is used to unwind the stack to an exception handler's call frame.
3029 It will be assigned zero on code paths that return normally.
3031 Typically this is a call-clobbered hard register that is otherwise
3032 untouched by the epilogue, but could also be a stack slot.
3034 Do not define this macro if the stack pointer is saved and restored
3035 by the regular prolog and epilog code in the call frame itself; in
3036 this case, the exception handling library routines will update the
3037 stack location to be restored in place. Otherwise, you must define
3038 this macro if you want to support call frame exception handling like
3039 that provided by DWARF 2.
3042 @defmac EH_RETURN_HANDLER_RTX
3043 A C expression whose value is RTL representing a location in which
3044 to store the address of an exception handler to which we should
3045 return. It will not be assigned on code paths that return normally.
3047 Typically this is the location in the call frame at which the normal
3048 return address is stored. For targets that return by popping an
3049 address off the stack, this might be a memory address just below
3050 the @emph{target} call frame rather than inside the current call
3051 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3052 been assigned, so it may be used to calculate the location of the
3055 Some targets have more complex requirements than storing to an
3056 address calculable during initial code generation. In that case
3057 the @code{eh_return} instruction pattern should be used instead.
3059 If you want to support call frame exception handling, you must
3060 define either this macro or the @code{eh_return} instruction pattern.
3063 @defmac RETURN_ADDR_OFFSET
3064 If defined, an integer-valued C expression for which rtl will be generated
3065 to add it to the exception handler address before it is searched in the
3066 exception handling tables, and to subtract it again from the address before
3067 using it to return to the exception handler.
3070 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3071 This macro chooses the encoding of pointers embedded in the exception
3072 handling sections. If at all possible, this should be defined such
3073 that the exception handling section will not require dynamic relocations,
3074 and so may be read-only.
3076 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3077 @var{global} is true if the symbol may be affected by dynamic relocations.
3078 The macro should return a combination of the @code{DW_EH_PE_*} defines
3079 as found in @file{dwarf2.h}.
3081 If this macro is not defined, pointers will not be encoded but
3082 represented directly.
3085 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3086 This macro allows the target to emit whatever special magic is required
3087 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3088 Generic code takes care of pc-relative and indirect encodings; this must
3089 be defined if the target uses text-relative or data-relative encodings.
3091 This is a C statement that branches to @var{done} if the format was
3092 handled. @var{encoding} is the format chosen, @var{size} is the number
3093 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3097 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs}, @var{success})
3098 This macro allows the target to add cpu and operating system specific
3099 code to the call-frame unwinder for use when there is no unwind data
3100 available. The most common reason to implement this macro is to unwind
3101 through signal frames.
3103 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3104 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3105 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3106 for the address of the code being executed and @code{context->cfa} for
3107 the stack pointer value. If the frame can be decoded, the register save
3108 addresses should be updated in @var{fs} and the macro should branch to
3109 @var{success}. If the frame cannot be decoded, the macro should do
3112 For proper signal handling in Java this macro is accompanied by
3113 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3116 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3117 This macro allows the target to add operating system specific code to the
3118 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3119 usually used for signal or interrupt frames.
3121 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3122 @var{context} is an @code{_Unwind_Context};
3123 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3124 for the abi and context in the @code{.unwabi} directive. If the
3125 @code{.unwabi} directive can be handled, the register save addresses should
3126 be updated in @var{fs}.
3129 @defmac TARGET_USES_WEAK_UNWIND_INFO
3130 A C expression that evaluates to true if the target requires unwind
3131 info to be given comdat linkage. Define it to be @code{1} if comdat
3132 linkage is necessary. The default is @code{0}.
3135 @node Stack Checking
3136 @subsection Specifying How Stack Checking is Done
3138 GCC will check that stack references are within the boundaries of
3139 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3143 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3144 will assume that you have arranged for stack checking to be done at
3145 appropriate places in the configuration files, e.g., in
3146 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3150 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3151 called @code{check_stack} in your @file{md} file, GCC will call that
3152 pattern with one argument which is the address to compare the stack
3153 value against. You must arrange for this pattern to report an error if
3154 the stack pointer is out of range.
3157 If neither of the above are true, GCC will generate code to periodically
3158 ``probe'' the stack pointer using the values of the macros defined below.
3161 Normally, you will use the default values of these macros, so GCC
3162 will use the third approach.
3164 @defmac STACK_CHECK_BUILTIN
3165 A nonzero value if stack checking is done by the configuration files in a
3166 machine-dependent manner. You should define this macro if stack checking
3167 is require by the ABI of your machine or if you would like to have to stack
3168 checking in some more efficient way than GCC's portable approach.
3169 The default value of this macro is zero.
3172 @defmac STACK_CHECK_PROBE_INTERVAL
3173 An integer representing the interval at which GCC must generate stack
3174 probe instructions. You will normally define this macro to be no larger
3175 than the size of the ``guard pages'' at the end of a stack area. The
3176 default value of 4096 is suitable for most systems.
3179 @defmac STACK_CHECK_PROBE_LOAD
3180 A integer which is nonzero if GCC should perform the stack probe
3181 as a load instruction and zero if GCC should use a store instruction.
3182 The default is zero, which is the most efficient choice on most systems.
3185 @defmac STACK_CHECK_PROTECT
3186 The number of bytes of stack needed to recover from a stack overflow,
3187 for languages where such a recovery is supported. The default value of
3188 75 words should be adequate for most machines.
3191 @defmac STACK_CHECK_MAX_FRAME_SIZE
3192 The maximum size of a stack frame, in bytes. GCC will generate probe
3193 instructions in non-leaf functions to ensure at least this many bytes of
3194 stack are available. If a stack frame is larger than this size, stack
3195 checking will not be reliable and GCC will issue a warning. The
3196 default is chosen so that GCC only generates one instruction on most
3197 systems. You should normally not change the default value of this macro.
3200 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3201 GCC uses this value to generate the above warning message. It
3202 represents the amount of fixed frame used by a function, not including
3203 space for any callee-saved registers, temporaries and user variables.
3204 You need only specify an upper bound for this amount and will normally
3205 use the default of four words.
3208 @defmac STACK_CHECK_MAX_VAR_SIZE
3209 The maximum size, in bytes, of an object that GCC will place in the
3210 fixed area of the stack frame when the user specifies
3211 @option{-fstack-check}.
3212 GCC computed the default from the values of the above macros and you will
3213 normally not need to override that default.
3217 @node Frame Registers
3218 @subsection Registers That Address the Stack Frame
3220 @c prevent bad page break with this line
3221 This discusses registers that address the stack frame.
3223 @defmac STACK_POINTER_REGNUM
3224 The register number of the stack pointer register, which must also be a
3225 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3226 the hardware determines which register this is.
3229 @defmac FRAME_POINTER_REGNUM
3230 The register number of the frame pointer register, which is used to
3231 access automatic variables in the stack frame. On some machines, the
3232 hardware determines which register this is. On other machines, you can
3233 choose any register you wish for this purpose.
3236 @defmac HARD_FRAME_POINTER_REGNUM
3237 On some machines the offset between the frame pointer and starting
3238 offset of the automatic variables is not known until after register
3239 allocation has been done (for example, because the saved registers are
3240 between these two locations). On those machines, define
3241 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3242 be used internally until the offset is known, and define
3243 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3244 used for the frame pointer.
3246 You should define this macro only in the very rare circumstances when it
3247 is not possible to calculate the offset between the frame pointer and
3248 the automatic variables until after register allocation has been
3249 completed. When this macro is defined, you must also indicate in your
3250 definition of @code{ELIMINABLE_REGS} how to eliminate
3251 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3252 or @code{STACK_POINTER_REGNUM}.
3254 Do not define this macro if it would be the same as
3255 @code{FRAME_POINTER_REGNUM}.
3258 @defmac ARG_POINTER_REGNUM
3259 The register number of the arg pointer register, which is used to access
3260 the function's argument list. On some machines, this is the same as the
3261 frame pointer register. On some machines, the hardware determines which
3262 register this is. On other machines, you can choose any register you
3263 wish for this purpose. If this is not the same register as the frame
3264 pointer register, then you must mark it as a fixed register according to
3265 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3266 (@pxref{Elimination}).
3269 @defmac RETURN_ADDRESS_POINTER_REGNUM
3270 The register number of the return address pointer register, which is used to
3271 access the current function's return address from the stack. On some
3272 machines, the return address is not at a fixed offset from the frame
3273 pointer or stack pointer or argument pointer. This register can be defined
3274 to point to the return address on the stack, and then be converted by
3275 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3277 Do not define this macro unless there is no other way to get the return
3278 address from the stack.
3281 @defmac STATIC_CHAIN_REGNUM
3282 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3283 Register numbers used for passing a function's static chain pointer. If
3284 register windows are used, the register number as seen by the called
3285 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3286 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3287 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3290 The static chain register need not be a fixed register.
3292 If the static chain is passed in memory, these macros should not be
3293 defined; instead, the next two macros should be defined.
3296 @defmac STATIC_CHAIN
3297 @defmacx STATIC_CHAIN_INCOMING
3298 If the static chain is passed in memory, these macros provide rtx giving
3299 @code{mem} expressions that denote where they are stored.
3300 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3301 as seen by the calling and called functions, respectively. Often the former
3302 will be at an offset from the stack pointer and the latter at an offset from
3305 @findex stack_pointer_rtx
3306 @findex frame_pointer_rtx
3307 @findex arg_pointer_rtx
3308 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3309 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3310 macros and should be used to refer to those items.
3312 If the static chain is passed in a register, the two previous macros should
3316 @defmac DWARF_FRAME_REGISTERS
3317 This macro specifies the maximum number of hard registers that can be
3318 saved in a call frame. This is used to size data structures used in
3319 DWARF2 exception handling.
3321 Prior to GCC 3.0, this macro was needed in order to establish a stable
3322 exception handling ABI in the face of adding new hard registers for ISA
3323 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3324 in the number of hard registers. Nevertheless, this macro can still be
3325 used to reduce the runtime memory requirements of the exception handling
3326 routines, which can be substantial if the ISA contains a lot of
3327 registers that are not call-saved.
3329 If this macro is not defined, it defaults to
3330 @code{FIRST_PSEUDO_REGISTER}.
3333 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3335 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3336 for backward compatibility in pre GCC 3.0 compiled code.
3338 If this macro is not defined, it defaults to
3339 @code{DWARF_FRAME_REGISTERS}.
3342 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3344 Define this macro if the target's representation for dwarf registers
3345 is different than the internal representation for unwind column.
3346 Given a dwarf register, this macro should return the internal unwind
3347 column number to use instead.
3349 See the PowerPC's SPE target for an example.
3352 @defmac DWARF_FRAME_REGNUM (@var{regno})
3354 Define this macro if the target's representation for dwarf registers
3355 used in .eh_frame or .debug_frame is different from that used in other
3356 debug info sections. Given a GCC hard register number, this macro
3357 should return the .eh_frame register number. The default is
3358 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3362 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3364 Define this macro to map register numbers held in the call frame info
3365 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3366 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3367 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3368 return @code{@var{regno}}.
3373 @subsection Eliminating Frame Pointer and Arg Pointer
3375 @c prevent bad page break with this line
3376 This is about eliminating the frame pointer and arg pointer.
3378 @defmac FRAME_POINTER_REQUIRED
3379 A C expression which is nonzero if a function must have and use a frame
3380 pointer. This expression is evaluated in the reload pass. If its value is
3381 nonzero the function will have a frame pointer.
3383 The expression can in principle examine the current function and decide
3384 according to the facts, but on most machines the constant 0 or the
3385 constant 1 suffices. Use 0 when the machine allows code to be generated
3386 with no frame pointer, and doing so saves some time or space. Use 1
3387 when there is no possible advantage to avoiding a frame pointer.
3389 In certain cases, the compiler does not know how to produce valid code
3390 without a frame pointer. The compiler recognizes those cases and
3391 automatically gives the function a frame pointer regardless of what
3392 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3395 In a function that does not require a frame pointer, the frame pointer
3396 register can be allocated for ordinary usage, unless you mark it as a
3397 fixed register. See @code{FIXED_REGISTERS} for more information.
3400 @findex get_frame_size
3401 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3402 A C statement to store in the variable @var{depth-var} the difference
3403 between the frame pointer and the stack pointer values immediately after
3404 the function prologue. The value would be computed from information
3405 such as the result of @code{get_frame_size ()} and the tables of
3406 registers @code{regs_ever_live} and @code{call_used_regs}.
3408 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3409 need not be defined. Otherwise, it must be defined even if
3410 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3411 case, you may set @var{depth-var} to anything.
3414 @defmac ELIMINABLE_REGS
3415 If defined, this macro specifies a table of register pairs used to
3416 eliminate unneeded registers that point into the stack frame. If it is not
3417 defined, the only elimination attempted by the compiler is to replace
3418 references to the frame pointer with references to the stack pointer.
3420 The definition of this macro is a list of structure initializations, each
3421 of which specifies an original and replacement register.
3423 On some machines, the position of the argument pointer is not known until
3424 the compilation is completed. In such a case, a separate hard register
3425 must be used for the argument pointer. This register can be eliminated by
3426 replacing it with either the frame pointer or the argument pointer,
3427 depending on whether or not the frame pointer has been eliminated.
3429 In this case, you might specify:
3431 #define ELIMINABLE_REGS \
3432 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3433 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3434 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3437 Note that the elimination of the argument pointer with the stack pointer is
3438 specified first since that is the preferred elimination.
3441 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3442 A C expression that returns nonzero if the compiler is allowed to try
3443 to replace register number @var{from-reg} with register number
3444 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3445 is defined, and will usually be the constant 1, since most of the cases
3446 preventing register elimination are things that the compiler already
3450 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3451 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3452 specifies the initial difference between the specified pair of
3453 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3457 @node Stack Arguments
3458 @subsection Passing Function Arguments on the Stack
3459 @cindex arguments on stack
3460 @cindex stack arguments
3462 The macros in this section control how arguments are passed
3463 on the stack. See the following section for other macros that
3464 control passing certain arguments in registers.
3466 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3467 This target hook returns @code{true} if an argument declared in a
3468 prototype as an integral type smaller than @code{int} should actually be
3469 passed as an @code{int}. In addition to avoiding errors in certain
3470 cases of mismatch, it also makes for better code on certain machines.
3471 The default is to not promote prototypes.
3475 A C expression. If nonzero, push insns will be used to pass
3477 If the target machine does not have a push instruction, set it to zero.
3478 That directs GCC to use an alternate strategy: to
3479 allocate the entire argument block and then store the arguments into
3480 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3483 @defmac PUSH_ARGS_REVERSED
3484 A C expression. If nonzero, function arguments will be evaluated from
3485 last to first, rather than from first to last. If this macro is not
3486 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3487 and args grow in opposite directions, and 0 otherwise.
3490 @defmac PUSH_ROUNDING (@var{npushed})
3491 A C expression that is the number of bytes actually pushed onto the
3492 stack when an instruction attempts to push @var{npushed} bytes.
3494 On some machines, the definition
3497 #define PUSH_ROUNDING(BYTES) (BYTES)
3501 will suffice. But on other machines, instructions that appear
3502 to push one byte actually push two bytes in an attempt to maintain
3503 alignment. Then the definition should be
3506 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3510 @findex current_function_outgoing_args_size
3511 @defmac ACCUMULATE_OUTGOING_ARGS
3512 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3513 will be computed and placed into the variable
3514 @code{current_function_outgoing_args_size}. No space will be pushed
3515 onto the stack for each call; instead, the function prologue should
3516 increase the stack frame size by this amount.
3518 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3522 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3523 Define this macro if functions should assume that stack space has been
3524 allocated for arguments even when their values are passed in
3527 The value of this macro is the size, in bytes, of the area reserved for
3528 arguments passed in registers for the function represented by @var{fndecl},
3529 which can be zero if GCC is calling a library function.
3531 This space can be allocated by the caller, or be a part of the
3532 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3535 @c above is overfull. not sure what to do. --mew 5feb93 did
3536 @c something, not sure if it looks good. --mew 10feb93
3538 @defmac OUTGOING_REG_PARM_STACK_SPACE
3539 Define this if it is the responsibility of the caller to allocate the area
3540 reserved for arguments passed in registers.
3542 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3543 whether the space for these arguments counts in the value of
3544 @code{current_function_outgoing_args_size}.
3547 @defmac STACK_PARMS_IN_REG_PARM_AREA
3548 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3549 stack parameters don't skip the area specified by it.
3550 @c i changed this, makes more sens and it should have taken care of the
3551 @c overfull.. not as specific, tho. --mew 5feb93
3553 Normally, when a parameter is not passed in registers, it is placed on the
3554 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3555 suppresses this behavior and causes the parameter to be passed on the
3556 stack in its natural location.
3559 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3560 A C expression that should indicate the number of bytes of its own
3561 arguments that a function pops on returning, or 0 if the
3562 function pops no arguments and the caller must therefore pop them all
3563 after the function returns.
3565 @var{fundecl} is a C variable whose value is a tree node that describes
3566 the function in question. Normally it is a node of type
3567 @code{FUNCTION_DECL} that describes the declaration of the function.
3568 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3570 @var{funtype} is a C variable whose value is a tree node that
3571 describes the function in question. Normally it is a node of type
3572 @code{FUNCTION_TYPE} that describes the data type of the function.
3573 From this it is possible to obtain the data types of the value and
3574 arguments (if known).
3576 When a call to a library function is being considered, @var{fundecl}
3577 will contain an identifier node for the library function. Thus, if
3578 you need to distinguish among various library functions, you can do so
3579 by their names. Note that ``library function'' in this context means
3580 a function used to perform arithmetic, whose name is known specially
3581 in the compiler and was not mentioned in the C code being compiled.
3583 @var{stack-size} is the number of bytes of arguments passed on the
3584 stack. If a variable number of bytes is passed, it is zero, and
3585 argument popping will always be the responsibility of the calling function.
3587 On the VAX, all functions always pop their arguments, so the definition
3588 of this macro is @var{stack-size}. On the 68000, using the standard
3589 calling convention, no functions pop their arguments, so the value of
3590 the macro is always 0 in this case. But an alternative calling
3591 convention is available in which functions that take a fixed number of
3592 arguments pop them but other functions (such as @code{printf}) pop
3593 nothing (the caller pops all). When this convention is in use,
3594 @var{funtype} is examined to determine whether a function takes a fixed
3595 number of arguments.
3598 @defmac CALL_POPS_ARGS (@var{cum})
3599 A C expression that should indicate the number of bytes a call sequence
3600 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3601 when compiling a function call.
3603 @var{cum} is the variable in which all arguments to the called function
3604 have been accumulated.
3606 On certain architectures, such as the SH5, a call trampoline is used
3607 that pops certain registers off the stack, depending on the arguments
3608 that have been passed to the function. Since this is a property of the
3609 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3613 @node Register Arguments
3614 @subsection Passing Arguments in Registers
3615 @cindex arguments in registers
3616 @cindex registers arguments
3618 This section describes the macros which let you control how various
3619 types of arguments are passed in registers or how they are arranged in
3622 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3623 A C expression that controls whether a function argument is passed
3624 in a register, and which register.
3626 The arguments are @var{cum}, which summarizes all the previous
3627 arguments; @var{mode}, the machine mode of the argument; @var{type},
3628 the data type of the argument as a tree node or 0 if that is not known
3629 (which happens for C support library functions); and @var{named},
3630 which is 1 for an ordinary argument and 0 for nameless arguments that
3631 correspond to @samp{@dots{}} in the called function's prototype.
3632 @var{type} can be an incomplete type if a syntax error has previously
3635 The value of the expression is usually either a @code{reg} RTX for the
3636 hard register in which to pass the argument, or zero to pass the
3637 argument on the stack.
3639 For machines like the VAX and 68000, where normally all arguments are
3640 pushed, zero suffices as a definition.
3642 The value of the expression can also be a @code{parallel} RTX@. This is
3643 used when an argument is passed in multiple locations. The mode of the
3644 @code{parallel} should be the mode of the entire argument. The
3645 @code{parallel} holds any number of @code{expr_list} pairs; each one
3646 describes where part of the argument is passed. In each
3647 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3648 register in which to pass this part of the argument, and the mode of the
3649 register RTX indicates how large this part of the argument is. The
3650 second operand of the @code{expr_list} is a @code{const_int} which gives
3651 the offset in bytes into the entire argument of where this part starts.
3652 As a special exception the first @code{expr_list} in the @code{parallel}
3653 RTX may have a first operand of zero. This indicates that the entire
3654 argument is also stored on the stack.
3656 The last time this macro is called, it is called with @code{MODE ==
3657 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3658 pattern as operands 2 and 3 respectively.
3660 @cindex @file{stdarg.h} and register arguments
3661 The usual way to make the ISO library @file{stdarg.h} work on a machine
3662 where some arguments are usually passed in registers, is to cause
3663 nameless arguments to be passed on the stack instead. This is done
3664 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3666 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3667 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3668 You may use the hook @code{targetm.calls.must_pass_in_stack}
3669 in the definition of this macro to determine if this argument is of a
3670 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3671 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3672 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3673 defined, the argument will be computed in the stack and then loaded into
3677 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type})
3678 This target hook should return @code{true} if we should not pass @var{type}
3679 solely in registers. The file @file{expr.h} defines a
3680 definition that is usually appropriate, refer to @file{expr.h} for additional
3684 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3685 Define this macro if the target machine has ``register windows'', so
3686 that the register in which a function sees an arguments is not
3687 necessarily the same as the one in which the caller passed the
3690 For such machines, @code{FUNCTION_ARG} computes the register in which
3691 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3692 be defined in a similar fashion to tell the function being called
3693 where the arguments will arrive.
3695 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3696 serves both purposes.
3699 @defmac FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3700 A C expression for the number of words, at the beginning of an
3701 argument, that must be put in registers. The value must be zero for
3702 arguments that are passed entirely in registers or that are entirely
3703 pushed on the stack.
3705 On some machines, certain arguments must be passed partially in
3706 registers and partially in memory. On these machines, typically the
3707 first @var{n} words of arguments are passed in registers, and the rest
3708 on the stack. If a multi-word argument (a @code{double} or a
3709 structure) crosses that boundary, its first few words must be passed
3710 in registers and the rest must be pushed. This macro tells the
3711 compiler when this occurs, and how many of the words should go in
3714 @code{FUNCTION_ARG} for these arguments should return the first
3715 register to be used by the caller for this argument; likewise
3716 @code{FUNCTION_INCOMING_ARG}, for the called function.
3719 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3720 This target hook should return @code{true} if an argument at the
3721 position indicated by @var{cum} should be passed by reference. This
3722 predicate is queried after target independent reasons for being
3723 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
3725 If the hook returns true, a copy of that argument is made in memory and a
3726 pointer to the argument is passed instead of the argument itself.
3727 The pointer is passed in whatever way is appropriate for passing a pointer
3731 @defmac FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3732 If defined, a C expression that indicates when it is the called function's
3733 responsibility to make a copy of arguments passed by invisible reference.
3734 Normally, the caller makes a copy and passes the address of the copy to the
3735 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3736 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3737 ``live'' value. The called function must not modify this value. If it can be
3738 determined that the value won't be modified, it need not make a copy;
3739 otherwise a copy must be made.
3742 @defmac CUMULATIVE_ARGS
3743 A C type for declaring a variable that is used as the first argument of
3744 @code{FUNCTION_ARG} and other related values. For some target machines,
3745 the type @code{int} suffices and can hold the number of bytes of
3748 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3749 arguments that have been passed on the stack. The compiler has other
3750 variables to keep track of that. For target machines on which all
3751 arguments are passed on the stack, there is no need to store anything in
3752 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3753 should not be empty, so use @code{int}.
3756 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3757 A C statement (sans semicolon) for initializing the variable
3758 @var{cum} for the state at the beginning of the argument list. The
3759 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3760 is the tree node for the data type of the function which will receive
3761 the args, or 0 if the args are to a compiler support library function.
3762 For direct calls that are not libcalls, @var{fndecl} contain the
3763 declaration node of the function. @var{fndecl} is also set when
3764 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3765 being compiled. @var{n_named_args} is set to the number of named
3766 arguments, including a structure return address if it is passed as a
3767 parameter, when making a call. When processing incoming arguments,
3768 @var{n_named_args} is set to -1.
3770 When processing a call to a compiler support library function,
3771 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3772 contains the name of the function, as a string. @var{libname} is 0 when
3773 an ordinary C function call is being processed. Thus, each time this
3774 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3775 never both of them at once.
3778 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3779 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3780 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3781 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3782 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3783 0)} is used instead.
3786 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3787 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3788 finding the arguments for the function being compiled. If this macro is
3789 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3791 The value passed for @var{libname} is always 0, since library routines
3792 with special calling conventions are never compiled with GCC@. The
3793 argument @var{libname} exists for symmetry with
3794 @code{INIT_CUMULATIVE_ARGS}.
3795 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3796 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3799 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3800 A C statement (sans semicolon) to update the summarizer variable
3801 @var{cum} to advance past an argument in the argument list. The
3802 values @var{mode}, @var{type} and @var{named} describe that argument.
3803 Once this is done, the variable @var{cum} is suitable for analyzing
3804 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3806 This macro need not do anything if the argument in question was passed
3807 on the stack. The compiler knows how to track the amount of stack space
3808 used for arguments without any special help.
3811 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3812 If defined, a C expression which determines whether, and in which direction,
3813 to pad out an argument with extra space. The value should be of type
3814 @code{enum direction}: either @code{upward} to pad above the argument,
3815 @code{downward} to pad below, or @code{none} to inhibit padding.
3817 The @emph{amount} of padding is always just enough to reach the next
3818 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3821 This macro has a default definition which is right for most systems.
