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 MD_STARTFILE_PREFIX
473 If defined, this macro supplies an additional prefix to try after the
474 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
475 @option{-b} option is used, or when the compiler is built as a cross
479 @defmac MD_STARTFILE_PREFIX_1
480 If defined, this macro supplies yet another prefix to try after the
481 standard prefixes. It is not searched when the @option{-b} option is
482 used, or when the compiler is built as a cross compiler.
485 @defmac INIT_ENVIRONMENT
486 Define this macro as a C string constant if you wish to set environment
487 variables for programs called by the driver, such as the assembler and
488 loader. The driver passes the value of this macro to @code{putenv} to
489 initialize the necessary environment variables.
492 @defmac LOCAL_INCLUDE_DIR
493 Define this macro as a C string constant if you wish to override the
494 standard choice of @file{/usr/local/include} as the default prefix to
495 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
496 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
498 Cross compilers do not search either @file{/usr/local/include} or its
502 @defmac MODIFY_TARGET_NAME
503 Define this macro if you wish to define command-line switches that
504 modify the default target name.
506 For each switch, you can include a string to be appended to the first
507 part of the configuration name or a string to be deleted from the
508 configuration name, if present. The definition should be an initializer
509 for an array of structures. Each array element should have three
510 elements: the switch name (a string constant, including the initial
511 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
512 indicate whether the string should be inserted or deleted, and the string
513 to be inserted or deleted (a string constant).
515 For example, on a machine where @samp{64} at the end of the
516 configuration name denotes a 64-bit target and you want the @option{-32}
517 and @option{-64} switches to select between 32- and 64-bit targets, you would
521 #define MODIFY_TARGET_NAME \
522 @{ @{ "-32", DELETE, "64"@}, \
523 @{"-64", ADD, "64"@}@}
527 @defmac SYSTEM_INCLUDE_DIR
528 Define this macro as a C string constant if you wish to specify a
529 system-specific directory to search for header files before the standard
530 directory. @code{SYSTEM_INCLUDE_DIR} comes before
531 @code{STANDARD_INCLUDE_DIR} in the search order.
533 Cross compilers do not use this macro and do not search the directory
537 @defmac STANDARD_INCLUDE_DIR
538 Define this macro as a C string constant if you wish to override the
539 standard choice of @file{/usr/include} as the default prefix to
540 try when searching for header files.
542 Cross compilers ignore this macro and do not search either
543 @file{/usr/include} or its replacement.
546 @defmac STANDARD_INCLUDE_COMPONENT
547 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
548 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
549 If you do not define this macro, no component is used.
552 @defmac INCLUDE_DEFAULTS
553 Define this macro if you wish to override the entire default search path
554 for include files. For a native compiler, the default search path
555 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
556 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
557 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
558 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
559 and specify private search areas for GCC@. The directory
560 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
562 The definition should be an initializer for an array of structures.
563 Each array element should have four elements: the directory name (a
564 string constant), the component name (also a string constant), a flag
565 for C++-only directories,
566 and a flag showing that the includes in the directory don't need to be
567 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
568 the array with a null element.
570 The component name denotes what GNU package the include file is part of,
571 if any, in all uppercase letters. For example, it might be @samp{GCC}
572 or @samp{BINUTILS}. If the package is part of a vendor-supplied
573 operating system, code the component name as @samp{0}.
575 For example, here is the definition used for VAX/VMS:
578 #define INCLUDE_DEFAULTS \
580 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
581 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
582 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
589 Here is the order of prefixes tried for exec files:
593 Any prefixes specified by the user with @option{-B}.
596 The environment variable @code{GCC_EXEC_PREFIX}, if any.
599 The directories specified by the environment variable @code{COMPILER_PATH}.
602 The macro @code{STANDARD_EXEC_PREFIX}.
605 @file{/usr/lib/gcc/}.
608 The macro @code{MD_EXEC_PREFIX}, if any.
611 Here is the order of prefixes tried for startfiles:
615 Any prefixes specified by the user with @option{-B}.
618 The environment variable @code{GCC_EXEC_PREFIX}, if any.
621 The directories specified by the environment variable @code{LIBRARY_PATH}
622 (or port-specific name; native only, cross compilers do not use this).
625 The macro @code{STANDARD_EXEC_PREFIX}.
628 @file{/usr/lib/gcc/}.
631 The macro @code{MD_EXEC_PREFIX}, if any.
634 The macro @code{MD_STARTFILE_PREFIX}, if any.
637 The macro @code{STANDARD_STARTFILE_PREFIX}.
646 @node Run-time Target
647 @section Run-time Target Specification
648 @cindex run-time target specification
649 @cindex predefined macros
650 @cindex target specifications
652 @c prevent bad page break with this line
653 Here are run-time target specifications.
655 @defmac TARGET_CPU_CPP_BUILTINS ()
656 This function-like macro expands to a block of code that defines
657 built-in preprocessor macros and assertions for the target cpu, using
658 the functions @code{builtin_define}, @code{builtin_define_std} and
659 @code{builtin_assert}. When the front end
660 calls this macro it provides a trailing semicolon, and since it has
661 finished command line option processing your code can use those
664 @code{builtin_assert} takes a string in the form you pass to the
665 command-line option @option{-A}, such as @code{cpu=mips}, and creates
666 the assertion. @code{builtin_define} takes a string in the form
667 accepted by option @option{-D} and unconditionally defines the macro.
669 @code{builtin_define_std} takes a string representing the name of an
670 object-like macro. If it doesn't lie in the user's namespace,
671 @code{builtin_define_std} defines it unconditionally. Otherwise, it
672 defines a version with two leading underscores, and another version
673 with two leading and trailing underscores, and defines the original
674 only if an ISO standard was not requested on the command line. For
675 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
676 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
677 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
678 defines only @code{_ABI64}.
680 You can also test for the C dialect being compiled. The variable
681 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
682 or @code{clk_objective_c}. Note that if we are preprocessing
683 assembler, this variable will be @code{clk_c} but the function-like
684 macro @code{preprocessing_asm_p()} will return true, so you might want
685 to check for that first. If you need to check for strict ANSI, the
686 variable @code{flag_iso} can be used. The function-like macro
687 @code{preprocessing_trad_p()} can be used to check for traditional
691 @defmac TARGET_OS_CPP_BUILTINS ()
692 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
693 and is used for the target operating system instead.
696 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
697 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
698 and is used for the target object format. @file{elfos.h} uses this
699 macro to define @code{__ELF__}, so you probably do not need to define
703 @deftypevar {extern int} target_flags
704 This declaration should be present.
707 @cindex optional hardware or system features
708 @cindex features, optional, in system conventions
710 @defmac TARGET_@var{featurename}
711 This series of macros is to allow compiler command arguments to
712 enable or disable the use of optional features of the target machine.
713 For example, one machine description serves both the 68000 and
714 the 68020; a command argument tells the compiler whether it should
715 use 68020-only instructions or not. This command argument works
716 by means of a macro @code{TARGET_68020} that tests a bit in
719 Define a macro @code{TARGET_@var{featurename}} for each such option.
720 Its definition should test a bit in @code{target_flags}. It is
721 recommended that a helper macro @code{MASK_@var{featurename}}
722 is defined for each bit-value to test, and used in
723 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
727 #define TARGET_MASK_68020 1
728 #define TARGET_68020 (target_flags & MASK_68020)
731 One place where these macros are used is in the condition-expressions
732 of instruction patterns. Note how @code{TARGET_68020} appears
733 frequently in the 68000 machine description file, @file{m68k.md}.
734 Another place they are used is in the definitions of the other
735 macros in the @file{@var{machine}.h} file.
738 @defmac TARGET_SWITCHES
739 This macro defines names of command options to set and clear
740 bits in @code{target_flags}. Its definition is an initializer
741 with a subgrouping for each command option.
743 Each subgrouping contains a string constant, that defines the option
744 name, a number, which contains the bits to set in
745 @code{target_flags}, and a second string which is the description
746 displayed by @option{--help}. If the number is negative then the bits specified
747 by the number are cleared instead of being set. If the description
748 string is present but empty, then no help information will be displayed
749 for that option, but it will not count as an undocumented option. The
750 actual option name is made by appending @samp{-m} to the specified name.
751 Non-empty description strings should be marked with @code{N_(@dots{})} for
752 @command{xgettext}. Please do not mark empty strings because the empty
753 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
754 of the message catalog with meta information, not the empty string.
756 In addition to the description for @option{--help},
757 more detailed documentation for each option should be added to
760 One of the subgroupings should have a null string. The number in
761 this grouping is the default value for @code{target_flags}. Any
762 target options act starting with that value.
764 Here is an example which defines @option{-m68000} and @option{-m68020}
765 with opposite meanings, and picks the latter as the default:
768 #define TARGET_SWITCHES \
769 @{ @{ "68020", MASK_68020, "" @}, \
770 @{ "68000", -MASK_68020, \
771 N_("Compile for the 68000") @}, \
772 @{ "", MASK_68020, "" @}, \
777 @defmac TARGET_OPTIONS
778 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
779 options that have values. Its definition is an initializer with a
780 subgrouping for each command option.
782 Each subgrouping contains a string constant, that defines the option
783 name, the address of a variable, a description string, and a value.
784 Non-empty description strings should be marked with @code{N_(@dots{})}
785 for @command{xgettext}. Please do not mark empty strings because the
786 empty string is reserved by GNU gettext. @code{gettext("")} returns the
787 header entry of the message catalog with meta information, not the empty
790 If the value listed in the table is @code{NULL}, then the variable, type
791 @code{char *}, is set to the variable part of the given option if the
792 fixed part matches. In other words, if the first part of the option
793 matches what's in the table, the variable will be set to point to the
794 rest of the option. This allows the user to specify a value for that
795 option. The actual option name is made by appending @samp{-m} to the
796 specified name. Again, each option should also be documented in
799 If the value listed in the table is non-@code{NULL}, then the option
800 must match the option in the table exactly (with @samp{-m}), and the
801 variable is set to point to the value listed in the table.
803 Here is an example which defines @option{-mshort-data-@var{number}}. If the
804 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
805 will be set to the string @code{"512"}.
808 extern char *m88k_short_data;
809 #define TARGET_OPTIONS \
810 @{ @{ "short-data-", &m88k_short_data, \
811 N_("Specify the size of the short data section"), 0 @} @}
814 Here is a variant of the above that allows the user to also specify
815 just @option{-mshort-data} where a default of @code{"64"} is used.
818 extern char *m88k_short_data;
819 #define TARGET_OPTIONS \
820 @{ @{ "short-data-", &m88k_short_data, \
821 N_("Specify the size of the short data section"), 0 @} \
822 @{ "short-data", &m88k_short_data, "", "64" @},
826 Here is an example which defines @option{-mno-alu}, @option{-malu1}, and
827 @option{-malu2} as a three-state switch, along with suitable macros for
828 checking the state of the option (documentation is elided for brevity).
832 char *chip_alu = ""; /* Specify default here. */
835 extern char *chip_alu;
836 #define TARGET_OPTIONS \
837 @{ @{ "no-alu", &chip_alu, "", "" @}, \
838 @{ "alu1", &chip_alu, "", "1" @}, \
839 @{ "alu2", &chip_alu, "", "2" @}, @}
840 #define TARGET_ALU (chip_alu[0] != '\0')
841 #define TARGET_ALU1 (chip_alu[0] == '1')
842 #define TARGET_ALU2 (chip_alu[0] == '2')
846 @defmac TARGET_VERSION
847 This macro is a C statement to print on @code{stderr} a string
848 describing the particular machine description choice. Every machine
849 description should define @code{TARGET_VERSION}. For example:
853 #define TARGET_VERSION \
854 fprintf (stderr, " (68k, Motorola syntax)");
856 #define TARGET_VERSION \
857 fprintf (stderr, " (68k, MIT syntax)");
862 @defmac OVERRIDE_OPTIONS
863 Sometimes certain combinations of command options do not make sense on
864 a particular target machine. You can define a macro
865 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
866 defined, is executed once just after all the command options have been
869 Don't use this macro to turn on various extra optimizations for
870 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
873 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
874 Some machines may desire to change what optimizations are performed for
875 various optimization levels. This macro, if defined, is executed once
876 just after the optimization level is determined and before the remainder
877 of the command options have been parsed. Values set in this macro are
878 used as the default values for the other command line options.
880 @var{level} is the optimization level specified; 2 if @option{-O2} is
881 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
883 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
885 You should not use this macro to change options that are not
886 machine-specific. These should uniformly selected by the same
887 optimization level on all supported machines. Use this macro to enable
888 machine-specific optimizations.
890 @strong{Do not examine @code{write_symbols} in
891 this macro!} The debugging options are not supposed to alter the
895 @defmac CAN_DEBUG_WITHOUT_FP
896 Define this macro if debugging can be performed even without a frame
897 pointer. If this macro is defined, GCC will turn on the
898 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
901 @node Per-Function Data
902 @section Defining data structures for per-function information.
903 @cindex per-function data
904 @cindex data structures
906 If the target needs to store information on a per-function basis, GCC
907 provides a macro and a couple of variables to allow this. Note, just
908 using statics to store the information is a bad idea, since GCC supports
909 nested functions, so you can be halfway through encoding one function
910 when another one comes along.
912 GCC defines a data structure called @code{struct function} which
913 contains all of the data specific to an individual function. This
914 structure contains a field called @code{machine} whose type is
915 @code{struct machine_function *}, which can be used by targets to point
916 to their own specific data.
918 If a target needs per-function specific data it should define the type
919 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
920 This macro should be used to initialize the function pointer
921 @code{init_machine_status}. This pointer is explained below.
923 One typical use of per-function, target specific data is to create an
924 RTX to hold the register containing the function's return address. This
925 RTX can then be used to implement the @code{__builtin_return_address}
926 function, for level 0.
928 Note---earlier implementations of GCC used a single data area to hold
929 all of the per-function information. Thus when processing of a nested
930 function began the old per-function data had to be pushed onto a
931 stack, and when the processing was finished, it had to be popped off the
932 stack. GCC used to provide function pointers called
933 @code{save_machine_status} and @code{restore_machine_status} to handle
934 the saving and restoring of the target specific information. Since the
935 single data area approach is no longer used, these pointers are no
938 @defmac INIT_EXPANDERS
939 Macro called to initialize any target specific information. This macro
940 is called once per function, before generation of any RTL has begun.
941 The intention of this macro is to allow the initialization of the
942 function pointer @code{init_machine_status}.
945 @deftypevar {void (*)(struct function *)} init_machine_status
946 If this function pointer is non-@code{NULL} it will be called once per
947 function, before function compilation starts, in order to allow the
948 target to perform any target specific initialization of the
949 @code{struct function} structure. It is intended that this would be
950 used to initialize the @code{machine} of that structure.
952 @code{struct machine_function} structures are expected to be freed by GC.
953 Generally, any memory that they reference must be allocated by using
954 @code{ggc_alloc}, including the structure itself.
958 @section Storage Layout
959 @cindex storage layout
961 Note that the definitions of the macros in this table which are sizes or
962 alignments measured in bits do not need to be constant. They can be C
963 expressions that refer to static variables, such as the @code{target_flags}.
964 @xref{Run-time Target}.
966 @defmac BITS_BIG_ENDIAN
967 Define this macro to have the value 1 if the most significant bit in a
968 byte has the lowest number; otherwise define it to have the value zero.
969 This means that bit-field instructions count from the most significant
970 bit. If the machine has no bit-field instructions, then this must still
971 be defined, but it doesn't matter which value it is defined to. This
972 macro need not be a constant.
974 This macro does not affect the way structure fields are packed into
975 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
978 @defmac BYTES_BIG_ENDIAN
979 Define this macro to have the value 1 if the most significant byte in a
980 word has the lowest number. This macro need not be a constant.
983 @defmac WORDS_BIG_ENDIAN
984 Define this macro to have the value 1 if, in a multiword object, the
985 most significant word has the lowest number. This applies to both
986 memory locations and registers; GCC fundamentally assumes that the
987 order of words in memory is the same as the order in registers. This
988 macro need not be a constant.
991 @defmac LIBGCC2_WORDS_BIG_ENDIAN
992 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
993 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
994 used only when compiling @file{libgcc2.c}. Typically the value will be set
995 based on preprocessor defines.
998 @defmac FLOAT_WORDS_BIG_ENDIAN
999 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
1000 @code{TFmode} floating point numbers are stored in memory with the word
1001 containing the sign bit at the lowest address; otherwise define it to
1002 have the value 0. This macro need not be a constant.
1004 You need not define this macro if the ordering is the same as for
1005 multi-word integers.
1008 @defmac BITS_PER_UNIT
1009 Define this macro to be the number of bits in an addressable storage
1010 unit (byte). If you do not define this macro the default is 8.
1013 @defmac BITS_PER_WORD
1014 Number of bits in a word. If you do not define this macro, the default
1015 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
1018 @defmac MAX_BITS_PER_WORD
1019 Maximum number of bits in a word. If this is undefined, the default is
1020 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
1021 largest value that @code{BITS_PER_WORD} can have at run-time.
1024 @defmac UNITS_PER_WORD
1025 Number of storage units in a word; normally 4.
1028 @defmac MIN_UNITS_PER_WORD
1029 Minimum number of units in a word. If this is undefined, the default is
1030 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
1031 smallest value that @code{UNITS_PER_WORD} can have at run-time.
1034 @defmac POINTER_SIZE
1035 Width of a pointer, in bits. You must specify a value no wider than the
1036 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1037 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1038 a value the default is @code{BITS_PER_WORD}.
1041 @defmac POINTERS_EXTEND_UNSIGNED
1042 A C expression whose value is greater than zero if pointers that need to be
1043 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
1044 be zero-extended and zero if they are to be sign-extended. If the value
1045 is less then zero then there must be an "ptr_extend" instruction that
1046 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
1048 You need not define this macro if the @code{POINTER_SIZE} is equal
1049 to the width of @code{Pmode}.
1052 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1053 A macro to update @var{m} and @var{unsignedp} when an object whose type
1054 is @var{type} and which has the specified mode and signedness is to be
1055 stored in a register. This macro is only called when @var{type} is a
1058 On most RISC machines, which only have operations that operate on a full
1059 register, define this macro to set @var{m} to @code{word_mode} if
1060 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1061 cases, only integer modes should be widened because wider-precision
1062 floating-point operations are usually more expensive than their narrower
1065 For most machines, the macro definition does not change @var{unsignedp}.
1066 However, some machines, have instructions that preferentially handle
1067 either signed or unsigned quantities of certain modes. For example, on
1068 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1069 sign-extend the result to 64 bits. On such machines, set
1070 @var{unsignedp} according to which kind of extension is more efficient.
1072 Do not define this macro if it would never modify @var{m}.
1075 @defmac PROMOTE_FUNCTION_MODE
1076 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
1077 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1078 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1080 The default is @code{PROMOTE_MODE}.
1083 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1084 This target hook should return @code{true} if the promotion described by
1085 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1089 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1090 This target hook should return @code{true} if the promotion described by
1091 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1094 If this target hook returns @code{true}, @code{FUNCTION_VALUE} must
1095 perform the same promotions done by @code{PROMOTE_FUNCTON_MODE}.
1098 @defmac PARM_BOUNDARY
1099 Normal alignment required for function parameters on the stack, in
1100 bits. All stack parameters receive at least this much alignment
1101 regardless of data type. On most machines, this is the same as the
1105 @defmac STACK_BOUNDARY
1106 Define this macro to the minimum alignment enforced by hardware for the
1107 stack pointer on this machine. The definition is a C expression for the
1108 desired alignment (measured in bits). This value is used as a default
1109 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1110 this should be the same as @code{PARM_BOUNDARY}.
1113 @defmac PREFERRED_STACK_BOUNDARY
1114 Define this macro if you wish to preserve a certain alignment for the
1115 stack pointer, greater than what the hardware enforces. The definition
1116 is a C expression for the desired alignment (measured in bits). This
1117 macro must evaluate to a value equal to or larger than
1118 @code{STACK_BOUNDARY}.
1121 @defmac FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1122 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1123 not guaranteed by the runtime and we should emit code to align the stack
1124 at the beginning of @code{main}.
1126 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1127 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1128 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1129 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1130 be momentarily unaligned while pushing arguments.
1133 @defmac FUNCTION_BOUNDARY
1134 Alignment required for a function entry point, in bits.
1137 @defmac BIGGEST_ALIGNMENT
1138 Biggest alignment that any data type can require on this machine, in bits.
1141 @defmac MINIMUM_ATOMIC_ALIGNMENT
1142 If defined, the smallest alignment, in bits, that can be given to an
1143 object that can be referenced in one operation, without disturbing any
1144 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1145 on machines that don't have byte or half-word store operations.
1148 @defmac BIGGEST_FIELD_ALIGNMENT
1149 Biggest alignment that any structure or union field can require on this
1150 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1151 structure and union fields only, unless the field alignment has been set
1152 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1155 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1156 An expression for the alignment of a structure field @var{field} if the
1157 alignment computed in the usual way (including applying of
1158 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1159 alignment) is @var{computed}. It overrides alignment only if the
1160 field alignment has not been set by the
1161 @code{__attribute__ ((aligned (@var{n})))} construct.
1164 @defmac MAX_OFILE_ALIGNMENT
1165 Biggest alignment supported by the object file format of this machine.
1166 Use this macro to limit the alignment which can be specified using the
1167 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1168 the default value is @code{BIGGEST_ALIGNMENT}.
1171 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1172 If defined, a C expression to compute the alignment for a variable in
1173 the static store. @var{type} is the data type, and @var{basic-align} is
1174 the alignment that the object would ordinarily have. The value of this
1175 macro is used instead of that alignment to align the object.
1177 If this macro is not defined, then @var{basic-align} is used.
1180 One use of this macro is to increase alignment of medium-size data to
1181 make it all fit in fewer cache lines. Another is to cause character
1182 arrays to be word-aligned so that @code{strcpy} calls that copy
1183 constants to character arrays can be done inline.
1186 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1187 If defined, a C expression to compute the alignment given to a constant
1188 that is being placed in memory. @var{constant} is the constant and
1189 @var{basic-align} is the alignment that the object would ordinarily
1190 have. The value of this macro is used instead of that alignment to
1193 If this macro is not defined, then @var{basic-align} is used.
1195 The typical use of this macro is to increase alignment for string
1196 constants to be word aligned so that @code{strcpy} calls that copy
1197 constants can be done inline.
1200 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1201 If defined, a C expression to compute the alignment for a variable in
1202 the local store. @var{type} is the data type, and @var{basic-align} is
1203 the alignment that the object would ordinarily have. The value of this
1204 macro is used instead of that alignment to align the object.
1206 If this macro is not defined, then @var{basic-align} is used.
1208 One use of this macro is to increase alignment of medium-size data to
1209 make it all fit in fewer cache lines.
1212 @defmac EMPTY_FIELD_BOUNDARY
1213 Alignment in bits to be given to a structure bit-field that follows an
1214 empty field such as @code{int : 0;}.
1216 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1219 @defmac STRUCTURE_SIZE_BOUNDARY
1220 Number of bits which any structure or union's size must be a multiple of.
1221 Each structure or union's size is rounded up to a multiple of this.
1223 If you do not define this macro, the default is the same as
1224 @code{BITS_PER_UNIT}.
1227 @defmac STRICT_ALIGNMENT
1228 Define this macro to be the value 1 if instructions will fail to work
1229 if given data not on the nominal alignment. If instructions will merely
1230 go slower in that case, define this macro as 0.
1233 @defmac PCC_BITFIELD_TYPE_MATTERS
1234 Define this if you wish to imitate the way many other C compilers handle
1235 alignment of bit-fields and the structures that contain them.
1237 The behavior is that the type written for a named bit-field (@code{int},
1238 @code{short}, or other integer type) imposes an alignment for the entire
1239 structure, as if the structure really did contain an ordinary field of
1240 that type. In addition, the bit-field is placed within the structure so
1241 that it would fit within such a field, not crossing a boundary for it.
1243 Thus, on most machines, a named bit-field whose type is written as
1244 @code{int} would not cross a four-byte boundary, and would force
1245 four-byte alignment for the whole structure. (The alignment used may
1246 not be four bytes; it is controlled by the other alignment parameters.)
1248 An unnamed bit-field will not affect the alignment of the containing
1251 If the macro is defined, its definition should be a C expression;
1252 a nonzero value for the expression enables this behavior.
1254 Note that if this macro is not defined, or its value is zero, some
1255 bit-fields may cross more than one alignment boundary. The compiler can
1256 support such references if there are @samp{insv}, @samp{extv}, and
1257 @samp{extzv} insns that can directly reference memory.
1259 The other known way of making bit-fields work is to define
1260 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1261 Then every structure can be accessed with fullwords.
1263 Unless the machine has bit-field instructions or you define
1264 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1265 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1267 If your aim is to make GCC use the same conventions for laying out
1268 bit-fields as are used by another compiler, here is how to investigate
1269 what the other compiler does. Compile and run this program:
1288 printf ("Size of foo1 is %d\n",
1289 sizeof (struct foo1));
1290 printf ("Size of foo2 is %d\n",
1291 sizeof (struct foo2));
1296 If this prints 2 and 5, then the compiler's behavior is what you would
1297 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1300 @defmac BITFIELD_NBYTES_LIMITED
1301 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1302 to aligning a bit-field within the structure.
1305 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1306 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1307 whether unnamed bitfields affect the alignment of the containing
1308 structure. The hook should return true if the structure should inherit
1309 the alignment requirements of an unnamed bitfield's type.
1312 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1313 Return 1 if a structure or array containing @var{field} should be accessed using
1316 If @var{field} is the only field in the structure, @var{mode} is its
1317 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1318 case where structures of one field would require the structure's mode to
1319 retain the field's mode.
1321 Normally, this is not needed. See the file @file{c4x.h} for an example
1322 of how to use this macro to prevent a structure having a floating point
1323 field from being accessed in an integer mode.
1326 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1327 Define this macro as an expression for the alignment of a type (given
1328 by @var{type} as a tree node) if the alignment computed in the usual
1329 way is @var{computed} and the alignment explicitly specified was
1332 The default is to use @var{specified} if it is larger; otherwise, use
1333 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1336 @defmac MAX_FIXED_MODE_SIZE
1337 An integer expression for the size in bits of the largest integer
1338 machine mode that should actually be used. All integer machine modes of
1339 this size or smaller can be used for structures and unions with the
1340 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1341 (DImode)} is assumed.
1344 @defmac VECTOR_MODE_SUPPORTED_P (@var{mode})
1345 Define this macro to be nonzero if the port is prepared to handle insns
1346 involving vector mode @var{mode}. At the very least, it must have move
1347 patterns for this mode.
1350 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1351 If defined, an expression of type @code{enum machine_mode} that
1352 specifies the mode of the save area operand of a
1353 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1354 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1355 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1356 having its mode specified.
1358 You need not define this macro if it always returns @code{Pmode}. You
1359 would most commonly define this macro if the
1360 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1364 @defmac STACK_SIZE_MODE
1365 If defined, an expression of type @code{enum machine_mode} that
1366 specifies the mode of the size increment operand of an
1367 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1369 You need not define this macro if it always returns @code{word_mode}.
1370 You would most commonly define this macro if the @code{allocate_stack}
1371 pattern needs to support both a 32- and a 64-bit mode.
1374 @defmac TARGET_FLOAT_FORMAT
1375 A code distinguishing the floating point format of the target machine.
1376 There are four defined values:
1379 @item IEEE_FLOAT_FORMAT
1380 This code indicates IEEE floating point. It is the default; there is no
1381 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1383 @item VAX_FLOAT_FORMAT
1384 This code indicates the ``F float'' (for @code{float}) and ``D float''
1385 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1387 @item IBM_FLOAT_FORMAT
1388 This code indicates the format used on the IBM System/370.
1390 @item C4X_FLOAT_FORMAT
1391 This code indicates the format used on the TMS320C3x/C4x.
1394 If your target uses a floating point format other than these, you must
1395 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1396 it to @file{real.c}.
1398 The ordering of the component words of floating point values stored in
1399 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1402 @defmac MODE_HAS_NANS (@var{mode})
1403 When defined, this macro should be true if @var{mode} has a NaN
1404 representation. The compiler assumes that NaNs are not equal to
1405 anything (including themselves) and that addition, subtraction,
1406 multiplication and division all return NaNs when one operand is
1409 By default, this macro is true if @var{mode} is a floating-point
1410 mode and the target floating-point format is IEEE@.
1413 @defmac MODE_HAS_INFINITIES (@var{mode})
1414 This macro should be true if @var{mode} can represent infinity. At
1415 present, the compiler uses this macro to decide whether @samp{x - x}
1416 is always defined. By default, the macro is true when @var{mode}
1417 is a floating-point mode and the target format is IEEE@.
1420 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1421 True if @var{mode} distinguishes between positive and negative zero.
1422 The rules are expected to follow the IEEE standard:
1426 @samp{x + x} has the same sign as @samp{x}.
1429 If the sum of two values with opposite sign is zero, the result is
1430 positive for all rounding modes expect towards @minus{}infinity, for
1431 which it is negative.
1434 The sign of a product or quotient is negative when exactly one
1435 of the operands is negative.
1438 The default definition is true if @var{mode} is a floating-point
1439 mode and the target format is IEEE@.
1442 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1443 If defined, this macro should be true for @var{mode} if it has at
1444 least one rounding mode in which @samp{x} and @samp{-x} can be
1445 rounded to numbers of different magnitude. Two such modes are
1446 towards @minus{}infinity and towards +infinity.
1448 The default definition of this macro is true if @var{mode} is
1449 a floating-point mode and the target format is IEEE@.
1452 @defmac ROUND_TOWARDS_ZERO
1453 If defined, this macro should be true if the prevailing rounding
1454 mode is towards zero. A true value has the following effects:
1458 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1461 @file{libgcc.a}'s floating-point emulator will round towards zero
1462 rather than towards nearest.
1465 The compiler's floating-point emulator will round towards zero after
1466 doing arithmetic, and when converting from the internal float format to
1470 The macro does not affect the parsing of string literals. When the
1471 primary rounding mode is towards zero, library functions like
1472 @code{strtod} might still round towards nearest, and the compiler's
1473 parser should behave like the target's @code{strtod} where possible.
1475 Not defining this macro is equivalent to returning zero.
1478 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1479 This macro should return true if floats with @var{size}
1480 bits do not have a NaN or infinity representation, but use the largest
1481 exponent for normal numbers instead.
1483 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1484 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1485 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1486 floating-point arithmetic.