3822 For little-endian machines, the default is to pad upward. For
3823 big-endian machines, the default is to pad downward for an argument of
3824 constant size shorter than an @code{int}, and upward otherwise.
3827 @defmac PAD_VARARGS_DOWN
3828 If defined, a C expression which determines whether the default
3829 implementation of va_arg will attempt to pad down before reading the
3830 next argument, if that argument is smaller than its aligned space as
3831 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3832 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3835 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3836 Specify padding for the last element of a block move between registers and
3837 memory. @var{first} is nonzero if this is the only element. Defining this
3838 macro allows better control of register function parameters on big-endian
3839 machines, without using @code{PARALLEL} rtl. In particular,
3840 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3841 registers, as there is no longer a "wrong" part of a register; For example,
3842 a three byte aggregate may be passed in the high part of a register if so
3846 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3847 If defined, a C expression that gives the alignment boundary, in bits,
3848 of an argument with the specified mode and type. If it is not defined,
3849 @code{PARM_BOUNDARY} is used for all arguments.
3852 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3853 A C expression that is nonzero if @var{regno} is the number of a hard
3854 register in which function arguments are sometimes passed. This does
3855 @emph{not} include implicit arguments such as the static chain and
3856 the structure-value address. On many machines, no registers can be
3857 used for this purpose since all function arguments are pushed on the
3861 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
3862 This hook should return true if parameter of type @var{type} are passed
3863 as two scalar parameters. By default, GCC will attempt to pack complex
3864 arguments into the target's word size. Some ABIs require complex arguments
3865 to be split and treated as their individual components. For example, on
3866 AIX64, complex floats should be passed in a pair of floating point
3867 registers, even though a complex float would fit in one 64-bit floating
3870 The default value of this hook is @code{NULL}, which is treated as always
3874 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
3875 This hook performs target-specific gimplification of
3876 @code{VA_ARG_EXPR}. The first two parameters correspond to the
3877 arguments to @code{va_arg}; the latter two are as in
3878 @code{gimplify.c:gimplify_expr}.
3881 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
3882 Define this to return nonzero if the port is prepared to handle
3883 insns involving scalar mode @var{mode}. For a scalar mode to be
3884 considered supported, all the basic arithmetic and comparisons
3887 The default version of this hook returns true for any mode
3888 required to handle the basic C types (as defined by the port).
3889 Included here are the double-word arithmetic supported by the
3890 code in @file{optabs.c}.
3893 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
3894 Define this to return nonzero if the port is prepared to handle
3895 insns involving vector mode @var{mode}. At the very least, it
3896 must have move patterns for this mode.
3900 @subsection How Scalar Function Values Are Returned
3901 @cindex return values in registers
3902 @cindex values, returned by functions
3903 @cindex scalars, returned as values
3905 This section discusses the macros that control returning scalars as
3906 values---values that can fit in registers.
3908 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3909 A C expression to create an RTX representing the place where a
3910 function returns a value of data type @var{valtype}. @var{valtype} is
3911 a tree node representing a data type. Write @code{TYPE_MODE
3912 (@var{valtype})} to get the machine mode used to represent that type.
3913 On many machines, only the mode is relevant. (Actually, on most
3914 machines, scalar values are returned in the same place regardless of
3917 The value of the expression is usually a @code{reg} RTX for the hard
3918 register where the return value is stored. The value can also be a
3919 @code{parallel} RTX, if the return value is in multiple places. See
3920 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3922 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply the same
3923 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3926 If the precise function being called is known, @var{func} is a tree
3927 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3928 pointer. This makes it possible to use a different value-returning
3929 convention for specific functions when all their calls are
3932 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3933 types, because these are returned in another way. See
3934 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
3937 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3938 Define this macro if the target machine has ``register windows''
3939 so that the register in which a function returns its value is not
3940 the same as the one in which the caller sees the value.
3942 For such machines, @code{FUNCTION_VALUE} computes the register in which
3943 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3944 defined in a similar fashion to tell the function where to put the
3947 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3948 @code{FUNCTION_VALUE} serves both purposes.
3950 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3951 aggregate data types, because these are returned in another way. See
3952 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
3955 @defmac LIBCALL_VALUE (@var{mode})
3956 A C expression to create an RTX representing the place where a library
3957 function returns a value of mode @var{mode}. If the precise function
3958 being called is known, @var{func} is a tree node
3959 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3960 pointer. This makes it possible to use a different value-returning
3961 convention for specific functions when all their calls are
3964 Note that ``library function'' in this context means a compiler
3965 support routine, used to perform arithmetic, whose name is known
3966 specially by the compiler and was not mentioned in the C code being
3969 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3970 data types, because none of the library functions returns such types.
3973 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3974 A C expression that is nonzero if @var{regno} is the number of a hard
3975 register in which the values of called function may come back.
3977 A register whose use for returning values is limited to serving as the
3978 second of a pair (for a value of type @code{double}, say) need not be
3979 recognized by this macro. So for most machines, this definition
3983 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3986 If the machine has register windows, so that the caller and the called
3987 function use different registers for the return value, this macro
3988 should recognize only the caller's register numbers.
3991 @defmac APPLY_RESULT_SIZE
3992 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3993 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3994 saving and restoring an arbitrary return value.
3997 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
3998 This hook should return true if values of type @var{type} are returned
3999 at the most significant end of a register (in other words, if they are
4000 padded at the least significant end). You can assume that @var{type}
4001 is returned in a register; the caller is required to check this.
4003 Note that the register provided by @code{FUNCTION_VALUE} must be able
4004 to hold the complete return value. For example, if a 1-, 2- or 3-byte
4005 structure is returned at the most significant end of a 4-byte register,
4006 @code{FUNCTION_VALUE} should provide an @code{SImode} rtx.
4009 @node Aggregate Return
4010 @subsection How Large Values Are Returned
4011 @cindex aggregates as return values
4012 @cindex large return values
4013 @cindex returning aggregate values
4014 @cindex structure value address
4016 When a function value's mode is @code{BLKmode} (and in some other
4017 cases), the value is not returned according to @code{FUNCTION_VALUE}
4018 (@pxref{Scalar Return}). Instead, the caller passes the address of a
4019 block of memory in which the value should be stored. This address
4020 is called the @dfn{structure value address}.
4022 This section describes how to control returning structure values in
4025 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
4026 This target hook should return a nonzero value to say to return the
4027 function value in memory, just as large structures are always returned.
4028 Here @var{type} will be the data type of the value, and @var{fntype}
4029 will be the type of the function doing the returning, or @code{NULL} for
4032 Note that values of mode @code{BLKmode} must be explicitly handled
4033 by this function. Also, the option @option{-fpcc-struct-return}
4034 takes effect regardless of this macro. On most systems, it is
4035 possible to leave the hook undefined; this causes a default
4036 definition to be used, whose value is the constant 1 for @code{BLKmode}
4037 values, and 0 otherwise.
4039 Do not use this hook to indicate that structures and unions should always
4040 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4044 @defmac DEFAULT_PCC_STRUCT_RETURN
4045 Define this macro to be 1 if all structure and union return values must be
4046 in memory. Since this results in slower code, this should be defined
4047 only if needed for compatibility with other compilers or with an ABI@.
4048 If you define this macro to be 0, then the conventions used for structure
4049 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4052 If not defined, this defaults to the value 1.
4055 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4056 This target hook should return the location of the structure value
4057 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4058 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4059 be @code{NULL}, for libcalls. You do not need to define this target
4060 hook if the address is always passed as an ``invisible'' first
4063 On some architectures the place where the structure value address
4064 is found by the called function is not the same place that the
4065 caller put it. This can be due to register windows, or it could
4066 be because the function prologue moves it to a different place.
4067 @var{incoming} is @code{true} when the location is needed in
4068 the context of the called function, and @code{false} in the context of
4071 If @var{incoming} is @code{true} and the address is to be found on the
4072 stack, return a @code{mem} which refers to the frame pointer.
4075 @defmac PCC_STATIC_STRUCT_RETURN
4076 Define this macro if the usual system convention on the target machine
4077 for returning structures and unions is for the called function to return
4078 the address of a static variable containing the value.
4080 Do not define this if the usual system convention is for the caller to
4081 pass an address to the subroutine.
4083 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4084 nothing when you use @option{-freg-struct-return} mode.
4088 @subsection Caller-Saves Register Allocation
4090 If you enable it, GCC can save registers around function calls. This
4091 makes it possible to use call-clobbered registers to hold variables that
4092 must live across calls.
4094 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4095 A C expression to determine whether it is worthwhile to consider placing
4096 a pseudo-register in a call-clobbered hard register and saving and
4097 restoring it around each function call. The expression should be 1 when
4098 this is worth doing, and 0 otherwise.
4100 If you don't define this macro, a default is used which is good on most
4101 machines: @code{4 * @var{calls} < @var{refs}}.
4104 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4105 A C expression specifying which mode is required for saving @var{nregs}
4106 of a pseudo-register in call-clobbered hard register @var{regno}. If
4107 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4108 returned. For most machines this macro need not be defined since GCC
4109 will select the smallest suitable mode.
4112 @node Function Entry
4113 @subsection Function Entry and Exit
4114 @cindex function entry and exit
4118 This section describes the macros that output function entry
4119 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4121 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4122 If defined, a function that outputs the assembler code for entry to a
4123 function. The prologue is responsible for setting up the stack frame,
4124 initializing the frame pointer register, saving registers that must be
4125 saved, and allocating @var{size} additional bytes of storage for the
4126 local variables. @var{size} is an integer. @var{file} is a stdio
4127 stream to which the assembler code should be output.
4129 The label for the beginning of the function need not be output by this
4130 macro. That has already been done when the macro is run.
4132 @findex regs_ever_live
4133 To determine which registers to save, the macro can refer to the array
4134 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4135 @var{r} is used anywhere within the function. This implies the function
4136 prologue should save register @var{r}, provided it is not one of the
4137 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4138 @code{regs_ever_live}.)
4140 On machines that have ``register windows'', the function entry code does
4141 not save on the stack the registers that are in the windows, even if
4142 they are supposed to be preserved by function calls; instead it takes
4143 appropriate steps to ``push'' the register stack, if any non-call-used
4144 registers are used in the function.
4146 @findex frame_pointer_needed
4147 On machines where functions may or may not have frame-pointers, the
4148 function entry code must vary accordingly; it must set up the frame
4149 pointer if one is wanted, and not otherwise. To determine whether a
4150 frame pointer is in wanted, the macro can refer to the variable
4151 @code{frame_pointer_needed}. The variable's value will be 1 at run
4152 time in a function that needs a frame pointer. @xref{Elimination}.
4154 The function entry code is responsible for allocating any stack space
4155 required for the function. This stack space consists of the regions
4156 listed below. In most cases, these regions are allocated in the
4157 order listed, with the last listed region closest to the top of the
4158 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4159 the highest address if it is not defined). You can use a different order
4160 for a machine if doing so is more convenient or required for
4161 compatibility reasons. Except in cases where required by standard
4162 or by a debugger, there is no reason why the stack layout used by GCC
4163 need agree with that used by other compilers for a machine.
4166 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4167 If defined, a function that outputs assembler code at the end of a
4168 prologue. This should be used when the function prologue is being
4169 emitted as RTL, and you have some extra assembler that needs to be
4170 emitted. @xref{prologue instruction pattern}.
4173 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4174 If defined, a function that outputs assembler code at the start of an
4175 epilogue. This should be used when the function epilogue is being
4176 emitted as RTL, and you have some extra assembler that needs to be
4177 emitted. @xref{epilogue instruction pattern}.
4180 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4181 If defined, a function that outputs the assembler code for exit from a
4182 function. The epilogue is responsible for restoring the saved
4183 registers and stack pointer to their values when the function was
4184 called, and returning control to the caller. This macro takes the
4185 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4186 registers to restore are determined from @code{regs_ever_live} and
4187 @code{CALL_USED_REGISTERS} in the same way.
4189 On some machines, there is a single instruction that does all the work
4190 of returning from the function. On these machines, give that
4191 instruction the name @samp{return} and do not define the macro
4192 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4194 Do not define a pattern named @samp{return} if you want the
4195 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4196 switches to control whether return instructions or epilogues are used,
4197 define a @samp{return} pattern with a validity condition that tests the
4198 target switches appropriately. If the @samp{return} pattern's validity
4199 condition is false, epilogues will be used.
4201 On machines where functions may or may not have frame-pointers, the
4202 function exit code must vary accordingly. Sometimes the code for these
4203 two cases is completely different. To determine whether a frame pointer
4204 is wanted, the macro can refer to the variable
4205 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4206 a function that needs a frame pointer.
4208 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4209 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4210 The C variable @code{current_function_is_leaf} is nonzero for such a
4211 function. @xref{Leaf Functions}.
4213 On some machines, some functions pop their arguments on exit while
4214 others leave that for the caller to do. For example, the 68020 when
4215 given @option{-mrtd} pops arguments in functions that take a fixed
4216 number of arguments.
4218 @findex current_function_pops_args
4219 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4220 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4221 needs to know what was decided. The variable that is called
4222 @code{current_function_pops_args} is the number of bytes of its
4223 arguments that a function should pop. @xref{Scalar Return}.
4224 @c what is the "its arguments" in the above sentence referring to, pray
4225 @c tell? --mew 5feb93
4228 @deftypefn {Target Hook} bool TARGET_LATE_RTL_PROLOGUE_EPILOGUE
4229 If set to @code{true}, it instructs the compiler to emit the RTL prologue
4230 and epilogue later in the game than usual, namely after all passes that
4231 modify the instructions (and not merely reorder them) have been run. In
4232 particular, the C variable @code{current_function_uses_only_leaf_regs} is
4233 valid at that point. This can be used on machines that have "register
4234 windows" to optimize away the regular "push" on the register stack.
4235 @xref{Leaf Functions}.
4240 @findex current_function_pretend_args_size
4241 A region of @code{current_function_pretend_args_size} bytes of
4242 uninitialized space just underneath the first argument arriving on the
4243 stack. (This may not be at the very start of the allocated stack region
4244 if the calling sequence has pushed anything else since pushing the stack
4245 arguments. But usually, on such machines, nothing else has been pushed
4246 yet, because the function prologue itself does all the pushing.) This
4247 region is used on machines where an argument may be passed partly in
4248 registers and partly in memory, and, in some cases to support the
4249 features in @code{<stdarg.h>}.
4252 An area of memory used to save certain registers used by the function.
4253 The size of this area, which may also include space for such things as
4254 the return address and pointers to previous stack frames, is
4255 machine-specific and usually depends on which registers have been used
4256 in the function. Machines with register windows often do not require
4260 A region of at least @var{size} bytes, possibly rounded up to an allocation
4261 boundary, to contain the local variables of the function. On some machines,
4262 this region and the save area may occur in the opposite order, with the
4263 save area closer to the top of the stack.
4266 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4267 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4268 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4269 argument lists of the function. @xref{Stack Arguments}.
4272 @defmac EXIT_IGNORE_STACK
4273 Define this macro as a C expression that is nonzero if the return
4274 instruction or the function epilogue ignores the value of the stack
4275 pointer; in other words, if it is safe to delete an instruction to
4276 adjust the stack pointer before a return from the function. The
4279 Note that this macro's value is relevant only for functions for which
4280 frame pointers are maintained. It is never safe to delete a final
4281 stack adjustment in a function that has no frame pointer, and the
4282 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4285 @defmac EPILOGUE_USES (@var{regno})
4286 Define this macro as a C expression that is nonzero for registers that are
4287 used by the epilogue or the @samp{return} pattern. The stack and frame
4288 pointer registers are already be assumed to be used as needed.
4291 @defmac EH_USES (@var{regno})
4292 Define this macro as a C expression that is nonzero for registers that are
4293 used by the exception handling mechanism, and so should be considered live
4294 on entry to an exception edge.
4297 @defmac DELAY_SLOTS_FOR_EPILOGUE
4298 Define this macro if the function epilogue contains delay slots to which
4299 instructions from the rest of the function can be ``moved''. The
4300 definition should be a C expression whose value is an integer
4301 representing the number of delay slots there.
4304 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4305 A C expression that returns 1 if @var{insn} can be placed in delay
4306 slot number @var{n} of the epilogue.
4308 The argument @var{n} is an integer which identifies the delay slot now
4309 being considered (since different slots may have different rules of
4310 eligibility). It is never negative and is always less than the number
4311 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4312 If you reject a particular insn for a given delay slot, in principle, it
4313 may be reconsidered for a subsequent delay slot. Also, other insns may
4314 (at least in principle) be considered for the so far unfilled delay
4317 @findex current_function_epilogue_delay_list
4318 @findex final_scan_insn
4319 The insns accepted to fill the epilogue delay slots are put in an RTL
4320 list made with @code{insn_list} objects, stored in the variable
4321 @code{current_function_epilogue_delay_list}. The insn for the first
4322 delay slot comes first in the list. Your definition of the macro
4323 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4324 outputting the insns in this list, usually by calling
4325 @code{final_scan_insn}.
4327 You need not define this macro if you did not define
4328 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4331 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, tree @var{function})
4332 A function that outputs the assembler code for a thunk
4333 function, used to implement C++ virtual function calls with multiple
4334 inheritance. The thunk acts as a wrapper around a virtual function,
4335 adjusting the implicit object parameter before handing control off to
4338 First, emit code to add the integer @var{delta} to the location that
4339 contains the incoming first argument. Assume that this argument
4340 contains a pointer, and is the one used to pass the @code{this} pointer
4341 in C++. This is the incoming argument @emph{before} the function prologue,
4342 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4343 all other incoming arguments.
4345 After the addition, emit code to jump to @var{function}, which is a
4346 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4347 not touch the return address. Hence returning from @var{FUNCTION} will
4348 return to whoever called the current @samp{thunk}.
4350 The effect must be as if @var{function} had been called directly with
4351 the adjusted first argument. This macro is responsible for emitting all
4352 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4353 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4355 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4356 have already been extracted from it.) It might possibly be useful on
4357 some targets, but probably not.
4359 If you do not define this macro, the target-independent code in the C++
4360 front end will generate a less efficient heavyweight thunk that calls
4361 @var{function} instead of jumping to it. The generic approach does
4362 not support varargs.
4365 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_VCALL_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, int @var{vcall_offset}, tree @var{function})
4366 A function like @code{TARGET_ASM_OUTPUT_MI_THUNK}, except that if
4367 @var{vcall_offset} is nonzero, an additional adjustment should be made
4368 after adding @code{delta}. In particular, if @var{p} is the
4369 adjusted pointer, the following adjustment should be made:
4372 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4376 If this function is defined, it will always be used in place of
4377 @code{TARGET_ASM_OUTPUT_MI_THUNK}.
4381 @subsection Generating Code for Profiling
4382 @cindex profiling, code generation
4384 These macros will help you generate code for profiling.
4386 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4387 A C statement or compound statement to output to @var{file} some
4388 assembler code to call the profiling subroutine @code{mcount}.
4391 The details of how @code{mcount} expects to be called are determined by
4392 your operating system environment, not by GCC@. To figure them out,
4393 compile a small program for profiling using the system's installed C
4394 compiler and look at the assembler code that results.
4396 Older implementations of @code{mcount} expect the address of a counter
4397 variable to be loaded into some register. The name of this variable is
4398 @samp{LP} followed by the number @var{labelno}, so you would generate
4399 the name using @samp{LP%d} in a @code{fprintf}.
4402 @defmac PROFILE_HOOK
4403 A C statement or compound statement to output to @var{file} some assembly
4404 code to call the profiling subroutine @code{mcount} even the target does
4405 not support profiling.
4408 @defmac NO_PROFILE_COUNTERS
4409 Define this macro if the @code{mcount} subroutine on your system does
4410 not need a counter variable allocated for each function. This is true
4411 for almost all modern implementations. If you define this macro, you
4412 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4415 @defmac PROFILE_BEFORE_PROLOGUE
4416 Define this macro if the code for function profiling should come before
4417 the function prologue. Normally, the profiling code comes after.
4421 @subsection Permitting tail calls
4424 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4425 True if it is ok to do sibling call optimization for the specified
4426 call expression @var{exp}. @var{decl} will be the called function,
4427 or @code{NULL} if this is an indirect call.
4429 It is not uncommon for limitations of calling conventions to prevent
4430 tail calls to functions outside the current unit of translation, or
4431 during PIC compilation. The hook is used to enforce these restrictions,
4432 as the @code{sibcall} md pattern can not fail, or fall over to a
4433 ``normal'' call. The criteria for successful sibling call optimization
4434 may vary greatly between different architectures.
4438 @section Implementing the Varargs Macros
4439 @cindex varargs implementation
4441 GCC comes with an implementation of @code{<varargs.h>} and
4442 @code{<stdarg.h>} that work without change on machines that pass arguments
4443 on the stack. Other machines require their own implementations of
4444 varargs, and the two machine independent header files must have
4445 conditionals to include it.
4447 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4448 the calling convention for @code{va_start}. The traditional
4449 implementation takes just one argument, which is the variable in which
4450 to store the argument pointer. The ISO implementation of
4451 @code{va_start} takes an additional second argument. The user is
4452 supposed to write the last named argument of the function here.
4454 However, @code{va_start} should not use this argument. The way to find
4455 the end of the named arguments is with the built-in functions described
4458 @defmac __builtin_saveregs ()
4459 Use this built-in function to save the argument registers in memory so
4460 that the varargs mechanism can access them. Both ISO and traditional
4461 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4462 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4464 On some machines, @code{__builtin_saveregs} is open-coded under the
4465 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4466 other machines, it calls a routine written in assembler language,
4467 found in @file{libgcc2.c}.
4469 Code generated for the call to @code{__builtin_saveregs} appears at the
4470 beginning of the function, as opposed to where the call to
4471 @code{__builtin_saveregs} is written, regardless of what the code is.
4472 This is because the registers must be saved before the function starts
4473 to use them for its own purposes.
4474 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4478 @defmac __builtin_args_info (@var{category})
4479 Use this built-in function to find the first anonymous arguments in
4482 In general, a machine may have several categories of registers used for
4483 arguments, each for a particular category of data types. (For example,
4484 on some machines, floating-point registers are used for floating-point
4485 arguments while other arguments are passed in the general registers.)
4486 To make non-varargs functions use the proper calling convention, you
4487 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4488 registers in each category have been used so far
4490 @code{__builtin_args_info} accesses the same data structure of type
4491 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4492 with it, with @var{category} specifying which word to access. Thus, the
4493 value indicates the first unused register in a given category.
4495 Normally, you would use @code{__builtin_args_info} in the implementation
4496 of @code{va_start}, accessing each category just once and storing the
4497 value in the @code{va_list} object. This is because @code{va_list} will
4498 have to update the values, and there is no way to alter the
4499 values accessed by @code{__builtin_args_info}.
4502 @defmac __builtin_next_arg (@var{lastarg})
4503 This is the equivalent of @code{__builtin_args_info}, for stack
4504 arguments. It returns the address of the first anonymous stack
4505 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4506 returns the address of the location above the first anonymous stack
4507 argument. Use it in @code{va_start} to initialize the pointer for
4508 fetching arguments from the stack. Also use it in @code{va_start} to
4509 verify that the second parameter @var{lastarg} is the last named argument
4510 of the current function.
4513 @defmac __builtin_classify_type (@var{object})
4514 Since each machine has its own conventions for which data types are
4515 passed in which kind of register, your implementation of @code{va_arg}
4516 has to embody these conventions. The easiest way to categorize the
4517 specified data type is to use @code{__builtin_classify_type} together
4518 with @code{sizeof} and @code{__alignof__}.
4520 @code{__builtin_classify_type} ignores the value of @var{object},
4521 considering only its data type. It returns an integer describing what
4522 kind of type that is---integer, floating, pointer, structure, and so on.
4524 The file @file{typeclass.h} defines an enumeration that you can use to
4525 interpret the values of @code{__builtin_classify_type}.
4528 These machine description macros help implement varargs:
4530 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4531 If defined, this hook produces the machine-specific code for a call to
4532 @code{__builtin_saveregs}. This code will be moved to the very
4533 beginning of the function, before any parameter access are made. The
4534 return value of this function should be an RTX that contains the value
4535 to use as the return of @code{__builtin_saveregs}.
4538 @deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARGS (CUMULATIVE_ARGS *@var{args_so_far}, enum machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time})
4539 This target hook offers an alternative to using
4540 @code{__builtin_saveregs} and defining the hook
4541 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4542 register arguments into the stack so that all the arguments appear to
4543 have been passed consecutively on the stack. Once this is done, you can
4544 use the standard implementation of varargs that works for machines that
4545 pass all their arguments on the stack.
4547 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4548 structure, containing the values that are obtained after processing the
4549 named arguments. The arguments @var{mode} and @var{type} describe the
4550 last named argument---its machine mode and its data type as a tree node.
4552 The target hook should do two things: first, push onto the stack all the
4553 argument registers @emph{not} used for the named arguments, and second,
4554 store the size of the data thus pushed into the @code{int}-valued
4555 variable pointed to by @var{pretend_args_size}. The value that you
4556 store here will serve as additional offset for setting up the stack
4559 Because you must generate code to push the anonymous arguments at
4560 compile time without knowing their data types,
4561 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4562 have just a single category of argument register and use it uniformly
4565 If the argument @var{second_time} is nonzero, it means that the
4566 arguments of the function are being analyzed for the second time. This
4567 happens for an inline function, which is not actually compiled until the
4568 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4569 not generate any instructions in this case.
4572 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4573 Define this hook to return @code{true} if the location where a function
4574 argument is passed depends on whether or not it is a named argument.