1488 The default definition of this macro returns false for all sizes.
1491 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1492 This target hook should return @code{true} a vector is opaque. That
1493 is, if no cast is needed when copying a vector value of type
1494 @var{type} into another vector lvalue of the same size. Vector opaque
1495 types cannot be initialized. The default is that there are no such
1499 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1500 This target hook returns @code{true} if bit-fields in the given
1501 @var{record_type} are to be laid out following the rules of Microsoft
1502 Visual C/C++, namely: (i) a bit-field won't share the same storage
1503 unit with the previous bit-field if their underlying types have
1504 different sizes, and the bit-field will be aligned to the highest
1505 alignment of the underlying types of itself and of the previous
1506 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1507 the whole enclosing structure, even if it is unnamed; except that
1508 (iii) a zero-sized bit-field will be disregarded unless it follows
1509 another bit-field of nonzero size. If this hook returns @code{true},
1510 other macros that control bit-field layout are ignored.
1512 When a bit-field is inserted into a packed record, the whole size
1513 of the underlying type is used by one or more same-size adjacent
1514 bit-fields (that is, if its long:3, 32 bits is used in the record,
1515 and any additional adjacent long bit-fields are packed into the same
1516 chunk of 32 bits. However, if the size changes, a new field of that
1517 size is allocated). In an unpacked record, this is the same as using
1518 alignment, but not equivalent when packing.
1520 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1521 the latter will take precedence. If @samp{__attribute__((packed))} is
1522 used on a single field when MS bit-fields are in use, it will take
1523 precedence for that field, but the alignment of the rest of the structure
1524 may affect its placement.
1527 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_FUNDAMENTAL_TYPE (tree @var{type})
1528 If your target defines any fundamental types, define this hook to
1529 return the appropriate encoding for these types as part of a C++
1530 mangled name. The @var{type} argument is the tree structure
1531 representing the type to be mangled. The hook may be applied to trees
1532 which are not target-specific fundamental types; it should return
1533 @code{NULL} for all such types, as well as arguments it does not
1534 recognize. If the return value is not @code{NULL}, it must point to
1535 a statically-allocated string constant.
1537 Target-specific fundamental types might be new fundamental types or
1538 qualified versions of ordinary fundamental types. Encode new
1539 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1540 is the name used for the type in source code, and @var{n} is the
1541 length of @var{name} in decimal. Encode qualified versions of
1542 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1543 @var{name} is the name used for the type qualifier in source code,
1544 @var{n} is the length of @var{name} as above, and @var{code} is the
1545 code used to represent the unqualified version of this type. (See
1546 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1547 codes.) In both cases the spaces are for clarity; do not include any
1548 spaces in your string.
1550 The default version of this hook always returns @code{NULL}, which is
1551 appropriate for a target that does not define any new fundamental
1556 @section Layout of Source Language Data Types
1558 These macros define the sizes and other characteristics of the standard
1559 basic data types used in programs being compiled. Unlike the macros in
1560 the previous section, these apply to specific features of C and related
1561 languages, rather than to fundamental aspects of storage layout.
1563 @defmac INT_TYPE_SIZE
1564 A C expression for the size in bits of the type @code{int} on the
1565 target machine. If you don't define this, the default is one word.
1568 @defmac SHORT_TYPE_SIZE
1569 A C expression for the size in bits of the type @code{short} on the
1570 target machine. If you don't define this, the default is half a word.
1571 (If this would be less than one storage unit, it is rounded up to one
1575 @defmac LONG_TYPE_SIZE
1576 A C expression for the size in bits of the type @code{long} on the
1577 target machine. If you don't define this, the default is one word.
1580 @defmac ADA_LONG_TYPE_SIZE
1581 On some machines, the size used for the Ada equivalent of the type
1582 @code{long} by a native Ada compiler differs from that used by C. In
1583 that situation, define this macro to be a C expression to be used for
1584 the size of that type. If you don't define this, the default is the
1585 value of @code{LONG_TYPE_SIZE}.
1588 @defmac LONG_LONG_TYPE_SIZE
1589 A C expression for the size in bits of the type @code{long long} on the
1590 target machine. If you don't define this, the default is two
1591 words. If you want to support GNU Ada on your machine, the value of this
1592 macro must be at least 64.
1595 @defmac CHAR_TYPE_SIZE
1596 A C expression for the size in bits of the type @code{char} on the
1597 target machine. If you don't define this, the default is
1598 @code{BITS_PER_UNIT}.
1601 @defmac BOOL_TYPE_SIZE
1602 A C expression for the size in bits of the C++ type @code{bool} and
1603 C99 type @code{_Bool} on the target machine. If you don't define
1604 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1607 @defmac FLOAT_TYPE_SIZE
1608 A C expression for the size in bits of the type @code{float} on the
1609 target machine. If you don't define this, the default is one word.
1612 @defmac DOUBLE_TYPE_SIZE
1613 A C expression for the size in bits of the type @code{double} on the
1614 target machine. If you don't define this, the default is two
1618 @defmac LONG_DOUBLE_TYPE_SIZE
1619 A C expression for the size in bits of the type @code{long double} on
1620 the target machine. If you don't define this, the default is two
1624 @defmac TARGET_FLT_EVAL_METHOD
1625 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1626 assuming, if applicable, that the floating-point control word is in its
1627 default state. If you do not define this macro the value of
1628 @code{FLT_EVAL_METHOD} will be zero.
1631 @defmac WIDEST_HARDWARE_FP_SIZE
1632 A C expression for the size in bits of the widest floating-point format
1633 supported by the hardware. If you define this macro, you must specify a
1634 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1635 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1639 @defmac DEFAULT_SIGNED_CHAR
1640 An expression whose value is 1 or 0, according to whether the type
1641 @code{char} should be signed or unsigned by default. The user can
1642 always override this default with the options @option{-fsigned-char}
1643 and @option{-funsigned-char}.
1646 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1647 This target hook should return true if the compiler should give an
1648 @code{enum} type only as many bytes as it takes to represent the range
1649 of possible values of that type. It should return false if all
1650 @code{enum} types should be allocated like @code{int}.
1652 The default is to return false.
1656 A C expression for a string describing the name of the data type to use
1657 for size values. The typedef name @code{size_t} is defined using the
1658 contents of the string.
1660 The string can contain more than one keyword. If so, separate them with
1661 spaces, and write first any length keyword, then @code{unsigned} if
1662 appropriate, and finally @code{int}. The string must exactly match one
1663 of the data type names defined in the function
1664 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1665 omit @code{int} or change the order---that would cause the compiler to
1668 If you don't define this macro, the default is @code{"long unsigned
1672 @defmac PTRDIFF_TYPE
1673 A C expression for a string describing the name of the data type to use
1674 for the result of subtracting two pointers. The typedef name
1675 @code{ptrdiff_t} is defined using the contents of the string. See
1676 @code{SIZE_TYPE} above for more information.
1678 If you don't define this macro, the default is @code{"long int"}.
1682 A C expression for a string describing the name of the data type to use
1683 for wide characters. The typedef name @code{wchar_t} is defined using
1684 the contents of the string. See @code{SIZE_TYPE} above for more
1687 If you don't define this macro, the default is @code{"int"}.
1690 @defmac WCHAR_TYPE_SIZE
1691 A C expression for the size in bits of the data type for wide
1692 characters. This is used in @code{cpp}, which cannot make use of
1697 A C expression for a string describing the name of the data type to
1698 use for wide characters passed to @code{printf} and returned from
1699 @code{getwc}. The typedef name @code{wint_t} is defined using the
1700 contents of the string. See @code{SIZE_TYPE} above for more
1703 If you don't define this macro, the default is @code{"unsigned int"}.
1707 A C expression for a string describing the name of the data type that
1708 can represent any value of any standard or extended signed integer type.
1709 The typedef name @code{intmax_t} is defined using the contents of the
1710 string. See @code{SIZE_TYPE} above for more information.
1712 If you don't define this macro, the default is the first of
1713 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1714 much precision as @code{long long int}.
1717 @defmac UINTMAX_TYPE
1718 A C expression for a string describing the name of the data type that
1719 can represent any value of any standard or extended unsigned integer
1720 type. The typedef name @code{uintmax_t} is defined using the contents
1721 of the string. See @code{SIZE_TYPE} above for more information.
1723 If you don't define this macro, the default is the first of
1724 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1725 unsigned int"} that has as much precision as @code{long long unsigned
1729 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1730 The C++ compiler represents a pointer-to-member-function with a struct
1737 ptrdiff_t vtable_index;
1744 The C++ compiler must use one bit to indicate whether the function that
1745 will be called through a pointer-to-member-function is virtual.
1746 Normally, we assume that the low-order bit of a function pointer must
1747 always be zero. Then, by ensuring that the vtable_index is odd, we can
1748 distinguish which variant of the union is in use. But, on some
1749 platforms function pointers can be odd, and so this doesn't work. In
1750 that case, we use the low-order bit of the @code{delta} field, and shift
1751 the remainder of the @code{delta} field to the left.
1753 GCC will automatically make the right selection about where to store
1754 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1755 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1756 set such that functions always start at even addresses, but the lowest
1757 bit of pointers to functions indicate whether the function at that
1758 address is in ARM or Thumb mode. If this is the case of your
1759 architecture, you should define this macro to
1760 @code{ptrmemfunc_vbit_in_delta}.
1762 In general, you should not have to define this macro. On architectures
1763 in which function addresses are always even, according to
1764 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1765 @code{ptrmemfunc_vbit_in_pfn}.
1768 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1769 Normally, the C++ compiler uses function pointers in vtables. This
1770 macro allows the target to change to use ``function descriptors''
1771 instead. Function descriptors are found on targets for whom a
1772 function pointer is actually a small data structure. Normally the
1773 data structure consists of the actual code address plus a data
1774 pointer to which the function's data is relative.
1776 If vtables are used, the value of this macro should be the number
1777 of words that the function descriptor occupies.
1780 @defmac TARGET_VTABLE_ENTRY_ALIGN
1781 By default, the vtable entries are void pointers, the so the alignment
1782 is the same as pointer alignment. The value of this macro specifies
1783 the alignment of the vtable entry in bits. It should be defined only
1784 when special alignment is necessary. */
1787 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1788 There are a few non-descriptor entries in the vtable at offsets below
1789 zero. If these entries must be padded (say, to preserve the alignment
1790 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1791 of words in each data entry.
1794 @node Escape Sequences
1795 @section Target Character Escape Sequences
1796 @cindex escape sequences
1798 By default, GCC assumes that the C character escape sequences and other
1799 characters take on their ASCII values for the target. If this is not
1800 correct, you must explicitly define all of the macros below. All of
1801 them must evaluate to constants; they are used in @code{case}
1807 @findex TARGET_DIGIT0
1810 @findex TARGET_NEWLINE
1813 @multitable {@code{TARGET_NEWLINE}} {Escape} {ASCII character}
1814 @item Macro @tab Escape @tab ASCII character
1815 @item @code{TARGET_BELL} @tab @kbd{\a} @tab @code{07}, @code{BEL}
1816 @item @code{TARGET_BS} @tab @kbd{\b} @tab @code{08}, @code{BS}
1817 @item @code{TARGET_CR} @tab @kbd{\r} @tab @code{0D}, @code{CR}
1818 @item @code{TARGET_DIGIT0} @tab @kbd{0} @tab @code{30}, @code{ZERO}
1819 @item @code{TARGET_ESC} @tab @kbd{\e}, @kbd{\E} @tab @code{1B}, @code{ESC}
1820 @item @code{TARGET_FF} @tab @kbd{\f} @tab @code{0C}, @code{FF}
1821 @item @code{TARGET_NEWLINE} @tab @kbd{\n} @tab @code{0A}, @code{LF}
1822 @item @code{TARGET_TAB} @tab @kbd{\t} @tab @code{09}, @code{HT}
1823 @item @code{TARGET_VT} @tab @kbd{\v} @tab @code{0B}, @code{VT}
1827 Note that the @kbd{\e} and @kbd{\E} escapes are GNU extensions, not
1828 part of the C standard.
1831 @section Register Usage
1832 @cindex register usage
1834 This section explains how to describe what registers the target machine
1835 has, and how (in general) they can be used.
1837 The description of which registers a specific instruction can use is
1838 done with register classes; see @ref{Register Classes}. For information
1839 on using registers to access a stack frame, see @ref{Frame Registers}.
1840 For passing values in registers, see @ref{Register Arguments}.
1841 For returning values in registers, see @ref{Scalar Return}.
1844 * Register Basics:: Number and kinds of registers.
1845 * Allocation Order:: Order in which registers are allocated.
1846 * Values in Registers:: What kinds of values each reg can hold.
1847 * Leaf Functions:: Renumbering registers for leaf functions.
1848 * Stack Registers:: Handling a register stack such as 80387.
1851 @node Register Basics
1852 @subsection Basic Characteristics of Registers
1854 @c prevent bad page break with this line
1855 Registers have various characteristics.
1857 @defmac FIRST_PSEUDO_REGISTER
1858 Number of hardware registers known to the compiler. They receive
1859 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1860 pseudo register's number really is assigned the number
1861 @code{FIRST_PSEUDO_REGISTER}.
1864 @defmac FIXED_REGISTERS
1865 @cindex fixed register
1866 An initializer that says which registers are used for fixed purposes
1867 all throughout the compiled code and are therefore not available for
1868 general allocation. These would include the stack pointer, the frame
1869 pointer (except on machines where that can be used as a general
1870 register when no frame pointer is needed), the program counter on
1871 machines where that is considered one of the addressable registers,
1872 and any other numbered register with a standard use.
1874 This information is expressed as a sequence of numbers, separated by
1875 commas and surrounded by braces. The @var{n}th number is 1 if
1876 register @var{n} is fixed, 0 otherwise.
1878 The table initialized from this macro, and the table initialized by
1879 the following one, may be overridden at run time either automatically,
1880 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1881 the user with the command options @option{-ffixed-@var{reg}},
1882 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1885 @defmac CALL_USED_REGISTERS
1886 @cindex call-used register
1887 @cindex call-clobbered register
1888 @cindex call-saved register
1889 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1890 clobbered (in general) by function calls as well as for fixed
1891 registers. This macro therefore identifies the registers that are not
1892 available for general allocation of values that must live across
1895 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1896 automatically saves it on function entry and restores it on function
1897 exit, if the register is used within the function.
1900 @defmac CALL_REALLY_USED_REGISTERS
1901 @cindex call-used register
1902 @cindex call-clobbered register
1903 @cindex call-saved register
1904 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1905 that the entire set of @code{FIXED_REGISTERS} be included.
1906 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1907 This macro is optional. If not specified, it defaults to the value
1908 of @code{CALL_USED_REGISTERS}.
1911 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1912 @cindex call-used register
1913 @cindex call-clobbered register
1914 @cindex call-saved register
1915 A C expression that is nonzero if it is not permissible to store a
1916 value of mode @var{mode} in hard register number @var{regno} across a
1917 call without some part of it being clobbered. For most machines this
1918 macro need not be defined. It is only required for machines that do not
1919 preserve the entire contents of a register across a call.
1923 @findex call_used_regs
1926 @findex reg_class_contents
1927 @defmac CONDITIONAL_REGISTER_USAGE
1928 Zero or more C statements that may conditionally modify five variables
1929 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1930 @code{reg_names}, and @code{reg_class_contents}, to take into account
1931 any dependence of these register sets on target flags. The first three
1932 of these are of type @code{char []} (interpreted as Boolean vectors).
1933 @code{global_regs} is a @code{const char *[]}, and
1934 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1935 called, @code{fixed_regs}, @code{call_used_regs},
1936 @code{reg_class_contents}, and @code{reg_names} have been initialized
1937 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1938 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1939 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1940 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1941 command options have been applied.
1943 You need not define this macro if it has no work to do.
1945 @cindex disabling certain registers
1946 @cindex controlling register usage
1947 If the usage of an entire class of registers depends on the target
1948 flags, you may indicate this to GCC by using this macro to modify
1949 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1950 registers in the classes which should not be used by GCC@. Also define
1951 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1952 to return @code{NO_REGS} if it
1953 is called with a letter for a class that shouldn't be used.
1955 (However, if this class is not included in @code{GENERAL_REGS} and all
1956 of the insn patterns whose constraints permit this class are
1957 controlled by target switches, then GCC will automatically avoid using
1958 these registers when the target switches are opposed to them.)
1961 @defmac NON_SAVING_SETJMP
1962 If this macro is defined and has a nonzero value, it means that
1963 @code{setjmp} and related functions fail to save the registers, or that
1964 @code{longjmp} fails to restore them. To compensate, the compiler
1965 avoids putting variables in registers in functions that use
1969 @defmac INCOMING_REGNO (@var{out})
1970 Define this macro if the target machine has register windows. This C
1971 expression returns the register number as seen by the called function
1972 corresponding to the register number @var{out} as seen by the calling
1973 function. Return @var{out} if register number @var{out} is not an
1977 @defmac OUTGOING_REGNO (@var{in})
1978 Define this macro if the target machine has register windows. This C
1979 expression returns the register number as seen by the calling function
1980 corresponding to the register number @var{in} as seen by the called
1981 function. Return @var{in} if register number @var{in} is not an inbound
1985 @defmac LOCAL_REGNO (@var{regno})
1986 Define this macro if the target machine has register windows. This C
1987 expression returns true if the register is call-saved but is in the
1988 register window. Unlike most call-saved registers, such registers
1989 need not be explicitly restored on function exit or during non-local
1994 If the program counter has a register number, define this as that
1995 register number. Otherwise, do not define it.
1998 @node Allocation Order
1999 @subsection Order of Allocation of Registers
2000 @cindex order of register allocation
2001 @cindex register allocation order
2003 @c prevent bad page break with this line
2004 Registers are allocated in order.
2006 @defmac REG_ALLOC_ORDER
2007 If defined, an initializer for a vector of integers, containing the
2008 numbers of hard registers in the order in which GCC should prefer
2009 to use them (from most preferred to least).
2011 If this macro is not defined, registers are used lowest numbered first
2012 (all else being equal).
2014 One use of this macro is on machines where the highest numbered
2015 registers must always be saved and the save-multiple-registers
2016 instruction supports only sequences of consecutive registers. On such
2017 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2018 the highest numbered allocable register first.
2021 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
2022 A C statement (sans semicolon) to choose the order in which to allocate
2023 hard registers for pseudo-registers local to a basic block.
2025 Store the desired register order in the array @code{reg_alloc_order}.
2026 Element 0 should be the register to allocate first; element 1, the next
2027 register; and so on.
2029 The macro body should not assume anything about the contents of
2030 @code{reg_alloc_order} before execution of the macro.
2032 On most machines, it is not necessary to define this macro.
2035 @node Values in Registers
2036 @subsection How Values Fit in Registers
2038 This section discusses the macros that describe which kinds of values
2039 (specifically, which machine modes) each register can hold, and how many
2040 consecutive registers are needed for a given mode.
2042 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2043 A C expression for the number of consecutive hard registers, starting
2044 at register number @var{regno}, required to hold a value of mode
2047 On a machine where all registers are exactly one word, a suitable
2048 definition of this macro is
2051 #define HARD_REGNO_NREGS(REGNO, MODE) \
2052 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2057 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2058 Define this macro if the natural size of registers that hold values
2059 of mode @var{mode} is not the word size. It is a C expression that
2060 should give the natural size in bytes for the specified mode. It is
2061 used by the register allocator to try to optimize its results. This
2062 happens for example on SPARC 64-bit where the natural size of
2063 floating-point registers is still 32-bit.
2066 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2067 A C expression that is nonzero if it is permissible to store a value
2068 of mode @var{mode} in hard register number @var{regno} (or in several
2069 registers starting with that one). For a machine where all registers
2070 are equivalent, a suitable definition is
2073 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2076 You need not include code to check for the numbers of fixed registers,
2077 because the allocation mechanism considers them to be always occupied.
2079 @cindex register pairs
2080 On some machines, double-precision values must be kept in even/odd
2081 register pairs. You can implement that by defining this macro to reject
2082 odd register numbers for such modes.
2084 The minimum requirement for a mode to be OK in a register is that the
2085 @samp{mov@var{mode}} instruction pattern support moves between the
2086 register and other hard register in the same class and that moving a
2087 value into the register and back out not alter it.
2089 Since the same instruction used to move @code{word_mode} will work for
2090 all narrower integer modes, it is not necessary on any machine for
2091 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2092 you define patterns @samp{movhi}, etc., to take advantage of this. This
2093 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2094 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2097 Many machines have special registers for floating point arithmetic.
2098 Often people assume that floating point machine modes are allowed only
2099 in floating point registers. This is not true. Any registers that
2100 can hold integers can safely @emph{hold} a floating point machine
2101 mode, whether or not floating arithmetic can be done on it in those
2102 registers. Integer move instructions can be used to move the values.
2104 On some machines, though, the converse is true: fixed-point machine
2105 modes may not go in floating registers. This is true if the floating
2106 registers normalize any value stored in them, because storing a
2107 non-floating value there would garble it. In this case,
2108 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2109 floating registers. But if the floating registers do not automatically
2110 normalize, if you can store any bit pattern in one and retrieve it
2111 unchanged without a trap, then any machine mode may go in a floating
2112 register, so you can define this macro to say so.
2114 The primary significance of special floating registers is rather that
2115 they are the registers acceptable in floating point arithmetic
2116 instructions. However, this is of no concern to
2117 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2118 constraints for those instructions.
2120 On some machines, the floating registers are especially slow to access,
2121 so that it is better to store a value in a stack frame than in such a
2122 register if floating point arithmetic is not being done. As long as the
2123 floating registers are not in class @code{GENERAL_REGS}, they will not
2124 be used unless some pattern's constraint asks for one.
2127 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2128 A C expression that is nonzero if it is OK to rename a hard register
2129 @var{from} to another hard register @var{to}.
2131 One common use of this macro is to prevent renaming of a register to
2132 another register that is not saved by a prologue in an interrupt
2135 The default is always nonzero.
2138 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2139 A C expression that is nonzero if a value of mode
2140 @var{mode1} is accessible in mode @var{mode2} without copying.
2142 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2143 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2144 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2145 should be nonzero. If they differ for any @var{r}, you should define
2146 this macro to return zero unless some other mechanism ensures the
2147 accessibility of the value in a narrower mode.
2149 You should define this macro to return nonzero in as many cases as
2150 possible since doing so will allow GCC to perform better register
2154 @defmac AVOID_CCMODE_COPIES
2155 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2156 registers. You should only define this macro if support for copying to/from
2157 @code{CCmode} is incomplete.
2160 @node Leaf Functions
2161 @subsection Handling Leaf Functions
2163 @cindex leaf functions
2164 @cindex functions, leaf
2165 On some machines, a leaf function (i.e., one which makes no calls) can run
2166 more efficiently if it does not make its own register window. Often this
2167 means it is required to receive its arguments in the registers where they
2168 are passed by the caller, instead of the registers where they would
2171 The special treatment for leaf functions generally applies only when
2172 other conditions are met; for example, often they may use only those
2173 registers for its own variables and temporaries. We use the term ``leaf
2174 function'' to mean a function that is suitable for this special
2175 handling, so that functions with no calls are not necessarily ``leaf
2178 GCC assigns register numbers before it knows whether the function is
2179 suitable for leaf function treatment. So it needs to renumber the
2180 registers in order to output a leaf function. The following macros
2183 @defmac LEAF_REGISTERS
2184 Name of a char vector, indexed by hard register number, which
2185 contains 1 for a register that is allowable in a candidate for leaf
2188 If leaf function treatment involves renumbering the registers, then the
2189 registers marked here should be the ones before renumbering---those that
2190 GCC would ordinarily allocate. The registers which will actually be
2191 used in the assembler code, after renumbering, should not be marked with 1
2194 Define this macro only if the target machine offers a way to optimize
2195 the treatment of leaf functions.
2198 @defmac LEAF_REG_REMAP (@var{regno})
2199 A C expression whose value is the register number to which @var{regno}
2200 should be renumbered, when a function is treated as a leaf function.
2202 If @var{regno} is a register number which should not appear in a leaf
2203 function before renumbering, then the expression should yield @minus{}1, which
2204 will cause the compiler to abort.
2206 Define this macro only if the target machine offers a way to optimize the
2207 treatment of leaf functions, and registers need to be renumbered to do
2211 @findex current_function_is_leaf
2212 @findex current_function_uses_only_leaf_regs
2213 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2214 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2215 specially. They can test the C variable @code{current_function_is_leaf}
2216 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2217 set prior to local register allocation and is valid for the remaining
2218 compiler passes. They can also test the C variable
2219 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2220 functions which only use leaf registers.
2221 @code{current_function_uses_only_leaf_regs} is valid after reload and is
2222 only useful if @code{LEAF_REGISTERS} is defined.
2223 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2224 @c of the next paragraph?! --mew 2feb93
2226 @node Stack Registers
2227 @subsection Registers That Form a Stack
2229 There are special features to handle computers where some of the
2230 ``registers'' form a stack. Stack registers are normally written by
2231 pushing onto the stack, and are numbered relative to the top of the
2234 Currently, GCC can only handle one group of stack-like registers, and
2235 they must be consecutively numbered. Furthermore, the existing
2236 support for stack-like registers is specific to the 80387 floating
2237 point coprocessor. If you have a new architecture that uses
2238 stack-like registers, you will need to do substantial work on
2239 @file{reg-stack.c} and write your machine description to cooperate
2240 with it, as well as defining these macros.
2243 Define this if the machine has any stack-like registers.
2246 @defmac FIRST_STACK_REG
2247 The number of the first stack-like register. This one is the top
2251 @defmac LAST_STACK_REG
2252 The number of the last stack-like register. This one is the bottom of
2256 @node Register Classes
2257 @section Register Classes
2258 @cindex register class definitions
2259 @cindex class definitions, register
2261 On many machines, the numbered registers are not all equivalent.
2262 For example, certain registers may not be allowed for indexed addressing;
2263 certain registers may not be allowed in some instructions. These machine
2264 restrictions are described to the compiler using @dfn{register classes}.
2266 You define a number of register classes, giving each one a name and saying
2267 which of the registers belong to it. Then you can specify register classes
2268 that are allowed as operands to particular instruction patterns.
2272 In general, each register will belong to several classes. In fact, one
2273 class must be named @code{ALL_REGS} and contain all the registers. Another
2274 class must be named @code{NO_REGS} and contain no registers. Often the
2275 union of two classes will be another class; however, this is not required.
2277 @findex GENERAL_REGS
2278 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2279 terribly special about the name, but the operand constraint letters
2280 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2281 the same as @code{ALL_REGS}, just define it as a macro which expands
2284 Order the classes so that if class @var{x} is contained in class @var{y}
2285 then @var{x} has a lower class number than @var{y}.
2287 The way classes other than @code{GENERAL_REGS} are specified in operand
2288 constraints is through machine-dependent operand constraint letters.
2289 You can define such letters to correspond to various classes, then use
2290 them in operand constraints.
2292 You should define a class for the union of two classes whenever some
2293 instruction allows both classes. For example, if an instruction allows
2294 either a floating point (coprocessor) register or a general register for a
2295 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2296 which includes both of them. Otherwise you will get suboptimal code.
2298 You must also specify certain redundant information about the register
2299 classes: for each class, which classes contain it and which ones are
2300 contained in it; for each pair of classes, the largest class contained
2303 When a value occupying several consecutive registers is expected in a
2304 certain class, all the registers used must belong to that class.
2305 Therefore, register classes cannot be used to enforce a requirement for
2306 a register pair to start with an even-numbered register. The way to
2307 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2309 Register classes used for input-operands of bitwise-and or shift
2310 instructions have a special requirement: each such class must have, for
2311 each fixed-point machine mode, a subclass whose registers can transfer that
2312 mode to or from memory. For example, on some machines, the operations for
2313 single-byte values (@code{QImode}) are limited to certain registers. When
2314 this is so, each register class that is used in a bitwise-and or shift
2315 instruction must have a subclass consisting of registers from which
2316 single-byte values can be loaded or stored. This is so that
2317 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2319 @deftp {Data type} {enum reg_class}
2320 An enumeral type that must be defined with all the register class names
2321 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
2322 must be the last register class, followed by one more enumeral value,
2323 @code{LIM_REG_CLASSES}, which is not a register class but rather
2324 tells how many classes there are.
2326 Each register class has a number, which is the value of casting
2327 the class name to type @code{int}. The number serves as an index
2328 in many of the tables described below.
2331 @defmac N_REG_CLASSES
2332 The number of distinct register classes, defined as follows:
2335 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2339 @defmac REG_CLASS_NAMES
2340 An initializer containing the names of the register classes as C string
2341 constants. These names are used in writing some of the debugging dumps.
2344 @defmac REG_CLASS_CONTENTS
2345 An initializer containing the contents of the register classes, as integers
2346 which are bit masks. The @var{n}th integer specifies the contents of class
2347 @var{n}. The way the integer @var{mask} is interpreted is that
2348 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2350 When the machine has more than 32 registers, an integer does not suffice.
2351 Then the integers are replaced by sub-initializers, braced groupings containing
2352 several integers. Each sub-initializer must be suitable as an initializer
2353 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2354 In this situation, the first integer in each sub-initializer corresponds to
2355 registers 0 through 31, the second integer to registers 32 through 63, and
2359 @defmac REGNO_REG_CLASS (@var{regno})
2360 A C expression whose value is a register class containing hard register
2361 @var{regno}. In general there is more than one such class; choose a class
2362 which is @dfn{minimal}, meaning that no smaller class also contains the
2366 @defmac BASE_REG_CLASS
2367 A macro whose definition is the name of the class to which a valid
2368 base register must belong. A base register is one used in an address
2369 which is the register value plus a displacement.
2372 @defmac MODE_BASE_REG_CLASS (@var{mode})
2373 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2374 the selection of a base register in a mode dependent manner. If
2375 @var{mode} is VOIDmode then it should return the same value as
2376 @code{BASE_REG_CLASS}.
2379 @defmac INDEX_REG_CLASS
2380 A macro whose definition is the name of the class to which a valid
2381 index register must belong. An index register is one used in an
2382 address where its value is either multiplied by a scale factor or
2383 added to another register (as well as added to a displacement).
2386 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2387 For the constraint at the start of @var{str}, which starts with the letter
2388 @var{c}, return the length. This allows you to have register class /
2389 constant / extra constraints that are longer than a single letter;
2390 you don't need to define this macro if you can do with single-letter
2391 constraints only. The definition of this macro should use
2392 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2393 to handle specially.
2394 There are some sanity checks in genoutput.c that check the constraint lengths
2395 for the md file, so you can also use this macro to help you while you are
2396 transitioning from a byzantine single-letter-constraint scheme: when you
2397 return a negative length for a constraint you want to re-use, genoutput
2398 will complain about every instance where it is used in the md file.
2401 @defmac REG_CLASS_FROM_LETTER (@var{char})
2402 A C expression which defines the machine-dependent operand constraint
2403 letters for register classes. If @var{char} is such a letter, the
2404 value should be the register class corresponding to it. Otherwise,
2405 the value should be @code{NO_REGS}. The register letter @samp{r},
2406 corresponding to class @code{GENERAL_REGS}, will not be passed
2407 to this macro; you do not need to handle it.
2410 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2411 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2412 passed in @var{str}, so that you can use suffixes to distinguish between
2416 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2417 A C expression which is nonzero if register number @var{num} is
2418 suitable for use as a base register in operand addresses. It may be
2419 either a suitable hard register or a pseudo register that has been
2420 allocated such a hard register.
2423 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2424 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2425 that expression may examine the mode of the memory reference in
2426 @var{mode}. You should define this macro if the mode of the memory
2427 reference affects whether a register may be used as a base register. If
2428 you define this macro, the compiler will use it instead of
2429 @code{REGNO_OK_FOR_BASE_P}.
2432 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2433 A C expression which is nonzero if register number @var{num} is
2434 suitable for use as an index register in operand addresses. It may be
2435 either a suitable hard register or a pseudo register that has been
2436 allocated such a hard register.
2438 The difference between an index register and a base register is that
2439 the index register may be scaled. If an address involves the sum of
2440 two registers, neither one of them scaled, then either one may be
2441 labeled the ``base'' and the other the ``index''; but whichever
2442 labeling is used must fit the machine's constraints of which registers
2443 may serve in each capacity. The compiler will try both labelings,
2444 looking for one that is valid, and will reload one or both registers
2445 only if neither labeling works.
2448 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2449 A C expression that places additional restrictions on the register class
2450 to use when it is necessary to copy value @var{x} into a register in class
2451 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2452 another, smaller class. On many machines, the following definition is
2456 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2459 Sometimes returning a more restrictive class makes better code. For
2460 example, on the 68000, when @var{x} is an integer constant that is in range
2461 for a @samp{moveq} instruction, the value of this macro is always
2462 @code{DATA_REGS} as long as @var{class} includes the data registers.
2463 Requiring a data register guarantees that a @samp{moveq} will be used.
2465 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2466 @var{class} is if @var{x} is a legitimate constant which cannot be
2467 loaded into some register class. By returning @code{NO_REGS} you can
2468 force @var{x} into a memory location. For example, rs6000 can load
2469 immediate values into general-purpose registers, but does not have an
2470 instruction for loading an immediate value into a floating-point
2471 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2472 @var{x} is a floating-point constant. If the constant can't be loaded
2473 into any kind of register, code generation will be better if
2474 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2475 of using @code{PREFERRED_RELOAD_CLASS}.
2478 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2479 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2480 input reloads. If you don't define this macro, the default is to use
2481 @var{class}, unchanged.
2484 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2485 A C expression that places additional restrictions on the register class
2486 to use when it is necessary to be able to hold a value of mode
2487 @var{mode} in a reload register for which class @var{class} would
2490 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2491 there are certain modes that simply can't go in certain reload classes.
2493 The value is a register class; perhaps @var{class}, or perhaps another,
2496 Don't define this macro unless the target machine has limitations which
2497 require the macro to do something nontrivial.
2500 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2501 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2502 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2503 Many machines have some registers that cannot be copied directly to or
2504 from memory or even from other types of registers. An example is the
2505 @samp{MQ} register, which on most machines, can only be copied to or
2506 from general registers, but not memory. Some machines allow copying all
2507 registers to and from memory, but require a scratch register for stores
2508 to some memory locations (e.g., those with symbolic address on the RT,
2509 and those with certain symbolic address on the SPARC when compiling
2510 PIC)@. In some cases, both an intermediate and a scratch register are
2513 You should define these macros to indicate to the reload phase that it may
2514 need to allocate at least one register for a reload in addition to the
2515 register to contain the data. Specifically, if copying @var{x} to a
2516 register @var{class} in @var{mode} requires an intermediate register,
2517 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2518 largest register class all of whose registers can be used as
2519 intermediate registers or scratch registers.
2521 If copying a register @var{class} in @var{mode} to @var{x} requires an
2522 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2523 should be defined to return the largest register class required. If the
2524 requirements for input and output reloads are the same, the macro
2525 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2528 The values returned by these macros are often @code{GENERAL_REGS}.
2529 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2530 can be directly copied to or from a register of @var{class} in
2531 @var{mode} without requiring a scratch register. Do not define this
2532 macro if it would always return @code{NO_REGS}.
2534 If a scratch register is required (either with or without an
2535 intermediate register), you should define patterns for
2536 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2537 (@pxref{Standard Names}. These patterns, which will normally be
2538 implemented with a @code{define_expand}, should be similar to the
2539 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2542 Define constraints for the reload register and scratch register that
2543 contain a single register class. If the original reload register (whose
2544 class is @var{class}) can meet the constraint given in the pattern, the
2545 value returned by these macros is used for the class of the scratch
2546 register. Otherwise, two additional reload registers are required.
2547 Their classes are obtained from the constraints in the insn pattern.
2549 @var{x} might be a pseudo-register or a @code{subreg} of a
2550 pseudo-register, which could either be in a hard register or in memory.
2551 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2552 in memory and the hard register number if it is in a register.
2554 These macros should not be used in the case where a particular class of
2555 registers can only be copied to memory and not to another class of
2556 registers. In that case, secondary reload registers are not needed and
2557 would not be helpful. Instead, a stack location must be used to perform
2558 the copy and the @code{mov@var{m}} pattern should use memory as an
2559 intermediate storage. This case often occurs between floating-point and
2563 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2564 Certain machines have the property that some registers cannot be copied
2565 to some other registers without using memory. Define this macro on
2566 those machines to be a C expression that is nonzero if objects of mode
2567 @var{m} in registers of @var{class1} can only be copied to registers of
2568 class @var{class2} by storing a register of @var{class1} into memory
2569 and loading that memory location into a register of @var{class2}.
2571 Do not define this macro if its value would always be zero.
2574 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2575 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2576 allocates a stack slot for a memory location needed for register copies.
2577 If this macro is defined, the compiler instead uses the memory location
2578 defined by this macro.
2580 Do not define this macro if you do not define
2581 @code{SECONDARY_MEMORY_NEEDED}.
2584 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2585 When the compiler needs a secondary memory location to copy between two
2586 registers of mode @var{mode}, it normally allocates sufficient memory to
2587 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2588 load operations in a mode that many bits wide and whose class is the
2589 same as that of @var{mode}.
2591 This is right thing to do on most machines because it ensures that all
2592 bits of the register are copied and prevents accesses to the registers
2593 in a narrower mode, which some machines prohibit for floating-point
2596 However, this default behavior is not correct on some machines, such as
2597 the DEC Alpha, that store short integers in floating-point registers
2598 differently than in integer registers. On those machines, the default
2599 widening will not work correctly and you must define this macro to
2600 suppress that widening in some cases. See the file @file{alpha.h} for
2603 Do not define this macro if you do not define
2604 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2605 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2608 @defmac SMALL_REGISTER_CLASSES
2609 On some machines, it is risky to let hard registers live across arbitrary
2610 insns. Typically, these machines have instructions that require values
2611 to be in specific registers (like an accumulator), and reload will fail
2612 if the required hard register is used for another purpose across such an
2615 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2616 value on these machines. When this macro has a nonzero value, the
2617 compiler will try to minimize the lifetime of hard registers.
2619 It is always safe to define this macro with a nonzero value, but if you
2620 unnecessarily define it, you will reduce the amount of optimizations
2621 that can be performed in some cases. If you do not define this macro
2622 with a nonzero value when it is required, the compiler will run out of
2623 spill registers and print a fatal error message. For most machines, you
2624 should not define this macro at all.
2627 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2628 A C expression whose value is nonzero if pseudos that have been assigned
2629 to registers of class @var{class} would likely be spilled because
2630 registers of @var{class} are needed for spill registers.
2632 The default value of this macro returns 1 if @var{class} has exactly one
2633 register and zero otherwise. On most machines, this default should be
2634 used. Only define this macro to some other expression if pseudos
2635 allocated by @file{local-alloc.c} end up in memory because their hard
2636 registers were needed for spill registers. If this macro returns nonzero
2637 for those classes, those pseudos will only be allocated by
2638 @file{global.c}, which knows how to reallocate the pseudo to another
2639 register. If there would not be another register available for
2640 reallocation, you should not change the definition of this macro since
2641 the only effect of such a definition would be to slow down register
2645 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2646 A C expression for the maximum number of consecutive registers
2647 of class @var{class} needed to hold a value of mode @var{mode}.
2649 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2650 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2651 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2652 @var{mode})} for all @var{regno} values in the class @var{class}.
2654 This macro helps control the handling of multiple-word values
2658 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2659 If defined, a C expression that returns nonzero for a @var{class} for which
2660 a change from mode @var{from} to mode @var{to} is invalid.
2662 For the example, loading 32-bit integer or floating-point objects into
2663 floating-point registers on the Alpha extends them to 64 bits.
2664 Therefore loading a 64-bit object and then storing it as a 32-bit object
2665 does not store the low-order 32 bits, as would be the case for a normal
2666 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2670 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2671 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2672 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2676 Three other special macros describe which operands fit which constraint
2679 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2680 A C expression that defines the machine-dependent operand constraint
2681 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2682 particular ranges of integer values. If @var{c} is one of those
2683 letters, the expression should check that @var{value}, an integer, is in
2684 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2685 not one of those letters, the value should be 0 regardless of
2689 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2690 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2691 string passed in @var{str}, so that you can use suffixes to distinguish
2692 between different variants.
2695 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2696 A C expression that defines the machine-dependent operand constraint
2697 letters that specify particular ranges of @code{const_double} values
2698 (@samp{G} or @samp{H}).
2700 If @var{c} is one of those letters, the expression should check that
2701 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2702 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2703 letters, the value should be 0 regardless of @var{value}.
2705 @code{const_double} is used for all floating-point constants and for
2706 @code{DImode} fixed-point constants. A given letter can accept either
2707 or both kinds of values. It can use @code{GET_MODE} to distinguish
2708 between these kinds.
2711 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2712 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2713 string passed in @var{str}, so that you can use suffixes to distinguish
2714 between different variants.
2717 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2718 A C expression that defines the optional machine-dependent constraint
2719 letters that can be used to segregate specific types of operands, usually
2720 memory references, for the target machine. Any letter that is not
2721 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2722 @code{REG_CLASS_FROM_CONSTRAINT}
2723 may be used. Normally this macro will not be defined.
2725 If it is required for a particular target machine, it should return 1
2726 if @var{value} corresponds to the operand type represented by the
2727 constraint letter @var{c}. If @var{c} is not defined as an extra
2728 constraint, the value returned should be 0 regardless of @var{value}.
2730 For example, on the ROMP, load instructions cannot have their output
2731 in r0 if the memory reference contains a symbolic address. Constraint
2732 letter @samp{Q} is defined as representing a memory address that does
2733 @emph{not} contain a symbolic address. An alternative is specified with
2734 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2735 alternative specifies @samp{m} on the input and a register class that
2736 does not include r0 on the output.
2739 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2740 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2741 in @var{str}, so that you can use suffixes to distinguish between different
2745 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2746 A C expression that defines the optional machine-dependent constraint
2747 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2748 be treated like memory constraints by the reload pass.
2750 It should return 1 if the operand type represented by the constraint
2751 at the start of @var{str}, the first letter of which is the letter @var{c},
2752 comprises a subset of all memory references including
2753 all those whose address is simply a base register. This allows the reload
2754 pass to reload an operand, if it does not directly correspond to the operand
2755 type of @var{c}, by copying its address into a base register.
2757 For example, on the S/390, some instructions do not accept arbitrary
2758 memory references, but only those that do not make use of an index
2759 register. The constraint letter @samp{Q} is defined via
2760 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2761 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2762 a @samp{Q} constraint can handle any memory operand, because the
2763 reload pass knows it can be reloaded by copying the memory address
2764 into a base register if required. This is analogous to the way
2765 a @samp{o} constraint can handle any memory operand.
2768 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2769 A C expression that defines the optional machine-dependent constraint
2770 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2771 @code{EXTRA_CONSTRAINT_STR}, that should
2772 be treated like address constraints by the reload pass.
2774 It should return 1 if the operand type represented by the constraint
2775 at the start of @var{str}, which starts with the letter @var{c}, comprises
2776 a subset of all memory addresses including
2777 all those that consist of just a base register. This allows the reload
2778 pass to reload an operand, if it does not directly correspond to the operand
2779 type of @var{str}, by copying it into a base register.
2781 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2782 be used with the @code{address_operand} predicate. It is treated
2783 analogously to the @samp{p} constraint.
2786 @node Stack and Calling
2787 @section Stack Layout and Calling Conventions
2788 @cindex calling conventions
2790 @c prevent bad page break with this line
2791 This describes the stack layout and calling conventions.
2795 * Exception Handling::
2800 * Register Arguments::
2802 * Aggregate Return::
2810 @subsection Basic Stack Layout
2811 @cindex stack frame layout
2812 @cindex frame layout
2814 @c prevent bad page break with this line
2815 Here is the basic stack layout.
2817 @defmac STACK_GROWS_DOWNWARD
2818 Define this macro if pushing a word onto the stack moves the stack
2819 pointer to a smaller address.
2821 When we say, ``define this macro if @dots{},'' it means that the
2822 compiler checks this macro only with @code{#ifdef} so the precise
2823 definition used does not matter.
2826 @defmac STACK_PUSH_CODE
2827 This macro defines the operation used when something is pushed
2828 on the stack. In RTL, a push operation will be
2829 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2831 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2832 and @code{POST_INC}. Which of these is correct depends on
2833 the stack direction and on whether the stack pointer points
2834 to the last item on the stack or whether it points to the
2835 space for the next item on the stack.
2837 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2838 defined, which is almost always right, and @code{PRE_INC} otherwise,
2839 which is often wrong.
2842 @defmac FRAME_GROWS_DOWNWARD
2843 Define this macro if the addresses of local variable slots are at negative
2844 offsets from the frame pointer.
2847 @defmac ARGS_GROW_DOWNWARD
2848 Define this macro if successive arguments to a function occupy decreasing
2849 addresses on the stack.
2852 @defmac STARTING_FRAME_OFFSET
2853 Offset from the frame pointer to the first local variable slot to be allocated.
2855 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2856 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2857 Otherwise, it is found by adding the length of the first slot to the
2858 value @code{STARTING_FRAME_OFFSET}.
2859 @c i'm not sure if the above is still correct.. had to change it to get
2860 @c rid of an overfull. --mew 2feb93
2863 @defmac STACK_ALIGNMENT_NEEDED
2864 Define to zero to disable final alignment of the stack during reload.
2865 The nonzero default for this macro is suitable for most ports.
2867 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2868 is a register save block following the local block that doesn't require
2869 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2870 stack alignment and do it in the backend.
2873 @defmac STACK_POINTER_OFFSET
2874 Offset from the stack pointer register to the first location at which
2875 outgoing arguments are placed. If not specified, the default value of
2876 zero is used. This is the proper value for most machines.
2878 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2879 the first location at which outgoing arguments are placed.
2882 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2883 Offset from the argument pointer register to the first argument's
2884 address. On some machines it may depend on the data type of the
2887 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2888 the first argument's address.
2891 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2892 Offset from the stack pointer register to an item dynamically allocated
2893 on the stack, e.g., by @code{alloca}.
2895 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2896 length of the outgoing arguments. The default is correct for most
2897 machines. See @file{function.c} for details.
2900 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2901 A C expression whose value is RTL representing the address in a stack
2902 frame where the pointer to the caller's frame is stored. Assume that
2903 @var{frameaddr} is an RTL expression for the address of the stack frame
2906 If you don't define this macro, the default is to return the value
2907 of @var{frameaddr}---that is, the stack frame address is also the
2908 address of the stack word that points to the previous frame.
2911 @defmac SETUP_FRAME_ADDRESSES
2912 If defined, a C expression that produces the machine-specific code to
2913 setup the stack so that arbitrary frames can be accessed. For example,
2914 on the SPARC, we must flush all of the register windows to the stack
2915 before we can access arbitrary stack frames. You will seldom need to
2919 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
2920 This target hook should return an rtx that is used to store
2921 the address of the current frame into the built in @code{setjmp} buffer.
2922 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2923 machines. One reason you may need to define this target hook is if
2924 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2927 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2928 A C expression whose value is RTL representing the value of the return
2929 address for the frame @var{count} steps up from the current frame, after
2930 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2931 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2932 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2934 The value of the expression must always be the correct address when
2935 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2936 determine the return address of other frames.
2939 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2940 Define this if the return address of a particular stack frame is accessed
2941 from the frame pointer of the previous stack frame.
2944 @defmac INCOMING_RETURN_ADDR_RTX
2945 A C expression whose value is RTL representing the location of the
2946 incoming return address at the beginning of any function, before the
2947 prologue. This RTL is either a @code{REG}, indicating that the return
2948 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2951 You only need to define this macro if you want to support call frame
2952 debugging information like that provided by DWARF 2.
2954 If this RTL is a @code{REG}, you should also define
2955 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2958 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2959 A C expression whose value is an integer giving a DWARF 2 column
2960 number that may be used as an alternate return column. This should
2961 be defined only if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2962 general register, but an alternate column needs to be used for
2966 @defmac INCOMING_FRAME_SP_OFFSET
2967 A C expression whose value is an integer giving the offset, in bytes,
2968 from the value of the stack pointer register to the top of the stack
2969 frame at the beginning of any function, before the prologue. The top of
2970 the frame is defined to be the value of the stack pointer in the
2971 previous frame, just before the call instruction.
2973 You only need to define this macro if you want to support call frame
2974 debugging information like that provided by DWARF 2.
2977 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2978 A C expression whose value is an integer giving the offset, in bytes,
2979 from the argument pointer to the canonical frame address (cfa). The
2980 final value should coincide with that calculated by
2981 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2982 during virtual register instantiation.
2984 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2985 which is correct for most machines; in general, the arguments are found
2986 immediately before the stack frame. Note that this is not the case on
2987 some targets that save registers into the caller's frame, such as SPARC
2988 and rs6000, and so such targets need to define this macro.
2990 You only need to define this macro if the default is incorrect, and you
2991 want to support call frame debugging information like that provided by
2995 @node Exception Handling
2996 @subsection Exception Handling Support
2997 @cindex exception handling
2999 @defmac EH_RETURN_DATA_REGNO (@var{N})
3000 A C expression whose value is the @var{N}th register number used for
3001 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3002 @var{N} registers are usable.
3004 The exception handling library routines communicate with the exception
3005 handlers via a set of agreed upon registers. Ideally these registers
3006 should be call-clobbered; it is possible to use call-saved registers,
3007 but may negatively impact code size. The target must support at least
3008 2 data registers, but should define 4 if there are enough free registers.
3010 You must define this macro if you want to support call frame exception
3011 handling like that provided by DWARF 2.
3014 @defmac EH_RETURN_STACKADJ_RTX
3015 A C expression whose value is RTL representing a location in which
3016 to store a stack adjustment to be applied before function return.
3017 This is used to unwind the stack to an exception handler's call frame.
3018 It will be assigned zero on code paths that return normally.
3020 Typically this is a call-clobbered hard register that is otherwise
3021 untouched by the epilogue, but could also be a stack slot.
3023 Do not define this macro if the stack pointer is saved and restored
3024 by the regular prolog and epilog code in the call frame itself; in
3025 this case, the exception handling library routines will update the
3026 stack location to be restored in place. Otherwise, you must define
3027 this macro if you want to support call frame exception handling like
3028 that provided by DWARF 2.
3031 @defmac EH_RETURN_HANDLER_RTX
3032 A C expression whose value is RTL representing a location in which
3033 to store the address of an exception handler to which we should
3034 return. It will not be assigned on code paths that return normally.
3036 Typically this is the location in the call frame at which the normal
3037 return address is stored. For targets that return by popping an
3038 address off the stack, this might be a memory address just below
3039 the @emph{target} call frame rather than inside the current call
3040 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3041 been assigned, so it may be used to calculate the location of the
3044 Some targets have more complex requirements than storing to an
3045 address calculable during initial code generation. In that case
3046 the @code{eh_return} instruction pattern should be used instead.
3048 If you want to support call frame exception handling, you must
3049 define either this macro or the @code{eh_return} instruction pattern.
3052 @defmac RETURN_ADDR_OFFSET
3053 If defined, an integer-valued C expression for which rtl will be generated
3054 to add it to the exception handler address before it is searched in the
3055 exception handling tables, and to subtract it again from the address before
3056 using it to return to the exception handler.
3059 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3060 This macro chooses the encoding of pointers embedded in the exception
3061 handling sections. If at all possible, this should be defined such
3062 that the exception handling section will not require dynamic relocations,
3063 and so may be read-only.
3065 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3066 @var{global} is true if the symbol may be affected by dynamic relocations.
3067 The macro should return a combination of the @code{DW_EH_PE_*} defines
3068 as found in @file{dwarf2.h}.
3070 If this macro is not defined, pointers will not be encoded but
3071 represented directly.
3074 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3075 This macro allows the target to emit whatever special magic is required
3076 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3077 Generic code takes care of pc-relative and indirect encodings; this must
3078 be defined if the target uses text-relative or data-relative encodings.
3080 This is a C statement that branches to @var{done} if the format was
3081 handled. @var{encoding} is the format chosen, @var{size} is the number
3082 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3086 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs}, @var{success})
3087 This macro allows the target to add cpu and operating system specific
3088 code to the call-frame unwinder for use when there is no unwind data
3089 available. The most common reason to implement this macro is to unwind
3090 through signal frames.
3092 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3093 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3094 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3095 for the address of the code being executed and @code{context->cfa} for
3096 the stack pointer value. If the frame can be decoded, the register save
3097 addresses should be updated in @var{fs} and the macro should branch to
3098 @var{success}. If the frame cannot be decoded, the macro should do
3101 For proper signal handling in Java this macro is accompanied by
3102 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3105 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3106 This macro allows the target to add operating system specific code to the
3107 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3108 usually used for signal or interrupt frames.
3110 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3111 @var{context} is an @code{_Unwind_Context};
3112 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3113 for the abi and context in the @code{.unwabi} directive. If the
3114 @code{.unwabi} directive can be handled, the register save addresses should
3115 be updated in @var{fs}.
3118 @defmac TARGET_USES_WEAK_UNWIND_INFO
3119 A C expression that evaluates to true if the target requires unwind
3120 info to be given comdat linkage. Define it to be @code{1} if comdat
3121 linkage is necessary. The default is @code{0}.
3124 @node Stack Checking
3125 @subsection Specifying How Stack Checking is Done
3127 GCC will check that stack references are within the boundaries of
3128 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3132 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3133 will assume that you have arranged for stack checking to be done at
3134 appropriate places in the configuration files, e.g., in
3135 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3139 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3140 called @code{check_stack} in your @file{md} file, GCC will call that
3141 pattern with one argument which is the address to compare the stack
3142 value against. You must arrange for this pattern to report an error if
3143 the stack pointer is out of range.
3146 If neither of the above are true, GCC will generate code to periodically
3147 ``probe'' the stack pointer using the values of the macros defined below.
3150 Normally, you will use the default values of these macros, so GCC
3151 will use the third approach.
3153 @defmac STACK_CHECK_BUILTIN
3154 A nonzero value if stack checking is done by the configuration files in a
3155 machine-dependent manner. You should define this macro if stack checking
3156 is require by the ABI of your machine or if you would like to have to stack
3157 checking in some more efficient way than GCC's portable approach.
3158 The default value of this macro is zero.
3161 @defmac STACK_CHECK_PROBE_INTERVAL
3162 An integer representing the interval at which GCC must generate stack
3163 probe instructions. You will normally define this macro to be no larger
3164 than the size of the ``guard pages'' at the end of a stack area. The
3165 default value of 4096 is suitable for most systems.
3168 @defmac STACK_CHECK_PROBE_LOAD
3169 A integer which is nonzero if GCC should perform the stack probe
3170 as a load instruction and zero if GCC should use a store instruction.
3171 The default is zero, which is the most efficient choice on most systems.
3174 @defmac STACK_CHECK_PROTECT
3175 The number of bytes of stack needed to recover from a stack overflow,
3176 for languages where such a recovery is supported. The default value of
3177 75 words should be adequate for most machines.
3180 @defmac STACK_CHECK_MAX_FRAME_SIZE
3181 The maximum size of a stack frame, in bytes. GCC will generate probe
3182 instructions in non-leaf functions to ensure at least this many bytes of
3183 stack are available. If a stack frame is larger than this size, stack
3184 checking will not be reliable and GCC will issue a warning. The
3185 default is chosen so that GCC only generates one instruction on most
3186 systems. You should normally not change the default value of this macro.
3189 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3190 GCC uses this value to generate the above warning message. It
3191 represents the amount of fixed frame used by a function, not including
3192 space for any callee-saved registers, temporaries and user variables.
3193 You need only specify an upper bound for this amount and will normally
3194 use the default of four words.
3197 @defmac STACK_CHECK_MAX_VAR_SIZE
3198 The maximum size, in bytes, of an object that GCC will place in the
3199 fixed area of the stack frame when the user specifies
3200 @option{-fstack-check}.
3201 GCC computed the default from the values of the above macros and you will
3202 normally not need to override that default.
3206 @node Frame Registers
3207 @subsection Registers That Address the Stack Frame
3209 @c prevent bad page break with this line
3210 This discusses registers that address the stack frame.
3212 @defmac STACK_POINTER_REGNUM
3213 The register number of the stack pointer register, which must also be a
3214 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3215 the hardware determines which register this is.
3218 @defmac FRAME_POINTER_REGNUM
3219 The register number of the frame pointer register, which is used to
3220 access automatic variables in the stack frame. On some machines, the
3221 hardware determines which register this is. On other machines, you can
3222 choose any register you wish for this purpose.
3225 @defmac HARD_FRAME_POINTER_REGNUM
3226 On some machines the offset between the frame pointer and starting
3227 offset of the automatic variables is not known until after register
3228 allocation has been done (for example, because the saved registers are
3229 between these two locations). On those machines, define
3230 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3231 be used internally until the offset is known, and define
3232 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3233 used for the frame pointer.
3235 You should define this macro only in the very rare circumstances when it
3236 is not possible to calculate the offset between the frame pointer and
3237 the automatic variables until after register allocation has been
3238 completed. When this macro is defined, you must also indicate in your
3239 definition of @code{ELIMINABLE_REGS} how to eliminate
3240 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3241 or @code{STACK_POINTER_REGNUM}.
3243 Do not define this macro if it would be the same as
3244 @code{FRAME_POINTER_REGNUM}.
3247 @defmac ARG_POINTER_REGNUM
3248 The register number of the arg pointer register, which is used to access
3249 the function's argument list. On some machines, this is the same as the
3250 frame pointer register. On some machines, the hardware determines which
3251 register this is. On other machines, you can choose any register you
3252 wish for this purpose. If this is not the same register as the frame
3253 pointer register, then you must mark it as a fixed register according to
3254 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3255 (@pxref{Elimination}).
3258 @defmac RETURN_ADDRESS_POINTER_REGNUM
3259 The register number of the return address pointer register, which is used to
3260 access the current function's return address from the stack. On some
3261 machines, the return address is not at a fixed offset from the frame
3262 pointer or stack pointer or argument pointer. This register can be defined
3263 to point to the return address on the stack, and then be converted by
3264 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3266 Do not define this macro unless there is no other way to get the return
3267 address from the stack.
3270 @defmac STATIC_CHAIN_REGNUM
3271 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3272 Register numbers used for passing a function's static chain pointer. If
3273 register windows are used, the register number as seen by the called
3274 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3275 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3276 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3279 The static chain register need not be a fixed register.
3281 If the static chain is passed in memory, these macros should not be
3282 defined; instead, the next two macros should be defined.
3285 @defmac STATIC_CHAIN
3286 @defmacx STATIC_CHAIN_INCOMING
3287 If the static chain is passed in memory, these macros provide rtx giving
3288 @code{mem} expressions that denote where they are stored.
3289 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3290 as seen by the calling and called functions, respectively. Often the former
3291 will be at an offset from the stack pointer and the latter at an offset from
3294 @findex stack_pointer_rtx
3295 @findex frame_pointer_rtx
3296 @findex arg_pointer_rtx
3297 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3298 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3299 macros and should be used to refer to those items.
3301 If the static chain is passed in a register, the two previous macros should
3305 @defmac DWARF_FRAME_REGISTERS
3306 This macro specifies the maximum number of hard registers that can be
3307 saved in a call frame. This is used to size data structures used in
3308 DWARF2 exception handling.
3310 Prior to GCC 3.0, this macro was needed in order to establish a stable
3311 exception handling ABI in the face of adding new hard registers for ISA
3312 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3313 in the number of hard registers. Nevertheless, this macro can still be
3314 used to reduce the runtime memory requirements of the exception handling
3315 routines, which can be substantial if the ISA contains a lot of
3316 registers that are not call-saved.
3318 If this macro is not defined, it defaults to
3319 @code{FIRST_PSEUDO_REGISTER}.
3322 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3324 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3325 for backward compatibility in pre GCC 3.0 compiled code.
3327 If this macro is not defined, it defaults to
3328 @code{DWARF_FRAME_REGISTERS}.
3331 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3333 Define this macro if the target's representation for dwarf registers
3334 is different than the internal representation for unwind column.
3335 Given a dwarf register, this macro should return the internal unwind
3336 column number to use instead.
3338 See the PowerPC's SPE target for an example.
3341 @defmac DWARF_FRAME_REGNUM (@var{regno})
3343 Define this macro if the target's representation for dwarf registers
3344 used in .eh_frame or .debug_frame is different from that used in other
3345 debug info sections. Given a GCC hard register number, this macro
3346 should return the .eh_frame register number. The default is
3347 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3351 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3353 Define this macro to map register numbers held in the call frame info
3354 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3355 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3356 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3357 return @code{@var{regno}}.
3362 @subsection Eliminating Frame Pointer and Arg Pointer
3364 @c prevent bad page break with this line
3365 This is about eliminating the frame pointer and arg pointer.
3367 @defmac FRAME_POINTER_REQUIRED
3368 A C expression which is nonzero if a function must have and use a frame
3369 pointer. This expression is evaluated in the reload pass. If its value is
3370 nonzero the function will have a frame pointer.
3372 The expression can in principle examine the current function and decide
3373 according to the facts, but on most machines the constant 0 or the
3374 constant 1 suffices. Use 0 when the machine allows code to be generated
3375 with no frame pointer, and doing so saves some time or space. Use 1
3376 when there is no possible advantage to avoiding a frame pointer.
3378 In certain cases, the compiler does not know how to produce valid code
3379 without a frame pointer. The compiler recognizes those cases and
3380 automatically gives the function a frame pointer regardless of what
3381 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3384 In a function that does not require a frame pointer, the frame pointer
3385 register can be allocated for ordinary usage, unless you mark it as a
3386 fixed register. See @code{FIXED_REGISTERS} for more information.