4576 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4577 is set for varargs and stdarg functions. If this hook returns
4578 @code{true}, the @var{named} argument is always true for named
4579 arguments, and false for unnamed arguments. If it returns @code{false},
4580 but @code{TARGET_PRETEND_OUTOGOING_VARARGS_NAMED} returns @code{true},
4581 then all arguments are treated as named. Otherwise, all named arguments
4582 except the last are treated as named.
4584 You need not define this hook if it always returns zero.
4587 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4588 If you need to conditionally change ABIs so that one works with
4589 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4590 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4591 defined, then define this hook to return @code{true} if
4592 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4593 Otherwise, you should not define this hook.
4597 @section Trampolines for Nested Functions
4598 @cindex trampolines for nested functions
4599 @cindex nested functions, trampolines for
4601 A @dfn{trampoline} is a small piece of code that is created at run time
4602 when the address of a nested function is taken. It normally resides on
4603 the stack, in the stack frame of the containing function. These macros
4604 tell GCC how to generate code to allocate and initialize a
4607 The instructions in the trampoline must do two things: load a constant
4608 address into the static chain register, and jump to the real address of
4609 the nested function. On CISC machines such as the m68k, this requires
4610 two instructions, a move immediate and a jump. Then the two addresses
4611 exist in the trampoline as word-long immediate operands. On RISC
4612 machines, it is often necessary to load each address into a register in
4613 two parts. Then pieces of each address form separate immediate
4616 The code generated to initialize the trampoline must store the variable
4617 parts---the static chain value and the function address---into the
4618 immediate operands of the instructions. On a CISC machine, this is
4619 simply a matter of copying each address to a memory reference at the
4620 proper offset from the start of the trampoline. On a RISC machine, it
4621 may be necessary to take out pieces of the address and store them
4624 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4625 A C statement to output, on the stream @var{file}, assembler code for a
4626 block of data that contains the constant parts of a trampoline. This
4627 code should not include a label---the label is taken care of
4630 If you do not define this macro, it means no template is needed
4631 for the target. Do not define this macro on systems where the block move
4632 code to copy the trampoline into place would be larger than the code
4633 to generate it on the spot.
4636 @defmac TRAMPOLINE_SECTION
4637 The name of a subroutine to switch to the section in which the
4638 trampoline template is to be placed (@pxref{Sections}). The default is
4639 a value of @samp{readonly_data_section}, which places the trampoline in
4640 the section containing read-only data.
4643 @defmac TRAMPOLINE_SIZE
4644 A C expression for the size in bytes of the trampoline, as an integer.
4647 @defmac TRAMPOLINE_ALIGNMENT
4648 Alignment required for trampolines, in bits.
4650 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4651 is used for aligning trampolines.
4654 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4655 A C statement to initialize the variable parts of a trampoline.
4656 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4657 an RTX for the address of the nested function; @var{static_chain} is an
4658 RTX for the static chain value that should be passed to the function
4662 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4663 A C statement that should perform any machine-specific adjustment in
4664 the address of the trampoline. Its argument contains the address that
4665 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4666 used for a function call should be different from the address in which
4667 the template was stored, the different address should be assigned to
4668 @var{addr}. If this macro is not defined, @var{addr} will be used for
4671 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4672 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4673 If this macro is not defined, by default the trampoline is allocated as
4674 a stack slot. This default is right for most machines. The exceptions
4675 are machines where it is impossible to execute instructions in the stack
4676 area. On such machines, you may have to implement a separate stack,
4677 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4678 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4680 @var{fp} points to a data structure, a @code{struct function}, which
4681 describes the compilation status of the immediate containing function of
4682 the function which the trampoline is for. The stack slot for the
4683 trampoline is in the stack frame of this containing function. Other
4684 allocation strategies probably must do something analogous with this
4688 Implementing trampolines is difficult on many machines because they have
4689 separate instruction and data caches. Writing into a stack location
4690 fails to clear the memory in the instruction cache, so when the program
4691 jumps to that location, it executes the old contents.
4693 Here are two possible solutions. One is to clear the relevant parts of
4694 the instruction cache whenever a trampoline is set up. The other is to
4695 make all trampolines identical, by having them jump to a standard
4696 subroutine. The former technique makes trampoline execution faster; the
4697 latter makes initialization faster.
4699 To clear the instruction cache when a trampoline is initialized, define
4700 the following macro.
4702 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4703 If defined, expands to a C expression clearing the @emph{instruction
4704 cache} in the specified interval. The definition of this macro would
4705 typically be a series of @code{asm} statements. Both @var{beg} and
4706 @var{end} are both pointer expressions.
4709 The operating system may also require the stack to be made executable
4710 before calling the trampoline. To implement this requirement, define
4711 the following macro.
4713 @defmac ENABLE_EXECUTE_STACK
4714 Define this macro if certain operations must be performed before executing
4715 code located on the stack. The macro should expand to a series of C
4716 file-scope constructs (e.g. functions) and provide a unique entry point
4717 named @code{__enable_execute_stack}. The target is responsible for
4718 emitting calls to the entry point in the code, for example from the
4719 @code{INITIALIZE_TRAMPOLINE} macro.
4722 To use a standard subroutine, define the following macro. In addition,
4723 you must make sure that the instructions in a trampoline fill an entire
4724 cache line with identical instructions, or else ensure that the
4725 beginning of the trampoline code is always aligned at the same point in
4726 its cache line. Look in @file{m68k.h} as a guide.
4728 @defmac TRANSFER_FROM_TRAMPOLINE
4729 Define this macro if trampolines need a special subroutine to do their
4730 work. The macro should expand to a series of @code{asm} statements
4731 which will be compiled with GCC@. They go in a library function named
4732 @code{__transfer_from_trampoline}.
4734 If you need to avoid executing the ordinary prologue code of a compiled
4735 C function when you jump to the subroutine, you can do so by placing a
4736 special label of your own in the assembler code. Use one @code{asm}
4737 statement to generate an assembler label, and another to make the label
4738 global. Then trampolines can use that label to jump directly to your
4739 special assembler code.
4743 @section Implicit Calls to Library Routines
4744 @cindex library subroutine names
4745 @cindex @file{libgcc.a}
4747 @c prevent bad page break with this line
4748 Here is an explanation of implicit calls to library routines.
4750 @defmac DECLARE_LIBRARY_RENAMES
4751 This macro, if defined, should expand to a piece of C code that will get
4752 expanded when compiling functions for libgcc.a. It can be used to
4753 provide alternate names for GCC's internal library functions if there
4754 are ABI-mandated names that the compiler should provide.
4757 @findex init_one_libfunc
4758 @findex set_optab_libfunc
4759 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
4760 This hook should declare additional library routines or rename
4761 existing ones, using the functions @code{set_optab_libfunc} and
4762 @code{init_one_libfunc} defined in @file{optabs.c}.
4763 @code{init_optabs} calls this macro after initializing all the normal
4766 The default is to do nothing. Most ports don't need to define this hook.
4769 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4770 This macro should return @code{true} if the library routine that
4771 implements the floating point comparison operator @var{comparison} in
4772 mode @var{mode} will return a boolean, and @var{false} if it will
4775 GCC's own floating point libraries return tristates from the
4776 comparison operators, so the default returns false always. Most ports
4777 don't need to define this macro.
4780 @defmac TARGET_LIB_INT_CMP_BIASED
4781 This macro should evaluate to @code{true} if the integer comparison
4782 functions (like @code{__cmpdi2}) return 0 to indicate that the first
4783 operand is smaller than the second, 1 to indicate that they are equal,
4784 and 2 to indicate that the first operand is greater than the second.
4785 If this macro evalutes to @code{false} the comparison functions return
4786 -1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
4787 in @file{libgcc.a}, you do not need to define this macro.
4790 @cindex US Software GOFAST, floating point emulation library
4791 @cindex floating point emulation library, US Software GOFAST
4792 @cindex GOFAST, floating point emulation library
4793 @findex gofast_maybe_init_libfuncs
4794 @defmac US_SOFTWARE_GOFAST
4795 Define this macro if your system C library uses the US Software GOFAST
4796 library to provide floating point emulation.
4798 In addition to defining this macro, your architecture must set
4799 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
4800 else call that function from its version of that hook. It is defined
4801 in @file{config/gofast.h}, which must be included by your
4802 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
4805 If this macro is defined, the
4806 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
4807 false for @code{SFmode} and @code{DFmode} comparisons.
4810 @cindex @code{EDOM}, implicit usage
4813 The value of @code{EDOM} on the target machine, as a C integer constant
4814 expression. If you don't define this macro, GCC does not attempt to
4815 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4816 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4819 If you do not define @code{TARGET_EDOM}, then compiled code reports
4820 domain errors by calling the library function and letting it report the
4821 error. If mathematical functions on your system use @code{matherr} when
4822 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4823 that @code{matherr} is used normally.
4826 @cindex @code{errno}, implicit usage
4827 @defmac GEN_ERRNO_RTX
4828 Define this macro as a C expression to create an rtl expression that
4829 refers to the global ``variable'' @code{errno}. (On certain systems,
4830 @code{errno} may not actually be a variable.) If you don't define this
4831 macro, a reasonable default is used.
4834 @cindex C99 math functions, implicit usage
4835 @defmac TARGET_C99_FUNCTIONS
4836 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4837 @code{sinf} and similarly for other functions defined by C99 standard. The
4838 default is nonzero that should be proper value for most modern systems, however
4839 number of existing systems lacks support for these functions in the runtime so
4840 they needs this macro to be redefined to 0.
4843 @defmac NEXT_OBJC_RUNTIME
4844 Define this macro to generate code for Objective-C message sending using
4845 the calling convention of the NeXT system. This calling convention
4846 involves passing the object, the selector and the method arguments all
4847 at once to the method-lookup library function.
4849 The default calling convention passes just the object and the selector
4850 to the lookup function, which returns a pointer to the method.
4853 @node Addressing Modes
4854 @section Addressing Modes
4855 @cindex addressing modes
4857 @c prevent bad page break with this line
4858 This is about addressing modes.
4860 @defmac HAVE_PRE_INCREMENT
4861 @defmacx HAVE_PRE_DECREMENT
4862 @defmacx HAVE_POST_INCREMENT
4863 @defmacx HAVE_POST_DECREMENT
4864 A C expression that is nonzero if the machine supports pre-increment,
4865 pre-decrement, post-increment, or post-decrement addressing respectively.
4868 @defmac HAVE_PRE_MODIFY_DISP
4869 @defmacx HAVE_POST_MODIFY_DISP
4870 A C expression that is nonzero if the machine supports pre- or
4871 post-address side-effect generation involving constants other than
4872 the size of the memory operand.
4875 @defmac HAVE_PRE_MODIFY_REG
4876 @defmacx HAVE_POST_MODIFY_REG
4877 A C expression that is nonzero if the machine supports pre- or
4878 post-address side-effect generation involving a register displacement.
4881 @defmac CONSTANT_ADDRESS_P (@var{x})
4882 A C expression that is 1 if the RTX @var{x} is a constant which
4883 is a valid address. On most machines, this can be defined as
4884 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4885 in which constant addresses are supported.
4888 @defmac CONSTANT_P (@var{x})
4889 @code{CONSTANT_P}, which is defined by target-independent code,
4890 accepts integer-values expressions whose values are not explicitly
4891 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4892 expressions and @code{const} arithmetic expressions, in addition to
4893 @code{const_int} and @code{const_double} expressions.
4896 @defmac MAX_REGS_PER_ADDRESS
4897 A number, the maximum number of registers that can appear in a valid
4898 memory address. Note that it is up to you to specify a value equal to
4899 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4903 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4904 A C compound statement with a conditional @code{goto @var{label};}
4905 executed if @var{x} (an RTX) is a legitimate memory address on the
4906 target machine for a memory operand of mode @var{mode}.
4908 It usually pays to define several simpler macros to serve as
4909 subroutines for this one. Otherwise it may be too complicated to
4912 This macro must exist in two variants: a strict variant and a
4913 non-strict one. The strict variant is used in the reload pass. It
4914 must be defined so that any pseudo-register that has not been
4915 allocated a hard register is considered a memory reference. In
4916 contexts where some kind of register is required, a pseudo-register
4917 with no hard register must be rejected.
4919 The non-strict variant is used in other passes. It must be defined to
4920 accept all pseudo-registers in every context where some kind of
4921 register is required.
4923 @findex REG_OK_STRICT
4924 Compiler source files that want to use the strict variant of this
4925 macro define the macro @code{REG_OK_STRICT}. You should use an
4926 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4927 in that case and the non-strict variant otherwise.
4929 Subroutines to check for acceptable registers for various purposes (one
4930 for base registers, one for index registers, and so on) are typically
4931 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4932 Then only these subroutine macros need have two variants; the higher
4933 levels of macros may be the same whether strict or not.
4935 Normally, constant addresses which are the sum of a @code{symbol_ref}
4936 and an integer are stored inside a @code{const} RTX to mark them as
4937 constant. Therefore, there is no need to recognize such sums
4938 specifically as legitimate addresses. Normally you would simply
4939 recognize any @code{const} as legitimate.
4941 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4942 sums that are not marked with @code{const}. It assumes that a naked
4943 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4944 naked constant sums as illegitimate addresses, so that none of them will
4945 be given to @code{PRINT_OPERAND_ADDRESS}.
4947 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4948 On some machines, whether a symbolic address is legitimate depends on
4949 the section that the address refers to. On these machines, define the
4950 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4951 into the @code{symbol_ref}, and then check for it here. When you see a
4952 @code{const}, you will have to look inside it to find the
4953 @code{symbol_ref} in order to determine the section. @xref{Assembler
4957 @defmac REG_OK_FOR_BASE_P (@var{x})
4958 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4959 RTX) is valid for use as a base register. For hard registers, it
4960 should always accept those which the hardware permits and reject the
4961 others. Whether the macro accepts or rejects pseudo registers must be
4962 controlled by @code{REG_OK_STRICT} as described above. This usually
4963 requires two variant definitions, of which @code{REG_OK_STRICT}
4964 controls the one actually used.
4967 @defmac REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4968 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4969 that expression may examine the mode of the memory reference in
4970 @var{mode}. You should define this macro if the mode of the memory
4971 reference affects whether a register may be used as a base register. If
4972 you define this macro, the compiler will use it instead of
4973 @code{REG_OK_FOR_BASE_P}.
4976 @defmac REG_OK_FOR_INDEX_P (@var{x})
4977 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4978 RTX) is valid for use as an index register.
4980 The difference between an index register and a base register is that
4981 the index register may be scaled. If an address involves the sum of
4982 two registers, neither one of them scaled, then either one may be
4983 labeled the ``base'' and the other the ``index''; but whichever
4984 labeling is used must fit the machine's constraints of which registers
4985 may serve in each capacity. The compiler will try both labelings,
4986 looking for one that is valid, and will reload one or both registers
4987 only if neither labeling works.
4990 @defmac FIND_BASE_TERM (@var{x})
4991 A C expression to determine the base term of address @var{x}.
4992 This macro is used in only one place: `find_base_term' in alias.c.
4994 It is always safe for this macro to not be defined. It exists so
4995 that alias analysis can understand machine-dependent addresses.
4997 The typical use of this macro is to handle addresses containing
4998 a label_ref or symbol_ref within an UNSPEC@.
5001 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
5002 A C compound statement that attempts to replace @var{x} with a valid
5003 memory address for an operand of mode @var{mode}. @var{win} will be a
5004 C statement label elsewhere in the code; the macro definition may use
5007 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
5011 to avoid further processing if the address has become legitimate.
5013 @findex break_out_memory_refs
5014 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5015 and @var{oldx} will be the operand that was given to that function to produce
5018 The code generated by this macro should not alter the substructure of
5019 @var{x}. If it transforms @var{x} into a more legitimate form, it
5020 should assign @var{x} (which will always be a C variable) a new value.
5022 It is not necessary for this macro to come up with a legitimate
5023 address. The compiler has standard ways of doing so in all cases. In
5024 fact, it is safe to omit this macro. But often a
5025 machine-dependent strategy can generate better code.
5028 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5029 A C compound statement that attempts to replace @var{x}, which is an address
5030 that needs reloading, with a valid memory address for an operand of mode
5031 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5032 It is not necessary to define this macro, but it might be useful for
5033 performance reasons.
5035 For example, on the i386, it is sometimes possible to use a single
5036 reload register instead of two by reloading a sum of two pseudo
5037 registers into a register. On the other hand, for number of RISC
5038 processors offsets are limited so that often an intermediate address
5039 needs to be generated in order to address a stack slot. By defining
5040 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5041 generated for adjacent some stack slots can be made identical, and thus
5044 @emph{Note}: This macro should be used with caution. It is necessary
5045 to know something of how reload works in order to effectively use this,
5046 and it is quite easy to produce macros that build in too much knowledge
5047 of reload internals.
5049 @emph{Note}: This macro must be able to reload an address created by a
5050 previous invocation of this macro. If it fails to handle such addresses
5051 then the compiler may generate incorrect code or abort.
5054 The macro definition should use @code{push_reload} to indicate parts that
5055 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5056 suitable to be passed unaltered to @code{push_reload}.
5058 The code generated by this macro must not alter the substructure of
5059 @var{x}. If it transforms @var{x} into a more legitimate form, it
5060 should assign @var{x} (which will always be a C variable) a new value.
5061 This also applies to parts that you change indirectly by calling
5064 @findex strict_memory_address_p
5065 The macro definition may use @code{strict_memory_address_p} to test if
5066 the address has become legitimate.
5069 If you want to change only a part of @var{x}, one standard way of doing
5070 this is to use @code{copy_rtx}. Note, however, that is unshares only a
5071 single level of rtl. Thus, if the part to be changed is not at the
5072 top level, you'll need to replace first the top level.
5073 It is not necessary for this macro to come up with a legitimate
5074 address; but often a machine-dependent strategy can generate better code.
5077 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5078 A C statement or compound statement with a conditional @code{goto
5079 @var{label};} executed if memory address @var{x} (an RTX) can have
5080 different meanings depending on the machine mode of the memory
5081 reference it is used for or if the address is valid for some modes
5084 Autoincrement and autodecrement addresses typically have mode-dependent
5085 effects because the amount of the increment or decrement is the size
5086 of the operand being addressed. Some machines have other mode-dependent
5087 addresses. Many RISC machines have no mode-dependent addresses.
5089 You may assume that @var{addr} is a valid address for the machine.
5092 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5093 A C expression that is nonzero if @var{x} is a legitimate constant for
5094 an immediate operand on the target machine. You can assume that
5095 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5096 @samp{1} is a suitable definition for this macro on machines where
5097 anything @code{CONSTANT_P} is valid.
5100 @node Condition Code
5101 @section Condition Code Status
5102 @cindex condition code status
5104 @c prevent bad page break with this line
5105 This describes the condition code status.
5108 The file @file{conditions.h} defines a variable @code{cc_status} to
5109 describe how the condition code was computed (in case the interpretation of
5110 the condition code depends on the instruction that it was set by). This
5111 variable contains the RTL expressions on which the condition code is
5112 currently based, and several standard flags.
5114 Sometimes additional machine-specific flags must be defined in the machine
5115 description header file. It can also add additional machine-specific
5116 information by defining @code{CC_STATUS_MDEP}.
5118 @defmac CC_STATUS_MDEP
5119 C code for a data type which is used for declaring the @code{mdep}
5120 component of @code{cc_status}. It defaults to @code{int}.
5122 This macro is not used on machines that do not use @code{cc0}.
5125 @defmac CC_STATUS_MDEP_INIT
5126 A C expression to initialize the @code{mdep} field to ``empty''.
5127 The default definition does nothing, since most machines don't use
5128 the field anyway. If you want to use the field, you should probably
5129 define this macro to initialize it.
5131 This macro is not used on machines that do not use @code{cc0}.
5134 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5135 A C compound statement to set the components of @code{cc_status}
5136 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5137 this macro's responsibility to recognize insns that set the condition
5138 code as a byproduct of other activity as well as those that explicitly
5141 This macro is not used on machines that do not use @code{cc0}.
5143 If there are insns that do not set the condition code but do alter
5144 other machine registers, this macro must check to see whether they
5145 invalidate the expressions that the condition code is recorded as
5146 reflecting. For example, on the 68000, insns that store in address
5147 registers do not set the condition code, which means that usually
5148 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5149 insns. But suppose that the previous insn set the condition code
5150 based on location @samp{a4@@(102)} and the current insn stores a new
5151 value in @samp{a4}. Although the condition code is not changed by
5152 this, it will no longer be true that it reflects the contents of
5153 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5154 @code{cc_status} in this case to say that nothing is known about the
5155 condition code value.
5157 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5158 with the results of peephole optimization: insns whose patterns are
5159 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5160 constants which are just the operands. The RTL structure of these
5161 insns is not sufficient to indicate what the insns actually do. What
5162 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5163 @code{CC_STATUS_INIT}.
5165 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5166 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5167 @samp{cc}. This avoids having detailed information about patterns in
5168 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5171 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5172 Returns a mode from class @code{MODE_CC} to be used when comparison
5173 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5174 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5175 @pxref{Jump Patterns} for a description of the reason for this
5179 #define SELECT_CC_MODE(OP,X,Y) \
5180 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5181 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5182 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5183 || GET_CODE (X) == NEG) \
5184 ? CC_NOOVmode : CCmode))
5187 You should define this macro if and only if you define extra CC modes
5188 in @file{@var{machine}-modes.def}.
5191 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5192 On some machines not all possible comparisons are defined, but you can
5193 convert an invalid comparison into a valid one. For example, the Alpha
5194 does not have a @code{GT} comparison, but you can use an @code{LT}
5195 comparison instead and swap the order of the operands.
5197 On such machines, define this macro to be a C statement to do any
5198 required conversions. @var{code} is the initial comparison code
5199 and @var{op0} and @var{op1} are the left and right operands of the
5200 comparison, respectively. You should modify @var{code}, @var{op0}, and
5201 @var{op1} as required.
5203 GCC will not assume that the comparison resulting from this macro is
5204 valid but will see if the resulting insn matches a pattern in the
5207 You need not define this macro if it would never change the comparison
5211 @defmac REVERSIBLE_CC_MODE (@var{mode})
5212 A C expression whose value is one if it is always safe to reverse a
5213 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5214 can ever return @var{mode} for a floating-point inequality comparison,
5215 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5217 You need not define this macro if it would always returns zero or if the
5218 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5219 For example, here is the definition used on the SPARC, where floating-point
5220 inequality comparisons are always given @code{CCFPEmode}:
5223 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5227 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5228 A C expression whose value is reversed condition code of the @var{code} for
5229 comparison done in CC_MODE @var{mode}. The macro is used only in case
5230 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5231 machine has some non-standard way how to reverse certain conditionals. For
5232 instance in case all floating point conditions are non-trapping, compiler may
5233 freely convert unordered compares to ordered one. Then definition may look
5237 #define REVERSE_CONDITION(CODE, MODE) \
5238 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5239 : reverse_condition_maybe_unordered (CODE))
5243 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
5244 A C expression that returns true if the conditional execution predicate
5245 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
5246 versa. Define this to return 0 if the target has conditional execution
5247 predicates that cannot be reversed safely. There is no need to validate
5248 that the arguments of op1 and op2 are the same, this is done separately.
5249 If no expansion is specified, this macro is defined as follows:
5252 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5253 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
5257 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5258 On targets which do not use @code{(cc0)}, and which use a hard
5259 register rather than a pseudo-register to hold condition codes, the
5260 regular CSE passes are often not able to identify cases in which the
5261 hard register is set to a common value. Use this hook to enable a
5262 small pass which optimizes such cases. This hook should return true
5263 to enable this pass, and it should set the integers to which its
5264 arguments point to the hard register numbers used for condition codes.
5265 When there is only one such register, as is true on most systems, the
5266 integer pointed to by the second argument should be set to
5267 @code{INVALID_REGNUM}.
5269 The default version of this hook returns false.
5272 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5273 On targets which use multiple condition code modes in class
5274 @code{MODE_CC}, it is sometimes the case that a comparison can be
5275 validly done in more than one mode. On such a system, define this
5276 target hook to take two mode arguments and to return a mode in which
5277 both comparisons may be validly done. If there is no such mode,
5278 return @code{VOIDmode}.
5280 The default version of this hook checks whether the modes are the
5281 same. If they are, it returns that mode. If they are different, it
5282 returns @code{VOIDmode}.
5286 @section Describing Relative Costs of Operations
5287 @cindex costs of instructions
5288 @cindex relative costs
5289 @cindex speed of instructions
5291 These macros let you describe the relative speed of various operations
5292 on the target machine.
5294 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5295 A C expression for the cost of moving data of mode @var{mode} from a
5296 register in class @var{from} to one in class @var{to}. The classes are
5297 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5298 value of 2 is the default; other values are interpreted relative to
5301 It is not required that the cost always equal 2 when @var{from} is the
5302 same as @var{to}; on some machines it is expensive to move between
5303 registers if they are not general registers.
5305 If reload sees an insn consisting of a single @code{set} between two
5306 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5307 classes returns a value of 2, reload does not check to ensure that the
5308 constraints of the insn are met. Setting a cost of other than 2 will
5309 allow reload to verify that the constraints are met. You should do this
5310 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5313 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5314 A C expression for the cost of moving data of mode @var{mode} between a
5315 register of class @var{class} and memory; @var{in} is zero if the value
5316 is to be written to memory, nonzero if it is to be read in. This cost
5317 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5318 registers and memory is more expensive than between two registers, you
5319 should define this macro to express the relative cost.
5321 If you do not define this macro, GCC uses a default cost of 4 plus
5322 the cost of copying via a secondary reload register, if one is
5323 needed. If your machine requires a secondary reload register to copy
5324 between memory and a register of @var{class} but the reload mechanism is
5325 more complex than copying via an intermediate, define this macro to
5326 reflect the actual cost of the move.
5328 GCC defines the function @code{memory_move_secondary_cost} if
5329 secondary reloads are needed. It computes the costs due to copying via
5330 a secondary register. If your machine copies from memory using a
5331 secondary register in the conventional way but the default base value of
5332 4 is not correct for your machine, define this macro to add some other
5333 value to the result of that function. The arguments to that function
5334 are the same as to this macro.
5338 A C expression for the cost of a branch instruction. A value of 1 is
5339 the default; other values are interpreted relative to that.
5342 Here are additional macros which do not specify precise relative costs,
5343 but only that certain actions are more expensive than GCC would
5346 @defmac SLOW_BYTE_ACCESS
5347 Define this macro as a C expression which is nonzero if accessing less
5348 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5349 faster than accessing a word of memory, i.e., if such access
5350 require more than one instruction or if there is no difference in cost
5351 between byte and (aligned) word loads.
5353 When this macro is not defined, the compiler will access a field by
5354 finding the smallest containing object; when it is defined, a fullword
5355 load will be used if alignment permits. Unless bytes accesses are
5356 faster than word accesses, using word accesses is preferable since it
5357 may eliminate subsequent memory access if subsequent accesses occur to
5358 other fields in the same word of the structure, but to different bytes.
5361 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5362 Define this macro to be the value 1 if memory accesses described by the
5363 @var{mode} and @var{alignment} parameters have a cost many times greater
5364 than aligned accesses, for example if they are emulated in a trap
5367 When this macro is nonzero, the compiler will act as if
5368 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5369 moves. This can cause significantly more instructions to be produced.
5370 Therefore, do not set this macro nonzero if unaligned accesses only add a
5371 cycle or two to the time for a memory access.
5373 If the value of this macro is always zero, it need not be defined. If
5374 this macro is defined, it should produce a nonzero value when
5375 @code{STRICT_ALIGNMENT} is nonzero.
5379 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5380 which a sequence of insns should be generated instead of a
5381 string move insn or a library call. Increasing the value will always
5382 make code faster, but eventually incurs high cost in increased code size.
5384 Note that on machines where the corresponding move insn is a
5385 @code{define_expand} that emits a sequence of insns, this macro counts
5386 the number of such sequences.
5388 If you don't define this, a reasonable default is used.
5391 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5392 A C expression used to determine whether @code{move_by_pieces} will be used to
5393 copy a chunk of memory, or whether some other block move mechanism
5394 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5395 than @code{MOVE_RATIO}.
5398 @defmac MOVE_MAX_PIECES
5399 A C expression used by @code{move_by_pieces} to determine the largest unit
5400 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5404 The threshold of number of scalar move insns, @emph{below} which a sequence
5405 of insns should be generated to clear memory instead of a string clear insn
5406 or a library call. Increasing the value will always make code faster, but
5407 eventually incurs high cost in increased code size.
5409 If you don't define this, a reasonable default is used.
5412 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5413 A C expression used to determine whether @code{clear_by_pieces} will be used
5414 to clear a chunk of memory, or whether some other block clear mechanism
5415 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5416 than @code{CLEAR_RATIO}.
5419 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5420 A C expression used to determine whether @code{store_by_pieces} will be
5421 used to set a chunk of memory to a constant value, or whether some other
5422 mechanism will be used. Used by @code{__builtin_memset} when storing
5423 values other than constant zero and by @code{__builtin_strcpy} when
5424 when called with a constant source string.
5425 Defaults to to 1 if @code{move_by_pieces_ninsns} returns less
5426 than @code{MOVE_RATIO}.
5429 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5430 A C expression used to determine whether a load postincrement is a good
5431 thing to use for a given mode. Defaults to the value of
5432 @code{HAVE_POST_INCREMENT}.
5435 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5436 A C expression used to determine whether a load postdecrement is a good
5437 thing to use for a given mode. Defaults to the value of
5438 @code{HAVE_POST_DECREMENT}.
5441 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5442 A C expression used to determine whether a load preincrement is a good
5443 thing to use for a given mode. Defaults to the value of
5444 @code{HAVE_PRE_INCREMENT}.
5447 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5448 A C expression used to determine whether a load predecrement is a good
5449 thing to use for a given mode. Defaults to the value of
5450 @code{HAVE_PRE_DECREMENT}.
5453 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5454 A C expression used to determine whether a store postincrement is a good
5455 thing to use for a given mode. Defaults to the value of
5456 @code{HAVE_POST_INCREMENT}.
5459 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5460 A C expression used to determine whether a store postdecrement is a good
5461 thing to use for a given mode. Defaults to the value of
5462 @code{HAVE_POST_DECREMENT}.
5465 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5466 This macro is used to determine whether a store preincrement is a good
5467 thing to use for a given mode. Defaults to the value of
5468 @code{HAVE_PRE_INCREMENT}.
5471 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5472 This macro is used to determine whether a store predecrement is a good
5473 thing to use for a given mode. Defaults to the value of
5474 @code{HAVE_PRE_DECREMENT}.
5477 @defmac NO_FUNCTION_CSE
5478 Define this macro if it is as good or better to call a constant
5479 function address than to call an address kept in a register.
5482 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5483 Define this macro if a non-short-circuit operation produced by
5484 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5485 @code{BRANCH_COST} is greater than or equal to the value 2.
5488 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5489 This target hook describes the relative costs of RTL expressions.
5491 The cost may depend on the precise form of the expression, which is
5492 available for examination in @var{x}, and the rtx code of the expression
5493 in which it is contained, found in @var{outer_code}. @var{code} is the
5494 expression code---redundant, since it can be obtained with
5495 @code{GET_CODE (@var{x})}.
5497 In implementing this hook, you can use the construct
5498 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5501 On entry to the hook, @code{*@var{total}} contains a default estimate
5502 for the cost of the expression. The hook should modify this value as
5503 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
5504 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
5505 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
5507 When optimizing for code size, i.e@. when @code{optimize_size} is
5508 non-zero, this target hook should be used to estimate the relative
5509 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
5511 The hook returns true when all subexpressions of @var{x} have been
5512 processed, and false when @code{rtx_cost} should recurse.
5515 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5516 This hook computes the cost of an addressing mode that contains
5517 @var{address}. If not defined, the cost is computed from
5518 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5520 For most CISC machines, the default cost is a good approximation of the
5521 true cost of the addressing mode. However, on RISC machines, all
5522 instructions normally have the same length and execution time. Hence
5523 all addresses will have equal costs.
5525 In cases where more than one form of an address is known, the form with
5526 the lowest cost will be used. If multiple forms have the same, lowest,
5527 cost, the one that is the most complex will be used.
5529 For example, suppose an address that is equal to the sum of a register
5530 and a constant is used twice in the same basic block. When this macro
5531 is not defined, the address will be computed in a register and memory
5532 references will be indirect through that register. On machines where
5533 the cost of the addressing mode containing the sum is no higher than
5534 that of a simple indirect reference, this will produce an additional
5535 instruction and possibly require an additional register. Proper
5536 specification of this macro eliminates this overhead for such machines.
5538 This hook is never called with an invalid address.
5540 On machines where an address involving more than one register is as
5541 cheap as an address computation involving only one register, defining
5542 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5543 be live over a region of code where only one would have been if
5544 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5545 should be considered in the definition of this macro. Equivalent costs
5546 should probably only be given to addresses with different numbers of
5547 registers on machines with lots of registers.
5551 @section Adjusting the Instruction Scheduler
5553 The instruction scheduler may need a fair amount of machine-specific
5554 adjustment in order to produce good code. GCC provides several target
5555 hooks for this purpose. It is usually enough to define just a few of
5556 them: try the first ones in this list first.
5558 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5559 This hook returns the maximum number of instructions that can ever
5560 issue at the same time on the target machine. The default is one.
5561 Although the insn scheduler can define itself the possibility of issue
5562 an insn on the same cycle, the value can serve as an additional
5563 constraint to issue insns on the same simulated processor cycle (see
5564 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5565 This value must be constant over the entire compilation. If you need
5566 it to vary depending on what the instructions are, you must use
5567 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5569 You could define this hook to return the value of the macro
5570 @code{MAX_DFA_ISSUE_RATE}.
5573 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5574 This hook is executed by the scheduler after it has scheduled an insn
5575 from the ready list. It should return the number of insns which can
5576 still be issued in the current cycle. The default is
5577 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5578 @code{USE}, which normally are not counted against the issue rate.
5579 You should define this hook if some insns take more machine resources
5580 than others, so that fewer insns can follow them in the same cycle.
5581 @var{file} is either a null pointer, or a stdio stream to write any
5582 debug output to. @var{verbose} is the verbose level provided by
5583 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5587 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5588 This function corrects the value of @var{cost} based on the
5589 relationship between @var{insn} and @var{dep_insn} through the
5590 dependence @var{link}. It should return the new value. The default
5591 is to make no adjustment to @var{cost}. This can be used for example
5592 to specify to the scheduler using the traditional pipeline description
5593 that an output- or anti-dependence does not incur the same cost as a
5594 data-dependence. If the scheduler using the automaton based pipeline
5595 description, the cost of anti-dependence is zero and the cost of
5596 output-dependence is maximum of one and the difference of latency
5597 times of the first and the second insns. If these values are not
5598 acceptable, you could use the hook to modify them too. See also
5599 @pxref{Processor pipeline description}.
5602 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5603 This hook adjusts the integer scheduling priority @var{priority} of
5604 @var{insn}. It should return the new priority. Reduce the priority to
5605 execute @var{insn} earlier, increase the priority to execute @var{insn}
5606 later. Do not define this hook if you do not need to adjust the
5607 scheduling priorities of insns.
5610 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5611 This hook is executed by the scheduler after it has scheduled the ready
5612 list, to allow the machine description to reorder it (for example to
5613 combine two small instructions together on @samp{VLIW} machines).
5614 @var{file} is either a null pointer, or a stdio stream to write any
5615 debug output to. @var{verbose} is the verbose level provided by
5616 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5617 list of instructions that are ready to be scheduled. @var{n_readyp} is
5618 a pointer to the number of elements in the ready list. The scheduler
5619 reads the ready list in reverse order, starting with
5620 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5621 is the timer tick of the scheduler. You may modify the ready list and
5622 the number of ready insns. The return value is the number of insns that
5623 can issue this cycle; normally this is just @code{issue_rate}. See also
5624 @samp{TARGET_SCHED_REORDER2}.
5627 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5628 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5629 function is called whenever the scheduler starts a new cycle. This one
5630 is called once per iteration over a cycle, immediately after
5631 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5632 return the number of insns to be scheduled in the same cycle. Defining
5633 this hook can be useful if there are frequent situations where
5634 scheduling one insn causes other insns to become ready in the same
5635 cycle. These other insns can then be taken into account properly.
5638 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5639 This hook is called after evaluation forward dependencies of insns in
5640 chain given by two parameter values (@var{head} and @var{tail}
5641 correspondingly) but before insns scheduling of the insn chain. For
5642 example, it can be used for better insn classification if it requires
5643 analysis of dependencies. This hook can use backward and forward
5644 dependencies of the insn scheduler because they are already
5648 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5649 This hook is executed by the scheduler at the beginning of each block of
5650 instructions that are to be scheduled. @var{file} is either a null
5651 pointer, or a stdio stream to write any debug output to. @var{verbose}
5652 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5653 @var{max_ready} is the maximum number of insns in the current scheduling
5654 region that can be live at the same time. This can be used to allocate
5655 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5658 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5659 This hook is executed by the scheduler at the end of each block of
5660 instructions that are to be scheduled. It can be used to perform
5661 cleanup of any actions done by the other scheduling hooks. @var{file}
5662 is either a null pointer, or a stdio stream to write any debug output
5663 to. @var{verbose} is the verbose level provided by
5664 @option{-fsched-verbose-@var{n}}.
5667 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
5668 This hook is executed by the scheduler after function level initializations.
5669 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5670 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5671 @var{old_max_uid} is the maximum insn uid when scheduling begins.
5674 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
5675 This is the cleanup hook corresponding to TARGET_SCHED_INIT_GLOBAL.
5676 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5677 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5680 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5681 The hook returns an RTL insn. The automaton state used in the
5682 pipeline hazard recognizer is changed as if the insn were scheduled
5683 when the new simulated processor cycle starts. Usage of the hook may
5684 simplify the automaton pipeline description for some @acronym{VLIW}
5685 processors. If the hook is defined, it is used only for the automaton
5686 based pipeline description. The default is not to change the state
5687 when the new simulated processor cycle starts.
5690 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5691 The hook can be used to initialize data used by the previous hook.
5694 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5695 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5696 to changed the state as if the insn were scheduled when the new
5697 simulated processor cycle finishes.
5700 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5701 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5702 used to initialize data used by the previous hook.
5705 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5706 This hook controls better choosing an insn from the ready insn queue
5707 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5708 chooses the first insn from the queue. If the hook returns a positive
5709 value, an additional scheduler code tries all permutations of
5710 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5711 subsequent ready insns to choose an insn whose issue will result in
5712 maximal number of issued insns on the same cycle. For the
5713 @acronym{VLIW} processor, the code could actually solve the problem of
5714 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5715 rules of @acronym{VLIW} packing are described in the automaton.
5717 This code also could be used for superscalar @acronym{RISC}
5718 processors. Let us consider a superscalar @acronym{RISC} processor
5719 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5720 @var{B}, some insns can be executed only in pipelines @var{B} or
5721 @var{C}, and one insn can be executed in pipeline @var{B}. The
5722 processor may issue the 1st insn into @var{A} and the 2nd one into
5723 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5724 until the next cycle. If the scheduler issues the 3rd insn the first,
5725 the processor could issue all 3 insns per cycle.
5727 Actually this code demonstrates advantages of the automaton based
5728 pipeline hazard recognizer. We try quickly and easy many insn
5729 schedules to choose the best one.
5731 The default is no multipass scheduling.
5734 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5736 This hook controls what insns from the ready insn queue will be
5737 considered for the multipass insn scheduling. If the hook returns
5738 zero for insn passed as the parameter, the insn will be not chosen to
5741 The default is that any ready insns can be chosen to be issued.
5744 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5746 This hook is called by the insn scheduler before issuing insn passed
5747 as the third parameter on given cycle. If the hook returns nonzero,
5748 the insn is not issued on given processors cycle. Instead of that,
5749 the processor cycle is advanced. If the value passed through the last
5750 parameter is zero, the insn ready queue is not sorted on the new cycle
5751 start as usually. The first parameter passes file for debugging
5752 output. The second one passes the scheduler verbose level of the
5753 debugging output. The forth and the fifth parameter values are
5754 correspondingly processor cycle on which the previous insn has been
5755 issued and the current processor cycle.
5758 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (rtx @var{insn1}, rtx @var{insn2}, rtx @var{dep_link}, int @var{dep_cost}, int @var{distance})
5759 This hook is used to define which dependences are considered costly by
5760 the target, so costly that it is not advisable to schedule the insns that
5761 are involved in the dependence too close to one another. The parameters
5762 to this hook are as follows: The second parameter @var{insn2} is dependent
5763 upon the first parameter @var{insn1}. The dependence between @var{insn1}
5764 and @var{insn2} is represented by the third parameter @var{dep_link}. The
5765 fourth parameter @var{cost} is the cost of the dependence, and the fifth
5766 parameter @var{distance} is the distance in cycles between the two insns.
5767 The hook returns @code{true} if considering the distance between the two
5768 insns the dependence between them is considered costly by the target,
5769 and @code{false} otherwise.
5771 Defining this hook can be useful in multiple-issue out-of-order machines,
5772 where (a) it's practically hopeless to predict the actual data/resource
5773 delays, however: (b) there's a better chance to predict the actual grouping
5774 that will be formed, and (c) correctly emulating the grouping can be very
5775 important. In such targets one may want to allow issuing dependent insns
5776 closer to one another - i.e, closer than the dependence distance; however,
5777 not in cases of "costly dependences", which this hooks allows to define.
5780 Macros in the following table are generated by the program
5781 @file{genattr} and can be useful for writing the hooks.
5783 @defmac MAX_DFA_ISSUE_RATE
5784 The macro definition is generated in the automaton based pipeline
5785 description interface. Its value is calculated from the automaton
5786 based pipeline description and is equal to maximal number of all insns
5787 described in constructions @samp{define_insn_reservation} which can be
5788 issued on the same processor cycle.
5792 @section Dividing the Output into Sections (Texts, Data, @dots{})
5793 @c the above section title is WAY too long. maybe cut the part between
5794 @c the (...)? --mew 10feb93
5796 An object file is divided into sections containing different types of
5797 data. In the most common case, there are three sections: the @dfn{text
5798 section}, which holds instructions and read-only data; the @dfn{data
5799 section}, which holds initialized writable data; and the @dfn{bss
5800 section}, which holds uninitialized data. Some systems have other kinds
5803 The compiler must tell the assembler when to switch sections. These
5804 macros control what commands to output to tell the assembler this. You
5805 can also define additional sections.
5807 @defmac TEXT_SECTION_ASM_OP
5808 A C expression whose value is a string, including spacing, containing the
5809 assembler operation that should precede instructions and read-only data.
5810 Normally @code{"\t.text"} is right.
5813 @defmac HOT_TEXT_SECTION_NAME
5814 If defined, a C string constant for the name of the section containing most
5815 frequently executed functions of the program. If not defined, GCC will provide
5816 a default definition if the target supports named sections.
5819 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5820 If defined, a C string constant for the name of the section containing unlikely
5821 executed functions in the program.
5824 @defmac DATA_SECTION_ASM_OP
5825 A C expression whose value is a string, including spacing, containing the
5826 assembler operation to identify the following data as writable initialized
5827 data. Normally @code{"\t.data"} is right.
5830 @defmac READONLY_DATA_SECTION_ASM_OP
5831 A C expression whose value is a string, including spacing, containing the
5832 assembler operation to identify the following data as read-only initialized
5836 @defmac READONLY_DATA_SECTION
5837 A macro naming a function to call to switch to the proper section for
5838 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5839 if defined, else fall back to @code{text_section}.
5841 The most common definition will be @code{data_section}, if the target
5842 does not have a special read-only data section, and does not put data
5843 in the text section.
5846 @defmac BSS_SECTION_ASM_OP
5847 If defined, a C expression whose value is a string, including spacing,
5848 containing the assembler operation to identify the following data as
5849 uninitialized global data. If not defined, and neither
5850 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5851 uninitialized global data will be output in the data section if
5852 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5856 @defmac INIT_SECTION_ASM_OP
5857 If defined, a C expression whose value is a string, including spacing,
5858 containing the assembler operation to identify the following data as
5859 initialization code. If not defined, GCC will assume such a section does
5863 @defmac FINI_SECTION_ASM_OP
5864 If defined, a C expression whose value is a string, including spacing,
5865 containing the assembler operation to identify the following data as
5866 finalization code. If not defined, GCC will assume such a section does
5870 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5871 If defined, an ASM statement that switches to a different section
5872 via @var{section_op}, calls @var{function}, and switches back to
5873 the text section. This is used in @file{crtstuff.c} if
5874 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5875 to initialization and finalization functions from the init and fini
5876 sections. By default, this macro uses a simple function call. Some
5877 ports need hand-crafted assembly code to avoid dependencies on
5878 registers initialized in the function prologue or to ensure that
5879 constant pools don't end up too far way in the text section.
5882 @defmac FORCE_CODE_SECTION_ALIGN
5883 If defined, an ASM statement that aligns a code section to some
5884 arbitrary boundary. This is used to force all fragments of the
5885 @code{.init} and @code{.fini} sections to have to same alignment
5886 and thus prevent the linker from having to add any padding.
5891 @defmac EXTRA_SECTIONS
5892 A list of names for sections other than the standard two, which are
5893 @code{in_text} and @code{in_data}. You need not define this macro
5894 on a system with no other sections (that GCC needs to use).
5897 @findex text_section
5898 @findex data_section
5899 @defmac EXTRA_SECTION_FUNCTIONS
5900 One or more functions to be defined in @file{varasm.c}. These
5901 functions should do jobs analogous to those of @code{text_section} and
5902 @code{data_section}, for your additional sections. Do not define this
5903 macro if you do not define @code{EXTRA_SECTIONS}.
5906 @defmac JUMP_TABLES_IN_TEXT_SECTION
5907 Define this macro to be an expression with a nonzero value if jump
5908 tables (for @code{tablejump} insns) should be output in the text
5909 section, along with the assembler instructions. Otherwise, the
5910 readonly data section is used.
5912 This macro is irrelevant if there is no separate readonly data section.
5915 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5916 Switches to the appropriate section for output of @var{exp}. You can
5917 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5918 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5919 requires link-time relocations. Bit 0 is set when variable contains
5920 local relocations only, while bit 1 is set for global relocations.
5921 Select the section by calling @code{data_section} or one of the
5922 alternatives for other sections. @var{align} is the constant alignment
5925 The default version of this function takes care of putting read-only
5926 variables in @code{readonly_data_section}.
5929 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5930 Build up a unique section name, expressed as a @code{STRING_CST} node,
5931 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5932 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5933 the initial value of @var{exp} requires link-time relocations.
5935 The default version of this function appends the symbol name to the
5936 ELF section name that would normally be used for the symbol. For
5937 example, the function @code{foo} would be placed in @code{.text.foo}.
5938 Whatever the actual target object format, this is often good enough.
5941 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
5942 Switches to a readonly data section associated with
5943 @samp{DECL_SECTION_NAME (@var{decl})}.
5944 The default version of this function switches to @code{.gnu.linkonce.r.name}
5945 section if function's section is @code{.gnu.linkonce.t.name}, to
5946 @code{.rodata.name} if function is in @code{.text.name} section
5947 and otherwise switches to the normal readonly data section.
5950 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
5951 Switches to the appropriate section for output of constant pool entry
5952 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
5953 constant in RTL@. The argument @var{mode} is redundant except in the
5954 case of a @code{const_int} rtx. Select the section by calling
5955 @code{readonly_data_section} or one of the alternatives for other
5956 sections. @var{align} is the constant alignment in bits.
5958 The default version of this function takes care of putting symbolic
5959 constants in @code{flag_pic} mode in @code{data_section} and everything
5960 else in @code{readonly_data_section}.
5963 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
5964 Define this hook if references to a symbol or a constant must be
5965 treated differently depending on something about the variable or
5966 function named by the symbol (such as what section it is in).
5968 The hook is executed immediately after rtl has been created for
5969 @var{decl}, which may be a variable or function declaration or
5970 an entry in the constant pool. In either case, @var{rtl} is the
5971 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
5972 in this hook; that field may not have been initialized yet.
5974 In the case of a constant, it is safe to assume that the rtl is
5975 a @code{mem} whose address is a @code{symbol_ref}. Most decls
5976 will also have this form, but that is not guaranteed. Global
5977 register variables, for instance, will have a @code{reg} for their
5978 rtl. (Normally the right thing to do with such unusual rtl is
5981 The @var{new_decl_p} argument will be true if this is the first time
5982 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
5983 be false for subsequent invocations, which will happen for duplicate
5984 declarations. Whether or not anything must be done for the duplicate
5985 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
5986 @var{new_decl_p} is always true when the hook is called for a constant.
5988 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
5989 The usual thing for this hook to do is to record flags in the
5990 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
5991 Historically, the name string was modified if it was necessary to
5992 encode more than one bit of information, but this practice is now
5993 discouraged; use @code{SYMBOL_REF_FLAGS}.
5995 The default definition of this hook, @code{default_encode_section_info}
5996 in @file{varasm.c}, sets a number of commonly-useful bits in
5997 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
5998 before overriding it.
6001 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6002 Decode @var{name} and return the real name part, sans
6003 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6007 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6008 Returns true if @var{exp} should be placed into a ``small data'' section.
6009 The default version of this hook always returns false.
6012 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6013 Contains the value true if the target places read-only
6014 ``small data'' into a separate section. The default value is false.
6017 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6018 Returns true if @var{exp} names an object for which name resolution
6019 rules must resolve to the current ``module'' (dynamic shared library
6020 or executable image).
6022 The default version of this hook implements the name resolution rules
6023 for ELF, which has a looser model of global name binding than other
6024 currently supported object file formats.
6027 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6028 Contains the value true if the target supports thread-local storage.
6029 The default value is false.
6034 @section Position Independent Code
6035 @cindex position independent code
6038 This section describes macros that help implement generation of position
6039 independent code. Simply defining these macros is not enough to
6040 generate valid PIC; you must also add support to the macros
6041 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6042 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6043 @samp{movsi} to do something appropriate when the source operand
6044 contains a symbolic address. You may also need to alter the handling of
6045 switch statements so that they use relative addresses.
6046 @c i rearranged the order of the macros above to try to force one of
6047 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6049 @defmac PIC_OFFSET_TABLE_REGNUM
6050 The register number of the register used to address a table of static
6051 data addresses in memory. In some cases this register is defined by a
6052 processor's ``application binary interface'' (ABI)@. When this macro
6053 is defined, RTL is generated for this register once, as with the stack
6054 pointer and frame pointer registers. If this macro is not defined, it
6055 is up to the machine-dependent files to allocate such a register (if
6056 necessary). Note that this register must be fixed when in use (e.g.@:
6057 when @code{flag_pic} is true).
6060 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6061 Define this macro if the register defined by
6062 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6063 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6066 @defmac FINALIZE_PIC
6067 By generating position-independent code, when two different programs (A
6068 and B) share a common library (libC.a), the text of the library can be
6069 shared whether or not the library is linked at the same address for both
6070 programs. In some of these environments, position-independent code
6071 requires not only the use of different addressing modes, but also
6072 special code to enable the use of these addressing modes.