3389 @findex get_frame_size
3390 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3391 A C statement to store in the variable @var{depth-var} the difference
3392 between the frame pointer and the stack pointer values immediately after
3393 the function prologue. The value would be computed from information
3394 such as the result of @code{get_frame_size ()} and the tables of
3395 registers @code{regs_ever_live} and @code{call_used_regs}.
3397 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3398 need not be defined. Otherwise, it must be defined even if
3399 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3400 case, you may set @var{depth-var} to anything.
3403 @defmac ELIMINABLE_REGS
3404 If defined, this macro specifies a table of register pairs used to
3405 eliminate unneeded registers that point into the stack frame. If it is not
3406 defined, the only elimination attempted by the compiler is to replace
3407 references to the frame pointer with references to the stack pointer.
3409 The definition of this macro is a list of structure initializations, each
3410 of which specifies an original and replacement register.
3412 On some machines, the position of the argument pointer is not known until
3413 the compilation is completed. In such a case, a separate hard register
3414 must be used for the argument pointer. This register can be eliminated by
3415 replacing it with either the frame pointer or the argument pointer,
3416 depending on whether or not the frame pointer has been eliminated.
3418 In this case, you might specify:
3420 #define ELIMINABLE_REGS \
3421 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3422 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3423 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3426 Note that the elimination of the argument pointer with the stack pointer is
3427 specified first since that is the preferred elimination.
3430 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3431 A C expression that returns nonzero if the compiler is allowed to try
3432 to replace register number @var{from-reg} with register number
3433 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3434 is defined, and will usually be the constant 1, since most of the cases
3435 preventing register elimination are things that the compiler already
3439 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3440 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3441 specifies the initial difference between the specified pair of
3442 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3446 @node Stack Arguments
3447 @subsection Passing Function Arguments on the Stack
3448 @cindex arguments on stack
3449 @cindex stack arguments
3451 The macros in this section control how arguments are passed
3452 on the stack. See the following section for other macros that
3453 control passing certain arguments in registers.
3455 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3456 This target hook returns @code{true} if an argument declared in a
3457 prototype as an integral type smaller than @code{int} should actually be
3458 passed as an @code{int}. In addition to avoiding errors in certain
3459 cases of mismatch, it also makes for better code on certain machines.
3460 The default is to not promote prototypes.
3464 A C expression. If nonzero, push insns will be used to pass
3466 If the target machine does not have a push instruction, set it to zero.
3467 That directs GCC to use an alternate strategy: to
3468 allocate the entire argument block and then store the arguments into
3469 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3472 @defmac PUSH_ARGS_REVERSED
3473 A C expression. If nonzero, function arguments will be evaluated from
3474 last to first, rather than from first to last. If this macro is not
3475 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3476 and args grow in opposite directions, and 0 otherwise.
3479 @defmac PUSH_ROUNDING (@var{npushed})
3480 A C expression that is the number of bytes actually pushed onto the
3481 stack when an instruction attempts to push @var{npushed} bytes.
3483 On some machines, the definition
3486 #define PUSH_ROUNDING(BYTES) (BYTES)
3490 will suffice. But on other machines, instructions that appear
3491 to push one byte actually push two bytes in an attempt to maintain
3492 alignment. Then the definition should be
3495 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3499 @findex current_function_outgoing_args_size
3500 @defmac ACCUMULATE_OUTGOING_ARGS
3501 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3502 will be computed and placed into the variable
3503 @code{current_function_outgoing_args_size}. No space will be pushed
3504 onto the stack for each call; instead, the function prologue should
3505 increase the stack frame size by this amount.
3507 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3511 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3512 Define this macro if functions should assume that stack space has been
3513 allocated for arguments even when their values are passed in
3516 The value of this macro is the size, in bytes, of the area reserved for
3517 arguments passed in registers for the function represented by @var{fndecl},
3518 which can be zero if GCC is calling a library function.
3520 This space can be allocated by the caller, or be a part of the
3521 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3524 @c above is overfull. not sure what to do. --mew 5feb93 did
3525 @c something, not sure if it looks good. --mew 10feb93
3527 @defmac OUTGOING_REG_PARM_STACK_SPACE
3528 Define this if it is the responsibility of the caller to allocate the area
3529 reserved for arguments passed in registers.
3531 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3532 whether the space for these arguments counts in the value of
3533 @code{current_function_outgoing_args_size}.
3536 @defmac STACK_PARMS_IN_REG_PARM_AREA
3537 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3538 stack parameters don't skip the area specified by it.
3539 @c i changed this, makes more sens and it should have taken care of the
3540 @c overfull.. not as specific, tho. --mew 5feb93
3542 Normally, when a parameter is not passed in registers, it is placed on the
3543 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3544 suppresses this behavior and causes the parameter to be passed on the
3545 stack in its natural location.
3548 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3549 A C expression that should indicate the number of bytes of its own
3550 arguments that a function pops on returning, or 0 if the
3551 function pops no arguments and the caller must therefore pop them all
3552 after the function returns.
3554 @var{fundecl} is a C variable whose value is a tree node that describes
3555 the function in question. Normally it is a node of type
3556 @code{FUNCTION_DECL} that describes the declaration of the function.
3557 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3559 @var{funtype} is a C variable whose value is a tree node that
3560 describes the function in question. Normally it is a node of type
3561 @code{FUNCTION_TYPE} that describes the data type of the function.
3562 From this it is possible to obtain the data types of the value and
3563 arguments (if known).
3565 When a call to a library function is being considered, @var{fundecl}
3566 will contain an identifier node for the library function. Thus, if
3567 you need to distinguish among various library functions, you can do so
3568 by their names. Note that ``library function'' in this context means
3569 a function used to perform arithmetic, whose name is known specially
3570 in the compiler and was not mentioned in the C code being compiled.
3572 @var{stack-size} is the number of bytes of arguments passed on the
3573 stack. If a variable number of bytes is passed, it is zero, and
3574 argument popping will always be the responsibility of the calling function.
3576 On the VAX, all functions always pop their arguments, so the definition
3577 of this macro is @var{stack-size}. On the 68000, using the standard
3578 calling convention, no functions pop their arguments, so the value of
3579 the macro is always 0 in this case. But an alternative calling
3580 convention is available in which functions that take a fixed number of
3581 arguments pop them but other functions (such as @code{printf}) pop
3582 nothing (the caller pops all). When this convention is in use,
3583 @var{funtype} is examined to determine whether a function takes a fixed
3584 number of arguments.
3587 @defmac CALL_POPS_ARGS (@var{cum})
3588 A C expression that should indicate the number of bytes a call sequence
3589 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3590 when compiling a function call.
3592 @var{cum} is the variable in which all arguments to the called function
3593 have been accumulated.
3595 On certain architectures, such as the SH5, a call trampoline is used
3596 that pops certain registers off the stack, depending on the arguments
3597 that have been passed to the function. Since this is a property of the
3598 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3602 @node Register Arguments
3603 @subsection Passing Arguments in Registers
3604 @cindex arguments in registers
3605 @cindex registers arguments
3607 This section describes the macros which let you control how various
3608 types of arguments are passed in registers or how they are arranged in
3611 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3612 A C expression that controls whether a function argument is passed
3613 in a register, and which register.
3615 The arguments are @var{cum}, which summarizes all the previous
3616 arguments; @var{mode}, the machine mode of the argument; @var{type},
3617 the data type of the argument as a tree node or 0 if that is not known
3618 (which happens for C support library functions); and @var{named},
3619 which is 1 for an ordinary argument and 0 for nameless arguments that
3620 correspond to @samp{@dots{}} in the called function's prototype.
3621 @var{type} can be an incomplete type if a syntax error has previously
3624 The value of the expression is usually either a @code{reg} RTX for the
3625 hard register in which to pass the argument, or zero to pass the
3626 argument on the stack.
3628 For machines like the VAX and 68000, where normally all arguments are
3629 pushed, zero suffices as a definition.
3631 The value of the expression can also be a @code{parallel} RTX@. This is
3632 used when an argument is passed in multiple locations. The mode of the
3633 @code{parallel} should be the mode of the entire argument. The
3634 @code{parallel} holds any number of @code{expr_list} pairs; each one
3635 describes where part of the argument is passed. In each
3636 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3637 register in which to pass this part of the argument, and the mode of the
3638 register RTX indicates how large this part of the argument is. The
3639 second operand of the @code{expr_list} is a @code{const_int} which gives
3640 the offset in bytes into the entire argument of where this part starts.
3641 As a special exception the first @code{expr_list} in the @code{parallel}
3642 RTX may have a first operand of zero. This indicates that the entire
3643 argument is also stored on the stack.
3645 The last time this macro is called, it is called with @code{MODE ==
3646 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3647 pattern as operands 2 and 3 respectively.
3649 @cindex @file{stdarg.h} and register arguments
3650 The usual way to make the ISO library @file{stdarg.h} work on a machine
3651 where some arguments are usually passed in registers, is to cause
3652 nameless arguments to be passed on the stack instead. This is done
3653 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3655 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3656 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3657 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3658 in the definition of this macro to determine if this argument is of a
3659 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3660 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3661 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3662 defined, the argument will be computed in the stack and then loaded into
3666 @defmac MUST_PASS_IN_STACK (@var{mode}, @var{type})
3667 Define as a C expression that evaluates to nonzero if we do not know how
3668 to pass TYPE solely in registers. The file @file{expr.h} defines a
3669 definition that is usually appropriate, refer to @file{expr.h} for additional
3673 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3674 Define this macro if the target machine has ``register windows'', so
3675 that the register in which a function sees an arguments is not
3676 necessarily the same as the one in which the caller passed the
3679 For such machines, @code{FUNCTION_ARG} computes the register in which
3680 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3681 be defined in a similar fashion to tell the function being called
3682 where the arguments will arrive.
3684 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3685 serves both purposes.
3688 @defmac FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3689 A C expression for the number of words, at the beginning of an
3690 argument, that must be put in registers. The value must be zero for
3691 arguments that are passed entirely in registers or that are entirely
3692 pushed on the stack.
3694 On some machines, certain arguments must be passed partially in
3695 registers and partially in memory. On these machines, typically the
3696 first @var{n} words of arguments are passed in registers, and the rest
3697 on the stack. If a multi-word argument (a @code{double} or a
3698 structure) crosses that boundary, its first few words must be passed
3699 in registers and the rest must be pushed. This macro tells the
3700 compiler when this occurs, and how many of the words should go in
3703 @code{FUNCTION_ARG} for these arguments should return the first
3704 register to be used by the caller for this argument; likewise
3705 @code{FUNCTION_INCOMING_ARG}, for the called function.
3708 @defmac FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3709 A C expression that indicates when an argument must be passed by reference.
3710 If nonzero for an argument, a copy of that argument is made in memory and a
3711 pointer to the argument is passed instead of the argument itself.
3712 The pointer is passed in whatever way is appropriate for passing a pointer
3715 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3716 definition of this macro might be
3718 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3719 (CUM, MODE, TYPE, NAMED) \
3720 MUST_PASS_IN_STACK (MODE, TYPE)
3722 @c this is *still* too long. --mew 5feb93
3725 @defmac FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3726 If defined, a C expression that indicates when it is the called function's
3727 responsibility to make a copy of arguments passed by invisible reference.
3728 Normally, the caller makes a copy and passes the address of the copy to the
3729 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3730 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3731 ``live'' value. The called function must not modify this value. If it can be
3732 determined that the value won't be modified, it need not make a copy;
3733 otherwise a copy must be made.
3736 @defmac CUMULATIVE_ARGS
3737 A C type for declaring a variable that is used as the first argument of
3738 @code{FUNCTION_ARG} and other related values. For some target machines,
3739 the type @code{int} suffices and can hold the number of bytes of
3742 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3743 arguments that have been passed on the stack. The compiler has other
3744 variables to keep track of that. For target machines on which all
3745 arguments are passed on the stack, there is no need to store anything in
3746 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3747 should not be empty, so use @code{int}.
3750 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3751 A C statement (sans semicolon) for initializing the variable
3752 @var{cum} for the state at the beginning of the argument list. The
3753 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3754 is the tree node for the data type of the function which will receive
3755 the args, or 0 if the args are to a compiler support library function.
3756 For direct calls that are not libcalls, @var{fndecl} contain the
3757 declaration node of the function. @var{fndecl} is also set when
3758 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3759 being compiled. @var{n_named_args} is set to the number of named
3760 arguments, including a structure return address if it is passed as a
3761 parameter, when making a call. When processing incoming arguments,
3762 @var{n_named_args} is set to -1.
3764 When processing a call to a compiler support library function,
3765 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3766 contains the name of the function, as a string. @var{libname} is 0 when
3767 an ordinary C function call is being processed. Thus, each time this
3768 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3769 never both of them at once.
3772 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3773 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3774 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3775 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3776 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3777 0)} is used instead.
3780 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3781 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3782 finding the arguments for the function being compiled. If this macro is
3783 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3785 The value passed for @var{libname} is always 0, since library routines
3786 with special calling conventions are never compiled with GCC@. The
3787 argument @var{libname} exists for symmetry with
3788 @code{INIT_CUMULATIVE_ARGS}.
3789 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3790 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3793 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3794 A C statement (sans semicolon) to update the summarizer variable
3795 @var{cum} to advance past an argument in the argument list. The
3796 values @var{mode}, @var{type} and @var{named} describe that argument.
3797 Once this is done, the variable @var{cum} is suitable for analyzing
3798 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3800 This macro need not do anything if the argument in question was passed
3801 on the stack. The compiler knows how to track the amount of stack space
3802 used for arguments without any special help.
3805 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3806 If defined, a C expression which determines whether, and in which direction,
3807 to pad out an argument with extra space. The value should be of type
3808 @code{enum direction}: either @code{upward} to pad above the argument,
3809 @code{downward} to pad below, or @code{none} to inhibit padding.
3811 The @emph{amount} of padding is always just enough to reach the next
3812 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3815 This macro has a default definition which is right for most systems.
3816 For little-endian machines, the default is to pad upward. For
3817 big-endian machines, the default is to pad downward for an argument of
3818 constant size shorter than an @code{int}, and upward otherwise.
3821 @defmac PAD_VARARGS_DOWN
3822 If defined, a C expression which determines whether the default
3823 implementation of va_arg will attempt to pad down before reading the
3824 next argument, if that argument is smaller than its aligned space as
3825 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3826 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3829 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3830 Specify padding for the last element of a block move between registers and
3831 memory. @var{first} is nonzero if this is the only element. Defining this
3832 macro allows better control of register function parameters on big-endian
3833 machines, without using @code{PARALLEL} rtl. In particular,
3834 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3835 registers, as there is no longer a "wrong" part of a register; For example,
3836 a three byte aggregate may be passed in the high part of a register if so
3840 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3841 If defined, a C expression that gives the alignment boundary, in bits,
3842 of an argument with the specified mode and type. If it is not defined,
3843 @code{PARM_BOUNDARY} is used for all arguments.
3846 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3847 A C expression that is nonzero if @var{regno} is the number of a hard
3848 register in which function arguments are sometimes passed. This does
3849 @emph{not} include implicit arguments such as the static chain and
3850 the structure-value address. On many machines, no registers can be
3851 used for this purpose since all function arguments are pushed on the
3855 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
3856 This hook should return true if parameter of type @var{type} are passed
3857 as two scalar parameters. By default, GCC will attempt to pack complex
3858 arguments into the target's word size. Some ABIs require complex arguments
3859 to be split and treated as their individual components. For example, on
3860 AIX64, complex floats should be passed in a pair of floating point
3861 registers, even though a complex float would fit in one 64-bit floating
3864 The default value of this hook is @code{NULL}, which is treated as always
3868 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
3869 This hook performs target-specific gimplification of
3870 @code{VA_ARG_EXPR}. The first two parameters correspond to the
3871 arguments to @code{va_arg}; the latter two are as in
3872 @code{gimplify.c:gimplify_expr}.
3874 You only need to define this hook if you previously defined
3875 @code{EXPAND_BUILTIN_VA_ARG}; it is pretty easy to reuse the same code
3876 for both. One significant difference is that
3877 @code{EXPAND_BUILTIN_VA_ARG} returns an address, whereas this hook
3878 produces an expression of type @var{type}, usually an @code{INDIRECT_REF}.
3880 Once you define this macro, you can change
3881 @code{EXPAND_BUILTIN_VA_ARG} to just abort, as it should never be
3886 @subsection How Scalar Function Values Are Returned
3887 @cindex return values in registers
3888 @cindex values, returned by functions
3889 @cindex scalars, returned as values
3891 This section discusses the macros that control returning scalars as
3892 values---values that can fit in registers.
3894 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3895 A C expression to create an RTX representing the place where a
3896 function returns a value of data type @var{valtype}. @var{valtype} is
3897 a tree node representing a data type. Write @code{TYPE_MODE
3898 (@var{valtype})} to get the machine mode used to represent that type.
3899 On many machines, only the mode is relevant. (Actually, on most
3900 machines, scalar values are returned in the same place regardless of
3903 The value of the expression is usually a @code{reg} RTX for the hard
3904 register where the return value is stored. The value can also be a
3905 @code{parallel} RTX, if the return value is in multiple places. See
3906 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3908 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply the same
3909 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3912 If the precise function being called is known, @var{func} is a tree
3913 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3914 pointer. This makes it possible to use a different value-returning
3915 convention for specific functions when all their calls are
3918 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3919 types, because these are returned in another way. See
3920 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
3923 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3924 Define this macro if the target machine has ``register windows''
3925 so that the register in which a function returns its value is not
3926 the same as the one in which the caller sees the value.
3928 For such machines, @code{FUNCTION_VALUE} computes the register in which
3929 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3930 defined in a similar fashion to tell the function where to put the
3933 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3934 @code{FUNCTION_VALUE} serves both purposes.
3936 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3937 aggregate data types, because these are returned in another way. See
3938 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
3941 @defmac LIBCALL_VALUE (@var{mode})
3942 A C expression to create an RTX representing the place where a library
3943 function returns a value of mode @var{mode}. If the precise function
3944 being called is known, @var{func} is a tree node
3945 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3946 pointer. This makes it possible to use a different value-returning
3947 convention for specific functions when all their calls are
3950 Note that ``library function'' in this context means a compiler
3951 support routine, used to perform arithmetic, whose name is known
3952 specially by the compiler and was not mentioned in the C code being
3955 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3956 data types, because none of the library functions returns such types.
3959 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3960 A C expression that is nonzero if @var{regno} is the number of a hard
3961 register in which the values of called function may come back.
3963 A register whose use for returning values is limited to serving as the
3964 second of a pair (for a value of type @code{double}, say) need not be
3965 recognized by this macro. So for most machines, this definition
3969 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3972 If the machine has register windows, so that the caller and the called
3973 function use different registers for the return value, this macro
3974 should recognize only the caller's register numbers.
3977 @defmac APPLY_RESULT_SIZE
3978 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3979 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3980 saving and restoring an arbitrary return value.
3983 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
3984 This hook should return true if values of type @var{type} are returned
3985 at the most significant end of a register (in other words, if they are
3986 padded at the least significant end). You can assume that @var{type}
3987 is returned in a register; the caller is required to check this.
3989 Note that the register provided by @code{FUNCTION_VALUE} must be able
3990 to hold the complete return value. For example, if a 1-, 2- or 3-byte
3991 structure is returned at the most significant end of a 4-byte register,
3992 @code{FUNCTION_VALUE} should provide an @code{SImode} rtx.
3995 @node Aggregate Return
3996 @subsection How Large Values Are Returned
3997 @cindex aggregates as return values
3998 @cindex large return values
3999 @cindex returning aggregate values
4000 @cindex structure value address
4002 When a function value's mode is @code{BLKmode} (and in some other
4003 cases), the value is not returned according to @code{FUNCTION_VALUE}
4004 (@pxref{Scalar Return}). Instead, the caller passes the address of a
4005 block of memory in which the value should be stored. This address
4006 is called the @dfn{structure value address}.
4008 This section describes how to control returning structure values in
4011 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
4012 This target hook should return a nonzero value to say to return the
4013 function value in memory, just as large structures are always returned.
4014 Here @var{type} will be the data type of the value, and @var{fntype}
4015 will be the type of the function doing the returning, or @code{NULL} for
4018 Note that values of mode @code{BLKmode} must be explicitly handled
4019 by this function. Also, the option @option{-fpcc-struct-return}
4020 takes effect regardless of this macro. On most systems, it is
4021 possible to leave the hook undefined; this causes a default
4022 definition to be used, whose value is the constant 1 for @code{BLKmode}
4023 values, and 0 otherwise.
4025 Do not use this hook to indicate that structures and unions should always
4026 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4030 @defmac DEFAULT_PCC_STRUCT_RETURN
4031 Define this macro to be 1 if all structure and union return values must be
4032 in memory. Since this results in slower code, this should be defined
4033 only if needed for compatibility with other compilers or with an ABI@.
4034 If you define this macro to be 0, then the conventions used for structure
4035 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4038 If not defined, this defaults to the value 1.
4041 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4042 This target hook should return the location of the structure value
4043 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4044 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4045 be @code{NULL}, for libcalls. You do not need to define this target
4046 hook if the address is always passed as an ``invisible'' first
4049 On some architectures the place where the structure value address
4050 is found by the called function is not the same place that the
4051 caller put it. This can be due to register windows, or it could
4052 be because the function prologue moves it to a different place.
4053 @var{incoming} is @code{true} when the location is needed in
4054 the context of the called function, and @code{false} in the context of
4057 If @var{incoming} is @code{true} and the address is to be found on the
4058 stack, return a @code{mem} which refers to the frame pointer.
4061 @defmac PCC_STATIC_STRUCT_RETURN
4062 Define this macro if the usual system convention on the target machine
4063 for returning structures and unions is for the called function to return
4064 the address of a static variable containing the value.
4066 Do not define this if the usual system convention is for the caller to
4067 pass an address to the subroutine.
4069 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4070 nothing when you use @option{-freg-struct-return} mode.
4074 @subsection Caller-Saves Register Allocation
4076 If you enable it, GCC can save registers around function calls. This
4077 makes it possible to use call-clobbered registers to hold variables that
4078 must live across calls.
4080 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4081 A C expression to determine whether it is worthwhile to consider placing
4082 a pseudo-register in a call-clobbered hard register and saving and
4083 restoring it around each function call. The expression should be 1 when
4084 this is worth doing, and 0 otherwise.
4086 If you don't define this macro, a default is used which is good on most
4087 machines: @code{4 * @var{calls} < @var{refs}}.
4090 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4091 A C expression specifying which mode is required for saving @var{nregs}
4092 of a pseudo-register in call-clobbered hard register @var{regno}. If
4093 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4094 returned. For most machines this macro need not be defined since GCC
4095 will select the smallest suitable mode.
4098 @node Function Entry
4099 @subsection Function Entry and Exit
4100 @cindex function entry and exit
4104 This section describes the macros that output function entry
4105 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4107 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4108 If defined, a function that outputs the assembler code for entry to a
4109 function. The prologue is responsible for setting up the stack frame,
4110 initializing the frame pointer register, saving registers that must be
4111 saved, and allocating @var{size} additional bytes of storage for the
4112 local variables. @var{size} is an integer. @var{file} is a stdio
4113 stream to which the assembler code should be output.
4115 The label for the beginning of the function need not be output by this
4116 macro. That has already been done when the macro is run.
4118 @findex regs_ever_live
4119 To determine which registers to save, the macro can refer to the array
4120 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4121 @var{r} is used anywhere within the function. This implies the function
4122 prologue should save register @var{r}, provided it is not one of the
4123 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4124 @code{regs_ever_live}.)
4126 On machines that have ``register windows'', the function entry code does
4127 not save on the stack the registers that are in the windows, even if
4128 they are supposed to be preserved by function calls; instead it takes
4129 appropriate steps to ``push'' the register stack, if any non-call-used
4130 registers are used in the function.
4132 @findex frame_pointer_needed
4133 On machines where functions may or may not have frame-pointers, the
4134 function entry code must vary accordingly; it must set up the frame
4135 pointer if one is wanted, and not otherwise. To determine whether a
4136 frame pointer is in wanted, the macro can refer to the variable
4137 @code{frame_pointer_needed}. The variable's value will be 1 at run
4138 time in a function that needs a frame pointer. @xref{Elimination}.
4140 The function entry code is responsible for allocating any stack space
4141 required for the function. This stack space consists of the regions
4142 listed below. In most cases, these regions are allocated in the
4143 order listed, with the last listed region closest to the top of the
4144 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4145 the highest address if it is not defined). You can use a different order
4146 for a machine if doing so is more convenient or required for
4147 compatibility reasons. Except in cases where required by standard
4148 or by a debugger, there is no reason why the stack layout used by GCC
4149 need agree with that used by other compilers for a machine.
4152 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4153 If defined, a function that outputs assembler code at the end of a
4154 prologue. This should be used when the function prologue is being
4155 emitted as RTL, and you have some extra assembler that needs to be
4156 emitted. @xref{prologue instruction pattern}.
4159 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4160 If defined, a function that outputs assembler code at the start of an
4161 epilogue. This should be used when the function epilogue is being
4162 emitted as RTL, and you have some extra assembler that needs to be
4163 emitted. @xref{epilogue instruction pattern}.
4166 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4167 If defined, a function that outputs the assembler code for exit from a
4168 function. The epilogue is responsible for restoring the saved
4169 registers and stack pointer to their values when the function was
4170 called, and returning control to the caller. This macro takes the
4171 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4172 registers to restore are determined from @code{regs_ever_live} and
4173 @code{CALL_USED_REGISTERS} in the same way.
4175 On some machines, there is a single instruction that does all the work
4176 of returning from the function. On these machines, give that
4177 instruction the name @samp{return} and do not define the macro
4178 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4180 Do not define a pattern named @samp{return} if you want the
4181 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4182 switches to control whether return instructions or epilogues are used,
4183 define a @samp{return} pattern with a validity condition that tests the
4184 target switches appropriately. If the @samp{return} pattern's validity
4185 condition is false, epilogues will be used.
4187 On machines where functions may or may not have frame-pointers, the
4188 function exit code must vary accordingly. Sometimes the code for these
4189 two cases is completely different. To determine whether a frame pointer
4190 is wanted, the macro can refer to the variable
4191 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4192 a function that needs a frame pointer.
4194 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4195 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4196 The C variable @code{current_function_is_leaf} is nonzero for such a
4197 function. @xref{Leaf Functions}.
4199 On some machines, some functions pop their arguments on exit while
4200 others leave that for the caller to do. For example, the 68020 when
4201 given @option{-mrtd} pops arguments in functions that take a fixed
4202 number of arguments.
4204 @findex current_function_pops_args
4205 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4206 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4207 needs to know what was decided. The variable that is called
4208 @code{current_function_pops_args} is the number of bytes of its
4209 arguments that a function should pop. @xref{Scalar Return}.
4210 @c what is the "its arguments" in the above sentence referring to, pray
4211 @c tell? --mew 5feb93
4216 @findex current_function_pretend_args_size
4217 A region of @code{current_function_pretend_args_size} bytes of
4218 uninitialized space just underneath the first argument arriving on the
4219 stack. (This may not be at the very start of the allocated stack region
4220 if the calling sequence has pushed anything else since pushing the stack
4221 arguments. But usually, on such machines, nothing else has been pushed
4222 yet, because the function prologue itself does all the pushing.) This
4223 region is used on machines where an argument may be passed partly in
4224 registers and partly in memory, and, in some cases to support the
4225 features in @code{<stdarg.h>}.
4228 An area of memory used to save certain registers used by the function.
4229 The size of this area, which may also include space for such things as
4230 the return address and pointers to previous stack frames, is
4231 machine-specific and usually depends on which registers have been used
4232 in the function. Machines with register windows often do not require
4236 A region of at least @var{size} bytes, possibly rounded up to an allocation
4237 boundary, to contain the local variables of the function. On some machines,
4238 this region and the save area may occur in the opposite order, with the
4239 save area closer to the top of the stack.
4242 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4243 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4244 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4245 argument lists of the function. @xref{Stack Arguments}.
4248 Normally, it is necessary for the macros
4249 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4250 @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially.
4251 The C variable @code{current_function_is_leaf} is nonzero for such a
4254 @defmac EXIT_IGNORE_STACK
4255 Define this macro as a C expression that is nonzero if the return
4256 instruction or the function epilogue ignores the value of the stack
4257 pointer; in other words, if it is safe to delete an instruction to
4258 adjust the stack pointer before a return from the function. The
4261 Note that this macro's value is relevant only for functions for which
4262 frame pointers are maintained. It is never safe to delete a final
4263 stack adjustment in a function that has no frame pointer, and the
4264 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4267 @defmac EPILOGUE_USES (@var{regno})
4268 Define this macro as a C expression that is nonzero for registers that are
4269 used by the epilogue or the @samp{return} pattern. The stack and frame
4270 pointer registers are already be assumed to be used as needed.
4273 @defmac EH_USES (@var{regno})
4274 Define this macro as a C expression that is nonzero for registers that are
4275 used by the exception handling mechanism, and so should be considered live
4276 on entry to an exception edge.
4279 @defmac DELAY_SLOTS_FOR_EPILOGUE
4280 Define this macro if the function epilogue contains delay slots to which
4281 instructions from the rest of the function can be ``moved''. The
4282 definition should be a C expression whose value is an integer
4283 representing the number of delay slots there.
4286 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4287 A C expression that returns 1 if @var{insn} can be placed in delay
4288 slot number @var{n} of the epilogue.
4290 The argument @var{n} is an integer which identifies the delay slot now
4291 being considered (since different slots may have different rules of
4292 eligibility). It is never negative and is always less than the number
4293 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4294 If you reject a particular insn for a given delay slot, in principle, it
4295 may be reconsidered for a subsequent delay slot. Also, other insns may
4296 (at least in principle) be considered for the so far unfilled delay
4299 @findex current_function_epilogue_delay_list
4300 @findex final_scan_insn
4301 The insns accepted to fill the epilogue delay slots are put in an RTL
4302 list made with @code{insn_list} objects, stored in the variable
4303 @code{current_function_epilogue_delay_list}. The insn for the first
4304 delay slot comes first in the list. Your definition of the macro
4305 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4306 outputting the insns in this list, usually by calling
4307 @code{final_scan_insn}.
4309 You need not define this macro if you did not define
4310 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4313 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, tree @var{function})
4314 A function that outputs the assembler code for a thunk
4315 function, used to implement C++ virtual function calls with multiple
4316 inheritance. The thunk acts as a wrapper around a virtual function,
4317 adjusting the implicit object parameter before handing control off to
4320 First, emit code to add the integer @var{delta} to the location that
4321 contains the incoming first argument. Assume that this argument
4322 contains a pointer, and is the one used to pass the @code{this} pointer
4323 in C++. This is the incoming argument @emph{before} the function prologue,
4324 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4325 all other incoming arguments.