6074 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
6075 codes once the function is being compiled into assembly code, but not
6076 before. (It is not done before, because in the case of compiling an
6077 inline function, it would lead to multiple PIC prologues being
6078 included in functions which used inline functions and were compiled to
6082 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6083 A C expression that is nonzero if @var{x} is a legitimate immediate
6084 operand on the target machine when generating position independent code.
6085 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6086 check this. You can also assume @var{flag_pic} is true, so you need not
6087 check it either. You need not define this macro if all constants
6088 (including @code{SYMBOL_REF}) can be immediate operands when generating
6089 position independent code.
6092 @node Assembler Format
6093 @section Defining the Output Assembler Language
6095 This section describes macros whose principal purpose is to describe how
6096 to write instructions in assembler language---rather than what the
6100 * File Framework:: Structural information for the assembler file.
6101 * Data Output:: Output of constants (numbers, strings, addresses).
6102 * Uninitialized Data:: Output of uninitialized variables.
6103 * Label Output:: Output and generation of labels.
6104 * Initialization:: General principles of initialization
6105 and termination routines.
6106 * Macros for Initialization::
6107 Specific macros that control the handling of
6108 initialization and termination routines.
6109 * Instruction Output:: Output of actual instructions.
6110 * Dispatch Tables:: Output of jump tables.
6111 * Exception Region Output:: Output of exception region code.
6112 * Alignment Output:: Pseudo ops for alignment and skipping data.
6115 @node File Framework
6116 @subsection The Overall Framework of an Assembler File
6117 @cindex assembler format
6118 @cindex output of assembler code
6120 @c prevent bad page break with this line
6121 This describes the overall framework of an assembly file.
6123 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6124 @findex default_file_start
6125 Output to @code{asm_out_file} any text which the assembler expects to
6126 find at the beginning of a file. The default behavior is controlled
6127 by two flags, documented below. Unless your target's assembler is
6128 quite unusual, if you override the default, you should call
6129 @code{default_file_start} at some point in your target hook. This
6130 lets other target files rely on these variables.
6133 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6134 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6135 printed as the very first line in the assembly file, unless
6136 @option{-fverbose-asm} is in effect. (If that macro has been defined
6137 to the empty string, this variable has no effect.) With the normal
6138 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6139 assembler that it need not bother stripping comments or extra
6140 whitespace from its input. This allows it to work a bit faster.
6142 The default is false. You should not set it to true unless you have
6143 verified that your port does not generate any extra whitespace or
6144 comments that will cause GAS to issue errors in NO_APP mode.
6147 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6148 If this flag is true, @code{output_file_directive} will be called
6149 for the primary source file, immediately after printing
6150 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6151 this to be done. The default is false.
6154 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6155 Output to @code{asm_out_file} any text which the assembler expects
6156 to find at the end of a file. The default is to output nothing.
6159 @deftypefun void file_end_indicate_exec_stack ()
6160 Some systems use a common convention, the @samp{.note.GNU-stack}
6161 special section, to indicate whether or not an object file relies on
6162 the stack being executable. If your system uses this convention, you
6163 should define @code{TARGET_ASM_FILE_END} to this function. If you
6164 need to do other things in that hook, have your hook function call
6168 @defmac ASM_COMMENT_START
6169 A C string constant describing how to begin a comment in the target
6170 assembler language. The compiler assumes that the comment will end at
6171 the end of the line.
6175 A C string constant for text to be output before each @code{asm}
6176 statement or group of consecutive ones. Normally this is
6177 @code{"#APP"}, which is a comment that has no effect on most
6178 assemblers but tells the GNU assembler that it must check the lines
6179 that follow for all valid assembler constructs.
6183 A C string constant for text to be output after each @code{asm}
6184 statement or group of consecutive ones. Normally this is
6185 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6186 time-saving assumptions that are valid for ordinary compiler output.
6189 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6190 A C statement to output COFF information or DWARF debugging information
6191 which indicates that filename @var{name} is the current source file to
6192 the stdio stream @var{stream}.
6194 This macro need not be defined if the standard form of output
6195 for the file format in use is appropriate.
6198 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6199 A C statement to output the string @var{string} to the stdio stream
6200 @var{stream}. If you do not call the function @code{output_quoted_string}
6201 in your config files, GCC will only call it to output filenames to
6202 the assembler source. So you can use it to canonicalize the format
6203 of the filename using this macro.
6206 @defmac ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
6207 A C statement to output DBX or SDB debugging information before code
6208 for line number @var{line} of the current source file to the
6209 stdio stream @var{stream}. @var{counter} is the number of time the
6210 macro was invoked, including the current invocation; it is intended
6211 to generate unique labels in the assembly output.
6213 This macro need not be defined if the standard form of debugging
6214 information for the debugger in use is appropriate.
6217 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6218 A C statement to output something to the assembler file to handle a
6219 @samp{#ident} directive containing the text @var{string}. If this
6220 macro is not defined, nothing is output for a @samp{#ident} directive.
6223 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6224 Output assembly directives to switch to section @var{name}. The section
6225 should have attributes as specified by @var{flags}, which is a bit mask
6226 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6227 is nonzero, it contains an alignment in bytes to be used for the section,
6228 otherwise some target default should be used. Only targets that must
6229 specify an alignment within the section directive need pay attention to
6230 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6233 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6234 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6237 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6238 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6239 based on a variable or function decl, a section name, and whether or not the
6240 declaration's initializer may contain runtime relocations. @var{decl} may be
6241 null, in which case read-write data should be assumed.
6243 The default version if this function handles choosing code vs data,
6244 read-only vs read-write data, and @code{flag_pic}. You should only
6245 need to override this if your target has special flags that might be
6246 set via @code{__attribute__}.
6251 @subsection Output of Data
6254 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6255 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6256 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6257 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6258 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6259 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6260 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6261 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6262 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6263 These hooks specify assembly directives for creating certain kinds
6264 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6265 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6266 aligned two-byte object, and so on. Any of the hooks may be
6267 @code{NULL}, indicating that no suitable directive is available.
6269 The compiler will print these strings at the start of a new line,
6270 followed immediately by the object's initial value. In most cases,
6271 the string should contain a tab, a pseudo-op, and then another tab.
6274 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6275 The @code{assemble_integer} function uses this hook to output an
6276 integer object. @var{x} is the object's value, @var{size} is its size
6277 in bytes and @var{aligned_p} indicates whether it is aligned. The
6278 function should return @code{true} if it was able to output the
6279 object. If it returns false, @code{assemble_integer} will try to
6280 split the object into smaller parts.
6282 The default implementation of this hook will use the
6283 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6284 when the relevant string is @code{NULL}.
6287 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6288 A C statement to recognize @var{rtx} patterns that
6289 @code{output_addr_const} can't deal with, and output assembly code to
6290 @var{stream} corresponding to the pattern @var{x}. This may be used to
6291 allow machine-dependent @code{UNSPEC}s to appear within constants.
6293 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6294 @code{goto fail}, so that a standard error message is printed. If it
6295 prints an error message itself, by calling, for example,
6296 @code{output_operand_lossage}, it may just complete normally.
6299 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6300 A C statement to output to the stdio stream @var{stream} an assembler
6301 instruction to assemble a string constant containing the @var{len}
6302 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6303 @code{char *} and @var{len} a C expression of type @code{int}.
6305 If the assembler has a @code{.ascii} pseudo-op as found in the
6306 Berkeley Unix assembler, do not define the macro
6307 @code{ASM_OUTPUT_ASCII}.
6310 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6311 A C statement to output word @var{n} of a function descriptor for
6312 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6313 is defined, and is otherwise unused.
6316 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6317 You may define this macro as a C expression. You should define the
6318 expression to have a nonzero value if GCC should output the constant
6319 pool for a function before the code for the function, or a zero value if
6320 GCC should output the constant pool after the function. If you do
6321 not define this macro, the usual case, GCC will output the constant
6322 pool before the function.
6325 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6326 A C statement to output assembler commands to define the start of the
6327 constant pool for a function. @var{funname} is a string giving
6328 the name of the function. Should the return type of the function
6329 be required, it can be obtained via @var{fundecl}. @var{size}
6330 is the size, in bytes, of the constant pool that will be written
6331 immediately after this call.
6333 If no constant-pool prefix is required, the usual case, this macro need
6337 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6338 A C statement (with or without semicolon) to output a constant in the
6339 constant pool, if it needs special treatment. (This macro need not do
6340 anything for RTL expressions that can be output normally.)
6342 The argument @var{file} is the standard I/O stream to output the
6343 assembler code on. @var{x} is the RTL expression for the constant to
6344 output, and @var{mode} is the machine mode (in case @var{x} is a
6345 @samp{const_int}). @var{align} is the required alignment for the value
6346 @var{x}; you should output an assembler directive to force this much
6349 The argument @var{labelno} is a number to use in an internal label for
6350 the address of this pool entry. The definition of this macro is
6351 responsible for outputting the label definition at the proper place.
6352 Here is how to do this:
6355 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6358 When you output a pool entry specially, you should end with a
6359 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6360 entry from being output a second time in the usual manner.
6362 You need not define this macro if it would do nothing.
6365 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6366 A C statement to output assembler commands to at the end of the constant
6367 pool for a function. @var{funname} is a string giving the name of the
6368 function. Should the return type of the function be required, you can
6369 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6370 constant pool that GCC wrote immediately before this call.
6372 If no constant-pool epilogue is required, the usual case, you need not
6376 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6377 Define this macro as a C expression which is nonzero if @var{C} is
6378 used as a logical line separator by the assembler.
6380 If you do not define this macro, the default is that only
6381 the character @samp{;} is treated as a logical line separator.
6384 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6385 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6386 These target hooks are C string constants, describing the syntax in the
6387 assembler for grouping arithmetic expressions. If not overridden, they
6388 default to normal parentheses, which is correct for most assemblers.
6391 These macros are provided by @file{real.h} for writing the definitions
6392 of @code{ASM_OUTPUT_DOUBLE} and the like:
6394 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6395 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6396 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6397 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6398 floating point representation, and store its bit pattern in the variable
6399 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6400 be a simple @code{long int}. For the others, it should be an array of
6401 @code{long int}. The number of elements in this array is determined by
6402 the size of the desired target floating point data type: 32 bits of it
6403 go in each @code{long int} array element. Each array element holds 32
6404 bits of the result, even if @code{long int} is wider than 32 bits on the
6407 The array element values are designed so that you can print them out
6408 using @code{fprintf} in the order they should appear in the target
6412 @node Uninitialized Data
6413 @subsection Output of Uninitialized Variables
6415 Each of the macros in this section is used to do the whole job of
6416 outputting a single uninitialized variable.
6418 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6419 A C statement (sans semicolon) to output to the stdio stream
6420 @var{stream} the assembler definition of a common-label named
6421 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6422 is the size rounded up to whatever alignment the caller wants.
6424 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6425 output the name itself; before and after that, output the additional
6426 assembler syntax for defining the name, and a newline.
6428 This macro controls how the assembler definitions of uninitialized
6429 common global variables are output.
6432 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6433 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6434 separate, explicit argument. If you define this macro, it is used in
6435 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6436 handling the required alignment of the variable. The alignment is specified
6437 as the number of bits.
6440 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6441 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6442 variable to be output, if there is one, or @code{NULL_TREE} if there
6443 is no corresponding variable. If you define this macro, GCC will use it
6444 in place of both @code{ASM_OUTPUT_COMMON} and
6445 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6446 the variable's decl in order to chose what to output.
6449 @defmac ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6450 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6451 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6455 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6456 A C statement (sans semicolon) to output to the stdio stream
6457 @var{stream} the assembler definition of uninitialized global @var{decl} named
6458 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6459 is the size rounded up to whatever alignment the caller wants.
6461 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6462 defining this macro. If unable, use the expression
6463 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6464 before and after that, output the additional assembler syntax for defining
6465 the name, and a newline.
6467 This macro controls how the assembler definitions of uninitialized global
6468 variables are output. This macro exists to properly support languages like
6469 C++ which do not have @code{common} data. However, this macro currently
6470 is not defined for all targets. If this macro and
6471 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6472 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6473 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6476 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6477 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6478 separate, explicit argument. If you define this macro, it is used in
6479 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6480 handling the required alignment of the variable. The alignment is specified
6481 as the number of bits.
6483 Try to use function @code{asm_output_aligned_bss} defined in file
6484 @file{varasm.c} when defining this macro.
6487 @defmac ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6488 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6489 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6493 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6494 A C statement (sans semicolon) to output to the stdio stream
6495 @var{stream} the assembler definition of a local-common-label named
6496 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6497 is the size rounded up to whatever alignment the caller wants.
6499 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6500 output the name itself; before and after that, output the additional
6501 assembler syntax for defining the name, and a newline.
6503 This macro controls how the assembler definitions of uninitialized
6504 static variables are output.
6507 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6508 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6509 separate, explicit argument. If you define this macro, it is used in
6510 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6511 handling the required alignment of the variable. The alignment is specified
6512 as the number of bits.
6515 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6516 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6517 variable to be output, if there is one, or @code{NULL_TREE} if there
6518 is no corresponding variable. If you define this macro, GCC will use it
6519 in place of both @code{ASM_OUTPUT_DECL} and
6520 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6521 the variable's decl in order to chose what to output.
6524 @defmac ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6525 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6526 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6531 @subsection Output and Generation of Labels
6533 @c prevent bad page break with this line
6534 This is about outputting labels.
6536 @findex assemble_name
6537 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6538 A C statement (sans semicolon) to output to the stdio stream
6539 @var{stream} the assembler definition of a label named @var{name}.
6540 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6541 output the name itself; before and after that, output the additional
6542 assembler syntax for defining the name, and a newline. A default
6543 definition of this macro is provided which is correct for most systems.
6547 A C string containing the appropriate assembler directive to specify the
6548 size of a symbol, without any arguments. On systems that use ELF, the
6549 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6550 systems, the default is not to define this macro.
6552 Define this macro only if it is correct to use the default definitions
6553 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6554 for your system. If you need your own custom definitions of those
6555 macros, or if you do not need explicit symbol sizes at all, do not
6559 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6560 A C statement (sans semicolon) to output to the stdio stream
6561 @var{stream} a directive telling the assembler that the size of the
6562 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6563 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6567 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6568 A C statement (sans semicolon) to output to the stdio stream
6569 @var{stream} a directive telling the assembler to calculate the size of
6570 the symbol @var{name} by subtracting its address from the current
6573 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6574 provided. The default assumes that the assembler recognizes a special
6575 @samp{.} symbol as referring to the current address, and can calculate
6576 the difference between this and another symbol. If your assembler does
6577 not recognize @samp{.} or cannot do calculations with it, you will need
6578 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6582 A C string containing the appropriate assembler directive to specify the
6583 type of a symbol, without any arguments. On systems that use ELF, the
6584 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6585 systems, the default is not to define this macro.
6587 Define this macro only if it is correct to use the default definition of
6588 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6589 custom definition of this macro, or if you do not need explicit symbol
6590 types at all, do not define this macro.
6593 @defmac TYPE_OPERAND_FMT
6594 A C string which specifies (using @code{printf} syntax) the format of
6595 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6596 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6597 the default is not to define this macro.
6599 Define this macro only if it is correct to use the default definition of
6600 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6601 custom definition of this macro, or if you do not need explicit symbol
6602 types at all, do not define this macro.
6605 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6606 A C statement (sans semicolon) to output to the stdio stream
6607 @var{stream} a directive telling the assembler that the type of the
6608 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6609 that string is always either @samp{"function"} or @samp{"object"}, but
6610 you should not count on this.
6612 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6613 definition of this macro is provided.
6616 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6617 A C statement (sans semicolon) to output to the stdio stream
6618 @var{stream} any text necessary for declaring the name @var{name} of a
6619 function which is being defined. This macro is responsible for
6620 outputting the label definition (perhaps using
6621 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6622 @code{FUNCTION_DECL} tree node representing the function.
6624 If this macro is not defined, then the function name is defined in the
6625 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6627 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6631 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6632 A C statement (sans semicolon) to output to the stdio stream
6633 @var{stream} any text necessary for declaring the size of a function
6634 which is being defined. The argument @var{name} is the name of the
6635 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6636 representing the function.
6638 If this macro is not defined, then the function size is not defined.
6640 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6644 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6645 A C statement (sans semicolon) to output to the stdio stream
6646 @var{stream} any text necessary for declaring the name @var{name} of an
6647 initialized variable which is being defined. This macro must output the
6648 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6649 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6651 If this macro is not defined, then the variable name is defined in the
6652 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6654 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6655 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6658 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
6659 A C statement (sans semicolon) to output to the stdio stream
6660 @var{stream} any text necessary for declaring the name @var{name} of a
6661 constant which is being defined. This macro is responsible for
6662 outputting the label definition (perhaps using
6663 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
6664 value of the constant, and @var{size} is the size of the constant
6665 in bytes. @var{name} will be an internal label.
6667 If this macro is not defined, then the @var{name} is defined in the
6668 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6670 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6674 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6675 A C statement (sans semicolon) to output to the stdio stream
6676 @var{stream} any text necessary for claiming a register @var{regno}
6677 for a global variable @var{decl} with name @var{name}.
6679 If you don't define this macro, that is equivalent to defining it to do
6683 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6684 A C statement (sans semicolon) to finish up declaring a variable name
6685 once the compiler has processed its initializer fully and thus has had a
6686 chance to determine the size of an array when controlled by an
6687 initializer. This is used on systems where it's necessary to declare
6688 something about the size of the object.
6690 If you don't define this macro, that is equivalent to defining it to do
6693 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6694 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6697 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6698 This target hook is a function to output to the stdio stream
6699 @var{stream} some commands that will make the label @var{name} global;
6700 that is, available for reference from other files.
6702 The default implementation relies on a proper definition of
6703 @code{GLOBAL_ASM_OP}.
6706 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6707 A C statement (sans semicolon) to output to the stdio stream
6708 @var{stream} some commands that will make the label @var{name} weak;
6709 that is, available for reference from other files but only used if
6710 no other definition is available. Use the expression
6711 @code{assemble_name (@var{stream}, @var{name})} to output the name
6712 itself; before and after that, output the additional assembler syntax
6713 for making that name weak, and a newline.
6715 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6716 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6720 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6721 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6722 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6723 or variable decl. If @var{value} is not @code{NULL}, this C statement
6724 should output to the stdio stream @var{stream} assembler code which
6725 defines (equates) the weak symbol @var{name} to have the value
6726 @var{value}. If @var{value} is @code{NULL}, it should output commands
6727 to make @var{name} weak.
6730 @defmac SUPPORTS_WEAK
6731 A C expression which evaluates to true if the target supports weak symbols.
6733 If you don't define this macro, @file{defaults.h} provides a default
6734 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6735 is defined, the default definition is @samp{1}; otherwise, it is
6736 @samp{0}. Define this macro if you want to control weak symbol support
6737 with a compiler flag such as @option{-melf}.
6740 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
6741 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6742 public symbol such that extra copies in multiple translation units will
6743 be discarded by the linker. Define this macro if your object file
6744 format provides support for this concept, such as the @samp{COMDAT}
6745 section flags in the Microsoft Windows PE/COFF format, and this support
6746 requires changes to @var{decl}, such as putting it in a separate section.
6749 @defmac SUPPORTS_ONE_ONLY
6750 A C expression which evaluates to true if the target supports one-only
6753 If you don't define this macro, @file{varasm.c} provides a default
6754 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6755 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6756 you want to control one-only symbol support with a compiler flag, or if
6757 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6758 be emitted as one-only.
6761 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6762 This target hook is a function to output to @var{asm_out_file} some
6763 commands that will make the symbol(s) associated with @var{decl} have
6764 hidden, protected or internal visibility as specified by @var{visibility}.
6767 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
6768 A C expression that evaluates to true if the target's linker expects
6769 that weak symbols do not appear in a static archive's table of contents.
6770 The default is @code{0}.
6772 Leaving weak symbols out of an archive's table of contents means that,
6773 if a symbol will only have a definition in one translation unit and
6774 will have undefined references from other translation units, that
6775 symbol should not be weak. Defining this macro to be nonzero will
6776 thus have the effect that certain symbols that would normally be weak
6777 (explicit template instantiations, and vtables for polymorphic classes
6778 with noninline key methods) will instead be nonweak.
6780 The C++ ABI requires this macro to be zero. Define this macro for
6781 targets where full C++ ABI compliance is impossible and where linker
6782 restrictions require weak symbols to be left out of a static archive's
6786 @defmac TARGET_SUPPORTS_HIDDEN
6787 A C expression that evaluates to true if the target supports hidden
6788 visibility. By default this expression is true if and only if
6789 @code{HAS_GAS_HIDDEN} is defined. Set this macro if the
6790 @code{HAS_GAS_HIDDEN} macro gives the wrong answer for this
6791 target. (For example, if the target's mechanism for supporting
6792 hidden visibility is not the same as GAS's.)
6795 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6796 A C statement (sans semicolon) to output to the stdio stream
6797 @var{stream} any text necessary for declaring the name of an external
6798 symbol named @var{name} which is referenced in this compilation but
6799 not defined. The value of @var{decl} is the tree node for the
6802 This macro need not be defined if it does not need to output anything.
6803 The GNU assembler and most Unix assemblers don't require anything.
6806 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
6807 This target hook is a function to output to @var{asm_out_file} an assembler
6808 pseudo-op to declare a library function name external. The name of the
6809 library function is given by @var{symref}, which is a @code{symbol_ref}.
6812 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl})
6813 This target hook is a function to output to @var{asm_out_file} an assembler
6814 directive to annotate used symbol. Darwin target use .no_dead_code_strip
6818 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6819 A C statement (sans semicolon) to output to the stdio stream
6820 @var{stream} a reference in assembler syntax to a label named
6821 @var{name}. This should add @samp{_} to the front of the name, if that
6822 is customary on your operating system, as it is in most Berkeley Unix
6823 systems. This macro is used in @code{assemble_name}.
6826 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6827 A C statement (sans semicolon) to output a reference to
6828 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6829 will be used to output the name of the symbol. This macro may be used
6830 to modify the way a symbol is referenced depending on information
6831 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6834 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6835 A C statement (sans semicolon) to output a reference to @var{buf}, the
6836 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6837 @code{assemble_name} will be used to output the name of the symbol.
6838 This macro is not used by @code{output_asm_label}, or the @code{%l}
6839 specifier that calls it; the intention is that this macro should be set
6840 when it is necessary to output a label differently when its address is
6844 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6845 A function to output to the stdio stream @var{stream} a label whose
6846 name is made from the string @var{prefix} and the number @var{labelno}.
6848 It is absolutely essential that these labels be distinct from the labels
6849 used for user-level functions and variables. Otherwise, certain programs
6850 will have name conflicts with internal labels.
6852 It is desirable to exclude internal labels from the symbol table of the
6853 object file. Most assemblers have a naming convention for labels that
6854 should be excluded; on many systems, the letter @samp{L} at the
6855 beginning of a label has this effect. You should find out what
6856 convention your system uses, and follow it.
6858 The default version of this function utilizes ASM_GENERATE_INTERNAL_LABEL.
6861 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6862 A C statement to output to the stdio stream @var{stream} a debug info
6863 label whose name is made from the string @var{prefix} and the number
6864 @var{num}. This is useful for VLIW targets, where debug info labels
6865 may need to be treated differently than branch target labels. On some
6866 systems, branch target labels must be at the beginning of instruction
6867 bundles, but debug info labels can occur in the middle of instruction
6870 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
6874 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6875 A C statement to store into the string @var{string} a label whose name
6876 is made from the string @var{prefix} and the number @var{num}.
6878 This string, when output subsequently by @code{assemble_name}, should
6879 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
6880 with the same @var{prefix} and @var{num}.
6882 If the string begins with @samp{*}, then @code{assemble_name} will
6883 output the rest of the string unchanged. It is often convenient for
6884 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6885 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6886 to output the string, and may change it. (Of course,
6887 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6888 you should know what it does on your machine.)
6891 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6892 A C expression to assign to @var{outvar} (which is a variable of type
6893 @code{char *}) a newly allocated string made from the string
6894 @var{name} and the number @var{number}, with some suitable punctuation
6895 added. Use @code{alloca} to get space for the string.
6897 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6898 produce an assembler label for an internal static variable whose name is
6899 @var{name}. Therefore, the string must be such as to result in valid
6900 assembler code. The argument @var{number} is different each time this
6901 macro is executed; it prevents conflicts between similarly-named
6902 internal static variables in different scopes.
6904 Ideally this string should not be a valid C identifier, to prevent any
6905 conflict with the user's own symbols. Most assemblers allow periods
6906 or percent signs in assembler symbols; putting at least one of these
6907 between the name and the number will suffice.
6909 If this macro is not defined, a default definition will be provided
6910 which is correct for most systems.
6913 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6914 A C statement to output to the stdio stream @var{stream} assembler code
6915 which defines (equates) the symbol @var{name} to have the value @var{value}.
6918 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6919 correct for most systems.
6922 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6923 A C statement to output to the stdio stream @var{stream} assembler code
6924 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6925 to have the value of the tree node @var{decl_of_value}. This macro will
6926 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6927 the tree nodes are available.
6930 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6931 correct for most systems.
6934 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6935 A C statement to output to the stdio stream @var{stream} assembler code
6936 which defines (equates) the weak symbol @var{name} to have the value
6937 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6938 an undefined weak symbol.
6940 Define this macro if the target only supports weak aliases; define
6941 @code{ASM_OUTPUT_DEF} instead if possible.
6944 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6945 Define this macro to override the default assembler names used for
6946 Objective-C methods.
6948 The default name is a unique method number followed by the name of the
6949 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6950 the category is also included in the assembler name (e.g.@:
6953 These names are safe on most systems, but make debugging difficult since
6954 the method's selector is not present in the name. Therefore, particular
6955 systems define other ways of computing names.
6957 @var{buf} is an expression of type @code{char *} which gives you a
6958 buffer in which to store the name; its length is as long as
6959 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6960 50 characters extra.
6962 The argument @var{is_inst} specifies whether the method is an instance
6963 method or a class method; @var{class_name} is the name of the class;
6964 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6965 in a category); and @var{sel_name} is the name of the selector.
6967 On systems where the assembler can handle quoted names, you can use this
6968 macro to provide more human-readable names.
6971 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
6972 A C statement (sans semicolon) to output to the stdio stream
6973 @var{stream} commands to declare that the label @var{name} is an
6974 Objective-C class reference. This is only needed for targets whose
6975 linkers have special support for NeXT-style runtimes.
6978 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6979 A C statement (sans semicolon) to output to the stdio stream
6980 @var{stream} commands to declare that the label @var{name} is an
6981 unresolved Objective-C class reference. This is only needed for targets
6982 whose linkers have special support for NeXT-style runtimes.
6985 @node Initialization
6986 @subsection How Initialization Functions Are Handled
6987 @cindex initialization routines
6988 @cindex termination routines
6989 @cindex constructors, output of
6990 @cindex destructors, output of
6992 The compiled code for certain languages includes @dfn{constructors}
6993 (also called @dfn{initialization routines})---functions to initialize
6994 data in the program when the program is started. These functions need
6995 to be called before the program is ``started''---that is to say, before
6996 @code{main} is called.
6998 Compiling some languages generates @dfn{destructors} (also called
6999 @dfn{termination routines}) that should be called when the program
7002 To make the initialization and termination functions work, the compiler
7003 must output something in the assembler code to cause those functions to
7004 be called at the appropriate time. When you port the compiler to a new
7005 system, you need to specify how to do this.
7007 There are two major ways that GCC currently supports the execution of
7008 initialization and termination functions. Each way has two variants.
7009 Much of the structure is common to all four variations.
7011 @findex __CTOR_LIST__
7012 @findex __DTOR_LIST__
7013 The linker must build two lists of these functions---a list of
7014 initialization functions, called @code{__CTOR_LIST__}, and a list of
7015 termination functions, called @code{__DTOR_LIST__}.
7017 Each list always begins with an ignored function pointer (which may hold
7018 0, @minus{}1, or a count of the function pointers after it, depending on
7019 the environment). This is followed by a series of zero or more function
7020 pointers to constructors (or destructors), followed by a function
7021 pointer containing zero.
7023 Depending on the operating system and its executable file format, either
7024 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7025 time and exit time. Constructors are called in reverse order of the
7026 list; destructors in forward order.
7028 The best way to handle static constructors works only for object file
7029 formats which provide arbitrarily-named sections. A section is set
7030 aside for a list of constructors, and another for a list of destructors.
7031 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7032 object file that defines an initialization function also puts a word in
7033 the constructor section to point to that function. The linker
7034 accumulates all these words into one contiguous @samp{.ctors} section.
7035 Termination functions are handled similarly.
7037 This method will be chosen as the default by @file{target-def.h} if
7038 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7039 support arbitrary sections, but does support special designated
7040 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7041 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7043 When arbitrary sections are available, there are two variants, depending
7044 upon how the code in @file{crtstuff.c} is called. On systems that
7045 support a @dfn{.init} section which is executed at program startup,
7046 parts of @file{crtstuff.c} are compiled into that section. The
7047 program is linked by the @command{gcc} driver like this:
7050 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7053 The prologue of a function (@code{__init}) appears in the @code{.init}
7054 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7055 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7056 files are provided by the operating system or by the GNU C library, but
7057 are provided by GCC for a few targets.
7059 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7060 compiled from @file{crtstuff.c}. They contain, among other things, code
7061 fragments within the @code{.init} and @code{.fini} sections that branch
7062 to routines in the @code{.text} section. The linker will pull all parts
7063 of a section together, which results in a complete @code{__init} function
7064 that invokes the routines we need at startup.
7066 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7069 If no init section is available, when GCC compiles any function called
7070 @code{main} (or more accurately, any function designated as a program
7071 entry point by the language front end calling @code{expand_main_function}),
7072 it inserts a procedure call to @code{__main} as the first executable code
7073 after the function prologue. The @code{__main} function is defined
7074 in @file{libgcc2.c} and runs the global constructors.
7076 In file formats that don't support arbitrary sections, there are again
7077 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7078 and an `a.out' format must be used. In this case,
7079 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7080 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7081 and with the address of the void function containing the initialization
7082 code as its value. The GNU linker recognizes this as a request to add
7083 the value to a @dfn{set}; the values are accumulated, and are eventually
7084 placed in the executable as a vector in the format described above, with
7085 a leading (ignored) count and a trailing zero element.
7086 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7087 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7088 the compilation of @code{main} to call @code{__main} as above, starting
7089 the initialization process.
7091 The last variant uses neither arbitrary sections nor the GNU linker.
7092 This is preferable when you want to do dynamic linking and when using
7093 file formats which the GNU linker does not support, such as `ECOFF'@. In
7094 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7095 termination functions are recognized simply by their names. This requires
7096 an extra program in the linkage step, called @command{collect2}. This program
7097 pretends to be the linker, for use with GCC; it does its job by running
7098 the ordinary linker, but also arranges to include the vectors of
7099 initialization and termination functions. These functions are called
7100 via @code{__main} as described above. In order to use this method,
7101 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7104 The following section describes the specific macros that control and
7105 customize the handling of initialization and termination functions.
7108 @node Macros for Initialization
7109 @subsection Macros Controlling Initialization Routines
7111 Here are the macros that control how the compiler handles initialization
7112 and termination functions:
7114 @defmac INIT_SECTION_ASM_OP
7115 If defined, a C string constant, including spacing, for the assembler
7116 operation to identify the following data as initialization code. If not
7117 defined, GCC will assume such a section does not exist. When you are
7118 using special sections for initialization and termination functions, this
7119 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7120 run the initialization functions.
7123 @defmac HAS_INIT_SECTION
7124 If defined, @code{main} will not call @code{__main} as described above.
7125 This macro should be defined for systems that control start-up code
7126 on a symbol-by-symbol basis, such as OSF/1, and should not
7127 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7130 @defmac LD_INIT_SWITCH
7131 If defined, a C string constant for a switch that tells the linker that
7132 the following symbol is an initialization routine.
7135 @defmac LD_FINI_SWITCH
7136 If defined, a C string constant for a switch that tells the linker that
7137 the following symbol is a finalization routine.
7140 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7141 If defined, a C statement that will write a function that can be
7142 automatically called when a shared library is loaded. The function
7143 should call @var{func}, which takes no arguments. If not defined, and
7144 the object format requires an explicit initialization function, then a
7145 function called @code{_GLOBAL__DI} will be generated.
7147 This function and the following one are used by collect2 when linking a
7148 shared library that needs constructors or destructors, or has DWARF2
7149 exception tables embedded in the code.
7152 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7153 If defined, a C statement that will write a function that can be
7154 automatically called when a shared library is unloaded. The function
7155 should call @var{func}, which takes no arguments. If not defined, and
7156 the object format requires an explicit finalization function, then a
7157 function called @code{_GLOBAL__DD} will be generated.
7160 @defmac INVOKE__main
7161 If defined, @code{main} will call @code{__main} despite the presence of
7162 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7163 where the init section is not actually run automatically, but is still
7164 useful for collecting the lists of constructors and destructors.
7167 @defmac SUPPORTS_INIT_PRIORITY
7168 If nonzero, the C++ @code{init_priority} attribute is supported and the
7169 compiler should emit instructions to control the order of initialization
7170 of objects. If zero, the compiler will issue an error message upon
7171 encountering an @code{init_priority} attribute.
7174 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7175 This value is true if the target supports some ``native'' method of
7176 collecting constructors and destructors to be run at startup and exit.
7177 It is false if we must use @command{collect2}.
7180 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7181 If defined, a function that outputs assembler code to arrange to call
7182 the function referenced by @var{symbol} at initialization time.
7184 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7185 no arguments and with no return value. If the target supports initialization
7186 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7187 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7189 If this macro is not defined by the target, a suitable default will
7190 be chosen if (1) the target supports arbitrary section names, (2) the
7191 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7195 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7196 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7197 functions rather than initialization functions.
7200 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7201 generated for the generated object file will have static linkage.
7203 If your system uses @command{collect2} as the means of processing
7204 constructors, then that program normally uses @command{nm} to scan
7205 an object file for constructor functions to be called.
7207 On certain kinds of systems, you can define this macro to make
7208 @command{collect2} work faster (and, in some cases, make it work at all):
7210 @defmac OBJECT_FORMAT_COFF
7211 Define this macro if the system uses COFF (Common Object File Format)
7212 object files, so that @command{collect2} can assume this format and scan
7213 object files directly for dynamic constructor/destructor functions.
7215 This macro is effective only in a native compiler; @command{collect2} as
7216 part of a cross compiler always uses @command{nm} for the target machine.
7219 @defmac COLLECT_PARSE_FLAG (@var{flag})
7220 Define this macro to be C code that examines @command{collect2} command
7221 line option @var{flag} and performs special actions if
7222 @command{collect2} needs to behave differently depending on @var{flag}.
7225 @defmac REAL_NM_FILE_NAME
7226 Define this macro as a C string constant containing the file name to use
7227 to execute @command{nm}. The default is to search the path normally for
7230 If your system supports shared libraries and has a program to list the
7231 dynamic dependencies of a given library or executable, you can define
7232 these macros to enable support for running initialization and
7233 termination functions in shared libraries:
7237 Define this macro to a C string constant containing the name of the program
7238 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7241 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7242 Define this macro to be C code that extracts filenames from the output
7243 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7244 of type @code{char *} that points to the beginning of a line of output
7245 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7246 code must advance @var{ptr} to the beginning of the filename on that
7247 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7250 @node Instruction Output
7251 @subsection Output of Assembler Instructions
7253 @c prevent bad page break with this line
7254 This describes assembler instruction output.
7256 @defmac REGISTER_NAMES
7257 A C initializer containing the assembler's names for the machine
7258 registers, each one as a C string constant. This is what translates
7259 register numbers in the compiler into assembler language.
7262 @defmac ADDITIONAL_REGISTER_NAMES
7263 If defined, a C initializer for an array of structures containing a name
7264 and a register number. This macro defines additional names for hard
7265 registers, thus allowing the @code{asm} option in declarations to refer
7266 to registers using alternate names.
7269 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7270 Define this macro if you are using an unusual assembler that
7271 requires different names for the machine instructions.
7273 The definition is a C statement or statements which output an
7274 assembler instruction opcode to the stdio stream @var{stream}. The
7275 macro-operand @var{ptr} is a variable of type @code{char *} which
7276 points to the opcode name in its ``internal'' form---the form that is
7277 written in the machine description. The definition should output the
7278 opcode name to @var{stream}, performing any translation you desire, and
7279 increment the variable @var{ptr} to point at the end of the opcode
7280 so that it will not be output twice.
7282 In fact, your macro definition may process less than the entire opcode
7283 name, or more than the opcode name; but if you want to process text
7284 that includes @samp{%}-sequences to substitute operands, you must take
7285 care of the substitution yourself. Just be sure to increment
7286 @var{ptr} over whatever text should not be output normally.
7288 @findex recog_data.operand
7289 If you need to look at the operand values, they can be found as the
7290 elements of @code{recog_data.operand}.
7292 If the macro definition does nothing, the instruction is output
7296 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7297 If defined, a C statement to be executed just prior to the output of
7298 assembler code for @var{insn}, to modify the extracted operands so
7299 they will be output differently.
7301 Here the argument @var{opvec} is the vector containing the operands
7302 extracted from @var{insn}, and @var{noperands} is the number of
7303 elements of the vector which contain meaningful data for this insn.
7304 The contents of this vector are what will be used to convert the insn
7305 template into assembler code, so you can change the assembler output
7306 by changing the contents of the vector.
7308 This macro is useful when various assembler syntaxes share a single
7309 file of instruction patterns; by defining this macro differently, you
7310 can cause a large class of instructions to be output differently (such
7311 as with rearranged operands). Naturally, variations in assembler
7312 syntax affecting individual insn patterns ought to be handled by
7313 writing conditional output routines in those patterns.
7315 If this macro is not defined, it is equivalent to a null statement.
7318 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7319 A C compound statement to output to stdio stream @var{stream} the
7320 assembler syntax for an instruction operand @var{x}. @var{x} is an
7323 @var{code} is a value that can be used to specify one of several ways
7324 of printing the operand. It is used when identical operands must be
7325 printed differently depending on the context. @var{code} comes from
7326 the @samp{%} specification that was used to request printing of the
7327 operand. If the specification was just @samp{%@var{digit}} then
7328 @var{code} is 0; if the specification was @samp{%@var{ltr}
7329 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7332 If @var{x} is a register, this macro should print the register's name.
7333 The names can be found in an array @code{reg_names} whose type is
7334 @code{char *[]}. @code{reg_names} is initialized from
7335 @code{REGISTER_NAMES}.
7337 When the machine description has a specification @samp{%@var{punct}}
7338 (a @samp{%} followed by a punctuation character), this macro is called
7339 with a null pointer for @var{x} and the punctuation character for
7343 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7344 A C expression which evaluates to true if @var{code} is a valid
7345 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7346 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7347 punctuation characters (except for the standard one, @samp{%}) are used
7351 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7352 A C compound statement to output to stdio stream @var{stream} the
7353 assembler syntax for an instruction operand that is a memory reference
7354 whose address is @var{x}. @var{x} is an RTL expression.
7356 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7357 On some machines, the syntax for a symbolic address depends on the
7358 section that the address refers to. On these machines, define the hook
7359 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7360 @code{symbol_ref}, and then check for it here. @xref{Assembler
7364 @findex dbr_sequence_length
7365 @defmac DBR_OUTPUT_SEQEND (@var{file})
7366 A C statement, to be executed after all slot-filler instructions have
7367 been output. If necessary, call @code{dbr_sequence_length} to
7368 determine the number of slots filled in a sequence (zero if not
7369 currently outputting a sequence), to decide how many no-ops to output,
7372 Don't define this macro if it has nothing to do, but it is helpful in
7373 reading assembly output if the extent of the delay sequence is made
7374 explicit (e.g.@: with white space).
7377 @findex final_sequence
7378 Note that output routines for instructions with delay slots must be
7379 prepared to deal with not being output as part of a sequence
7380 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7381 found.) The variable @code{final_sequence} is null when not
7382 processing a sequence, otherwise it contains the @code{sequence} rtx
7386 @defmac REGISTER_PREFIX
7387 @defmacx LOCAL_LABEL_PREFIX
7388 @defmacx USER_LABEL_PREFIX
7389 @defmacx IMMEDIATE_PREFIX
7390 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7391 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7392 @file{final.c}). These are useful when a single @file{md} file must
7393 support multiple assembler formats. In that case, the various @file{tm.h}
7394 files can define these macros differently.
7397 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
7398 If defined this macro should expand to a series of @code{case}
7399 statements which will be parsed inside the @code{switch} statement of
7400 the @code{asm_fprintf} function. This allows targets to define extra
7401 printf formats which may useful when generating their assembler
7402 statements. Note that uppercase letters are reserved for future
7403 generic extensions to asm_fprintf, and so are not available to target
7404 specific code. The output file is given by the parameter @var{file}.
7405 The varargs input pointer is @var{argptr} and the rest of the format
7406 string, starting the character after the one that is being switched
7407 upon, is pointed to by @var{format}.
7410 @defmac ASSEMBLER_DIALECT
7411 If your target supports multiple dialects of assembler language (such as
7412 different opcodes), define this macro as a C expression that gives the
7413 numeric index of the assembler language dialect to use, with zero as the
7416 If this macro is defined, you may use constructs of the form
7418 @samp{@{option0|option1|option2@dots{}@}}
7421 in the output templates of patterns (@pxref{Output Template}) or in the
7422 first argument of @code{asm_fprintf}. This construct outputs
7423 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7424 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7425 within these strings retain their usual meaning. If there are fewer
7426 alternatives within the braces than the value of
7427 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7429 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7430 @samp{@}} do not have any special meaning when used in templates or
7431 operands to @code{asm_fprintf}.
7433 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7434 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7435 the variations in assembler language syntax with that mechanism. Define
7436 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7437 if the syntax variant are larger and involve such things as different
7438 opcodes or operand order.
7441 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7442 A C expression to output to @var{stream} some assembler code
7443 which will push hard register number @var{regno} onto the stack.
7444 The code need not be optimal, since this macro is used only when
7448 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7449 A C expression to output to @var{stream} some assembler code
7450 which will pop hard register number @var{regno} off of the stack.
7451 The code need not be optimal, since this macro is used only when
7455 @node Dispatch Tables
7456 @subsection Output of Dispatch Tables
7458 @c prevent bad page break with this line
7459 This concerns dispatch tables.
7461 @cindex dispatch table
7462 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7463 A C statement to output to the stdio stream @var{stream} an assembler
7464 pseudo-instruction to generate a difference between two labels.
7465 @var{value} and @var{rel} are the numbers of two internal labels. The
7466 definitions of these labels are output using
7467 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7468 way here. For example,
7471 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7472 @var{value}, @var{rel})
7475 You must provide this macro on machines where the addresses in a
7476 dispatch table are relative to the table's own address. If defined, GCC
7477 will also use this macro on all machines when producing PIC@.
7478 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7479 mode and flags can be read.
7482 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7483 This macro should be provided on machines where the addresses
7484 in a dispatch table are absolute.
7486 The definition should be a C statement to output to the stdio stream
7487 @var{stream} an assembler pseudo-instruction to generate a reference to
7488 a label. @var{value} is the number of an internal label whose
7489 definition is output using @code{(*targetm.asm_out.internal_label)}.
7493 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7497 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7498 Define this if the label before a jump-table needs to be output
7499 specially. The first three arguments are the same as for
7500 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7501 jump-table which follows (a @code{jump_insn} containing an
7502 @code{addr_vec} or @code{addr_diff_vec}).
7504 This feature is used on system V to output a @code{swbeg} statement
7507 If this macro is not defined, these labels are output with
7508 @code{(*targetm.asm_out.internal_label)}.
7511 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7512 Define this if something special must be output at the end of a
7513 jump-table. The definition should be a C statement to be executed
7514 after the assembler code for the table is written. It should write
7515 the appropriate code to stdio stream @var{stream}. The argument
7516 @var{table} is the jump-table insn, and @var{num} is the label-number
7517 of the preceding label.
7519 If this macro is not defined, nothing special is output at the end of
7523 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
7524 This target hook emits a label at the beginning of each FDE. It
7525 should be defined on targets where FDEs need special labels, and it
7526 should write the appropriate label, for the FDE associated with the
7527 function declaration @var{decl}, to the stdio stream @var{stream}.
7528 The third argument, @var{for_eh}, is a boolean: true if this is for an
7529 exception table. The fourth argument, @var{empty}, is a boolean:
7530 true if this is a placeholder label for an omitted FDE.
7532 The default is that FDEs are not given nonlocal labels.
7535 @deftypefn {Taget Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn})
7536 This target hook emits and assembly directives required to unwind the
7537 given instruction. This is only used when TARGET_UNWIND_INFO is set.
7540 @node Exception Region Output
7541 @subsection Assembler Commands for Exception Regions
7543 @c prevent bad page break with this line
7545 This describes commands marking the start and the end of an exception
7548 @defmac EH_FRAME_SECTION_NAME
7549 If defined, a C string constant for the name of the section containing
7550 exception handling frame unwind information. If not defined, GCC will
7551 provide a default definition if the target supports named sections.
7552 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7554 You should define this symbol if your target supports DWARF 2 frame
7555 unwind information and the default definition does not work.
7558 @defmac EH_FRAME_IN_DATA_SECTION
7559 If defined, DWARF 2 frame unwind information will be placed in the
7560 data section even though the target supports named sections. This
7561 might be necessary, for instance, if the system linker does garbage
7562 collection and sections cannot be marked as not to be collected.
7564 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7568 @defmac MASK_RETURN_ADDR
7569 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7570 that it does not contain any extraneous set bits in it.
7573 @defmac DWARF2_UNWIND_INFO
7574 Define this macro to 0 if your target supports DWARF 2 frame unwind
7575 information, but it does not yet work with exception handling.
7576 Otherwise, if your target supports this information (if it defines
7577 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7578 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7581 If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder
7582 will be used in all cases. Defining this macro will enable the generation
7583 of DWARF 2 frame debugging information.
7585 If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1,
7586 the DWARF 2 unwinder will be the default exception handling mechanism;
7587 otherwise, @code{setjmp}/@code{longjmp} will be used by default.
7590 @defmac TARGET_UNWIND_INFO
7591 Define this macro if your target has ABI specified unwind tables. Usually
7592 these will be output by @code{TARGET_UNWIND_EMIT}.
7595 @defmac MUST_USE_SJLJ_EXCEPTIONS
7596 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
7597 runtime-variable. In that case, @file{except.h} cannot correctly
7598 determine the corresponding definition of
7599 @code{MUST_USE_SJLJ_EXCEPTIONS}, so the target must provide it directly.
7602 @defmac DWARF_CIE_DATA_ALIGNMENT
7603 This macro need only be defined if the target might save registers in the
7604 function prologue at an offset to the stack pointer that is not aligned to
7605 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7606 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7607 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7608 the target supports DWARF 2 frame unwind information.
7611 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7612 If defined, a function that switches to the section in which the main
7613 exception table is to be placed (@pxref{Sections}). The default is a
7614 function that switches to a section named @code{.gcc_except_table} on
7615 machines that support named sections via
7616 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7617 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7618 @code{readonly_data_section}.
7621 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7622 If defined, a function that switches to the section in which the DWARF 2
7623 frame unwind information to be placed (@pxref{Sections}). The default
7624 is a function that outputs a standard GAS section directive, if
7625 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7626 directive followed by a synthetic label.
7629 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7630 Contains the value true if the target should add a zero word onto the
7631 end of a Dwarf-2 frame info section when used for exception handling.
7632 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7636 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
7637 Given a register, this hook should return a parallel of registers to
7638 represent where to find the register pieces. Define this hook if the
7639 register and its mode are represented in Dwarf in non-contiguous
7640 locations, or if the register should be represented in more than one
7641 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
7642 If not defined, the default is to return @code{NULL_RTX}.
7645 @node Alignment Output
7646 @subsection Assembler Commands for Alignment
7648 @c prevent bad page break with this line
7649 This describes commands for alignment.
7651 @defmac JUMP_ALIGN (@var{label})
7652 The alignment (log base 2) to put in front of @var{label}, which is
7653 a common destination of jumps and has no fallthru incoming edge.
7655 This macro need not be defined if you don't want any special alignment
7656 to be done at such a time. Most machine descriptions do not currently
7659 Unless it's necessary to inspect the @var{label} parameter, it is better
7660 to set the variable @var{align_jumps} in the target's
7661 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7662 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7665 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
7666 The alignment (log base 2) to put in front of @var{label}, which follows
7669 This macro need not be defined if you don't want any special alignment
7670 to be done at such a time. Most machine descriptions do not currently
7674 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7675 The maximum number of bytes to skip when applying
7676 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7677 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7680 @defmac LOOP_ALIGN (@var{label})
7681 The alignment (log base 2) to put in front of @var{label}, which follows
7682 a @code{NOTE_INSN_LOOP_BEG} note.
7684 This macro need not be defined if you don't want any special alignment
7685 to be done at such a time. Most machine descriptions do not currently
7688 Unless it's necessary to inspect the @var{label} parameter, it is better
7689 to set the variable @code{align_loops} in the target's
7690 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7691 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7694 @defmac LOOP_ALIGN_MAX_SKIP
7695 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7696 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7699 @defmac LABEL_ALIGN (@var{label})
7700 The alignment (log base 2) to put in front of @var{label}.
7701 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7702 the maximum of the specified values is used.
7704 Unless it's necessary to inspect the @var{label} parameter, it is better
7705 to set the variable @code{align_labels} in the target's
7706 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7707 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7710 @defmac LABEL_ALIGN_MAX_SKIP
7711 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7712 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7715 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7716 A C statement to output to the stdio stream @var{stream} an assembler
7717 instruction to advance the location counter by @var{nbytes} bytes.
7718 Those bytes should be zero when loaded. @var{nbytes} will be a C
7719 expression of type @code{int}.
7722 @defmac ASM_NO_SKIP_IN_TEXT
7723 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7724 text section because it fails to put zeros in the bytes that are skipped.
7725 This is true on many Unix systems, where the pseudo--op to skip bytes
7726 produces no-op instructions rather than zeros when used in the text
7730 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7731 A C statement to output to the stdio stream @var{stream} an assembler
7732 command to advance the location counter to a multiple of 2 to the
7733 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7736 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7737 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7738 for padding, if necessary.
7741 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7742 A C statement to output to the stdio stream @var{stream} an assembler
7743 command to advance the location counter to a multiple of 2 to the
7744 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7745 satisfy the alignment request. @var{power} and @var{max_skip} will be
7746 a C expression of type @code{int}.
7750 @node Debugging Info
7751 @section Controlling Debugging Information Format
7753 @c prevent bad page break with this line
7754 This describes how to specify debugging information.
7757 * All Debuggers:: Macros that affect all debugging formats uniformly.
7758 * DBX Options:: Macros enabling specific options in DBX format.
7759 * DBX Hooks:: Hook macros for varying DBX format.
7760 * File Names and DBX:: Macros controlling output of file names in DBX format.
7761 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7762 * VMS Debug:: Macros for VMS debug format.