4327 After the addition, emit code to jump to @var{function}, which is a
4328 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4329 not touch the return address. Hence returning from @var{FUNCTION} will
4330 return to whoever called the current @samp{thunk}.
4332 The effect must be as if @var{function} had been called directly with
4333 the adjusted first argument. This macro is responsible for emitting all
4334 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4335 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4337 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4338 have already been extracted from it.) It might possibly be useful on
4339 some targets, but probably not.
4341 If you do not define this macro, the target-independent code in the C++
4342 front end will generate a less efficient heavyweight thunk that calls
4343 @var{function} instead of jumping to it. The generic approach does
4344 not support varargs.
4347 @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})
4348 A function like @code{TARGET_ASM_OUTPUT_MI_THUNK}, except that if
4349 @var{vcall_offset} is nonzero, an additional adjustment should be made
4350 after adding @code{delta}. In particular, if @var{p} is the
4351 adjusted pointer, the following adjustment should be made:
4354 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4358 If this function is defined, it will always be used in place of
4359 @code{TARGET_ASM_OUTPUT_MI_THUNK}.
4363 @subsection Generating Code for Profiling
4364 @cindex profiling, code generation
4366 These macros will help you generate code for profiling.
4368 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4369 A C statement or compound statement to output to @var{file} some
4370 assembler code to call the profiling subroutine @code{mcount}.
4373 The details of how @code{mcount} expects to be called are determined by
4374 your operating system environment, not by GCC@. To figure them out,
4375 compile a small program for profiling using the system's installed C
4376 compiler and look at the assembler code that results.
4378 Older implementations of @code{mcount} expect the address of a counter
4379 variable to be loaded into some register. The name of this variable is
4380 @samp{LP} followed by the number @var{labelno}, so you would generate
4381 the name using @samp{LP%d} in a @code{fprintf}.
4384 @defmac PROFILE_HOOK
4385 A C statement or compound statement to output to @var{file} some assembly
4386 code to call the profiling subroutine @code{mcount} even the target does
4387 not support profiling.
4390 @defmac NO_PROFILE_COUNTERS
4391 Define this macro if the @code{mcount} subroutine on your system does
4392 not need a counter variable allocated for each function. This is true
4393 for almost all modern implementations. If you define this macro, you
4394 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4397 @defmac PROFILE_BEFORE_PROLOGUE
4398 Define this macro if the code for function profiling should come before
4399 the function prologue. Normally, the profiling code comes after.
4403 @subsection Permitting tail calls
4406 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4407 True if it is ok to do sibling call optimization for the specified
4408 call expression @var{exp}. @var{decl} will be the called function,
4409 or @code{NULL} if this is an indirect call.
4411 It is not uncommon for limitations of calling conventions to prevent
4412 tail calls to functions outside the current unit of translation, or
4413 during PIC compilation. The hook is used to enforce these restrictions,
4414 as the @code{sibcall} md pattern can not fail, or fall over to a
4415 ``normal'' call. The criteria for successful sibling call optimization
4416 may vary greatly between different architectures.
4420 @section Implementing the Varargs Macros
4421 @cindex varargs implementation
4423 GCC comes with an implementation of @code{<varargs.h>} and
4424 @code{<stdarg.h>} that work without change on machines that pass arguments
4425 on the stack. Other machines require their own implementations of
4426 varargs, and the two machine independent header files must have
4427 conditionals to include it.
4429 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4430 the calling convention for @code{va_start}. The traditional
4431 implementation takes just one argument, which is the variable in which
4432 to store the argument pointer. The ISO implementation of
4433 @code{va_start} takes an additional second argument. The user is
4434 supposed to write the last named argument of the function here.
4436 However, @code{va_start} should not use this argument. The way to find
4437 the end of the named arguments is with the built-in functions described
4440 @defmac __builtin_saveregs ()
4441 Use this built-in function to save the argument registers in memory so
4442 that the varargs mechanism can access them. Both ISO and traditional
4443 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4444 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4446 On some machines, @code{__builtin_saveregs} is open-coded under the
4447 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4448 other machines, it calls a routine written in assembler language,
4449 found in @file{libgcc2.c}.
4451 Code generated for the call to @code{__builtin_saveregs} appears at the
4452 beginning of the function, as opposed to where the call to
4453 @code{__builtin_saveregs} is written, regardless of what the code is.
4454 This is because the registers must be saved before the function starts
4455 to use them for its own purposes.
4456 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4460 @defmac __builtin_args_info (@var{category})
4461 Use this built-in function to find the first anonymous arguments in
4464 In general, a machine may have several categories of registers used for
4465 arguments, each for a particular category of data types. (For example,
4466 on some machines, floating-point registers are used for floating-point
4467 arguments while other arguments are passed in the general registers.)
4468 To make non-varargs functions use the proper calling convention, you
4469 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4470 registers in each category have been used so far
4472 @code{__builtin_args_info} accesses the same data structure of type
4473 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4474 with it, with @var{category} specifying which word to access. Thus, the
4475 value indicates the first unused register in a given category.
4477 Normally, you would use @code{__builtin_args_info} in the implementation
4478 of @code{va_start}, accessing each category just once and storing the
4479 value in the @code{va_list} object. This is because @code{va_list} will
4480 have to update the values, and there is no way to alter the
4481 values accessed by @code{__builtin_args_info}.
4484 @defmac __builtin_next_arg (@var{lastarg})
4485 This is the equivalent of @code{__builtin_args_info}, for stack
4486 arguments. It returns the address of the first anonymous stack
4487 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4488 returns the address of the location above the first anonymous stack
4489 argument. Use it in @code{va_start} to initialize the pointer for
4490 fetching arguments from the stack. Also use it in @code{va_start} to
4491 verify that the second parameter @var{lastarg} is the last named argument
4492 of the current function.
4495 @defmac __builtin_classify_type (@var{object})
4496 Since each machine has its own conventions for which data types are
4497 passed in which kind of register, your implementation of @code{va_arg}
4498 has to embody these conventions. The easiest way to categorize the
4499 specified data type is to use @code{__builtin_classify_type} together
4500 with @code{sizeof} and @code{__alignof__}.
4502 @code{__builtin_classify_type} ignores the value of @var{object},
4503 considering only its data type. It returns an integer describing what
4504 kind of type that is---integer, floating, pointer, structure, and so on.
4506 The file @file{typeclass.h} defines an enumeration that you can use to
4507 interpret the values of @code{__builtin_classify_type}.
4510 These machine description macros help implement varargs:
4512 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4513 If defined, this hook produces the machine-specific code for a call to
4514 @code{__builtin_saveregs}. This code will be moved to the very
4515 beginning of the function, before any parameter access are made. The
4516 return value of this function should be an RTX that contains the value
4517 to use as the return of @code{__builtin_saveregs}.
4520 @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})
4521 This target hook offers an alternative to using
4522 @code{__builtin_saveregs} and defining the hook
4523 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4524 register arguments into the stack so that all the arguments appear to
4525 have been passed consecutively on the stack. Once this is done, you can
4526 use the standard implementation of varargs that works for machines that
4527 pass all their arguments on the stack.
4529 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4530 structure, containing the values that are obtained after processing the
4531 named arguments. The arguments @var{mode} and @var{type} describe the
4532 last named argument---its machine mode and its data type as a tree node.
4534 The target hook should do two things: first, push onto the stack all the
4535 argument registers @emph{not} used for the named arguments, and second,
4536 store the size of the data thus pushed into the @code{int}-valued
4537 variable pointed to by @var{pretend_args_size}. The value that you
4538 store here will serve as additional offset for setting up the stack
4541 Because you must generate code to push the anonymous arguments at
4542 compile time without knowing their data types,
4543 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4544 have just a single category of argument register and use it uniformly
4547 If the argument @var{second_time} is nonzero, it means that the
4548 arguments of the function are being analyzed for the second time. This
4549 happens for an inline function, which is not actually compiled until the
4550 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4551 not generate any instructions in this case.
4554 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4555 Define this hook to return @code{true} if the location where a function
4556 argument is passed depends on whether or not it is a named argument.
4558 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4559 is set for varargs and stdarg functions. If this hook returns
4560 @code{true}, the @var{named} argument is always true for named
4561 arguments, and false for unnamed arguments. If it returns @code{false},
4562 but @code{TARGET_PRETEND_OUTOGOING_VARARGS_NAMED} returns @code{true},
4563 then all arguments are treated as named. Otherwise, all named arguments
4564 except the last are treated as named.
4566 You need not define this hook if it always returns zero.
4569 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4570 If you need to conditionally change ABIs so that one works with
4571 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4572 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4573 defined, then define this hook to return @code{true} if
4574 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4575 Otherwise, you should not define this hook.
4579 @section Trampolines for Nested Functions
4580 @cindex trampolines for nested functions
4581 @cindex nested functions, trampolines for
4583 A @dfn{trampoline} is a small piece of code that is created at run time
4584 when the address of a nested function is taken. It normally resides on
4585 the stack, in the stack frame of the containing function. These macros
4586 tell GCC how to generate code to allocate and initialize a
4589 The instructions in the trampoline must do two things: load a constant
4590 address into the static chain register, and jump to the real address of
4591 the nested function. On CISC machines such as the m68k, this requires
4592 two instructions, a move immediate and a jump. Then the two addresses
4593 exist in the trampoline as word-long immediate operands. On RISC
4594 machines, it is often necessary to load each address into a register in
4595 two parts. Then pieces of each address form separate immediate
4598 The code generated to initialize the trampoline must store the variable
4599 parts---the static chain value and the function address---into the
4600 immediate operands of the instructions. On a CISC machine, this is
4601 simply a matter of copying each address to a memory reference at the
4602 proper offset from the start of the trampoline. On a RISC machine, it
4603 may be necessary to take out pieces of the address and store them
4606 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4607 A C statement to output, on the stream @var{file}, assembler code for a
4608 block of data that contains the constant parts of a trampoline. This
4609 code should not include a label---the label is taken care of
4612 If you do not define this macro, it means no template is needed
4613 for the target. Do not define this macro on systems where the block move
4614 code to copy the trampoline into place would be larger than the code
4615 to generate it on the spot.
4618 @defmac TRAMPOLINE_SECTION
4619 The name of a subroutine to switch to the section in which the
4620 trampoline template is to be placed (@pxref{Sections}). The default is
4621 a value of @samp{readonly_data_section}, which places the trampoline in
4622 the section containing read-only data.
4625 @defmac TRAMPOLINE_SIZE
4626 A C expression for the size in bytes of the trampoline, as an integer.
4629 @defmac TRAMPOLINE_ALIGNMENT
4630 Alignment required for trampolines, in bits.
4632 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4633 is used for aligning trampolines.
4636 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4637 A C statement to initialize the variable parts of a trampoline.
4638 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4639 an RTX for the address of the nested function; @var{static_chain} is an
4640 RTX for the static chain value that should be passed to the function
4644 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4645 A C statement that should perform any machine-specific adjustment in
4646 the address of the trampoline. Its argument contains the address that
4647 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4648 used for a function call should be different from the address in which
4649 the template was stored, the different address should be assigned to
4650 @var{addr}. If this macro is not defined, @var{addr} will be used for
4653 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4654 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4655 If this macro is not defined, by default the trampoline is allocated as
4656 a stack slot. This default is right for most machines. The exceptions
4657 are machines where it is impossible to execute instructions in the stack
4658 area. On such machines, you may have to implement a separate stack,
4659 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4660 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4662 @var{fp} points to a data structure, a @code{struct function}, which
4663 describes the compilation status of the immediate containing function of
4664 the function which the trampoline is for. The stack slot for the
4665 trampoline is in the stack frame of this containing function. Other
4666 allocation strategies probably must do something analogous with this
4670 Implementing trampolines is difficult on many machines because they have
4671 separate instruction and data caches. Writing into a stack location
4672 fails to clear the memory in the instruction cache, so when the program
4673 jumps to that location, it executes the old contents.
4675 Here are two possible solutions. One is to clear the relevant parts of
4676 the instruction cache whenever a trampoline is set up. The other is to
4677 make all trampolines identical, by having them jump to a standard
4678 subroutine. The former technique makes trampoline execution faster; the
4679 latter makes initialization faster.
4681 To clear the instruction cache when a trampoline is initialized, define
4682 the following macro.
4684 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4685 If defined, expands to a C expression clearing the @emph{instruction
4686 cache} in the specified interval. The definition of this macro would
4687 typically be a series of @code{asm} statements. Both @var{beg} and
4688 @var{end} are both pointer expressions.
4691 To use a standard subroutine, define the following macro. In addition,
4692 you must make sure that the instructions in a trampoline fill an entire
4693 cache line with identical instructions, or else ensure that the
4694 beginning of the trampoline code is always aligned at the same point in
4695 its cache line. Look in @file{m68k.h} as a guide.
4697 @defmac TRANSFER_FROM_TRAMPOLINE
4698 Define this macro if trampolines need a special subroutine to do their
4699 work. The macro should expand to a series of @code{asm} statements
4700 which will be compiled with GCC@. They go in a library function named
4701 @code{__transfer_from_trampoline}.
4703 If you need to avoid executing the ordinary prologue code of a compiled
4704 C function when you jump to the subroutine, you can do so by placing a
4705 special label of your own in the assembler code. Use one @code{asm}
4706 statement to generate an assembler label, and another to make the label
4707 global. Then trampolines can use that label to jump directly to your
4708 special assembler code.
4712 @section Implicit Calls to Library Routines
4713 @cindex library subroutine names
4714 @cindex @file{libgcc.a}
4716 @c prevent bad page break with this line
4717 Here is an explanation of implicit calls to library routines.
4719 @defmac DECLARE_LIBRARY_RENAMES
4720 This macro, if defined, should expand to a piece of C code that will get
4721 expanded when compiling functions for libgcc.a. It can be used to
4722 provide alternate names for GCC's internal library functions if there
4723 are ABI-mandated names that the compiler should provide.
4726 @findex init_one_libfunc
4727 @findex set_optab_libfunc
4728 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
4729 This hook should declare additional library routines or rename
4730 existing ones, using the functions @code{set_optab_libfunc} and
4731 @code{init_one_libfunc} defined in @file{optabs.c}.
4732 @code{init_optabs} calls this macro after initializing all the normal
4735 The default is to do nothing. Most ports don't need to define this hook.
4738 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4739 This macro should return @code{true} if the library routine that
4740 implements the floating point comparison operator @var{comparison} in
4741 mode @var{mode} will return a boolean, and @var{false} if it will
4744 GCC's own floating point libraries return tristates from the
4745 comparison operators, so the default returns false always. Most ports
4746 don't need to define this macro.
4749 @cindex US Software GOFAST, floating point emulation library
4750 @cindex floating point emulation library, US Software GOFAST
4751 @cindex GOFAST, floating point emulation library
4752 @findex gofast_maybe_init_libfuncs
4753 @defmac US_SOFTWARE_GOFAST
4754 Define this macro if your system C library uses the US Software GOFAST
4755 library to provide floating point emulation.
4757 In addition to defining this macro, your architecture must set
4758 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
4759 else call that function from its version of that hook. It is defined
4760 in @file{config/gofast.h}, which must be included by your
4761 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
4764 If this macro is defined, the
4765 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
4766 false for @code{SFmode} and @code{DFmode} comparisons.
4769 @cindex @code{EDOM}, implicit usage
4772 The value of @code{EDOM} on the target machine, as a C integer constant
4773 expression. If you don't define this macro, GCC does not attempt to
4774 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4775 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4778 If you do not define @code{TARGET_EDOM}, then compiled code reports
4779 domain errors by calling the library function and letting it report the
4780 error. If mathematical functions on your system use @code{matherr} when
4781 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4782 that @code{matherr} is used normally.
4785 @cindex @code{errno}, implicit usage
4786 @defmac GEN_ERRNO_RTX
4787 Define this macro as a C expression to create an rtl expression that
4788 refers to the global ``variable'' @code{errno}. (On certain systems,
4789 @code{errno} may not actually be a variable.) If you don't define this
4790 macro, a reasonable default is used.
4793 @cindex @code{bcopy}, implicit usage
4794 @cindex @code{memcpy}, implicit usage
4795 @cindex @code{memmove}, implicit usage
4796 @cindex @code{bzero}, implicit usage
4797 @cindex @code{memset}, implicit usage
4798 @defmac TARGET_MEM_FUNCTIONS
4799 Define this macro if GCC should generate calls to the ISO C
4800 (and System V) library functions @code{memcpy}, @code{memmove} and
4801 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4804 @cindex C99 math functions, implicit usage
4805 @defmac TARGET_C99_FUNCTIONS
4806 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4807 @code{sinf} and similarly for other functions defined by C99 standard. The
4808 default is nonzero that should be proper value for most modern systems, however
4809 number of existing systems lacks support for these functions in the runtime so
4810 they needs this macro to be redefined to 0.
4813 @defmac NEXT_OBJC_RUNTIME
4814 Define this macro to generate code for Objective-C message sending using
4815 the calling convention of the NeXT system. This calling convention
4816 involves passing the object, the selector and the method arguments all
4817 at once to the method-lookup library function.
4819 The default calling convention passes just the object and the selector
4820 to the lookup function, which returns a pointer to the method.
4823 @node Addressing Modes
4824 @section Addressing Modes
4825 @cindex addressing modes
4827 @c prevent bad page break with this line
4828 This is about addressing modes.
4830 @defmac HAVE_PRE_INCREMENT
4831 @defmacx HAVE_PRE_DECREMENT
4832 @defmacx HAVE_POST_INCREMENT
4833 @defmacx HAVE_POST_DECREMENT
4834 A C expression that is nonzero if the machine supports pre-increment,
4835 pre-decrement, post-increment, or post-decrement addressing respectively.
4838 @defmac HAVE_PRE_MODIFY_DISP
4839 @defmacx HAVE_POST_MODIFY_DISP
4840 A C expression that is nonzero if the machine supports pre- or
4841 post-address side-effect generation involving constants other than
4842 the size of the memory operand.
4845 @defmac HAVE_PRE_MODIFY_REG
4846 @defmacx HAVE_POST_MODIFY_REG
4847 A C expression that is nonzero if the machine supports pre- or
4848 post-address side-effect generation involving a register displacement.
4851 @defmac CONSTANT_ADDRESS_P (@var{x})
4852 A C expression that is 1 if the RTX @var{x} is a constant which
4853 is a valid address. On most machines, this can be defined as
4854 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4855 in which constant addresses are supported.
4858 @defmac CONSTANT_P (@var{x})
4859 @code{CONSTANT_P}, which is defined by target-independent code,
4860 accepts integer-values expressions whose values are not explicitly
4861 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4862 expressions and @code{const} arithmetic expressions, in addition to
4863 @code{const_int} and @code{const_double} expressions.
4866 @defmac MAX_REGS_PER_ADDRESS
4867 A number, the maximum number of registers that can appear in a valid
4868 memory address. Note that it is up to you to specify a value equal to
4869 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4873 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4874 A C compound statement with a conditional @code{goto @var{label};}
4875 executed if @var{x} (an RTX) is a legitimate memory address on the
4876 target machine for a memory operand of mode @var{mode}.
4878 It usually pays to define several simpler macros to serve as
4879 subroutines for this one. Otherwise it may be too complicated to
4882 This macro must exist in two variants: a strict variant and a
4883 non-strict one. The strict variant is used in the reload pass. It
4884 must be defined so that any pseudo-register that has not been
4885 allocated a hard register is considered a memory reference. In
4886 contexts where some kind of register is required, a pseudo-register
4887 with no hard register must be rejected.
4889 The non-strict variant is used in other passes. It must be defined to
4890 accept all pseudo-registers in every context where some kind of
4891 register is required.
4893 @findex REG_OK_STRICT
4894 Compiler source files that want to use the strict variant of this
4895 macro define the macro @code{REG_OK_STRICT}. You should use an
4896 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4897 in that case and the non-strict variant otherwise.
4899 Subroutines to check for acceptable registers for various purposes (one
4900 for base registers, one for index registers, and so on) are typically
4901 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4902 Then only these subroutine macros need have two variants; the higher
4903 levels of macros may be the same whether strict or not.
4905 Normally, constant addresses which are the sum of a @code{symbol_ref}
4906 and an integer are stored inside a @code{const} RTX to mark them as
4907 constant. Therefore, there is no need to recognize such sums
4908 specifically as legitimate addresses. Normally you would simply
4909 recognize any @code{const} as legitimate.
4911 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4912 sums that are not marked with @code{const}. It assumes that a naked
4913 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4914 naked constant sums as illegitimate addresses, so that none of them will
4915 be given to @code{PRINT_OPERAND_ADDRESS}.
4917 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4918 On some machines, whether a symbolic address is legitimate depends on
4919 the section that the address refers to. On these machines, define the
4920 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4921 into the @code{symbol_ref}, and then check for it here. When you see a
4922 @code{const}, you will have to look inside it to find the
4923 @code{symbol_ref} in order to determine the section. @xref{Assembler
4927 @defmac REG_OK_FOR_BASE_P (@var{x})
4928 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4929 RTX) is valid for use as a base register. For hard registers, it
4930 should always accept those which the hardware permits and reject the
4931 others. Whether the macro accepts or rejects pseudo registers must be
4932 controlled by @code{REG_OK_STRICT} as described above. This usually
4933 requires two variant definitions, of which @code{REG_OK_STRICT}
4934 controls the one actually used.
4937 @defmac REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4938 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4939 that expression may examine the mode of the memory reference in
4940 @var{mode}. You should define this macro if the mode of the memory
4941 reference affects whether a register may be used as a base register. If
4942 you define this macro, the compiler will use it instead of
4943 @code{REG_OK_FOR_BASE_P}.
4946 @defmac REG_OK_FOR_INDEX_P (@var{x})
4947 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4948 RTX) is valid for use as an index register.
4950 The difference between an index register and a base register is that
4951 the index register may be scaled. If an address involves the sum of
4952 two registers, neither one of them scaled, then either one may be
4953 labeled the ``base'' and the other the ``index''; but whichever
4954 labeling is used must fit the machine's constraints of which registers
4955 may serve in each capacity. The compiler will try both labelings,
4956 looking for one that is valid, and will reload one or both registers
4957 only if neither labeling works.
4960 @defmac FIND_BASE_TERM (@var{x})
4961 A C expression to determine the base term of address @var{x}.
4962 This macro is used in only one place: `find_base_term' in alias.c.
4964 It is always safe for this macro to not be defined. It exists so
4965 that alias analysis can understand machine-dependent addresses.
4967 The typical use of this macro is to handle addresses containing
4968 a label_ref or symbol_ref within an UNSPEC@.
4971 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4972 A C compound statement that attempts to replace @var{x} with a valid
4973 memory address for an operand of mode @var{mode}. @var{win} will be a
4974 C statement label elsewhere in the code; the macro definition may use
4977 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4981 to avoid further processing if the address has become legitimate.
4983 @findex break_out_memory_refs
4984 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4985 and @var{oldx} will be the operand that was given to that function to produce
4988 The code generated by this macro should not alter the substructure of
4989 @var{x}. If it transforms @var{x} into a more legitimate form, it
4990 should assign @var{x} (which will always be a C variable) a new value.
4992 It is not necessary for this macro to come up with a legitimate
4993 address. The compiler has standard ways of doing so in all cases. In
4994 fact, it is safe to omit this macro. But often a
4995 machine-dependent strategy can generate better code.
4998 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4999 A C compound statement that attempts to replace @var{x}, which is an address
5000 that needs reloading, with a valid memory address for an operand of mode
5001 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5002 It is not necessary to define this macro, but it might be useful for
5003 performance reasons.
5005 For example, on the i386, it is sometimes possible to use a single
5006 reload register instead of two by reloading a sum of two pseudo
5007 registers into a register. On the other hand, for number of RISC
5008 processors offsets are limited so that often an intermediate address
5009 needs to be generated in order to address a stack slot. By defining
5010 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5011 generated for adjacent some stack slots can be made identical, and thus
5014 @emph{Note}: This macro should be used with caution. It is necessary
5015 to know something of how reload works in order to effectively use this,
5016 and it is quite easy to produce macros that build in too much knowledge
5017 of reload internals.
5019 @emph{Note}: This macro must be able to reload an address created by a
5020 previous invocation of this macro. If it fails to handle such addresses
5021 then the compiler may generate incorrect code or abort.
5024 The macro definition should use @code{push_reload} to indicate parts that
5025 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5026 suitable to be passed unaltered to @code{push_reload}.
5028 The code generated by this macro must not alter the substructure of
5029 @var{x}. If it transforms @var{x} into a more legitimate form, it
5030 should assign @var{x} (which will always be a C variable) a new value.
5031 This also applies to parts that you change indirectly by calling
5034 @findex strict_memory_address_p
5035 The macro definition may use @code{strict_memory_address_p} to test if
5036 the address has become legitimate.
5039 If you want to change only a part of @var{x}, one standard way of doing
5040 this is to use @code{copy_rtx}. Note, however, that is unshares only a
5041 single level of rtl. Thus, if the part to be changed is not at the
5042 top level, you'll need to replace first the top level.
5043 It is not necessary for this macro to come up with a legitimate
5044 address; but often a machine-dependent strategy can generate better code.
5047 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5048 A C statement or compound statement with a conditional @code{goto
5049 @var{label};} executed if memory address @var{x} (an RTX) can have
5050 different meanings depending on the machine mode of the memory
5051 reference it is used for or if the address is valid for some modes
5054 Autoincrement and autodecrement addresses typically have mode-dependent
5055 effects because the amount of the increment or decrement is the size
5056 of the operand being addressed. Some machines have other mode-dependent
5057 addresses. Many RISC machines have no mode-dependent addresses.
5059 You may assume that @var{addr} is a valid address for the machine.
5062 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5063 A C expression that is nonzero if @var{x} is a legitimate constant for
5064 an immediate operand on the target machine. You can assume that
5065 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5066 @samp{1} is a suitable definition for this macro on machines where
5067 anything @code{CONSTANT_P} is valid.
5070 @node Condition Code
5071 @section Condition Code Status
5072 @cindex condition code status
5074 @c prevent bad page break with this line
5075 This describes the condition code status.
5078 The file @file{conditions.h} defines a variable @code{cc_status} to
5079 describe how the condition code was computed (in case the interpretation of
5080 the condition code depends on the instruction that it was set by). This
5081 variable contains the RTL expressions on which the condition code is
5082 currently based, and several standard flags.
5084 Sometimes additional machine-specific flags must be defined in the machine
5085 description header file. It can also add additional machine-specific
5086 information by defining @code{CC_STATUS_MDEP}.
5088 @defmac CC_STATUS_MDEP
5089 C code for a data type which is used for declaring the @code{mdep}
5090 component of @code{cc_status}. It defaults to @code{int}.
5092 This macro is not used on machines that do not use @code{cc0}.
5095 @defmac CC_STATUS_MDEP_INIT
5096 A C expression to initialize the @code{mdep} field to ``empty''.
5097 The default definition does nothing, since most machines don't use
5098 the field anyway. If you want to use the field, you should probably
5099 define this macro to initialize it.
5101 This macro is not used on machines that do not use @code{cc0}.
5104 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5105 A C compound statement to set the components of @code{cc_status}
5106 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5107 this macro's responsibility to recognize insns that set the condition
5108 code as a byproduct of other activity as well as those that explicitly
5111 This macro is not used on machines that do not use @code{cc0}.
5113 If there are insns that do not set the condition code but do alter
5114 other machine registers, this macro must check to see whether they
5115 invalidate the expressions that the condition code is recorded as
5116 reflecting. For example, on the 68000, insns that store in address
5117 registers do not set the condition code, which means that usually
5118 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5119 insns. But suppose that the previous insn set the condition code
5120 based on location @samp{a4@@(102)} and the current insn stores a new
5121 value in @samp{a4}. Although the condition code is not changed by
5122 this, it will no longer be true that it reflects the contents of
5123 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5124 @code{cc_status} in this case to say that nothing is known about the
5125 condition code value.
5127 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5128 with the results of peephole optimization: insns whose patterns are
5129 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5130 constants which are just the operands. The RTL structure of these
5131 insns is not sufficient to indicate what the insns actually do. What
5132 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5133 @code{CC_STATUS_INIT}.
5135 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5136 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5137 @samp{cc}. This avoids having detailed information about patterns in
5138 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5141 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5142 Returns a mode from class @code{MODE_CC} to be used when comparison
5143 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5144 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5145 @pxref{Jump Patterns} for a description of the reason for this
5149 #define SELECT_CC_MODE(OP,X,Y) \
5150 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5151 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5152 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5153 || GET_CODE (X) == NEG) \
5154 ? CC_NOOVmode : CCmode))
5157 You should define this macro if and only if you define extra CC modes
5158 in @file{@var{machine}-modes.def}.
5161 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5162 On some machines not all possible comparisons are defined, but you can
5163 convert an invalid comparison into a valid one. For example, the Alpha
5164 does not have a @code{GT} comparison, but you can use an @code{LT}
5165 comparison instead and swap the order of the operands.
5167 On such machines, define this macro to be a C statement to do any
5168 required conversions. @var{code} is the initial comparison code
5169 and @var{op0} and @var{op1} are the left and right operands of the
5170 comparison, respectively. You should modify @var{code}, @var{op0}, and
5171 @var{op1} as required.
5173 GCC will not assume that the comparison resulting from this macro is
5174 valid but will see if the resulting insn matches a pattern in the
5177 You need not define this macro if it would never change the comparison
5181 @defmac REVERSIBLE_CC_MODE (@var{mode})
5182 A C expression whose value is one if it is always safe to reverse a
5183 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5184 can ever return @var{mode} for a floating-point inequality comparison,
5185 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5187 You need not define this macro if it would always returns zero or if the
5188 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5189 For example, here is the definition used on the SPARC, where floating-point
5190 inequality comparisons are always given @code{CCFPEmode}:
5193 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5197 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5198 A C expression whose value is reversed condition code of the @var{code} for
5199 comparison done in CC_MODE @var{mode}. The macro is used only in case
5200 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5201 machine has some non-standard way how to reverse certain conditionals. For
5202 instance in case all floating point conditions are non-trapping, compiler may
5203 freely convert unordered compares to ordered one. Then definition may look
5207 #define REVERSE_CONDITION(CODE, MODE) \
5208 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5209 : reverse_condition_maybe_unordered (CODE))
5213 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
5214 A C expression that returns true if the conditional execution predicate
5215 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
5216 return 0 if the target has conditional execution predicates that cannot be
5217 reversed safely. If no expansion is specified, this macro is defined as
5221 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5222 ((x) == reverse_condition (y))
5226 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5227 On targets which do not use @code{(cc0)}, and which use a hard
5228 register rather than a pseudo-register to hold condition codes, the
5229 regular CSE passes are often not able to identify cases in which the
5230 hard register is set to a common value. Use this hook to enable a
5231 small pass which optimizes such cases. This hook should return true
5232 to enable this pass, and it should set the integers to which its
5233 arguments point to the hard register numbers used for condition codes.