7766 @subsection Macros Affecting All Debugging Formats
7768 @c prevent bad page break with this line
7769 These macros affect all debugging formats.
7771 @defmac DBX_REGISTER_NUMBER (@var{regno})
7772 A C expression that returns the DBX register number for the compiler
7773 register number @var{regno}. In the default macro provided, the value
7774 of this expression will be @var{regno} itself. But sometimes there are
7775 some registers that the compiler knows about and DBX does not, or vice
7776 versa. In such cases, some register may need to have one number in the
7777 compiler and another for DBX@.
7779 If two registers have consecutive numbers inside GCC, and they can be
7780 used as a pair to hold a multiword value, then they @emph{must} have
7781 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7782 Otherwise, debuggers will be unable to access such a pair, because they
7783 expect register pairs to be consecutive in their own numbering scheme.
7785 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7786 does not preserve register pairs, then what you must do instead is
7787 redefine the actual register numbering scheme.
7790 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
7791 A C expression that returns the integer offset value for an automatic
7792 variable having address @var{x} (an RTL expression). The default
7793 computation assumes that @var{x} is based on the frame-pointer and
7794 gives the offset from the frame-pointer. This is required for targets
7795 that produce debugging output for DBX or COFF-style debugging output
7796 for SDB and allow the frame-pointer to be eliminated when the
7797 @option{-g} options is used.
7800 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7801 A C expression that returns the integer offset value for an argument
7802 having address @var{x} (an RTL expression). The nominal offset is
7806 @defmac PREFERRED_DEBUGGING_TYPE
7807 A C expression that returns the type of debugging output GCC should
7808 produce when the user specifies just @option{-g}. Define
7809 this if you have arranged for GCC to support more than one format of
7810 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7811 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7812 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7814 When the user specifies @option{-ggdb}, GCC normally also uses the
7815 value of this macro to select the debugging output format, but with two
7816 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
7817 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7818 defined, GCC uses @code{DBX_DEBUG}.
7820 The value of this macro only affects the default debugging output; the
7821 user can always get a specific type of output by using @option{-gstabs},
7822 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
7826 @subsection Specific Options for DBX Output
7828 @c prevent bad page break with this line
7829 These are specific options for DBX output.
7831 @defmac DBX_DEBUGGING_INFO
7832 Define this macro if GCC should produce debugging output for DBX
7833 in response to the @option{-g} option.
7836 @defmac XCOFF_DEBUGGING_INFO
7837 Define this macro if GCC should produce XCOFF format debugging output
7838 in response to the @option{-g} option. This is a variant of DBX format.
7841 @defmac DEFAULT_GDB_EXTENSIONS
7842 Define this macro to control whether GCC should by default generate
7843 GDB's extended version of DBX debugging information (assuming DBX-format
7844 debugging information is enabled at all). If you don't define the
7845 macro, the default is 1: always generate the extended information
7846 if there is any occasion to.
7849 @defmac DEBUG_SYMS_TEXT
7850 Define this macro if all @code{.stabs} commands should be output while
7851 in the text section.
7854 @defmac ASM_STABS_OP
7855 A C string constant, including spacing, naming the assembler pseudo op to
7856 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7857 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7858 applies only to DBX debugging information format.
7861 @defmac ASM_STABD_OP
7862 A C string constant, including spacing, naming the assembler pseudo op to
7863 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7864 value is the current location. If you don't define this macro,
7865 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7869 @defmac ASM_STABN_OP
7870 A C string constant, including spacing, naming the assembler pseudo op to
7871 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7872 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7873 macro applies only to DBX debugging information format.
7876 @defmac DBX_NO_XREFS
7877 Define this macro if DBX on your system does not support the construct
7878 @samp{xs@var{tagname}}. On some systems, this construct is used to
7879 describe a forward reference to a structure named @var{tagname}.
7880 On other systems, this construct is not supported at all.
7883 @defmac DBX_CONTIN_LENGTH
7884 A symbol name in DBX-format debugging information is normally
7885 continued (split into two separate @code{.stabs} directives) when it
7886 exceeds a certain length (by default, 80 characters). On some
7887 operating systems, DBX requires this splitting; on others, splitting
7888 must not be done. You can inhibit splitting by defining this macro
7889 with the value zero. You can override the default splitting-length by
7890 defining this macro as an expression for the length you desire.
7893 @defmac DBX_CONTIN_CHAR
7894 Normally continuation is indicated by adding a @samp{\} character to
7895 the end of a @code{.stabs} string when a continuation follows. To use
7896 a different character instead, define this macro as a character
7897 constant for the character you want to use. Do not define this macro
7898 if backslash is correct for your system.
7901 @defmac DBX_STATIC_STAB_DATA_SECTION
7902 Define this macro if it is necessary to go to the data section before
7903 outputting the @samp{.stabs} pseudo-op for a non-global static
7907 @defmac DBX_TYPE_DECL_STABS_CODE
7908 The value to use in the ``code'' field of the @code{.stabs} directive
7909 for a typedef. The default is @code{N_LSYM}.
7912 @defmac DBX_STATIC_CONST_VAR_CODE
7913 The value to use in the ``code'' field of the @code{.stabs} directive
7914 for a static variable located in the text section. DBX format does not
7915 provide any ``right'' way to do this. The default is @code{N_FUN}.
7918 @defmac DBX_REGPARM_STABS_CODE
7919 The value to use in the ``code'' field of the @code{.stabs} directive
7920 for a parameter passed in registers. DBX format does not provide any
7921 ``right'' way to do this. The default is @code{N_RSYM}.
7924 @defmac DBX_REGPARM_STABS_LETTER
7925 The letter to use in DBX symbol data to identify a symbol as a parameter
7926 passed in registers. DBX format does not customarily provide any way to
7927 do this. The default is @code{'P'}.
7930 @defmac DBX_MEMPARM_STABS_LETTER
7931 The letter to use in DBX symbol data to identify a symbol as a stack
7932 parameter. The default is @code{'p'}.
7935 @defmac DBX_FUNCTION_FIRST
7936 Define this macro if the DBX information for a function and its
7937 arguments should precede the assembler code for the function. Normally,
7938 in DBX format, the debugging information entirely follows the assembler
7942 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
7943 Define this macro if the value of a symbol describing the scope of a
7944 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7945 of the enclosing function. Normally, GCC uses an absolute address.
7948 @defmac DBX_USE_BINCL
7949 Define this macro if GCC should generate @code{N_BINCL} and
7950 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7951 macro also directs GCC to output a type number as a pair of a file
7952 number and a type number within the file. Normally, GCC does not
7953 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7954 number for a type number.
7958 @subsection Open-Ended Hooks for DBX Format
7960 @c prevent bad page break with this line
7961 These are hooks for DBX format.
7963 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7964 Define this macro to say how to output to @var{stream} the debugging
7965 information for the start of a scope level for variable names. The
7966 argument @var{name} is the name of an assembler symbol (for use with
7967 @code{assemble_name}) whose value is the address where the scope begins.
7970 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7971 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7974 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
7975 Define this macro if the target machine requires special handling to
7976 output an @code{N_FUN} entry for the function @var{decl}.
7979 @defmac DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7980 Define this macro if the target machine requires special output at the
7981 end of the debugging information for a function. The definition should
7982 be a C statement (sans semicolon) to output the appropriate information
7983 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7987 @defmac NO_DBX_FUNCTION_END
7988 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7989 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
7990 On those machines, define this macro to turn this feature off without
7991 disturbing the rest of the gdb extensions.
7994 @node File Names and DBX
7995 @subsection File Names in DBX Format
7997 @c prevent bad page break with this line
7998 This describes file names in DBX format.
8000 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8001 A C statement to output DBX debugging information to the stdio stream
8002 @var{stream} which indicates that file @var{name} is the main source
8003 file---the file specified as the input file for compilation.
8004 This macro is called only once, at the beginning of compilation.
8006 This macro need not be defined if the standard form of output
8007 for DBX debugging information is appropriate.
8010 @defmac DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
8011 A C statement to output DBX debugging information to the stdio stream
8012 @var{stream} which indicates that the current directory during
8013 compilation is named @var{name}.
8015 This macro need not be defined if the standard form of output
8016 for DBX debugging information is appropriate.
8019 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8020 A C statement to output DBX debugging information at the end of
8021 compilation of the main source file @var{name}.
8023 If you don't define this macro, nothing special is output at the end
8024 of compilation, which is correct for most machines.
8029 @subsection Macros for SDB and DWARF Output
8031 @c prevent bad page break with this line
8032 Here are macros for SDB and DWARF output.
8034 @defmac SDB_DEBUGGING_INFO
8035 Define this macro if GCC should produce COFF-style debugging output
8036 for SDB in response to the @option{-g} option.
8039 @defmac DWARF2_DEBUGGING_INFO
8040 Define this macro if GCC should produce dwarf version 2 format
8041 debugging output in response to the @option{-g} option.
8043 To support optional call frame debugging information, you must also
8044 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8045 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8046 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8047 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8050 @defmac DWARF2_FRAME_INFO
8051 Define this macro to a nonzero value if GCC should always output
8052 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8053 (@pxref{Exception Region Output} is nonzero, GCC will output this
8054 information not matter how you define @code{DWARF2_FRAME_INFO}.
8057 @defmac DWARF2_GENERATE_TEXT_SECTION_LABEL
8058 By default, the Dwarf 2 debugging information generator will generate a
8059 label to mark the beginning of the text section. If it is better simply
8060 to use the name of the text section itself, rather than an explicit label,
8061 to indicate the beginning of the text section, define this macro to zero.
8064 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8065 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8066 line debug info sections. This will result in much more compact line number
8067 tables, and hence is desirable if it works.
8070 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8071 A C statement to issue assembly directives that create a difference
8072 between the two given labels, using an integer of the given size.
8075 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8076 A C statement to issue assembly directives that create a
8077 section-relative reference to the given label, using an integer of the
8081 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8082 A C statement to issue assembly directives that create a self-relative
8083 reference to the given label, using an integer of the given size.
8086 @defmac PUT_SDB_@dots{}
8087 Define these macros to override the assembler syntax for the special
8088 SDB assembler directives. See @file{sdbout.c} for a list of these
8089 macros and their arguments. If the standard syntax is used, you need
8090 not define them yourself.
8094 Some assemblers do not support a semicolon as a delimiter, even between
8095 SDB assembler directives. In that case, define this macro to be the
8096 delimiter to use (usually @samp{\n}). It is not necessary to define
8097 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8101 @defmac SDB_GENERATE_FAKE
8102 Define this macro to override the usual method of constructing a dummy
8103 name for anonymous structure and union types. See @file{sdbout.c} for
8107 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8108 Define this macro to allow references to unknown structure,
8109 union, or enumeration tags to be emitted. Standard COFF does not
8110 allow handling of unknown references, MIPS ECOFF has support for
8114 @defmac SDB_ALLOW_FORWARD_REFERENCES
8115 Define this macro to allow references to structure, union, or
8116 enumeration tags that have not yet been seen to be handled. Some
8117 assemblers choke if forward tags are used, while some require it.
8122 @subsection Macros for VMS Debug Format
8124 @c prevent bad page break with this line
8125 Here are macros for VMS debug format.
8127 @defmac VMS_DEBUGGING_INFO
8128 Define this macro if GCC should produce debugging output for VMS
8129 in response to the @option{-g} option. The default behavior for VMS
8130 is to generate minimal debug info for a traceback in the absence of
8131 @option{-g} unless explicitly overridden with @option{-g0}. This
8132 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8133 @code{OVERRIDE_OPTIONS}.
8136 @node Floating Point
8137 @section Cross Compilation and Floating Point
8138 @cindex cross compilation and floating point
8139 @cindex floating point and cross compilation
8141 While all modern machines use twos-complement representation for integers,
8142 there are a variety of representations for floating point numbers. This
8143 means that in a cross-compiler the representation of floating point numbers
8144 in the compiled program may be different from that used in the machine
8145 doing the compilation.
8147 Because different representation systems may offer different amounts of
8148 range and precision, all floating point constants must be represented in
8149 the target machine's format. Therefore, the cross compiler cannot
8150 safely use the host machine's floating point arithmetic; it must emulate
8151 the target's arithmetic. To ensure consistency, GCC always uses
8152 emulation to work with floating point values, even when the host and
8153 target floating point formats are identical.
8155 The following macros are provided by @file{real.h} for the compiler to
8156 use. All parts of the compiler which generate or optimize
8157 floating-point calculations must use these macros. They may evaluate
8158 their operands more than once, so operands must not have side effects.
8160 @defmac REAL_VALUE_TYPE
8161 The C data type to be used to hold a floating point value in the target
8162 machine's format. Typically this is a @code{struct} containing an
8163 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8167 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8168 Compares for equality the two values, @var{x} and @var{y}. If the target
8169 floating point format supports negative zeroes and/or NaNs,
8170 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8171 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8174 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8175 Tests whether @var{x} is less than @var{y}.
8178 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8179 Truncates @var{x} to a signed integer, rounding toward zero.
8182 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8183 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8184 @var{x} is negative, returns zero.
8187 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8188 Converts @var{string} into a floating point number in the target machine's
8189 representation for mode @var{mode}. This routine can handle both
8190 decimal and hexadecimal floating point constants, using the syntax
8191 defined by the C language for both.
8194 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8195 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8198 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8199 Determines whether @var{x} represents infinity (positive or negative).
8202 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8203 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8206 @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})
8207 Calculates an arithmetic operation on the two floating point values
8208 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8211 The operation to be performed is specified by @var{code}. Only the
8212 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8213 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8215 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8216 target's floating point format cannot represent infinity, it will call
8217 @code{abort}. Callers should check for this situation first, using
8218 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8221 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8222 Returns the negative of the floating point value @var{x}.
8225 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8226 Returns the absolute value of @var{x}.
8229 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8230 Truncates the floating point value @var{x} to fit in @var{mode}. The
8231 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8232 appropriate bit pattern to be output asa floating constant whose
8233 precision accords with mode @var{mode}.
8236 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8237 Converts a floating point value @var{x} into a double-precision integer
8238 which is then stored into @var{low} and @var{high}. If the value is not
8239 integral, it is truncated.
8242 @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})
8243 Converts a double-precision integer found in @var{low} and @var{high},
8244 into a floating point value which is then stored into @var{x}. The
8245 value is truncated to fit in mode @var{mode}.
8248 @node Mode Switching
8249 @section Mode Switching Instructions
8250 @cindex mode switching
8251 The following macros control mode switching optimizations:
8253 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8254 Define this macro if the port needs extra instructions inserted for mode
8255 switching in an optimizing compilation.
8257 For an example, the SH4 can perform both single and double precision
8258 floating point operations, but to perform a single precision operation,
8259 the FPSCR PR bit has to be cleared, while for a double precision
8260 operation, this bit has to be set. Changing the PR bit requires a general
8261 purpose register as a scratch register, hence these FPSCR sets have to
8262 be inserted before reload, i.e.@: you can't put this into instruction emitting
8263 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
8265 You can have multiple entities that are mode-switched, and select at run time
8266 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8267 return nonzero for any @var{entity} that needs mode-switching.
8268 If you define this macro, you also have to define
8269 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8270 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8271 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
8275 @defmac NUM_MODES_FOR_MODE_SWITCHING
8276 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8277 initializer for an array of integers. Each initializer element
8278 N refers to an entity that needs mode switching, and specifies the number
8279 of different modes that might need to be set for this entity.
8280 The position of the initializer in the initializer - starting counting at
8281 zero - determines the integer that is used to refer to the mode-switched
8283 In macros that take mode arguments / yield a mode result, modes are
8284 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8285 switch is needed / supplied.
8288 @defmac MODE_NEEDED (@var{entity}, @var{insn})
8289 @var{entity} is an integer specifying a mode-switched entity. If
8290 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8291 return an integer value not larger than the corresponding element in
8292 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8293 be switched into prior to the execution of @var{insn}.
8296 @defmac MODE_AFTER (@var{mode}, @var{insn})
8297 If this macro is defined, it is evaluated for every @var{insn} during
8298 mode switching. It determines the mode that an insn results in (if
8299 different from the incoming mode).
8302 @defmac MODE_ENTRY (@var{entity})
8303 If this macro is defined, it is evaluated for every @var{entity} that needs
8304 mode switching. It should evaluate to an integer, which is a mode that
8305 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
8306 is defined then @code{MODE_EXIT} must be defined.
8309 @defmac MODE_EXIT (@var{entity})
8310 If this macro is defined, it is evaluated for every @var{entity} that needs
8311 mode switching. It should evaluate to an integer, which is a mode that
8312 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
8313 is defined then @code{MODE_ENTRY} must be defined.
8316 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8317 This macro specifies the order in which modes for @var{entity} are processed.
8318 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8319 lowest. The value of the macro should be an integer designating a mode
8320 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8321 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8322 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8325 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8326 Generate one or more insns to set @var{entity} to @var{mode}.
8327 @var{hard_reg_live} is the set of hard registers live at the point where
8328 the insn(s) are to be inserted.
8331 @node Target Attributes
8332 @section Defining target-specific uses of @code{__attribute__}
8333 @cindex target attributes
8334 @cindex machine attributes
8335 @cindex attributes, target-specific
8337 Target-specific attributes may be defined for functions, data and types.
8338 These are described using the following target hooks; they also need to
8339 be documented in @file{extend.texi}.
8341 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8342 If defined, this target hook points to an array of @samp{struct
8343 attribute_spec} (defined in @file{tree.h}) specifying the machine
8344 specific attributes for this target and some of the restrictions on the
8345 entities to which these attributes are applied and the arguments they
8349 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8350 If defined, this target hook is a function which returns zero if the attributes on
8351 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8352 and two if they are nearly compatible (which causes a warning to be
8353 generated). If this is not defined, machine-specific attributes are
8354 supposed always to be compatible.
8357 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8358 If defined, this target hook is a function which assigns default attributes to
8359 newly defined @var{type}.
8362 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8363 Define this target hook if the merging of type attributes needs special
8364 handling. If defined, the result is a list of the combined
8365 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8366 that @code{comptypes} has already been called and returned 1. This
8367 function may call @code{merge_attributes} to handle machine-independent
8371 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8372 Define this target hook if the merging of decl attributes needs special
8373 handling. If defined, the result is a list of the combined
8374 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8375 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8376 when this is needed are when one attribute overrides another, or when an
8377 attribute is nullified by a subsequent definition. This function may
8378 call @code{merge_attributes} to handle machine-independent merging.
8380 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8381 If the only target-specific handling you require is @samp{dllimport}
8382 for Microsoft Windows targets, you should define the macro
8383 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
8384 will then define a function called
8385 @code{merge_dllimport_decl_attributes} which can then be defined as
8386 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
8387 add @code{handle_dll_attribute} in the attribute table for your port
8388 to perform initial processing of the @samp{dllimport} and
8389 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
8390 @file{i386/i386.c}, for example.
8393 @defmac TARGET_DECLSPEC
8394 Define this macro to a non-zero value if you want to treat
8395 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
8396 default, this behavior is enabled only for targets that define
8397 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
8398 of @code{__declspec} is via a built-in macro, but you should not rely
8399 on this implementation detail.
8402 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8403 Define this target hook if you want to be able to add attributes to a decl
8404 when it is being created. This is normally useful for back ends which
8405 wish to implement a pragma by using the attributes which correspond to
8406 the pragma's effect. The @var{node} argument is the decl which is being
8407 created. The @var{attr_ptr} argument is a pointer to the attribute list
8408 for this decl. The list itself should not be modified, since it may be
8409 shared with other decls, but attributes may be chained on the head of
8410 the list and @code{*@var{attr_ptr}} modified to point to the new
8411 attributes, or a copy of the list may be made if further changes are
8415 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8417 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8418 into the current function, despite its having target-specific
8419 attributes, @code{false} otherwise. By default, if a function has a
8420 target specific attribute attached to it, it will not be inlined.
8423 @node MIPS Coprocessors
8424 @section Defining coprocessor specifics for MIPS targets.
8425 @cindex MIPS coprocessor-definition macros
8427 The MIPS specification allows MIPS implementations to have as many as 4
8428 coprocessors, each with as many as 32 private registers. GCC supports
8429 accessing these registers and transferring values between the registers
8430 and memory using asm-ized variables. For example:
8433 register unsigned int cp0count asm ("c0r1");
8439 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8440 names may be added as described below, or the default names may be
8441 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8443 Coprocessor registers are assumed to be epilogue-used; sets to them will
8444 be preserved even if it does not appear that the register is used again
8445 later in the function.
8447 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8448 the FPU. One accesses COP1 registers through standard mips
8449 floating-point support; they are not included in this mechanism.
8451 There is one macro used in defining the MIPS coprocessor interface which
8452 you may want to override in subtargets; it is described below.
8454 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
8455 A comma-separated list (with leading comma) of pairs describing the
8456 alternate names of coprocessor registers. The format of each entry should be
8458 @{ @var{alternatename}, @var{register_number}@}
8464 @section Parameters for Precompiled Header Validity Checking
8465 @cindex parameters, precompiled headers
8467 @deftypefn {Target Hook} void * TARGET_GET_PCH_VALIDITY (size_t * @var{sz})
8468 Define this hook if your target needs to check a different collection
8469 of flags than the default, which is every flag defined by
8470 @code{TARGET_SWITCHES} and @code{TARGET_OPTIONS}. It should return
8471 some data which will be saved in the PCH file and presented to
8472 @code{TARGET_PCH_VALID_P} later; it should set @code{SZ} to the size
8476 @deftypefn {Target Hook} const char * TARGET_PCH_VALID_P (const void * @var{data}, size_t @var{sz})
8477 Define this hook if your target needs to check a different collection of
8478 flags than the default, which is every flag defined by @code{TARGET_SWITCHES}
8479 and @code{TARGET_OPTIONS}. It is given data which came from
8480 @code{TARGET_GET_PCH_VALIDITY} (in this version of this compiler, so there
8481 is no need for extensive validity checking). It returns @code{NULL} if
8482 it is safe to load a PCH file with this data, or a suitable error message
8483 if not. The error message will be presented to the user, so it should
8488 @section C++ ABI parameters
8489 @cindex parameters, c++ abi
8491 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
8492 Define this hook to override the integer type used for guard variables.
8493 These are used to implement one-time construction of static objects. The
8494 default is long_long_integer_type_node.
8497 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
8498 This hook determines how guard variables are used. It should return
8499 @code{false} (the default) if first byte should be used. A return value of
8500 @code{true} indicates the least significant bit should be used.
8503 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
8504 This hook returns the size of the cookie to use when allocating an array
8505 whose elements have the indicated @var{type}. Assumes that it is already
8506 known that a cookie is needed. The default is
8507 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
8508 IA64/Generic C++ ABI.
8511 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
8512 This hook should return @code{true} if the element size should be stored in
8513 array cookies. The default is to return @code{false}.
8516 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
8517 If defined by a backend this hook allows the decision made to export
8518 class @var{type} to be overruled. Upon entry @var{import_export}
8519 will contain 1 if the class is going to be exported, -1 if it is going
8520 to be imported and 0 otherwise. This function should return the
8521 modified value and perform any other actions necessary to support the
8522 backend's targeted operating system.
8525 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
8526 This hook should return @code{true} if constructors and destructors return
8527 the address of the object created/destroyed. The default is to return
8531 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
8532 This hook returns true if the key method for a class (i.e., the method
8533 which, if defined in the current translation unit, causes the virtual
8534 table to be emitted) may be an inline function. Under the standard
8535 Itanium C++ ABI the key method may be an inline function so long as
8536 the function is not declared inline in the class definition. Under
8537 some variants of the ABI, an inline function can never be the key
8538 method. The default is to return @code{true}.
8541 @deftypefn {Target Hook} bool TARGET_CXX_EXPORT_CLASS_DATA (void)
8542 If this hook returns false (the default), then virtual tables and RTTI
8543 data structures will have the ELF visibility of their containing
8544 class. If this hook returns true, then these data structures will
8545 have ELF ``default'' visibility, independently of the visibility of
8546 the containing class.
8550 @section Miscellaneous Parameters
8551 @cindex parameters, miscellaneous
8553 @c prevent bad page break with this line
8554 Here are several miscellaneous parameters.
8556 @defmac PREDICATE_CODES
8557 Define this if you have defined special-purpose predicates in the file
8558 @file{@var{machine}.c}. This macro is called within an initializer of an
8559 array of structures. The first field in the structure is the name of a
8560 predicate and the second field is an array of rtl codes. For each
8561 predicate, list all rtl codes that can be in expressions matched by the
8562 predicate. The list should have a trailing comma. Here is an example
8563 of two entries in the list for a typical RISC machine:
8566 #define PREDICATE_CODES \
8567 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8568 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8571 Defining this macro does not affect the generated code (however,
8572 incorrect definitions that omit an rtl code that may be matched by the
8573 predicate can cause the compiler to malfunction). Instead, it allows
8574 the table built by @file{genrecog} to be more compact and efficient,
8575 thus speeding up the compiler. The most important predicates to include
8576 in the list specified by this macro are those used in the most insn
8579 For each predicate function named in @code{PREDICATE_CODES}, a
8580 declaration will be generated in @file{insn-codes.h}.
8582 Use of this macro is deprecated; use @code{define_predicate} instead.
8583 @xref{Defining Predicates}.
8586 @defmac SPECIAL_MODE_PREDICATES
8587 Define this if you have special predicates that know special things
8588 about modes. Genrecog will warn about certain forms of
8589 @code{match_operand} without a mode; if the operand predicate is
8590 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8593 Here is an example from the IA-32 port (@code{ext_register_operand}
8594 specially checks for @code{HImode} or @code{SImode} in preparation
8595 for a byte extraction from @code{%ah} etc.).
8598 #define SPECIAL_MODE_PREDICATES \
8599 "ext_register_operand",
8602 Use of this macro is deprecated; use @code{define_special_predicate}
8603 instead. @xref{Defining Predicates}.
8606 @defmac HAS_LONG_COND_BRANCH
8607 Define this boolean macro to indicate whether or not your architecture
8608 has conditional branches that can span all of memory. It is used in
8609 conjunction with an optimization that partitions hot and cold basic
8610 blocks into separate sections of the executable. If this macro is
8611 set to false, gcc will convert any conditional branches that attempt
8612 to cross between sections into unconditional branches or indirect jumps.
8615 @defmac HAS_LONG_UNCOND_BRANCH
8616 Define this boolean macro to indicate whether or not your architecture
8617 has unconditional branches that can span all of memory. It is used in
8618 conjunction with an optimization that partitions hot and cold basic
8619 blocks into separate sections of the executable. If this macro is
8620 set to false, gcc will convert any unconditional branches that attempt
8621 to cross between sections into indirect jumps.
8624 @defmac CASE_VECTOR_MODE
8625 An alias for a machine mode name. This is the machine mode that
8626 elements of a jump-table should have.
8629 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8630 Optional: return the preferred mode for an @code{addr_diff_vec}
8631 when the minimum and maximum offset are known. If you define this,
8632 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8633 To make this work, you also have to define @code{INSN_ALIGN} and
8634 make the alignment for @code{addr_diff_vec} explicit.
8635 The @var{body} argument is provided so that the offset_unsigned and scale
8636 flags can be updated.
8639 @defmac CASE_VECTOR_PC_RELATIVE
8640 Define this macro to be a C expression to indicate when jump-tables
8641 should contain relative addresses. You need not define this macro if
8642 jump-tables never contain relative addresses, or jump-tables should
8643 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
8647 @defmac CASE_DROPS_THROUGH
8648 Define this if control falls through a @code{case} insn when the index
8649 value is out of range. This means the specified default-label is
8650 actually ignored by the @code{case} insn proper.
8653 @defmac CASE_VALUES_THRESHOLD
8654 Define this to be the smallest number of different values for which it
8655 is best to use a jump-table instead of a tree of conditional branches.
8656 The default is four for machines with a @code{casesi} instruction and
8657 five otherwise. This is best for most machines.
8660 @defmac CASE_USE_BIT_TESTS
8661 Define this macro to be a C expression to indicate whether C switch
8662 statements may be implemented by a sequence of bit tests. This is
8663 advantageous on processors that can efficiently implement left shift
8664 of 1 by the number of bits held in a register, but inappropriate on
8665 targets that would require a loop. By default, this macro returns
8666 @code{true} if the target defines an @code{ashlsi3} pattern, and
8667 @code{false} otherwise.
8670 @defmac WORD_REGISTER_OPERATIONS
8671 Define this macro if operations between registers with integral mode
8672 smaller than a word are always performed on the entire register.
8673 Most RISC machines have this property and most CISC machines do not.
8676 @defmac LOAD_EXTEND_OP (@var{mem_mode})
8677 Define this macro to be a C expression indicating when insns that read
8678 memory in @var{mem_mode}, an integral mode narrower than a word, set the
8679 bits outside of @var{mem_mode} to be either the sign-extension or the
8680 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8681 of @var{mem_mode} for which the
8682 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8683 @code{UNKNOWN} for other modes.
8685 This macro is not called with @var{mem_mode} non-integral or with a width
8686 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8687 value in this case. Do not define this macro if it would always return
8688 @code{UNKNOWN}. On machines where this macro is defined, you will normally
8689 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8691 You may return a non-@code{UNKNOWN} value even if for some hard registers
8692 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
8693 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
8694 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
8695 integral mode larger than this but not larger than @code{word_mode}.
8697 You must return @code{UNKNOWN} if for some hard registers that allow this
8698 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
8699 @code{word_mode}, but that they can change to another integral mode that
8700 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
8703 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
8704 Define this macro if loading short immediate values into registers sign
8708 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
8709 Define this macro if the same instructions that convert a floating
8710 point number to a signed fixed point number also convert validly to an
8715 The maximum number of bytes that a single instruction can move quickly
8716 between memory and registers or between two memory locations.
8719 @defmac MAX_MOVE_MAX
8720 The maximum number of bytes that a single instruction can move quickly
8721 between memory and registers or between two memory locations. If this
8722 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8723 constant value that is the largest value that @code{MOVE_MAX} can have
8727 @defmac SHIFT_COUNT_TRUNCATED
8728 A C expression that is nonzero if on this machine the number of bits
8729 actually used for the count of a shift operation is equal to the number
8730 of bits needed to represent the size of the object being shifted. When
8731 this macro is nonzero, the compiler will assume that it is safe to omit
8732 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8733 truncates the count of a shift operation. On machines that have
8734 instructions that act on bit-fields at variable positions, which may
8735 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8736 also enables deletion of truncations of the values that serve as
8737 arguments to bit-field instructions.
8739 If both types of instructions truncate the count (for shifts) and
8740 position (for bit-field operations), or if no variable-position bit-field
8741 instructions exist, you should define this macro.
8743 However, on some machines, such as the 80386 and the 680x0, truncation
8744 only applies to shift operations and not the (real or pretended)
8745 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8746 such machines. Instead, add patterns to the @file{md} file that include
8747 the implied truncation of the shift instructions.
8749 You need not define this macro if it would always have the value of zero.
8752 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
8753 @deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
8754 This function describes how the standard shift patterns for @var{mode}
8755 deal with shifts by negative amounts or by more than the width of the mode.
8756 @xref{shift patterns}.
8758 On many machines, the shift patterns will apply a mask @var{m} to the
8759 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
8760 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
8761 this is true for mode @var{mode}, the function should return @var{m},
8762 otherwise it should return 0. A return value of 0 indicates that no
8763 particular behavior is guaranteed.
8765 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
8766 @emph{not} apply to general shift rtxes; it applies only to instructions
8767 that are generated by the named shift patterns.
8769 The default implementation of this function returns
8770 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
8771 and 0 otherwise. This definition is always safe, but if
8772 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
8773 nevertheless truncate the shift count, you may get better code
8777 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8778 A C expression which is nonzero if on this machine it is safe to
8779 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8780 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8781 operating on it as if it had only @var{outprec} bits.
8783 On many machines, this expression can be 1.
8785 @c rearranged this, removed the phrase "it is reported that". this was
8786 @c to fix an overfull hbox. --mew 10feb93
8787 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8788 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8789 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8790 such cases may improve things.
8793 @defmac STORE_FLAG_VALUE
8794 A C expression describing the value returned by a comparison operator
8795 with an integral mode and stored by a store-flag instruction
8796 (@samp{s@var{cond}}) when the condition is true. This description must
8797 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8798 comparison operators whose results have a @code{MODE_INT} mode.
8800 A value of 1 or @minus{}1 means that the instruction implementing the
8801 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8802 and 0 when the comparison is false. Otherwise, the value indicates
8803 which bits of the result are guaranteed to be 1 when the comparison is
8804 true. This value is interpreted in the mode of the comparison
8805 operation, which is given by the mode of the first operand in the
8806 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8807 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8810 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8811 generate code that depends only on the specified bits. It can also
8812 replace comparison operators with equivalent operations if they cause
8813 the required bits to be set, even if the remaining bits are undefined.
8814 For example, on a machine whose comparison operators return an
8815 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8816 @samp{0x80000000}, saying that just the sign bit is relevant, the
8820 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8827 (ashift:SI @var{x} (const_int @var{n}))
8831 where @var{n} is the appropriate shift count to move the bit being
8832 tested into the sign bit.
8834 There is no way to describe a machine that always sets the low-order bit
8835 for a true value, but does not guarantee the value of any other bits,
8836 but we do not know of any machine that has such an instruction. If you
8837 are trying to port GCC to such a machine, include an instruction to
8838 perform a logical-and of the result with 1 in the pattern for the
8839 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8841 Often, a machine will have multiple instructions that obtain a value
8842 from a comparison (or the condition codes). Here are rules to guide the
8843 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8848 Use the shortest sequence that yields a valid definition for
8849 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8850 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8851 comparison operators to do so because there may be opportunities to
8852 combine the normalization with other operations.
8855 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8856 slightly preferred on machines with expensive jumps and 1 preferred on
8860 As a second choice, choose a value of @samp{0x80000001} if instructions
8861 exist that set both the sign and low-order bits but do not define the
8865 Otherwise, use a value of @samp{0x80000000}.
8868 Many machines can produce both the value chosen for
8869 @code{STORE_FLAG_VALUE} and its negation in the same number of
8870 instructions. On those machines, you should also define a pattern for
8871 those cases, e.g., one matching
8874 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8877 Some machines can also perform @code{and} or @code{plus} operations on
8878 condition code values with less instructions than the corresponding
8879 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8880 machines, define the appropriate patterns. Use the names @code{incscc}
8881 and @code{decscc}, respectively, for the patterns which perform
8882 @code{plus} or @code{minus} operations on condition code values. See
8883 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8884 find such instruction sequences on other machines.
8886 If this macro is not defined, the default value, 1, is used. You need
8887 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8888 instructions, or if the value generated by these instructions is 1.
8891 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
8892 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8893 returned when comparison operators with floating-point results are true.
8894 Define this macro on machine that have comparison operations that return
8895 floating-point values. If there are no such operations, do not define
8899 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8900 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8901 A C expression that evaluates to true if the architecture defines a value
8902 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
8903 should be set to this value. If this macro is not defined, the value of
8904 @code{clz} or @code{ctz} is assumed to be undefined.
8906 This macro must be defined if the target's expansion for @code{ffs}
8907 relies on a particular value to get correct results. Otherwise it
8908 is not necessary, though it may be used to optimize some corner cases.
8910 Note that regardless of this macro the ``definedness'' of @code{clz}
8911 and @code{ctz} at zero do @emph{not} extend to the builtin functions
8912 visible to the user. Thus one may be free to adjust the value at will
8913 to match the target expansion of these operations without fear of
8918 An alias for the machine mode for pointers. On most machines, define
8919 this to be the integer mode corresponding to the width of a hardware
8920 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8921 On some machines you must define this to be one of the partial integer
8922 modes, such as @code{PSImode}.
8924 The width of @code{Pmode} must be at least as large as the value of
8925 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8926 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8930 @defmac FUNCTION_MODE
8931 An alias for the machine mode used for memory references to functions
8932 being called, in @code{call} RTL expressions. On most machines this
8933 should be @code{QImode}.
8936 @defmac STDC_0_IN_SYSTEM_HEADERS
8937 In normal operation, the preprocessor expands @code{__STDC__} to the
8938 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8939 hosts, like Solaris, the system compiler uses a different convention,
8940 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8941 strict conformance to the C Standard.
8943 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8944 convention when processing system header files, but when processing user
8945 files @code{__STDC__} will always expand to 1.
8948 @defmac NO_IMPLICIT_EXTERN_C
8949 Define this macro if the system header files support C++ as well as C@.
8950 This macro inhibits the usual method of using system header files in
8951 C++, which is to pretend that the file's contents are enclosed in
8952 @samp{extern "C" @{@dots{}@}}.
8957 @defmac REGISTER_TARGET_PRAGMAS ()
8958 Define this macro if you want to implement any target-specific pragmas.
8959 If defined, it is a C expression which makes a series of calls to
8960 @code{c_register_pragma} for each pragma. The macro may also do any
8961 setup required for the pragmas.
8963 The primary reason to define this macro is to provide compatibility with
8964 other compilers for the same target. In general, we discourage
8965 definition of target-specific pragmas for GCC@.
8967 If the pragma can be implemented by attributes then you should consider
8968 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8970 Preprocessor macros that appear on pragma lines are not expanded. All
8971 @samp{#pragma} directives that do not match any registered pragma are
8972 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8975 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
8977 Each call to @code{c_register_pragma} establishes one pragma. The
8978 @var{callback} routine will be called when the preprocessor encounters a
8982 #pragma [@var{space}] @var{name} @dots{}
8985 @var{space} is the case-sensitive namespace of the pragma, or
8986 @code{NULL} to put the pragma in the global namespace. The callback
8987 routine receives @var{pfile} as its first argument, which can be passed
8988 on to cpplib's functions if necessary. You can lex tokens after the
8989 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8990 callback will be silently ignored. The end of the line is indicated by
8991 a token of type @code{CPP_EOF}
8993 For an example use of this routine, see @file{c4x.h} and the callback
8994 routines defined in @file{c4x-c.c}.
8996 Note that the use of @code{c_lex} is specific to the C and C++
8997 compilers. It will not work in the Java or Fortran compilers, or any
8998 other language compilers for that matter. Thus if @code{c_lex} is going
8999 to be called from target-specific code, it must only be done so when
9000 building the C and C++ compilers. This can be done by defining the
9001 variables @code{c_target_objs} and @code{cxx_target_objs} in the
9002 target entry in the @file{config.gcc} file. These variables should name
9003 the target-specific, language-specific object file which contains the
9004 code that uses @code{c_lex}. Note it will also be necessary to add a
9005 rule to the makefile fragment pointed to by @code{tmake_file} that shows
9006 how to build this object file.
9011 @defmac HANDLE_SYSV_PRAGMA
9012 Define this macro (to a value of 1) if you want the System V style
9013 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
9014 [=<value>]} to be supported by gcc.
9016 The pack pragma specifies the maximum alignment (in bytes) of fields
9017 within a structure, in much the same way as the @samp{__aligned__} and
9018 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
9019 the behavior to the default.
9021 A subtlety for Microsoft Visual C/C++ style bit-field packing
9022 (e.g. -mms-bitfields) for targets that support it:
9023 When a bit-field is inserted into a packed record, the whole size
9024 of the underlying type is used by one or more same-size adjacent
9025 bit-fields (that is, if its long:3, 32 bits is used in the record,
9026 and any additional adjacent long bit-fields are packed into the same
9027 chunk of 32 bits. However, if the size changes, a new field of that
9030 If both MS bit-fields and @samp{__attribute__((packed))} are used,
9031 the latter will take precedence. If @samp{__attribute__((packed))} is
9032 used on a single field when MS bit-fields are in use, it will take
9033 precedence for that field, but the alignment of the rest of the structure
9034 may affect its placement.
9036 The weak pragma only works if @code{SUPPORTS_WEAK} and
9037 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
9038 of specifically named weak labels, optionally with a value.
9043 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
9044 Define this macro (to a value of 1) if you want to support the Win32
9045 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
9046 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
9047 alignment (in bytes) of fields within a structure, in much the same way as
9048 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
9049 pack value of zero resets the behavior to the default. Successive
9050 invocations of this pragma cause the previous values to be stacked, so
9051 that invocations of @samp{#pragma pack(pop)} will return to the previous
9055 @defmac TARGET_DEFAULT_PACK_STRUCT
9056 If your target requires a structure packing default other than 0 (meaning
9057 the machine default), define this macro the the necessary value (in bytes).
9058 This must be a value that would also valid to be used with
9059 @samp{#pragma pack()} (that is, a small power of two).
9062 @defmac DOLLARS_IN_IDENTIFIERS
9063 Define this macro to control use of the character @samp{$} in
9064 identifier names for the C family of languages. 0 means @samp{$} is
9065 not allowed by default; 1 means it is allowed. 1 is the default;
9066 there is no need to define this macro in that case.
9069 @defmac NO_DOLLAR_IN_LABEL
9070 Define this macro if the assembler does not accept the character
9071 @samp{$} in label names. By default constructors and destructors in
9072 G++ have @samp{$} in the identifiers. If this macro is defined,
9073 @samp{.} is used instead.
9076 @defmac NO_DOT_IN_LABEL
9077 Define this macro if the assembler does not accept the character
9078 @samp{.} in label names. By default constructors and destructors in G++
9079 have names that use @samp{.}. If this macro is defined, these names
9080 are rewritten to avoid @samp{.}.
9083 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
9084 Define this macro as a C expression that is nonzero if it is safe for the
9085 delay slot scheduler to place instructions in the delay slot of @var{insn},
9086 even if they appear to use a resource set or clobbered in @var{insn}.
9087 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
9088 every @code{call_insn} has this behavior. On machines where some @code{insn}
9089 or @code{jump_insn} is really a function call and hence has this behavior,
9090 you should define this macro.
9092 You need not define this macro if it would always return zero.
9095 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
9096 Define this macro as a C expression that is nonzero if it is safe for the
9097 delay slot scheduler to place instructions in the delay slot of @var{insn},
9098 even if they appear to set or clobber a resource referenced in @var{insn}.
9099 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
9100 some @code{insn} or @code{jump_insn} is really a function call and its operands
9101 are registers whose use is actually in the subroutine it calls, you should
9102 define this macro. Doing so allows the delay slot scheduler to move
9103 instructions which copy arguments into the argument registers into the delay
9106 You need not define this macro if it would always return zero.
9109 @defmac MULTIPLE_SYMBOL_SPACES
9110 Define this macro if in some cases global symbols from one translation
9111 unit may not be bound to undefined symbols in another translation unit
9112 without user intervention. For instance, under Microsoft Windows
9113 symbols must be explicitly imported from shared libraries (DLLs).
9116 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{clobbers})
9117 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
9118 any hard regs the port wishes to automatically clobber for all asms.
9119 It should return the result of the last @code{tree_cons} used to add a
9123 @defmac MATH_LIBRARY
9124 Define this macro as a C string constant for the linker argument to link
9125 in the system math library, or @samp{""} if the target does not have a
9126 separate math library.
9128 You need only define this macro if the default of @samp{"-lm"} is wrong.
9131 @defmac LIBRARY_PATH_ENV
9132 Define this macro as a C string constant for the environment variable that
9133 specifies where the linker should look for libraries.
9135 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9139 @defmac TARGET_HAS_F_SETLKW
9140 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
9141 Note that this functionality is part of POSIX@.
9142 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
9143 to use file locking when exiting a program, which avoids race conditions
9144 if the program has forked.
9147 @defmac MAX_CONDITIONAL_EXECUTE
9149 A C expression for the maximum number of instructions to execute via
9150 conditional execution instructions instead of a branch. A value of
9151 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9152 1 if it does use cc0.
9155 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
9156 Used if the target needs to perform machine-dependent modifications on the
9157 conditionals used for turning basic blocks into conditionally executed code.
9158 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9159 contains information about the currently processed blocks. @var{true_expr}
9160 and @var{false_expr} are the tests that are used for converting the
9161 then-block and the else-block, respectively. Set either @var{true_expr} or
9162 @var{false_expr} to a null pointer if the tests cannot be converted.
9165 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9166 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9167 if-statements into conditions combined by @code{and} and @code{or} operations.
9168 @var{bb} contains the basic block that contains the test that is currently
9169 being processed and about to be turned into a condition.
9172 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
9173 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9174 be converted to conditional execution format. @var{ce_info} points to
9175 a data structure, @code{struct ce_if_block}, which contains information
9176 about the currently processed blocks.
9179 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
9180 A C expression to perform any final machine dependent modifications in
9181 converting code to conditional execution. The involved basic blocks
9182 can be found in the @code{struct ce_if_block} structure that is pointed
9183 to by @var{ce_info}.
9186 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
9187 A C expression to cancel any machine dependent modifications in
9188 converting code to conditional execution. The involved basic blocks
9189 can be found in the @code{struct ce_if_block} structure that is pointed
9190 to by @var{ce_info}.
9193 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
9194 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9195 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9198 @defmac IFCVT_EXTRA_FIELDS
9199 If defined, it should expand to a set of field declarations that will be
9200 added to the @code{struct ce_if_block} structure. These should be initialized
9201 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9204 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
9205 If non-null, this hook performs a target-specific pass over the
9206 instruction stream. The compiler will run it at all optimization levels,
9207 just before the point at which it normally does delayed-branch scheduling.
9209 The exact purpose of the hook varies from target to target. Some use
9210 it to do transformations that are necessary for correctness, such as
9211 laying out in-function constant pools or avoiding hardware hazards.
9212 Others use it as an opportunity to do some machine-dependent optimizations.
9214 You need not implement the hook if it has nothing to do. The default
9218 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9219 Define this hook if you have any machine-specific built-in functions
9220 that need to be defined. It should be a function that performs the
9223 Machine specific built-in functions can be useful to expand special machine
9224 instructions that would otherwise not normally be generated because
9225 they have no equivalent in the source language (for example, SIMD vector
9226 instructions or prefetch instructions).
9228 To create a built-in function, call the function
9229 @code{lang_hooks.builtin_function}
9230 which is defined by the language front end. You can use any type nodes set
9231 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9232 only language front ends that use those two functions will call
9233 @samp{TARGET_INIT_BUILTINS}.
9236 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9238 Expand a call to a machine specific built-in function that was set up by
9239 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9240 function call; the result should go to @var{target} if that is
9241 convenient, and have mode @var{mode} if that is convenient.
9242 @var{subtarget} may be used as the target for computing one of
9243 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9244 ignored. This function should return the result of the call to the
9248 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
9250 Take a branch insn in @var{branch1} and another in @var{branch2}.
9251 Return true if redirecting @var{branch1} to the destination of
9252 @var{branch2} is possible.
9254 On some targets, branches may have a limited range. Optimizing the
9255 filling of delay slots can result in branches being redirected, and this
9256 may in turn cause a branch offset to overflow.
9259 @defmac ALLOCATE_INITIAL_VALUE (@var{hard_reg})
9261 When the initial value of a hard register has been copied in a pseudo
9262 register, it is often not necessary to actually allocate another register
9263 to this pseudo register, because the original hard register or a stack slot
9264 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9265 defined, is called at the start of register allocation once for each
9266 hard register that had its initial value copied by using
9267 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9268 Possible values are @code{NULL_RTX}, if you don't want
9269 to do any special allocation, a @code{REG} rtx---that would typically be
9270 the hard register itself, if it is known not to be clobbered---or a
9272 If you are returning a @code{MEM}, this is only a hint for the allocator;
9273 it might decide to use another register anyways.
9274 You may use @code{current_function_leaf_function} in the definition of the
9275 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9276 register in question will not be clobbered.
9279 @defmac TARGET_OBJECT_SUFFIX
9280 Define this macro to be a C string representing the suffix for object
9281 files on your target machine. If you do not define this macro, GCC will
9282 use @samp{.o} as the suffix for object files.
9285 @defmac TARGET_EXECUTABLE_SUFFIX
9286 Define this macro to be a C string representing the suffix to be
9287 automatically added to executable files on your target machine. If you
9288 do not define this macro, GCC will use the null string as the suffix for
9292 @defmac COLLECT_EXPORT_LIST
9293 If defined, @code{collect2} will scan the individual object files
9294 specified on its command line and create an export list for the linker.
9295 Define this macro for systems like AIX, where the linker discards
9296 object files that are not referenced from @code{main} and uses export
9300 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
9301 Define this macro to a C expression representing a variant of the
9302 method call @var{mdecl}, if Java Native Interface (JNI) methods
9303 must be invoked differently from other methods on your target.
9304 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
9305 the @code{stdcall} calling convention and this macro is then
9306 defined as this expression:
9309 build_type_attribute_variant (@var{mdecl},
9311 (get_identifier ("stdcall"),
9316 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9317 This target hook returns @code{true} past the point in which new jump
9318 instructions could be created. On machines that require a register for
9319 every jump such as the SHmedia ISA of SH5, this point would typically be
9320 reload, so this target hook should be defined to a function such as:
9324 cannot_modify_jumps_past_reload_p ()
9326 return (reload_completed || reload_in_progress);
9331 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
9332 This target hook returns a register class for which branch target register
9333 optimizations should be applied. All registers in this class should be
9334 usable interchangeably. After reload, registers in this class will be
9335 re-allocated and loads will be hoisted out of loops and be subjected
9336 to inter-block scheduling.
9339 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
9340 Branch target register optimization will by default exclude callee-saved
9342 that are not already live during the current function; if this target hook
9343 returns true, they will be included. The target code must than make sure
9344 that all target registers in the class returned by
9345 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
9346 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
9347 epilogues have already been generated. Note, even if you only return
9348 true when @var{after_prologue_epilogue_gen} is false, you still are likely
9349 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
9350 to reserve space for caller-saved target registers.
9353 @defmac POWI_MAX_MULTS
9354 If defined, this macro is interpreted as a signed integer C expression
9355 that specifies the maximum number of floating point multiplications
9356 that should be emitted when expanding exponentiation by an integer
9357 constant inline. When this value is defined, exponentiation requiring
9358 more than this number of multiplications is implemented by calling the
9359 system library's @code{pow}, @code{powf} or @code{powl} routines.
9360 The default value places no upper bound on the multiplication count.
9363 @deftypefn Macro void TARGET_EXTRA_INCLUDES (int @var{stdinc})
9364 This target hook should register any extra include files for the
9365 target. The parameter @var{stdinc} indicates if normal include files
9369 @deftypefn Macro void TARGET_OPTF (char *@var{path})
9370 This target hook should register special include paths for the target.
9371 The parameter @var{path} is the include to register. On Darwin
9372 systems, this is used for Framework includes, which have semantics
9373 that are different from @option{-I}.
9376 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
9377 This target hook returns @code{true} if it is safe to use a local alias
9378 for a virtual function @var{fndecl} when constructing thunks,
9379 @code{false} otherwise. By default, the hook returns @code{true} for all
9380 functions, if a target supports aliases (ie. defines
9381 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
9384 @defmac TARGET_FORMAT_TYPES
9385 If defined, this macro is the name of a global variable containing
9386 target-specific format checking information for the @option{-Wformat}
9387 option. The default is to have no target-specific format checks.
9390 @defmac TARGET_N_FORMAT_TYPES
9391 If defined, this macro is the number of entries in
9392 @code{TARGET_FORMAT_TYPES}.