5234 When there is only one such register, as is true on most systems, the
5235 integer pointed to by the second argument should be set to
5236 @code{INVALID_REGNUM}.
5238 The default version of this hook returns false.
5241 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5242 On targets which use multiple condition code modes in class
5243 @code{MODE_CC}, it is sometimes the case that a comparison can be
5244 validly done in more than one mode. On such a system, define this
5245 target hook to take two mode arguments and to return a mode in which
5246 both comparisons may be validly done. If there is no such mode,
5247 return @code{VOIDmode}.
5249 The default version of this hook checks whether the modes are the
5250 same. If they are, it returns that mode. If they are different, it
5251 returns @code{VOIDmode}.
5255 @section Describing Relative Costs of Operations
5256 @cindex costs of instructions
5257 @cindex relative costs
5258 @cindex speed of instructions
5260 These macros let you describe the relative speed of various operations
5261 on the target machine.
5263 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5264 A C expression for the cost of moving data of mode @var{mode} from a
5265 register in class @var{from} to one in class @var{to}. The classes are
5266 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5267 value of 2 is the default; other values are interpreted relative to
5270 It is not required that the cost always equal 2 when @var{from} is the
5271 same as @var{to}; on some machines it is expensive to move between
5272 registers if they are not general registers.
5274 If reload sees an insn consisting of a single @code{set} between two
5275 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5276 classes returns a value of 2, reload does not check to ensure that the
5277 constraints of the insn are met. Setting a cost of other than 2 will
5278 allow reload to verify that the constraints are met. You should do this
5279 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5282 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5283 A C expression for the cost of moving data of mode @var{mode} between a
5284 register of class @var{class} and memory; @var{in} is zero if the value
5285 is to be written to memory, nonzero if it is to be read in. This cost
5286 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5287 registers and memory is more expensive than between two registers, you
5288 should define this macro to express the relative cost.
5290 If you do not define this macro, GCC uses a default cost of 4 plus
5291 the cost of copying via a secondary reload register, if one is
5292 needed. If your machine requires a secondary reload register to copy
5293 between memory and a register of @var{class} but the reload mechanism is
5294 more complex than copying via an intermediate, define this macro to
5295 reflect the actual cost of the move.
5297 GCC defines the function @code{memory_move_secondary_cost} if
5298 secondary reloads are needed. It computes the costs due to copying via
5299 a secondary register. If your machine copies from memory using a
5300 secondary register in the conventional way but the default base value of
5301 4 is not correct for your machine, define this macro to add some other
5302 value to the result of that function. The arguments to that function
5303 are the same as to this macro.
5307 A C expression for the cost of a branch instruction. A value of 1 is
5308 the default; other values are interpreted relative to that.
5311 Here are additional macros which do not specify precise relative costs,
5312 but only that certain actions are more expensive than GCC would
5315 @defmac SLOW_BYTE_ACCESS
5316 Define this macro as a C expression which is nonzero if accessing less
5317 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5318 faster than accessing a word of memory, i.e., if such access
5319 require more than one instruction or if there is no difference in cost
5320 between byte and (aligned) word loads.
5322 When this macro is not defined, the compiler will access a field by
5323 finding the smallest containing object; when it is defined, a fullword
5324 load will be used if alignment permits. Unless bytes accesses are
5325 faster than word accesses, using word accesses is preferable since it
5326 may eliminate subsequent memory access if subsequent accesses occur to
5327 other fields in the same word of the structure, but to different bytes.
5330 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5331 Define this macro to be the value 1 if memory accesses described by the
5332 @var{mode} and @var{alignment} parameters have a cost many times greater
5333 than aligned accesses, for example if they are emulated in a trap
5336 When this macro is nonzero, the compiler will act as if
5337 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5338 moves. This can cause significantly more instructions to be produced.
5339 Therefore, do not set this macro nonzero if unaligned accesses only add a
5340 cycle or two to the time for a memory access.
5342 If the value of this macro is always zero, it need not be defined. If
5343 this macro is defined, it should produce a nonzero value when
5344 @code{STRICT_ALIGNMENT} is nonzero.
5348 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5349 which a sequence of insns should be generated instead of a
5350 string move insn or a library call. Increasing the value will always
5351 make code faster, but eventually incurs high cost in increased code size.
5353 Note that on machines where the corresponding move insn is a
5354 @code{define_expand} that emits a sequence of insns, this macro counts
5355 the number of such sequences.
5357 If you don't define this, a reasonable default is used.
5360 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5361 A C expression used to determine whether @code{move_by_pieces} will be used to
5362 copy a chunk of memory, or whether some other block move mechanism
5363 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5364 than @code{MOVE_RATIO}.
5367 @defmac MOVE_MAX_PIECES
5368 A C expression used by @code{move_by_pieces} to determine the largest unit
5369 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5373 The threshold of number of scalar move insns, @emph{below} which a sequence
5374 of insns should be generated to clear memory instead of a string clear insn
5375 or a library call. Increasing the value will always make code faster, but
5376 eventually incurs high cost in increased code size.
5378 If you don't define this, a reasonable default is used.
5381 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5382 A C expression used to determine whether @code{clear_by_pieces} will be used
5383 to clear a chunk of memory, or whether some other block clear mechanism
5384 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5385 than @code{CLEAR_RATIO}.
5388 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5389 A C expression used to determine whether @code{store_by_pieces} will be
5390 used to set a chunk of memory to a constant value, or whether some other
5391 mechanism will be used. Used by @code{__builtin_memset} when storing
5392 values other than constant zero and by @code{__builtin_strcpy} when
5393 when called with a constant source string.
5394 Defaults to @code{MOVE_BY_PIECES_P}.
5397 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5398 A C expression used to determine whether a load postincrement is a good
5399 thing to use for a given mode. Defaults to the value of
5400 @code{HAVE_POST_INCREMENT}.
5403 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5404 A C expression used to determine whether a load postdecrement is a good
5405 thing to use for a given mode. Defaults to the value of
5406 @code{HAVE_POST_DECREMENT}.
5409 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5410 A C expression used to determine whether a load preincrement is a good
5411 thing to use for a given mode. Defaults to the value of
5412 @code{HAVE_PRE_INCREMENT}.
5415 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5416 A C expression used to determine whether a load predecrement is a good
5417 thing to use for a given mode. Defaults to the value of
5418 @code{HAVE_PRE_DECREMENT}.
5421 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5422 A C expression used to determine whether a store postincrement is a good
5423 thing to use for a given mode. Defaults to the value of
5424 @code{HAVE_POST_INCREMENT}.
5427 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5428 A C expression used to determine whether a store postdecrement is a good
5429 thing to use for a given mode. Defaults to the value of
5430 @code{HAVE_POST_DECREMENT}.
5433 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5434 This macro is used to determine whether a store preincrement is a good
5435 thing to use for a given mode. Defaults to the value of
5436 @code{HAVE_PRE_INCREMENT}.
5439 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5440 This macro is used to determine whether a store predecrement is a good
5441 thing to use for a given mode. Defaults to the value of
5442 @code{HAVE_PRE_DECREMENT}.
5445 @defmac NO_FUNCTION_CSE
5446 Define this macro if it is as good or better to call a constant
5447 function address than to call an address kept in a register.
5450 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5451 Define this macro if a non-short-circuit operation produced by
5452 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5453 @code{BRANCH_COST} is greater than or equal to the value 2.
5456 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5457 This target hook describes the relative costs of RTL expressions.
5459 The cost may depend on the precise form of the expression, which is
5460 available for examination in @var{x}, and the rtx code of the expression
5461 in which it is contained, found in @var{outer_code}. @var{code} is the
5462 expression code---redundant, since it can be obtained with
5463 @code{GET_CODE (@var{x})}.
5465 In implementing this hook, you can use the construct
5466 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5469 On entry to the hook, @code{*@var{total}} contains a default estimate
5470 for the cost of the expression. The hook should modify this value as
5471 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
5472 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
5473 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
5475 When optimizing for code size, i.e@. when @code{optimize_size} is
5476 non-zero, this target hook should be used to estimate the relative
5477 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
5479 The hook returns true when all subexpressions of @var{x} have been
5480 processed, and false when @code{rtx_cost} should recurse.
5483 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5484 This hook computes the cost of an addressing mode that contains
5485 @var{address}. If not defined, the cost is computed from
5486 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5488 For most CISC machines, the default cost is a good approximation of the
5489 true cost of the addressing mode. However, on RISC machines, all
5490 instructions normally have the same length and execution time. Hence
5491 all addresses will have equal costs.
5493 In cases where more than one form of an address is known, the form with
5494 the lowest cost will be used. If multiple forms have the same, lowest,
5495 cost, the one that is the most complex will be used.
5497 For example, suppose an address that is equal to the sum of a register
5498 and a constant is used twice in the same basic block. When this macro
5499 is not defined, the address will be computed in a register and memory
5500 references will be indirect through that register. On machines where
5501 the cost of the addressing mode containing the sum is no higher than
5502 that of a simple indirect reference, this will produce an additional
5503 instruction and possibly require an additional register. Proper
5504 specification of this macro eliminates this overhead for such machines.
5506 This hook is never called with an invalid address.
5508 On machines where an address involving more than one register is as
5509 cheap as an address computation involving only one register, defining
5510 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5511 be live over a region of code where only one would have been if
5512 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5513 should be considered in the definition of this macro. Equivalent costs
5514 should probably only be given to addresses with different numbers of
5515 registers on machines with lots of registers.
5519 @section Adjusting the Instruction Scheduler
5521 The instruction scheduler may need a fair amount of machine-specific
5522 adjustment in order to produce good code. GCC provides several target
5523 hooks for this purpose. It is usually enough to define just a few of
5524 them: try the first ones in this list first.
5526 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5527 This hook returns the maximum number of instructions that can ever
5528 issue at the same time on the target machine. The default is one.
5529 Although the insn scheduler can define itself the possibility of issue
5530 an insn on the same cycle, the value can serve as an additional
5531 constraint to issue insns on the same simulated processor cycle (see
5532 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5533 This value must be constant over the entire compilation. If you need
5534 it to vary depending on what the instructions are, you must use
5535 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5537 For the automaton based pipeline interface, you could define this hook
5538 to return the value of the macro @code{MAX_DFA_ISSUE_RATE}.
5541 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5542 This hook is executed by the scheduler after it has scheduled an insn
5543 from the ready list. It should return the number of insns which can
5544 still be issued in the current cycle. The default is
5545 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5546 @code{USE}, which normally are not counted against the issue rate.
5547 You should define this hook if some insns take more machine resources
5548 than others, so that fewer insns can follow them in the same cycle.
5549 @var{file} is either a null pointer, or a stdio stream to write any
5550 debug output to. @var{verbose} is the verbose level provided by
5551 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5555 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5556 This function corrects the value of @var{cost} based on the
5557 relationship between @var{insn} and @var{dep_insn} through the
5558 dependence @var{link}. It should return the new value. The default
5559 is to make no adjustment to @var{cost}. This can be used for example
5560 to specify to the scheduler using the traditional pipeline description
5561 that an output- or anti-dependence does not incur the same cost as a
5562 data-dependence. If the scheduler using the automaton based pipeline
5563 description, the cost of anti-dependence is zero and the cost of
5564 output-dependence is maximum of one and the difference of latency
5565 times of the first and the second insns. If these values are not
5566 acceptable, you could use the hook to modify them too. See also
5567 @pxref{Automaton pipeline description}.
5570 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5571 This hook adjusts the integer scheduling priority @var{priority} of
5572 @var{insn}. It should return the new priority. Reduce the priority to
5573 execute @var{insn} earlier, increase the priority to execute @var{insn}
5574 later. Do not define this hook if you do not need to adjust the
5575 scheduling priorities of insns.
5578 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5579 This hook is executed by the scheduler after it has scheduled the ready
5580 list, to allow the machine description to reorder it (for example to
5581 combine two small instructions together on @samp{VLIW} machines).
5582 @var{file} is either a null pointer, or a stdio stream to write any
5583 debug output to. @var{verbose} is the verbose level provided by
5584 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5585 list of instructions that are ready to be scheduled. @var{n_readyp} is
5586 a pointer to the number of elements in the ready list. The scheduler
5587 reads the ready list in reverse order, starting with
5588 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5589 is the timer tick of the scheduler. You may modify the ready list and
5590 the number of ready insns. The return value is the number of insns that
5591 can issue this cycle; normally this is just @code{issue_rate}. See also
5592 @samp{TARGET_SCHED_REORDER2}.
5595 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5596 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5597 function is called whenever the scheduler starts a new cycle. This one
5598 is called once per iteration over a cycle, immediately after
5599 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5600 return the number of insns to be scheduled in the same cycle. Defining
5601 this hook can be useful if there are frequent situations where
5602 scheduling one insn causes other insns to become ready in the same
5603 cycle. These other insns can then be taken into account properly.
5606 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5607 This hook is called after evaluation forward dependencies of insns in
5608 chain given by two parameter values (@var{head} and @var{tail}
5609 correspondingly) but before insns scheduling of the insn chain. For
5610 example, it can be used for better insn classification if it requires
5611 analysis of dependencies. This hook can use backward and forward
5612 dependencies of the insn scheduler because they are already
5616 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5617 This hook is executed by the scheduler at the beginning of each block of
5618 instructions that are to be scheduled. @var{file} is either a null
5619 pointer, or a stdio stream to write any debug output to. @var{verbose}
5620 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5621 @var{max_ready} is the maximum number of insns in the current scheduling
5622 region that can be live at the same time. This can be used to allocate
5623 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5626 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5627 This hook is executed by the scheduler at the end of each block of
5628 instructions that are to be scheduled. It can be used to perform
5629 cleanup of any actions done by the other scheduling hooks. @var{file}
5630 is either a null pointer, or a stdio stream to write any debug output
5631 to. @var{verbose} is the verbose level provided by
5632 @option{-fsched-verbose-@var{n}}.
5635 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
5636 This hook is executed by the scheduler after function level initializations.
5637 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5638 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5639 @var{old_max_uid} is the maximum insn uid when scheduling begins.
5642 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
5643 This is the cleanup hook corresponding to TARGET_SCHED_INIT_GLOBAL.
5644 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5645 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5648 @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void)
5649 This hook is called many times during insn scheduling. If the hook
5650 returns nonzero, the automaton based pipeline description is used for
5651 insn scheduling. Otherwise the traditional pipeline description is
5652 used. The default is usage of the traditional pipeline description.
5654 You should also remember that to simplify the insn scheduler sources
5655 an empty traditional pipeline description interface is generated even
5656 if there is no a traditional pipeline description in the @file{.md}
5657 file. The same is true for the automaton based pipeline description.
5658 That means that you should be accurate in defining the hook.
5661 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5662 The hook returns an RTL insn. The automaton state used in the
5663 pipeline hazard recognizer is changed as if the insn were scheduled
5664 when the new simulated processor cycle starts. Usage of the hook may
5665 simplify the automaton pipeline description for some @acronym{VLIW}
5666 processors. If the hook is defined, it is used only for the automaton
5667 based pipeline description. The default is not to change the state
5668 when the new simulated processor cycle starts.
5671 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5672 The hook can be used to initialize data used by the previous hook.
5675 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5676 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5677 to changed the state as if the insn were scheduled when the new
5678 simulated processor cycle finishes.
5681 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5682 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5683 used to initialize data used by the previous hook.
5686 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5687 This hook controls better choosing an insn from the ready insn queue
5688 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5689 chooses the first insn from the queue. If the hook returns a positive
5690 value, an additional scheduler code tries all permutations of
5691 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5692 subsequent ready insns to choose an insn whose issue will result in
5693 maximal number of issued insns on the same cycle. For the
5694 @acronym{VLIW} processor, the code could actually solve the problem of
5695 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5696 rules of @acronym{VLIW} packing are described in the automaton.
5698 This code also could be used for superscalar @acronym{RISC}
5699 processors. Let us consider a superscalar @acronym{RISC} processor
5700 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5701 @var{B}, some insns can be executed only in pipelines @var{B} or
5702 @var{C}, and one insn can be executed in pipeline @var{B}. The
5703 processor may issue the 1st insn into @var{A} and the 2nd one into
5704 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5705 until the next cycle. If the scheduler issues the 3rd insn the first,
5706 the processor could issue all 3 insns per cycle.
5708 Actually this code demonstrates advantages of the automaton based
5709 pipeline hazard recognizer. We try quickly and easy many insn
5710 schedules to choose the best one.
5712 The default is no multipass scheduling.
5715 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5717 This hook controls what insns from the ready insn queue will be
5718 considered for the multipass insn scheduling. If the hook returns
5719 zero for insn passed as the parameter, the insn will be not chosen to
5722 The default is that any ready insns can be chosen to be issued.
5725 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5727 This hook is called by the insn scheduler before issuing insn passed
5728 as the third parameter on given cycle. If the hook returns nonzero,
5729 the insn is not issued on given processors cycle. Instead of that,
5730 the processor cycle is advanced. If the value passed through the last
5731 parameter is zero, the insn ready queue is not sorted on the new cycle
5732 start as usually. The first parameter passes file for debugging
5733 output. The second one passes the scheduler verbose level of the
5734 debugging output. The forth and the fifth parameter values are
5735 correspondingly processor cycle on which the previous insn has been
5736 issued and the current processor cycle.
5739 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void)
5740 The @acronym{DFA}-based scheduler could take the insertion of nop
5741 operations for better insn scheduling into account. It can be done
5742 only if the multi-pass insn scheduling works (see hook
5743 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}).
5745 Let us consider a @acronym{VLIW} processor insn with 3 slots. Each
5746 insn can be placed only in one of the three slots. We have 3 ready
5747 insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be
5748 placed only in the 1st slot, @var{B} can be placed only in the 3rd
5749 slot. We described the automaton which does not permit empty slot
5750 gaps between insns (usually such description is simpler). Without
5751 this code the scheduler would place each insn in 3 separate
5752 @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd
5753 slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is
5754 the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook
5755 @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or
5756 create the nop insns.
5758 You should remember that the scheduler does not insert the nop insns.
5759 It is not wise because of the following optimizations. The scheduler
5760 only considers such possibility to improve the result schedule. The
5761 nop insns should be inserted lately, e.g. on the final phase.
5764 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index})
5765 This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert
5766 nop operations for better insn scheduling when @acronym{DFA}-based
5767 scheduler makes multipass insn scheduling (see also description of
5768 hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook
5769 returns a nop insn with given @var{index}. The indexes start with
5770 zero. The hook should return @code{NULL} if there are no more nop
5771 insns with indexes greater than given index.
5774 @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})
5775 This hook is used to define which dependences are considered costly by
5776 the target, so costly that it is not advisable to schedule the insns that
5777 are involved in the dependence too close to one another. The parameters
5778 to this hook are as follows: The second parameter @var{insn2} is dependent
5779 upon the first parameter @var{insn1}. The dependence between @var{insn1}
5780 and @var{insn2} is represented by the third parameter @var{dep_link}. The
5781 fourth parameter @var{cost} is the cost of the dependence, and the fifth
5782 parameter @var{distance} is the distance in cycles between the two insns.
5783 The hook returns @code{true} if considering the distance between the two
5784 insns the dependence between them is considered costly by the target,
5785 and @code{false} otherwise.
5787 Defining this hook can be useful in multiple-issue out-of-order machines,
5788 where (a) it's practically hopeless to predict the actual data/resource
5789 delays, however: (b) there's a better chance to predict the actual grouping
5790 that will be formed, and (c) correctly emulating the grouping can be very
5791 important. In such targets one may want to allow issuing dependent insns
5792 closer to one another - i.e, closer than the dependence distance; however,
5793 not in cases of "costly dependences", which this hooks allows to define.
5796 Macros in the following table are generated by the program
5797 @file{genattr} and can be useful for writing the hooks.
5799 @defmac MAX_DFA_ISSUE_RATE
5800 The macro definition is generated in the automaton based pipeline
5801 description interface. Its value is calculated from the automaton
5802 based pipeline description and is equal to maximal number of all insns
5803 described in constructions @samp{define_insn_reservation} which can be
5804 issued on the same processor cycle.
5808 @section Dividing the Output into Sections (Texts, Data, @dots{})
5809 @c the above section title is WAY too long. maybe cut the part between
5810 @c the (...)? --mew 10feb93
5812 An object file is divided into sections containing different types of
5813 data. In the most common case, there are three sections: the @dfn{text
5814 section}, which holds instructions and read-only data; the @dfn{data
5815 section}, which holds initialized writable data; and the @dfn{bss
5816 section}, which holds uninitialized data. Some systems have other kinds
5819 The compiler must tell the assembler when to switch sections. These
5820 macros control what commands to output to tell the assembler this. You
5821 can also define additional sections.
5823 @defmac TEXT_SECTION_ASM_OP
5824 A C expression whose value is a string, including spacing, containing the
5825 assembler operation that should precede instructions and read-only data.
5826 Normally @code{"\t.text"} is right.
5829 @defmac HOT_TEXT_SECTION_NAME
5830 If defined, a C string constant for the name of the section containing most
5831 frequently executed functions of the program. If not defined, GCC will provide
5832 a default definition if the target supports named sections.
5835 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5836 If defined, a C string constant for the name of the section containing unlikely
5837 executed functions in the program.
5840 @defmac DATA_SECTION_ASM_OP
5841 A C expression whose value is a string, including spacing, containing the
5842 assembler operation to identify the following data as writable initialized
5843 data. Normally @code{"\t.data"} is right.
5846 @defmac READONLY_DATA_SECTION_ASM_OP
5847 A C expression whose value is a string, including spacing, containing the
5848 assembler operation to identify the following data as read-only initialized
5852 @defmac READONLY_DATA_SECTION
5853 A macro naming a function to call to switch to the proper section for
5854 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5855 if defined, else fall back to @code{text_section}.
5857 The most common definition will be @code{data_section}, if the target
5858 does not have a special read-only data section, and does not put data
5859 in the text section.
5862 @defmac BSS_SECTION_ASM_OP
5863 If defined, a C expression whose value is a string, including spacing,
5864 containing the assembler operation to identify the following data as
5865 uninitialized global data. If not defined, and neither
5866 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5867 uninitialized global data will be output in the data section if
5868 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5872 @defmac INIT_SECTION_ASM_OP
5873 If defined, a C expression whose value is a string, including spacing,
5874 containing the assembler operation to identify the following data as
5875 initialization code. If not defined, GCC will assume such a section does
5879 @defmac FINI_SECTION_ASM_OP
5880 If defined, a C expression whose value is a string, including spacing,
5881 containing the assembler operation to identify the following data as
5882 finalization code. If not defined, GCC will assume such a section does
5886 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5887 If defined, an ASM statement that switches to a different section
5888 via @var{section_op}, calls @var{function}, and switches back to
5889 the text section. This is used in @file{crtstuff.c} if
5890 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5891 to initialization and finalization functions from the init and fini
5892 sections. By default, this macro uses a simple function call. Some
5893 ports need hand-crafted assembly code to avoid dependencies on
5894 registers initialized in the function prologue or to ensure that
5895 constant pools don't end up too far way in the text section.
5898 @defmac FORCE_CODE_SECTION_ALIGN
5899 If defined, an ASM statement that aligns a code section to some
5900 arbitrary boundary. This is used to force all fragments of the
5901 @code{.init} and @code{.fini} sections to have to same alignment
5902 and thus prevent the linker from having to add any padding.
5907 @defmac EXTRA_SECTIONS
5908 A list of names for sections other than the standard two, which are
5909 @code{in_text} and @code{in_data}. You need not define this macro
5910 on a system with no other sections (that GCC needs to use).
5913 @findex text_section
5914 @findex data_section
5915 @defmac EXTRA_SECTION_FUNCTIONS
5916 One or more functions to be defined in @file{varasm.c}. These
5917 functions should do jobs analogous to those of @code{text_section} and
5918 @code{data_section}, for your additional sections. Do not define this
5919 macro if you do not define @code{EXTRA_SECTIONS}.
5922 @defmac JUMP_TABLES_IN_TEXT_SECTION
5923 Define this macro to be an expression with a nonzero value if jump
5924 tables (for @code{tablejump} insns) should be output in the text
5925 section, along with the assembler instructions. Otherwise, the
5926 readonly data section is used.
5928 This macro is irrelevant if there is no separate readonly data section.
5931 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5932 Switches to the appropriate section for output of @var{exp}. You can
5933 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5934 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5935 requires link-time relocations. Bit 0 is set when variable contains
5936 local relocations only, while bit 1 is set for global relocations.
5937 Select the section by calling @code{data_section} or one of the
5938 alternatives for other sections. @var{align} is the constant alignment
5941 The default version of this function takes care of putting read-only
5942 variables in @code{readonly_data_section}.
5945 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5946 Build up a unique section name, expressed as a @code{STRING_CST} node,
5947 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5948 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5949 the initial value of @var{exp} requires link-time relocations.
5951 The default version of this function appends the symbol name to the
5952 ELF section name that would normally be used for the symbol. For
5953 example, the function @code{foo} would be placed in @code{.text.foo}.
5954 Whatever the actual target object format, this is often good enough.
5957 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
5958 Switches to the appropriate section for output of constant pool entry
5959 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
5960 constant in RTL@. The argument @var{mode} is redundant except in the
5961 case of a @code{const_int} rtx. Select the section by calling
5962 @code{readonly_data_section} or one of the alternatives for other
5963 sections. @var{align} is the constant alignment in bits.
5965 The default version of this function takes care of putting symbolic
5966 constants in @code{flag_pic} mode in @code{data_section} and everything
5967 else in @code{readonly_data_section}.
5970 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
5971 Define this hook if references to a symbol or a constant must be
5972 treated differently depending on something about the variable or
5973 function named by the symbol (such as what section it is in).
5975 The hook is executed immediately after rtl has been created for
5976 @var{decl}, which may be a variable or function declaration or
5977 an entry in the constant pool. In either case, @var{rtl} is the
5978 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
5979 in this hook; that field may not have been initialized yet.
5981 In the case of a constant, it is safe to assume that the rtl is
5982 a @code{mem} whose address is a @code{symbol_ref}. Most decls
5983 will also have this form, but that is not guaranteed. Global
5984 register variables, for instance, will have a @code{reg} for their
5985 rtl. (Normally the right thing to do with such unusual rtl is
5988 The @var{new_decl_p} argument will be true if this is the first time
5989 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
5990 be false for subsequent invocations, which will happen for duplicate
5991 declarations. Whether or not anything must be done for the duplicate
5992 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
5993 @var{new_decl_p} is always true when the hook is called for a constant.
5995 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
5996 The usual thing for this hook to do is to record flags in the
5997 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
5998 Historically, the name string was modified if it was necessary to
5999 encode more than one bit of information, but this practice is now
6000 discouraged; use @code{SYMBOL_REF_FLAGS}.
6002 The default definition of this hook, @code{default_encode_section_info}
6003 in @file{varasm.c}, sets a number of commonly-useful bits in
6004 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6005 before overriding it.
6008 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6009 Decode @var{name} and return the real name part, sans
6010 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6014 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6015 Returns true if @var{exp} should be placed into a ``small data'' section.
6016 The default version of this hook always returns false.
6019 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6020 Contains the value true if the target places read-only
6021 ``small data'' into a separate section. The default value is false.
6024 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6025 Returns true if @var{exp} names an object for which name resolution
6026 rules must resolve to the current ``module'' (dynamic shared library
6027 or executable image).
6029 The default version of this hook implements the name resolution rules
6030 for ELF, which has a looser model of global name binding than other
6031 currently supported object file formats.
6034 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6035 Contains the value true if the target supports thread-local storage.
6036 The default value is false.
6041 @section Position Independent Code
6042 @cindex position independent code
6045 This section describes macros that help implement generation of position
6046 independent code. Simply defining these macros is not enough to
6047 generate valid PIC; you must also add support to the macros
6048 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6049 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6050 @samp{movsi} to do something appropriate when the source operand
6051 contains a symbolic address. You may also need to alter the handling of
6052 switch statements so that they use relative addresses.
6053 @c i rearranged the order of the macros above to try to force one of
6054 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6056 @defmac PIC_OFFSET_TABLE_REGNUM
6057 The register number of the register used to address a table of static
6058 data addresses in memory. In some cases this register is defined by a
6059 processor's ``application binary interface'' (ABI)@. When this macro
6060 is defined, RTL is generated for this register once, as with the stack
6061 pointer and frame pointer registers. If this macro is not defined, it
6062 is up to the machine-dependent files to allocate such a register (if
6063 necessary). Note that this register must be fixed when in use (e.g.@:
6064 when @code{flag_pic} is true).
6067 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6068 Define this macro if the register defined by
6069 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6070 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6073 @defmac FINALIZE_PIC
6074 By generating position-independent code, when two different programs (A
6075 and B) share a common library (libC.a), the text of the library can be
6076 shared whether or not the library is linked at the same address for both
6077 programs. In some of these environments, position-independent code
6078 requires not only the use of different addressing modes, but also
6079 special code to enable the use of these addressing modes.
6081 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
6082 codes once the function is being compiled into assembly code, but not
6083 before. (It is not done before, because in the case of compiling an
6084 inline function, it would lead to multiple PIC prologues being
6085 included in functions which used inline functions and were compiled to
6089 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6090 A C expression that is nonzero if @var{x} is a legitimate immediate
6091 operand on the target machine when generating position independent code.
6092 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6093 check this. You can also assume @var{flag_pic} is true, so you need not
6094 check it either. You need not define this macro if all constants
6095 (including @code{SYMBOL_REF}) can be immediate operands when generating
6096 position independent code.
6099 @node Assembler Format
6100 @section Defining the Output Assembler Language
6102 This section describes macros whose principal purpose is to describe how
6103 to write instructions in assembler language---rather than what the
6107 * File Framework:: Structural information for the assembler file.
6108 * Data Output:: Output of constants (numbers, strings, addresses).
6109 * Uninitialized Data:: Output of uninitialized variables.
6110 * Label Output:: Output and generation of labels.
6111 * Initialization:: General principles of initialization
6112 and termination routines.
6113 * Macros for Initialization::
6114 Specific macros that control the handling of
6115 initialization and termination routines.
6116 * Instruction Output:: Output of actual instructions.
6117 * Dispatch Tables:: Output of jump tables.
6118 * Exception Region Output:: Output of exception region code.
6119 * Alignment Output:: Pseudo ops for alignment and skipping data.
6122 @node File Framework
6123 @subsection The Overall Framework of an Assembler File
6124 @cindex assembler format
6125 @cindex output of assembler code
6127 @c prevent bad page break with this line
6128 This describes the overall framework of an assembly file.
6130 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6131 @findex default_file_start
6132 Output to @code{asm_out_file} any text which the assembler expects to
6133 find at the beginning of a file. The default behavior is controlled
6134 by two flags, documented below. Unless your target's assembler is
6135 quite unusual, if you override the default, you should call
6136 @code{default_file_start} at some point in your target hook. This
6137 lets other target files rely on these variables.
6140 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6141 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6142 printed as the very first line in the assembly file, unless
6143 @option{-fverbose-asm} is in effect. (If that macro has been defined
6144 to the empty string, this variable has no effect.) With the normal
6145 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6146 assembler that it need not bother stripping comments or extra
6147 whitespace from its input. This allows it to work a bit faster.
6149 The default is false. You should not set it to true unless you have
6150 verified that your port does not generate any extra whitespace or
6151 comments that will cause GAS to issue errors in NO_APP mode.
6154 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6155 If this flag is true, @code{output_file_directive} will be called
6156 for the primary source file, immediately after printing
6157 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6158 this to be done. The default is false.
6161 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6162 Output to @code{asm_out_file} any text which the assembler expects
6163 to find at the end of a file. The default is to output nothing.
6166 @deftypefun void file_end_indicate_exec_stack ()
6167 Some systems use a common convention, the @samp{.note.GNU-stack}
6168 special section, to indicate whether or not an object file relies on
6169 the stack being executable. If your system uses this convention, you
6170 should define @code{TARGET_ASM_FILE_END} to this function. If you
6171 need to do other things in that hook, have your hook function call
6175 @defmac ASM_COMMENT_START
6176 A C string constant describing how to begin a comment in the target
6177 assembler language. The compiler assumes that the comment will end at
6178 the end of the line.
6182 A C string constant for text to be output before each @code{asm}
6183 statement or group of consecutive ones. Normally this is
6184 @code{"#APP"}, which is a comment that has no effect on most
6185 assemblers but tells the GNU assembler that it must check the lines
6186 that follow for all valid assembler constructs.
6190 A C string constant for text to be output after each @code{asm}
6191 statement or group of consecutive ones. Normally this is
6192 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6193 time-saving assumptions that are valid for ordinary compiler output.
6196 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6197 A C statement to output COFF information or DWARF debugging information
6198 which indicates that filename @var{name} is the current source file to
6199 the stdio stream @var{stream}.
6201 This macro need not be defined if the standard form of output
6202 for the file format in use is appropriate.
6205 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6206 A C statement to output the string @var{string} to the stdio stream
6207 @var{stream}. If you do not call the function @code{output_quoted_string}
6208 in your config files, GCC will only call it to output filenames to
6209 the assembler source. So you can use it to canonicalize the format
6210 of the filename using this macro.
6213 @defmac ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
6214 A C statement to output DBX or SDB debugging information before code
6215 for line number @var{line} of the current source file to the
6216 stdio stream @var{stream}. @var{counter} is the number of time the
6217 macro was invoked, including the current invocation; it is intended
6218 to generate unique labels in the assembly output.
6220 This macro need not be defined if the standard form of debugging
6221 information for the debugger in use is appropriate.
6224 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6225 A C statement to output something to the assembler file to handle a
6226 @samp{#ident} directive containing the text @var{string}. If this
6227 macro is not defined, nothing is output for a @samp{#ident} directive.
6230 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6231 Output assembly directives to switch to section @var{name}. The section
6232 should have attributes as specified by @var{flags}, which is a bit mask
6233 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6234 is nonzero, it contains an alignment in bytes to be used for the section,
6235 otherwise some target default should be used. Only targets that must
6236 specify an alignment within the section directive need pay attention to
6237 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6240 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6241 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6244 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6245 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6246 based on a variable or function decl, a section name, and whether or not the
6247 declaration's initializer may contain runtime relocations. @var{decl} may be
6248 null, in which case read-write data should be assumed.
6250 The default version if this function handles choosing code vs data,
6251 read-only vs read-write data, and @code{flag_pic}. You should only
6252 need to override this if your target has special flags that might be
6253 set via @code{__attribute__}.
6258 @subsection Output of Data
6261 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6262 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6263 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6264 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6265 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6266 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6267 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6268 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6269 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6270 These hooks specify assembly directives for creating certain kinds
6271 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6272 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6273 aligned two-byte object, and so on. Any of the hooks may be
6274 @code{NULL}, indicating that no suitable directive is available.
6276 The compiler will print these strings at the start of a new line,
6277 followed immediately by the object's initial value. In most cases,
6278 the string should contain a tab, a pseudo-op, and then another tab.
6281 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6282 The @code{assemble_integer} function uses this hook to output an
6283 integer object. @var{x} is the object's value, @var{size} is its size
6284 in bytes and @var{aligned_p} indicates whether it is aligned. The
6285 function should return @code{true} if it was able to output the
6286 object. If it returns false, @code{assemble_integer} will try to
6287 split the object into smaller parts.
6289 The default implementation of this hook will use the
6290 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6291 when the relevant string is @code{NULL}.
6294 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6295 A C statement to recognize @var{rtx} patterns that
6296 @code{output_addr_const} can't deal with, and output assembly code to
6297 @var{stream} corresponding to the pattern @var{x}. This may be used to
6298 allow machine-dependent @code{UNSPEC}s to appear within constants.
6300 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6301 @code{goto fail}, so that a standard error message is printed. If it
6302 prints an error message itself, by calling, for example,
6303 @code{output_operand_lossage}, it may just complete normally.
6306 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6307 A C statement to output to the stdio stream @var{stream} an assembler
6308 instruction to assemble a string constant containing the @var{len}
6309 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6310 @code{char *} and @var{len} a C expression of type @code{int}.
6312 If the assembler has a @code{.ascii} pseudo-op as found in the
6313 Berkeley Unix assembler, do not define the macro
6314 @code{ASM_OUTPUT_ASCII}.
6317 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6318 A C statement to output word @var{n} of a function descriptor for
6319 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6320 is defined, and is otherwise unused.
6323 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6324 You may define this macro as a C expression. You should define the
6325 expression to have a nonzero value if GCC should output the constant
6326 pool for a function before the code for the function, or a zero value if
6327 GCC should output the constant pool after the function. If you do
6328 not define this macro, the usual case, GCC will output the constant
6329 pool before the function.
6332 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6333 A C statement to output assembler commands to define the start of the
6334 constant pool for a function. @var{funname} is a string giving
6335 the name of the function. Should the return type of the function
6336 be required, it can be obtained via @var{fundecl}. @var{size}
6337 is the size, in bytes, of the constant pool that will be written
6338 immediately after this call.
6340 If no constant-pool prefix is required, the usual case, this macro need
6344 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6345 A C statement (with or without semicolon) to output a constant in the
6346 constant pool, if it needs special treatment. (This macro need not do
6347 anything for RTL expressions that can be output normally.)
6349 The argument @var{file} is the standard I/O stream to output the
6350 assembler code on. @var{x} is the RTL expression for the constant to
6351 output, and @var{mode} is the machine mode (in case @var{x} is a
6352 @samp{const_int}). @var{align} is the required alignment for the value
6353 @var{x}; you should output an assembler directive to force this much
6356 The argument @var{labelno} is a number to use in an internal label for
6357 the address of this pool entry. The definition of this macro is
6358 responsible for outputting the label definition at the proper place.
6359 Here is how to do this:
6362 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6365 When you output a pool entry specially, you should end with a
6366 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6367 entry from being output a second time in the usual manner.
6369 You need not define this macro if it would do nothing.
6372 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6373 A C statement to output assembler commands to at the end of the constant
6374 pool for a function. @var{funname} is a string giving the name of the
6375 function. Should the return type of the function be required, you can
6376 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6377 constant pool that GCC wrote immediately before this call.
6379 If no constant-pool epilogue is required, the usual case, you need not
6383 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6384 Define this macro as a C expression which is nonzero if @var{C} is
6385 used as a logical line separator by the assembler.
6387 If you do not define this macro, the default is that only
6388 the character @samp{;} is treated as a logical line separator.
6391 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6392 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6393 These target hooks are C string constants, describing the syntax in the
6394 assembler for grouping arithmetic expressions. If not overridden, they
6395 default to normal parentheses, which is correct for most assemblers.
6398 These macros are provided by @file{real.h} for writing the definitions
6399 of @code{ASM_OUTPUT_DOUBLE} and the like:
6401 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6402 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6403 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6404 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6405 floating point representation, and store its bit pattern in the variable
6406 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6407 be a simple @code{long int}. For the others, it should be an array of
6408 @code{long int}. The number of elements in this array is determined by
6409 the size of the desired target floating point data type: 32 bits of it
6410 go in each @code{long int} array element. Each array element holds 32
6411 bits of the result, even if @code{long int} is wider than 32 bits on the
6414 The array element values are designed so that you can print them out
6415 using @code{fprintf} in the order they should appear in the target
6419 @node Uninitialized Data
6420 @subsection Output of Uninitialized Variables
6422 Each of the macros in this section is used to do the whole job of
6423 outputting a single uninitialized variable.
6425 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6426 A C statement (sans semicolon) to output to the stdio stream
6427 @var{stream} the assembler definition of a common-label named
6428 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6429 is the size rounded up to whatever alignment the caller wants.
6431 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6432 output the name itself; before and after that, output the additional
6433 assembler syntax for defining the name, and a newline.
6435 This macro controls how the assembler definitions of uninitialized
6436 common global variables are output.
6439 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6440 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6441 separate, explicit argument. If you define this macro, it is used in
6442 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6443 handling the required alignment of the variable. The alignment is specified
6444 as the number of bits.
6447 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6448 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6449 variable to be output, if there is one, or @code{NULL_TREE} if there
6450 is no corresponding variable. If you define this macro, GCC will use it
6451 in place of both @code{ASM_OUTPUT_COMMON} and
6452 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6453 the variable's decl in order to chose what to output.
6456 @defmac ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6457 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6458 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6462 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6463 A C statement (sans semicolon) to output to the stdio stream
6464 @var{stream} the assembler definition of uninitialized global @var{decl} named
6465 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6466 is the size rounded up to whatever alignment the caller wants.
6468 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6469 defining this macro. If unable, use the expression
6470 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6471 before and after that, output the additional assembler syntax for defining
6472 the name, and a newline.
6474 This macro controls how the assembler definitions of uninitialized global
6475 variables are output. This macro exists to properly support languages like
6476 C++ which do not have @code{common} data. However, this macro currently
6477 is not defined for all targets. If this macro and
6478 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6479 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6480 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6483 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6484 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6485 separate, explicit argument. If you define this macro, it is used in
6486 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6487 handling the required alignment of the variable. The alignment is specified
6488 as the number of bits.
6490 Try to use function @code{asm_output_aligned_bss} defined in file
6491 @file{varasm.c} when defining this macro.
6494 @defmac ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6495 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6496 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6500 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6501 A C statement (sans semicolon) to output to the stdio stream
6502 @var{stream} the assembler definition of a local-common-label named
6503 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6504 is the size rounded up to whatever alignment the caller wants.
6506 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6507 output the name itself; before and after that, output the additional
6508 assembler syntax for defining the name, and a newline.
6510 This macro controls how the assembler definitions of uninitialized
6511 static variables are output.
6514 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6515 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6516 separate, explicit argument. If you define this macro, it is used in
6517 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6518 handling the required alignment of the variable. The alignment is specified
6519 as the number of bits.
6522 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6523 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6524 variable to be output, if there is one, or @code{NULL_TREE} if there
6525 is no corresponding variable. If you define this macro, GCC will use it
6526 in place of both @code{ASM_OUTPUT_DECL} and
6527 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6528 the variable's decl in order to chose what to output.
6531 @defmac ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6532 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6533 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6538 @subsection Output and Generation of Labels
6540 @c prevent bad page break with this line
6541 This is about outputting labels.
6543 @findex assemble_name
6544 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6545 A C statement (sans semicolon) to output to the stdio stream
6546 @var{stream} the assembler definition of a label named @var{name}.
6547 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6548 output the name itself; before and after that, output the additional
6549 assembler syntax for defining the name, and a newline. A default
6550 definition of this macro is provided which is correct for most systems.
6554 A C string containing the appropriate assembler directive to specify the
6555 size of a symbol, without any arguments. On systems that use ELF, the
6556 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6557 systems, the default is not to define this macro.
6559 Define this macro only if it is correct to use the default definitions
6560 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6561 for your system. If you need your own custom definitions of those
6562 macros, or if you do not need explicit symbol sizes at all, do not
6566 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6567 A C statement (sans semicolon) to output to the stdio stream
6568 @var{stream} a directive telling the assembler that the size of the
6569 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6570 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6574 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6575 A C statement (sans semicolon) to output to the stdio stream
6576 @var{stream} a directive telling the assembler to calculate the size of
6577 the symbol @var{name} by subtracting its address from the current
6580 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6581 provided. The default assumes that the assembler recognizes a special
6582 @samp{.} symbol as referring to the current address, and can calculate
6583 the difference between this and another symbol. If your assembler does
6584 not recognize @samp{.} or cannot do calculations with it, you will need
6585 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6589 A C string containing the appropriate assembler directive to specify the
6590 type of a symbol, without any arguments. On systems that use ELF, the
6591 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6592 systems, the default is not to define this macro.
6594 Define this macro only if it is correct to use the default definition of
6595 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6596 custom definition of this macro, or if you do not need explicit symbol
6597 types at all, do not define this macro.
6600 @defmac TYPE_OPERAND_FMT
6601 A C string which specifies (using @code{printf} syntax) the format of
6602 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6603 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6604 the default is not to define this macro.
6606 Define this macro only if it is correct to use the default definition of
6607 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6608 custom definition of this macro, or if you do not need explicit symbol
6609 types at all, do not define this macro.
6612 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6613 A C statement (sans semicolon) to output to the stdio stream
6614 @var{stream} a directive telling the assembler that the type of the
6615 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6616 that string is always either @samp{"function"} or @samp{"object"}, but
6617 you should not count on this.
6619 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6620 definition of this macro is provided.
6623 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6624 A C statement (sans semicolon) to output to the stdio stream
6625 @var{stream} any text necessary for declaring the name @var{name} of a
6626 function which is being defined. This macro is responsible for
6627 outputting the label definition (perhaps using
6628 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6629 @code{FUNCTION_DECL} tree node representing the function.
6631 If this macro is not defined, then the function name is defined in the
6632 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6634 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6638 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6639 A C statement (sans semicolon) to output to the stdio stream
6640 @var{stream} any text necessary for declaring the size of a function
6641 which is being defined. The argument @var{name} is the name of the
6642 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6643 representing the function.
6645 If this macro is not defined, then the function size is not defined.
6647 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6651 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6652 A C statement (sans semicolon) to output to the stdio stream
6653 @var{stream} any text necessary for declaring the name @var{name} of an
6654 initialized variable which is being defined. This macro must output the
6655 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6656 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6658 If this macro is not defined, then the variable name is defined in the
6659 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6661 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6662 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6665 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
6666 A C statement (sans semicolon) to output to the stdio stream
6667 @var{stream} any text necessary for declaring the name @var{name} of a
6668 constant which is being defined. This macro is responsible for
6669 outputting the label definition (perhaps using
6670 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
6671 value of the constant, and @var{size} is the size of the constant
6672 in bytes. @var{name} will be an internal label.
6674 If this macro is not defined, then the @var{name} is defined in the
6675 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6677 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6681 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6682 A C statement (sans semicolon) to output to the stdio stream
6683 @var{stream} any text necessary for claiming a register @var{regno}
6684 for a global variable @var{decl} with name @var{name}.
6686 If you don't define this macro, that is equivalent to defining it to do
6690 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6691 A C statement (sans semicolon) to finish up declaring a variable name
6692 once the compiler has processed its initializer fully and thus has had a
6693 chance to determine the size of an array when controlled by an
6694 initializer. This is used on systems where it's necessary to declare
6695 something about the size of the object.
6697 If you don't define this macro, that is equivalent to defining it to do
6700 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6701 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6704 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6705 This target hook is a function to output to the stdio stream
6706 @var{stream} some commands that will make the label @var{name} global;
6707 that is, available for reference from other files.
6709 The default implementation relies on a proper definition of
6710 @code{GLOBAL_ASM_OP}.
6713 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6714 A C statement (sans semicolon) to output to the stdio stream
6715 @var{stream} some commands that will make the label @var{name} weak;
6716 that is, available for reference from other files but only used if
6717 no other definition is available. Use the expression
6718 @code{assemble_name (@var{stream}, @var{name})} to output the name
6719 itself; before and after that, output the additional assembler syntax
6720 for making that name weak, and a newline.
6722 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6723 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6727 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6728 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6729 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6730 or variable decl. If @var{value} is not @code{NULL}, this C statement
6731 should output to the stdio stream @var{stream} assembler code which
6732 defines (equates) the weak symbol @var{name} to have the value
6733 @var{value}. If @var{value} is @code{NULL}, it should output commands
6734 to make @var{name} weak.
6737 @defmac SUPPORTS_WEAK
6738 A C expression which evaluates to true if the target supports weak symbols.
6740 If you don't define this macro, @file{defaults.h} provides a default
6741 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6742 is defined, the default definition is @samp{1}; otherwise, it is
6743 @samp{0}. Define this macro if you want to control weak symbol support
6744 with a compiler flag such as @option{-melf}.
6747 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
6748 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6749 public symbol such that extra copies in multiple translation units will
6750 be discarded by the linker. Define this macro if your object file
6751 format provides support for this concept, such as the @samp{COMDAT}
6752 section flags in the Microsoft Windows PE/COFF format, and this support
6753 requires changes to @var{decl}, such as putting it in a separate section.
6756 @defmac SUPPORTS_ONE_ONLY
6757 A C expression which evaluates to true if the target supports one-only
6760 If you don't define this macro, @file{varasm.c} provides a default
6761 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6762 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6763 you want to control one-only symbol support with a compiler flag, or if
6764 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6765 be emitted as one-only.
6768 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6769 This target hook is a function to output to @var{asm_out_file} some
6770 commands that will make the symbol(s) associated with @var{decl} have
6771 hidden, protected or internal visibility as specified by @var{visibility}.
6774 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
6775 A C expression that evaluates to true if the target's linker expects
6776 that weak symbols do not appear in a static archive's table of contents.
6777 The default is @code{0}.
6779 Leaving weak symbols out of an archive's table of contents means that,
6780 if a symbol will only have a definition in one translation unit and
6781 will have undefined references from other translation units, that
6782 symbol should not be weak. Defining this macro to be nonzero will
6783 thus have the effect that certain symbols that would normally be weak
6784 (explicit template instantiations, and vtables for polymorphic classes
6785 with noninline key methods) will instead be nonweak.
6787 The C++ ABI requires this macro to be zero. Define this macro for
6788 targets where full C++ ABI compliance is impossible and where linker
6789 restrictions require weak symbols to be left out of a static archive's
6793 @defmac TARGET_SUPPORTS_HIDDEN
6794 A C expression that evaluates to true if the target supports hidden
6795 visibility. By default this expression is true if and only if
6796 @code{HAS_GAS_HIDDEN} is defined. Set this macro if the
6797 @code{HAS_GAS_HIDDEN} macro gives the wrong answer for this
6798 target. (For example, if the target's mechanism for supporting
6799 hidden visibility is not the same as GAS's.)
6802 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6803 A C statement (sans semicolon) to output to the stdio stream
6804 @var{stream} any text necessary for declaring the name of an external
6805 symbol named @var{name} which is referenced in this compilation but
6806 not defined. The value of @var{decl} is the tree node for the
6809 This macro need not be defined if it does not need to output anything.
6810 The GNU assembler and most Unix assemblers don't require anything.
6813 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
6814 This target hook is a function to output to @var{asm_out_file} an assembler
6815 pseudo-op to declare a library function name external. The name of the
6816 library function is given by @var{symref}, which is a @code{symbol_ref}.
6819 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6820 A C statement (sans semicolon) to output to the stdio stream
6821 @var{stream} a reference in assembler syntax to a label named
6822 @var{name}. This should add @samp{_} to the front of the name, if that
6823 is customary on your operating system, as it is in most Berkeley Unix
6824 systems. This macro is used in @code{assemble_name}.
6827 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6828 A C statement (sans semicolon) to output a reference to
6829 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6830 will be used to output the name of the symbol. This macro may be used
6831 to modify the way a symbol is referenced depending on information
6832 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6835 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6836 A C statement (sans semicolon) to output a reference to @var{buf}, the
6837 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6838 @code{assemble_name} will be used to output the name of the symbol.
6839 This macro is not used by @code{output_asm_label}, or the @code{%l}
6840 specifier that calls it; the intention is that this macro should be set
6841 when it is necessary to output a label differently when its address is
6845 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6846 A function to output to the stdio stream @var{stream} a label whose
6847 name is made from the string @var{prefix} and the number @var{labelno}.
6849 It is absolutely essential that these labels be distinct from the labels
6850 used for user-level functions and variables. Otherwise, certain programs
6851 will have name conflicts with internal labels.
6853 It is desirable to exclude internal labels from the symbol table of the
6854 object file. Most assemblers have a naming convention for labels that
6855 should be excluded; on many systems, the letter @samp{L} at the
6856 beginning of a label has this effect. You should find out what
6857 convention your system uses, and follow it.
6859 The default version of this function utilizes ASM_GENERATE_INTERNAL_LABEL.
6862 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6863 A C statement to output to the stdio stream @var{stream} a debug info
6864 label whose name is made from the string @var{prefix} and the number
6865 @var{num}. This is useful for VLIW targets, where debug info labels
6866 may need to be treated differently than branch target labels. On some
6867 systems, branch target labels must be at the beginning of instruction
6868 bundles, but debug info labels can occur in the middle of instruction
6871 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
6875 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6876 A C statement to store into the string @var{string} a label whose name
6877 is made from the string @var{prefix} and the number @var{num}.
6879 This string, when output subsequently by @code{assemble_name}, should
6880 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
6881 with the same @var{prefix} and @var{num}.
6883 If the string begins with @samp{*}, then @code{assemble_name} will
6884 output the rest of the string unchanged. It is often convenient for
6885 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6886 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6887 to output the string, and may change it. (Of course,
6888 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6889 you should know what it does on your machine.)
6892 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6893 A C expression to assign to @var{outvar} (which is a variable of type
6894 @code{char *}) a newly allocated string made from the string
6895 @var{name} and the number @var{number}, with some suitable punctuation
6896 added. Use @code{alloca} to get space for the string.
6898 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6899 produce an assembler label for an internal static variable whose name is
6900 @var{name}. Therefore, the string must be such as to result in valid
6901 assembler code. The argument @var{number} is different each time this
6902 macro is executed; it prevents conflicts between similarly-named
6903 internal static variables in different scopes.
6905 Ideally this string should not be a valid C identifier, to prevent any
6906 conflict with the user's own symbols. Most assemblers allow periods
6907 or percent signs in assembler symbols; putting at least one of these
6908 between the name and the number will suffice.
6910 If this macro is not defined, a default definition will be provided
6911 which is correct for most systems.
6914 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6915 A C statement to output to the stdio stream @var{stream} assembler code
6916 which defines (equates) the symbol @var{name} to have the value @var{value}.
6919 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6920 correct for most systems.
6923 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6924 A C statement to output to the stdio stream @var{stream} assembler code
6925 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6926 to have the value of the tree node @var{decl_of_value}. This macro will
6927 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6928 the tree nodes are available.
6931 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6932 correct for most systems.
6935 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6936 A C statement to output to the stdio stream @var{stream} assembler code
6937 which defines (equates) the weak symbol @var{name} to have the value
6938 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6939 an undefined weak symbol.
6941 Define this macro if the target only supports weak aliases; define
6942 @code{ASM_OUTPUT_DEF} instead if possible.
6945 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6946 Define this macro to override the default assembler names used for
6947 Objective-C methods.
6949 The default name is a unique method number followed by the name of the
6950 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6951 the category is also included in the assembler name (e.g.@:
6954 These names are safe on most systems, but make debugging difficult since
6955 the method's selector is not present in the name. Therefore, particular
6956 systems define other ways of computing names.
6958 @var{buf} is an expression of type @code{char *} which gives you a
6959 buffer in which to store the name; its length is as long as
6960 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6961 50 characters extra.
6963 The argument @var{is_inst} specifies whether the method is an instance
6964 method or a class method; @var{class_name} is the name of the class;
6965 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6966 in a category); and @var{sel_name} is the name of the selector.
6968 On systems where the assembler can handle quoted names, you can use this
6969 macro to provide more human-readable names.
6972 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
6973 A C statement (sans semicolon) to output to the stdio stream
6974 @var{stream} commands to declare that the label @var{name} is an
6975 Objective-C class reference. This is only needed for targets whose
6976 linkers have special support for NeXT-style runtimes.
6979 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6980 A C statement (sans semicolon) to output to the stdio stream
6981 @var{stream} commands to declare that the label @var{name} is an
6982 unresolved Objective-C class reference. This is only needed for targets
6983 whose linkers have special support for NeXT-style runtimes.
6986 @node Initialization
6987 @subsection How Initialization Functions Are Handled
6988 @cindex initialization routines
6989 @cindex termination routines
6990 @cindex constructors, output of
6991 @cindex destructors, output of
6993 The compiled code for certain languages includes @dfn{constructors}
6994 (also called @dfn{initialization routines})---functions to initialize
6995 data in the program when the program is started. These functions need
6996 to be called before the program is ``started''---that is to say, before
6997 @code{main} is called.
6999 Compiling some languages generates @dfn{destructors} (also called
7000 @dfn{termination routines}) that should be called when the program
7003 To make the initialization and termination functions work, the compiler
7004 must output something in the assembler code to cause those functions to
7005 be called at the appropriate time. When you port the compiler to a new
7006 system, you need to specify how to do this.
7008 There are two major ways that GCC currently supports the execution of
7009 initialization and termination functions. Each way has two variants.
7010 Much of the structure is common to all four variations.
7012 @findex __CTOR_LIST__
7013 @findex __DTOR_LIST__
7014 The linker must build two lists of these functions---a list of
7015 initialization functions, called @code{__CTOR_LIST__}, and a list of
7016 termination functions, called @code{__DTOR_LIST__}.
7018 Each list always begins with an ignored function pointer (which may hold
7019 0, @minus{}1, or a count of the function pointers after it, depending on
7020 the environment). This is followed by a series of zero or more function
7021 pointers to constructors (or destructors), followed by a function
7022 pointer containing zero.
7024 Depending on the operating system and its executable file format, either
7025 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7026 time and exit time. Constructors are called in reverse order of the
7027 list; destructors in forward order.
7029 The best way to handle static constructors works only for object file
7030 formats which provide arbitrarily-named sections. A section is set
7031 aside for a list of constructors, and another for a list of destructors.
7032 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7033 object file that defines an initialization function also puts a word in
7034 the constructor section to point to that function. The linker
7035 accumulates all these words into one contiguous @samp{.ctors} section.
7036 Termination functions are handled similarly.
7038 This method will be chosen as the default by @file{target-def.h} if
7039 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7040 support arbitrary sections, but does support special designated
7041 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7042 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7044 When arbitrary sections are available, there are two variants, depending
7045 upon how the code in @file{crtstuff.c} is called. On systems that
7046 support a @dfn{.init} section which is executed at program startup,
7047 parts of @file{crtstuff.c} are compiled into that section. The
7048 program is linked by the @command{gcc} driver like this:
7051 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7054 The prologue of a function (@code{__init}) appears in the @code{.init}
7055 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7056 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7057 files are provided by the operating system or by the GNU C library, but
7058 are provided by GCC for a few targets.
7060 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7061 compiled from @file{crtstuff.c}. They contain, among other things, code
7062 fragments within the @code{.init} and @code{.fini} sections that branch
7063 to routines in the @code{.text} section. The linker will pull all parts
7064 of a section together, which results in a complete @code{__init} function
7065 that invokes the routines we need at startup.
7067 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7070 If no init section is available, when GCC compiles any function called
7071 @code{main} (or more accurately, any function designated as a program
7072 entry point by the language front end calling @code{expand_main_function}),
7073 it inserts a procedure call to @code{__main} as the first executable code
7074 after the function prologue. The @code{__main} function is defined
7075 in @file{libgcc2.c} and runs the global constructors.
7077 In file formats that don't support arbitrary sections, there are again
7078 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7079 and an `a.out' format must be used. In this case,
7080 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7081 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7082 and with the address of the void function containing the initialization
7083 code as its value. The GNU linker recognizes this as a request to add
7084 the value to a @dfn{set}; the values are accumulated, and are eventually
7085 placed in the executable as a vector in the format described above, with
7086 a leading (ignored) count and a trailing zero element.
7087 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7088 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7089 the compilation of @code{main} to call @code{__main} as above, starting
7090 the initialization process.
7092 The last variant uses neither arbitrary sections nor the GNU linker.
7093 This is preferable when you want to do dynamic linking and when using
7094 file formats which the GNU linker does not support, such as `ECOFF'@. In
7095 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7096 termination functions are recognized simply by their names. This requires
7097 an extra program in the linkage step, called @command{collect2}. This program
7098 pretends to be the linker, for use with GCC; it does its job by running
7099 the ordinary linker, but also arranges to include the vectors of
7100 initialization and termination functions. These functions are called
7101 via @code{__main} as described above. In order to use this method,
7102 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7105 The following section describes the specific macros that control and
7106 customize the handling of initialization and termination functions.
7109 @node Macros for Initialization
7110 @subsection Macros Controlling Initialization Routines
7112 Here are the macros that control how the compiler handles initialization
7113 and termination functions:
7115 @defmac INIT_SECTION_ASM_OP
7116 If defined, a C string constant, including spacing, for the assembler
7117 operation to identify the following data as initialization code. If not
7118 defined, GCC will assume such a section does not exist. When you are
7119 using special sections for initialization and termination functions, this
7120 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7121 run the initialization functions.
7124 @defmac HAS_INIT_SECTION
7125 If defined, @code{main} will not call @code{__main} as described above.
7126 This macro should be defined for systems that control start-up code
7127 on a symbol-by-symbol basis, such as OSF/1, and should not
7128 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7131 @defmac LD_INIT_SWITCH
7132 If defined, a C string constant for a switch that tells the linker that
7133 the following symbol is an initialization routine.
7136 @defmac LD_FINI_SWITCH
7137 If defined, a C string constant for a switch that tells the linker that
7138 the following symbol is a finalization routine.
7141 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7142 If defined, a C statement that will write a function that can be
7143 automatically called when a shared library is loaded. The function
7144 should call @var{func}, which takes no arguments. If not defined, and
7145 the object format requires an explicit initialization function, then a
7146 function called @code{_GLOBAL__DI} will be generated.
7148 This function and the following one are used by collect2 when linking a
7149 shared library that needs constructors or destructors, or has DWARF2
7150 exception tables embedded in the code.
7153 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7154 If defined, a C statement that will write a function that can be
7155 automatically called when a shared library is unloaded. The function
7156 should call @var{func}, which takes no arguments. If not defined, and
7157 the object format requires an explicit finalization function, then a
7158 function called @code{_GLOBAL__DD} will be generated.
7161 @defmac INVOKE__main
7162 If defined, @code{main} will call @code{__main} despite the presence of
7163 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7164 where the init section is not actually run automatically, but is still
7165 useful for collecting the lists of constructors and destructors.
7168 @defmac SUPPORTS_INIT_PRIORITY
7169 If nonzero, the C++ @code{init_priority} attribute is supported and the
7170 compiler should emit instructions to control the order of initialization
7171 of objects. If zero, the compiler will issue an error message upon
7172 encountering an @code{init_priority} attribute.
7175 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7176 This value is true if the target supports some ``native'' method of
7177 collecting constructors and destructors to be run at startup and exit.
7178 It is false if we must use @command{collect2}.
7181 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7182 If defined, a function that outputs assembler code to arrange to call
7183 the function referenced by @var{symbol} at initialization time.
7185 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7186 no arguments and with no return value. If the target supports initialization
7187 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7188 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7190 If this macro is not defined by the target, a suitable default will
7191 be chosen if (1) the target supports arbitrary section names, (2) the
7192 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7196 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7197 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7198 functions rather than initialization functions.
7201 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7202 generated for the generated object file will have static linkage.
7204 If your system uses @command{collect2} as the means of processing
7205 constructors, then that program normally uses @command{nm} to scan
7206 an object file for constructor functions to be called.
7208 On certain kinds of systems, you can define this macro to make
7209 @command{collect2} work faster (and, in some cases, make it work at all):
7211 @defmac OBJECT_FORMAT_COFF
7212 Define this macro if the system uses COFF (Common Object File Format)
7213 object files, so that @command{collect2} can assume this format and scan
7214 object files directly for dynamic constructor/destructor functions.
7216 This macro is effective only in a native compiler; @command{collect2} as
7217 part of a cross compiler always uses @command{nm} for the target machine.
7220 @defmac COLLECT_PARSE_FLAG (@var{flag})
7221 Define this macro to be C code that examines @command{collect2} command
7222 line option @var{flag} and performs special actions if
7223 @command{collect2} needs to behave differently depending on @var{flag}.
7226 @defmac REAL_NM_FILE_NAME
7227 Define this macro as a C string constant containing the file name to use
7228 to execute @command{nm}. The default is to search the path normally for
7231 If your system supports shared libraries and has a program to list the
7232 dynamic dependencies of a given library or executable, you can define
7233 these macros to enable support for running initialization and
7234 termination functions in shared libraries:
7238 Define this macro to a C string constant containing the name of the program
7239 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7242 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7243 Define this macro to be C code that extracts filenames from the output
7244 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7245 of type @code{char *} that points to the beginning of a line of output
7246 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7247 code must advance @var{ptr} to the beginning of the filename on that
7248 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7251 @node Instruction Output
7252 @subsection Output of Assembler Instructions
7254 @c prevent bad page break with this line
7255 This describes assembler instruction output.
7257 @defmac REGISTER_NAMES
7258 A C initializer containing the assembler's names for the machine
7259 registers, each one as a C string constant. This is what translates
7260 register numbers in the compiler into assembler language.
7263 @defmac ADDITIONAL_REGISTER_NAMES
7264 If defined, a C initializer for an array of structures containing a name
7265 and a register number. This macro defines additional names for hard
7266 registers, thus allowing the @code{asm} option in declarations to refer
7267 to registers using alternate names.
7270 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7271 Define this macro if you are using an unusual assembler that
7272 requires different names for the machine instructions.
7274 The definition is a C statement or statements which output an
7275 assembler instruction opcode to the stdio stream @var{stream}. The
7276 macro-operand @var{ptr} is a variable of type @code{char *} which
7277 points to the opcode name in its ``internal'' form---the form that is
7278 written in the machine description. The definition should output the
7279 opcode name to @var{stream}, performing any translation you desire, and
7280 increment the variable @var{ptr} to point at the end of the opcode
7281 so that it will not be output twice.
7283 In fact, your macro definition may process less than the entire opcode
7284 name, or more than the opcode name; but if you want to process text
7285 that includes @samp{%}-sequences to substitute operands, you must take
7286 care of the substitution yourself. Just be sure to increment
7287 @var{ptr} over whatever text should not be output normally.
7289 @findex recog_data.operand
7290 If you need to look at the operand values, they can be found as the
7291 elements of @code{recog_data.operand}.
7293 If the macro definition does nothing, the instruction is output
7297 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7298 If defined, a C statement to be executed just prior to the output of
7299 assembler code for @var{insn}, to modify the extracted operands so
7300 they will be output differently.
7302 Here the argument @var{opvec} is the vector containing the operands
7303 extracted from @var{insn}, and @var{noperands} is the number of
7304 elements of the vector which contain meaningful data for this insn.
7305 The contents of this vector are what will be used to convert the insn
7306 template into assembler code, so you can change the assembler output
7307 by changing the contents of the vector.
7309 This macro is useful when various assembler syntaxes share a single
7310 file of instruction patterns; by defining this macro differently, you
7311 can cause a large class of instructions to be output differently (such
7312 as with rearranged operands). Naturally, variations in assembler
7313 syntax affecting individual insn patterns ought to be handled by
7314 writing conditional output routines in those patterns.
7316 If this macro is not defined, it is equivalent to a null statement.
7319 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7320 A C compound statement to output to stdio stream @var{stream} the
7321 assembler syntax for an instruction operand @var{x}. @var{x} is an
7324 @var{code} is a value that can be used to specify one of several ways
7325 of printing the operand. It is used when identical operands must be
7326 printed differently depending on the context. @var{code} comes from
7327 the @samp{%} specification that was used to request printing of the
7328 operand. If the specification was just @samp{%@var{digit}} then
7329 @var{code} is 0; if the specification was @samp{%@var{ltr}
7330 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7333 If @var{x} is a register, this macro should print the register's name.
7334 The names can be found in an array @code{reg_names} whose type is
7335 @code{char *[]}. @code{reg_names} is initialized from
7336 @code{REGISTER_NAMES}.
7338 When the machine description has a specification @samp{%@var{punct}}
7339 (a @samp{%} followed by a punctuation character), this macro is called
7340 with a null pointer for @var{x} and the punctuation character for
7344 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7345 A C expression which evaluates to true if @var{code} is a valid
7346 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7347 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7348 punctuation characters (except for the standard one, @samp{%}) are used
7352 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7353 A C compound statement to output to stdio stream @var{stream} the
7354 assembler syntax for an instruction operand that is a memory reference
7355 whose address is @var{x}. @var{x} is an RTL expression.
7357 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7358 On some machines, the syntax for a symbolic address depends on the
7359 section that the address refers to. On these machines, define the hook
7360 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7361 @code{symbol_ref}, and then check for it here. @xref{Assembler
7365 @findex dbr_sequence_length
7366 @defmac DBR_OUTPUT_SEQEND (@var{file})
7367 A C statement, to be executed after all slot-filler instructions have
7368 been output. If necessary, call @code{dbr_sequence_length} to
7369 determine the number of slots filled in a sequence (zero if not
7370 currently outputting a sequence), to decide how many no-ops to output,
7373 Don't define this macro if it has nothing to do, but it is helpful in
7374 reading assembly output if the extent of the delay sequence is made
7375 explicit (e.g.@: with white space).
7378 @findex final_sequence
7379 Note that output routines for instructions with delay slots must be
7380 prepared to deal with not being output as part of a sequence
7381 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7382 found.) The variable @code{final_sequence} is null when not
7383 processing a sequence, otherwise it contains the @code{sequence} rtx
7387 @defmac REGISTER_PREFIX
7388 @defmacx LOCAL_LABEL_PREFIX
7389 @defmacx USER_LABEL_PREFIX
7390 @defmacx IMMEDIATE_PREFIX
7391 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7392 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7393 @file{final.c}). These are useful when a single @file{md} file must
7394 support multiple assembler formats. In that case, the various @file{tm.h}
7395 files can define these macros differently.
7398 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
7399 If defined this macro should expand to a series of @code{case}
7400 statements which will be parsed inside the @code{switch} statement of
7401 the @code{asm_fprintf} function. This allows targets to define extra
7402 printf formats which may useful when generating their assembler
7403 statements. Note that uppercase letters are reserved for future
7404 generic extensions to asm_fprintf, and so are not available to target
7405 specific code. The output file is given by the parameter @var{file}.
7406 The varargs input pointer is @var{argptr} and the rest of the format
7407 string, starting the character after the one that is being switched
7408 upon, is pointed to by @var{format}.
7411 @defmac ASSEMBLER_DIALECT
7412 If your target supports multiple dialects of assembler language (such as
7413 different opcodes), define this macro as a C expression that gives the
7414 numeric index of the assembler language dialect to use, with zero as the
7417 If this macro is defined, you may use constructs of the form
7419 @samp{@{option0|option1|option2@dots{}@}}
7422 in the output templates of patterns (@pxref{Output Template}) or in the
7423 first argument of @code{asm_fprintf}. This construct outputs
7424 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7425 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7426 within these strings retain their usual meaning. If there are fewer
7427 alternatives within the braces than the value of
7428 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7430 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7431 @samp{@}} do not have any special meaning when used in templates or
7432 operands to @code{asm_fprintf}.
7434 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7435 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7436 the variations in assembler language syntax with that mechanism. Define
7437 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7438 if the syntax variant are larger and involve such things as different
7439 opcodes or operand order.
7442 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7443 A C expression to output to @var{stream} some assembler code
7444 which will push hard register number @var{regno} onto the stack.
7445 The code need not be optimal, since this macro is used only when
7449 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7450 A C expression to output to @var{stream} some assembler code
7451 which will pop hard register number @var{regno} off of the stack.
7452 The code need not be optimal, since this macro is used only when
7456 @node Dispatch Tables
7457 @subsection Output of Dispatch Tables
7459 @c prevent bad page break with this line
7460 This concerns dispatch tables.
7462 @cindex dispatch table
7463 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7464 A C statement to output to the stdio stream @var{stream} an assembler
7465 pseudo-instruction to generate a difference between two labels.
7466 @var{value} and @var{rel} are the numbers of two internal labels. The
7467 definitions of these labels are output using
7468 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7469 way here. For example,
7472 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7473 @var{value}, @var{rel})
7476 You must provide this macro on machines where the addresses in a
7477 dispatch table are relative to the table's own address. If defined, GCC
7478 will also use this macro on all machines when producing PIC@.
7479 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7480 mode and flags can be read.
7483 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7484 This macro should be provided on machines where the addresses
7485 in a dispatch table are absolute.
7487 The definition should be a C statement to output to the stdio stream
7488 @var{stream} an assembler pseudo-instruction to generate a reference to
7489 a label. @var{value} is the number of an internal label whose
7490 definition is output using @code{(*targetm.asm_out.internal_label)}.
7494 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7498 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7499 Define this if the label before a jump-table needs to be output
7500 specially. The first three arguments are the same as for
7501 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7502 jump-table which follows (a @code{jump_insn} containing an
7503 @code{addr_vec} or @code{addr_diff_vec}).
7505 This feature is used on system V to output a @code{swbeg} statement
7508 If this macro is not defined, these labels are output with
7509 @code{(*targetm.asm_out.internal_label)}.
7512 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7513 Define this if something special must be output at the end of a
7514 jump-table. The definition should be a C statement to be executed
7515 after the assembler code for the table is written. It should write
7516 the appropriate code to stdio stream @var{stream}. The argument
7517 @var{table} is the jump-table insn, and @var{num} is the label-number
7518 of the preceding label.
7520 If this macro is not defined, nothing special is output at the end of
7524 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
7525 This target hook emits a label at the beginning of each FDE. It
7526 should be defined on targets where FDEs need special labels, and it
7527 should write the appropriate label, for the FDE associated with the
7528 function declaration @var{decl}, to the stdio stream @var{stream}.
7529 The third argument, @var{for_eh}, is a boolean: true if this is for an
7530 exception table. The fourth argument, @var{empty}, is a boolean:
7531 true if this is a placeholder label for an omitted FDE.
7533 The default is that FDEs are not given nonlocal labels.
7536 @node Exception Region Output
7537 @subsection Assembler Commands for Exception Regions
7539 @c prevent bad page break with this line
7541 This describes commands marking the start and the end of an exception
7544 @defmac EH_FRAME_SECTION_NAME
7545 If defined, a C string constant for the name of the section containing
7546 exception handling frame unwind information. If not defined, GCC will
7547 provide a default definition if the target supports named sections.
7548 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7550 You should define this symbol if your target supports DWARF 2 frame
7551 unwind information and the default definition does not work.
7554 @defmac EH_FRAME_IN_DATA_SECTION
7555 If defined, DWARF 2 frame unwind information will be placed in the
7556 data section even though the target supports named sections. This
7557 might be necessary, for instance, if the system linker does garbage
7558 collection and sections cannot be marked as not to be collected.
7560 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7564 @defmac MASK_RETURN_ADDR
7565 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7566 that it does not contain any extraneous set bits in it.
7569 @defmac DWARF2_UNWIND_INFO
7570 Define this macro to 0 if your target supports DWARF 2 frame unwind
7571 information, but it does not yet work with exception handling.
7572 Otherwise, if your target supports this information (if it defines
7573 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7574 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7577 If this macro is defined to 1, the DWARF 2 unwinder will be the default
7578 exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by
7581 If this macro is defined to anything, the DWARF 2 unwinder will be used
7582 instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case.
7585 @defmac MUST_USE_SJLJ_EXCEPTIONS
7586 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
7587 runtime-variable. In that case, @file{except.h} cannot correctly
7588 determine the corresponding definition of
7589 @code{MUST_USE_SJLJ_EXCEPTIONS}, so the target must provide it directly.
7592 @defmac DWARF_CIE_DATA_ALIGNMENT
7593 This macro need only be defined if the target might save registers in the
7594 function prologue at an offset to the stack pointer that is not aligned to
7595 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7596 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7597 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7598 the target supports DWARF 2 frame unwind information.
7601 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7602 If defined, a function that switches to the section in which the main
7603 exception table is to be placed (@pxref{Sections}). The default is a
7604 function that switches to a section named @code{.gcc_except_table} on
7605 machines that support named sections via
7606 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7607 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7608 @code{readonly_data_section}.
7611 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7612 If defined, a function that switches to the section in which the DWARF 2
7613 frame unwind information to be placed (@pxref{Sections}). The default
7614 is a function that outputs a standard GAS section directive, if
7615 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7616 directive followed by a synthetic label.
7619 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7620 Contains the value true if the target should add a zero word onto the
7621 end of a Dwarf-2 frame info section when used for exception handling.
7622 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7626 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
7627 Given a register, this hook should return a parallel of registers to
7628 represent where to find the register pieces. Define this hook if the
7629 register and its mode are represented in Dwarf in non-contiguous
7630 locations, or if the register should be represented in more than one
7631 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
7632 If not defined, the default is to return @code{NULL_RTX}.
7635 @node Alignment Output
7636 @subsection Assembler Commands for Alignment
7638 @c prevent bad page break with this line
7639 This describes commands for alignment.
7641 @defmac JUMP_ALIGN (@var{label})
7642 The alignment (log base 2) to put in front of @var{label}, which is
7643 a common destination of jumps and has no fallthru incoming edge.
7645 This macro need not be defined if you don't want any special alignment
7646 to be done at such a time. Most machine descriptions do not currently
7649 Unless it's necessary to inspect the @var{label} parameter, it is better
7650 to set the variable @var{align_jumps} in the target's
7651 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7652 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7655 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
7656 The alignment (log base 2) to put in front of @var{label}, which follows
7659 This macro need not be defined if you don't want any special alignment
7660 to be done at such a time. Most machine descriptions do not currently
7664 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7665 The maximum number of bytes to skip when applying
7666 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7667 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7670 @defmac LOOP_ALIGN (@var{label})
7671 The alignment (log base 2) to put in front of @var{label}, which follows
7672 a @code{NOTE_INSN_LOOP_BEG} note.
7674 This macro need not be defined if you don't want any special alignment
7675 to be done at such a time. Most machine descriptions do not currently
7678 Unless it's necessary to inspect the @var{label} parameter, it is better
7679 to set the variable @code{align_loops} in the target's
7680 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7681 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7684 @defmac LOOP_ALIGN_MAX_SKIP
7685 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7686 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7689 @defmac LABEL_ALIGN (@var{label})
7690 The alignment (log base 2) to put in front of @var{label}.
7691 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7692 the maximum of the specified values is used.
7694 Unless it's necessary to inspect the @var{label} parameter, it is better
7695 to set the variable @code{align_labels} in the target's
7696 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7697 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7700 @defmac LABEL_ALIGN_MAX_SKIP
7701 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7702 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7705 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7706 A C statement to output to the stdio stream @var{stream} an assembler
7707 instruction to advance the location counter by @var{nbytes} bytes.
7708 Those bytes should be zero when loaded. @var{nbytes} will be a C
7709 expression of type @code{int}.
7712 @defmac ASM_NO_SKIP_IN_TEXT
7713 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7714 text section because it fails to put zeros in the bytes that are skipped.
7715 This is true on many Unix systems, where the pseudo--op to skip bytes
7716 produces no-op instructions rather than zeros when used in the text
7720 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7721 A C statement to output to the stdio stream @var{stream} an assembler
7722 command to advance the location counter to a multiple of 2 to the
7723 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7726 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7727 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7728 for padding, if necessary.
7731 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7732 A C statement to output to the stdio stream @var{stream} an assembler
7733 command to advance the location counter to a multiple of 2 to the
7734 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7735 satisfy the alignment request. @var{power} and @var{max_skip} will be
7736 a C expression of type @code{int}.
7740 @node Debugging Info
7741 @section Controlling Debugging Information Format
7743 @c prevent bad page break with this line
7744 This describes how to specify debugging information.
7747 * All Debuggers:: Macros that affect all debugging formats uniformly.
7748 * DBX Options:: Macros enabling specific options in DBX format.
7749 * DBX Hooks:: Hook macros for varying DBX format.
7750 * File Names and DBX:: Macros controlling output of file names in DBX format.
7751 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7752 * VMS Debug:: Macros for VMS debug format.
7756 @subsection Macros Affecting All Debugging Formats
7758 @c prevent bad page break with this line
7759 These macros affect all debugging formats.
7761 @defmac DBX_REGISTER_NUMBER (@var{regno})
7762 A C expression that returns the DBX register number for the compiler
7763 register number @var{regno}. In the default macro provided, the value
7764 of this expression will be @var{regno} itself. But sometimes there are
7765 some registers that the compiler knows about and DBX does not, or vice
7766 versa. In such cases, some register may need to have one number in the
7767 compiler and another for DBX@.
7769 If two registers have consecutive numbers inside GCC, and they can be
7770 used as a pair to hold a multiword value, then they @emph{must} have
7771 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7772 Otherwise, debuggers will be unable to access such a pair, because they
7773 expect register pairs to be consecutive in their own numbering scheme.
7775 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7776 does not preserve register pairs, then what you must do instead is
7777 redefine the actual register numbering scheme.
7780 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
7781 A C expression that returns the integer offset value for an automatic
7782 variable having address @var{x} (an RTL expression). The default
7783 computation assumes that @var{x} is based on the frame-pointer and
7784 gives the offset from the frame-pointer. This is required for targets
7785 that produce debugging output for DBX or COFF-style debugging output
7786 for SDB and allow the frame-pointer to be eliminated when the
7787 @option{-g} options is used.
7790 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7791 A C expression that returns the integer offset value for an argument
7792 having address @var{x} (an RTL expression). The nominal offset is
7796 @defmac PREFERRED_DEBUGGING_TYPE
7797 A C expression that returns the type of debugging output GCC should
7798 produce when the user specifies just @option{-g}. Define
7799 this if you have arranged for GCC to support more than one format of
7800 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7801 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7802 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7804 When the user specifies @option{-ggdb}, GCC normally also uses the
7805 value of this macro to select the debugging output format, but with two
7806 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
7807 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7808 defined, GCC uses @code{DBX_DEBUG}.
7810 The value of this macro only affects the default debugging output; the
7811 user can always get a specific type of output by using @option{-gstabs},
7812 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
7816 @subsection Specific Options for DBX Output
7818 @c prevent bad page break with this line
7819 These are specific options for DBX output.
7821 @defmac DBX_DEBUGGING_INFO
7822 Define this macro if GCC should produce debugging output for DBX
7823 in response to the @option{-g} option.
7826 @defmac XCOFF_DEBUGGING_INFO
7827 Define this macro if GCC should produce XCOFF format debugging output
7828 in response to the @option{-g} option. This is a variant of DBX format.
7831 @defmac DEFAULT_GDB_EXTENSIONS
7832 Define this macro to control whether GCC should by default generate
7833 GDB's extended version of DBX debugging information (assuming DBX-format
7834 debugging information is enabled at all). If you don't define the
7835 macro, the default is 1: always generate the extended information
7836 if there is any occasion to.
7839 @defmac DEBUG_SYMS_TEXT
7840 Define this macro if all @code{.stabs} commands should be output while
7841 in the text section.
7844 @defmac ASM_STABS_OP
7845 A C string constant, including spacing, naming the assembler pseudo op to
7846 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7847 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7848 applies only to DBX debugging information format.
7851 @defmac ASM_STABD_OP
7852 A C string constant, including spacing, naming the assembler pseudo op to
7853 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7854 value is the current location. If you don't define this macro,
7855 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7859 @defmac ASM_STABN_OP
7860 A C string constant, including spacing, naming the assembler pseudo op to
7861 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7862 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7863 macro applies only to DBX debugging information format.
7866 @defmac DBX_NO_XREFS
7867 Define this macro if DBX on your system does not support the construct
7868 @samp{xs@var{tagname}}. On some systems, this construct is used to
7869 describe a forward reference to a structure named @var{tagname}.
7870 On other systems, this construct is not supported at all.
7873 @defmac DBX_CONTIN_LENGTH
7874 A symbol name in DBX-format debugging information is normally
7875 continued (split into two separate @code{.stabs} directives) when it
7876 exceeds a certain length (by default, 80 characters). On some
7877 operating systems, DBX requires this splitting; on others, splitting
7878 must not be done. You can inhibit splitting by defining this macro
7879 with the value zero. You can override the default splitting-length by
7880 defining this macro as an expression for the length you desire.
7883 @defmac DBX_CONTIN_CHAR
7884 Normally continuation is indicated by adding a @samp{\} character to
7885 the end of a @code{.stabs} string when a continuation follows. To use
7886 a different character instead, define this macro as a character
7887 constant for the character you want to use. Do not define this macro
7888 if backslash is correct for your system.
7891 @defmac DBX_STATIC_STAB_DATA_SECTION
7892 Define this macro if it is necessary to go to the data section before
7893 outputting the @samp{.stabs} pseudo-op for a non-global static
7897 @defmac DBX_TYPE_DECL_STABS_CODE
7898 The value to use in the ``code'' field of the @code{.stabs} directive
7899 for a typedef. The default is @code{N_LSYM}.
7902 @defmac DBX_STATIC_CONST_VAR_CODE
7903 The value to use in the ``code'' field of the @code{.stabs} directive
7904 for a static variable located in the text section. DBX format does not
7905 provide any ``right'' way to do this. The default is @code{N_FUN}.
7908 @defmac DBX_REGPARM_STABS_CODE
7909 The value to use in the ``code'' field of the @code{.stabs} directive
7910 for a parameter passed in registers. DBX format does not provide any
7911 ``right'' way to do this. The default is @code{N_RSYM}.
7914 @defmac DBX_REGPARM_STABS_LETTER
7915 The letter to use in DBX symbol data to identify a symbol as a parameter
7916 passed in registers. DBX format does not customarily provide any way to
7917 do this. The default is @code{'P'}.
7920 @defmac DBX_MEMPARM_STABS_LETTER
7921 The letter to use in DBX symbol data to identify a symbol as a stack
7922 parameter. The default is @code{'p'}.
7925 @defmac DBX_FUNCTION_FIRST
7926 Define this macro if the DBX information for a function and its
7927 arguments should precede the assembler code for the function. Normally,
7928 in DBX format, the debugging information entirely follows the assembler
7932 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
7933 Define this macro if the value of a symbol describing the scope of a
7934 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7935 of the enclosing function. Normally, GCC uses an absolute address.
7938 @defmac DBX_USE_BINCL
7939 Define this macro if GCC should generate @code{N_BINCL} and
7940 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7941 macro also directs GCC to output a type number as a pair of a file
7942 number and a type number within the file. Normally, GCC does not
7943 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7944 number for a type number.
7948 @subsection Open-Ended Hooks for DBX Format
7950 @c prevent bad page break with this line
7951 These are hooks for DBX format.
7953 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7954 Define this macro to say how to output to @var{stream} the debugging
7955 information for the start of a scope level for variable names. The
7956 argument @var{name} is the name of an assembler symbol (for use with
7957 @code{assemble_name}) whose value is the address where the scope begins.
7960 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7961 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7964 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
7965 Define this macro if the target machine requires special handling to
7966 output an @code{N_FUN} entry for the function @var{decl}.
7969 @defmac DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7970 Define this macro if the target machine requires special output at the
7971 end of the debugging information for a function. The definition should
7972 be a C statement (sans semicolon) to output the appropriate information
7973 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7977 @defmac NO_DBX_FUNCTION_END
7978 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7979 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
7980 On those machines, define this macro to turn this feature off without
7981 disturbing the rest of the gdb extensions.
7984 @node File Names and DBX
7985 @subsection File Names in DBX Format
7987 @c prevent bad page break with this line
7988 This describes file names in DBX format.
7990 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7991 A C statement to output DBX debugging information to the stdio stream
7992 @var{stream} which indicates that file @var{name} is the main source
7993 file---the file specified as the input file for compilation.
7994 This macro is called only once, at the beginning of compilation.
7996 This macro need not be defined if the standard form of output
7997 for DBX debugging information is appropriate.
8000 @defmac DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
8001 A C statement to output DBX debugging information to the stdio stream
8002 @var{stream} which indicates that the current directory during
8003 compilation is named @var{name}.
8005 This macro need not be defined if the standard form of output
8006 for DBX debugging information is appropriate.
8009 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8010 A C statement to output DBX debugging information at the end of
8011 compilation of the main source file @var{name}.
8013 If you don't define this macro, nothing special is output at the end
8014 of compilation, which is correct for most machines.
8019 @subsection Macros for SDB and DWARF Output
8021 @c prevent bad page break with this line
8022 Here are macros for SDB and DWARF output.
8024 @defmac SDB_DEBUGGING_INFO
8025 Define this macro if GCC should produce COFF-style debugging output
8026 for SDB in response to the @option{-g} option.
8029 @defmac DWARF2_DEBUGGING_INFO
8030 Define this macro if GCC should produce dwarf version 2 format
8031 debugging output in response to the @option{-g} option.
8033 To support optional call frame debugging information, you must also
8034 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8035 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8036 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8037 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8040 @defmac DWARF2_FRAME_INFO
8041 Define this macro to a nonzero value if GCC should always output
8042 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8043 (@pxref{Exception Region Output} is nonzero, GCC will output this
8044 information not matter how you define @code{DWARF2_FRAME_INFO}.
8047 @defmac DWARF2_GENERATE_TEXT_SECTION_LABEL
8048 By default, the Dwarf 2 debugging information generator will generate a
8049 label to mark the beginning of the text section. If it is better simply
8050 to use the name of the text section itself, rather than an explicit label,
8051 to indicate the beginning of the text section, define this macro to zero.
8054 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8055 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8056 line debug info sections. This will result in much more compact line number
8057 tables, and hence is desirable if it works.
8060 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8061 A C statement to issue assembly directives that create a difference
8062 between the two given labels, using an integer of the given size.
8065 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8066 A C statement to issue assembly directives that create a
8067 section-relative reference to the given label, using an integer of the
8071 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8072 A C statement to issue assembly directives that create a self-relative
8073 reference to the given label, using an integer of the given size.
8076 @defmac PUT_SDB_@dots{}
8077 Define these macros to override the assembler syntax for the special
8078 SDB assembler directives. See @file{sdbout.c} for a list of these
8079 macros and their arguments. If the standard syntax is used, you need
8080 not define them yourself.
8084 Some assemblers do not support a semicolon as a delimiter, even between
8085 SDB assembler directives. In that case, define this macro to be the
8086 delimiter to use (usually @samp{\n}). It is not necessary to define
8087 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8091 @defmac SDB_GENERATE_FAKE
8092 Define this macro to override the usual method of constructing a dummy
8093 name for anonymous structure and union types. See @file{sdbout.c} for
8097 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8098 Define this macro to allow references to unknown structure,
8099 union, or enumeration tags to be emitted. Standard COFF does not
8100 allow handling of unknown references, MIPS ECOFF has support for
8104 @defmac SDB_ALLOW_FORWARD_REFERENCES
8105 Define this macro to allow references to structure, union, or
8106 enumeration tags that have not yet been seen to be handled. Some
8107 assemblers choke if forward tags are used, while some require it.
8112 @subsection Macros for VMS Debug Format
8114 @c prevent bad page break with this line
8115 Here are macros for VMS debug format.
8117 @defmac VMS_DEBUGGING_INFO
8118 Define this macro if GCC should produce debugging output for VMS
8119 in response to the @option{-g} option. The default behavior for VMS
8120 is to generate minimal debug info for a traceback in the absence of
8121 @option{-g} unless explicitly overridden with @option{-g0}. This
8122 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8123 @code{OVERRIDE_OPTIONS}.
8126 @node Floating Point
8127 @section Cross Compilation and Floating Point
8128 @cindex cross compilation and floating point
8129 @cindex floating point and cross compilation
8131 While all modern machines use twos-complement representation for integers,
8132 there are a variety of representations for floating point numbers. This
8133 means that in a cross-compiler the representation of floating point numbers
8134 in the compiled program may be different from that used in the machine
8135 doing the compilation.
8137 Because different representation systems may offer different amounts of
8138 range and precision, all floating point constants must be represented in
8139 the target machine's format. Therefore, the cross compiler cannot
8140 safely use the host machine's floating point arithmetic; it must emulate
8141 the target's arithmetic. To ensure consistency, GCC always uses
8142 emulation to work with floating point values, even when the host and
8143 target floating point formats are identical.
8145 The following macros are provided by @file{real.h} for the compiler to
8146 use. All parts of the compiler which generate or optimize
8147 floating-point calculations must use these macros. They may evaluate
8148 their operands more than once, so operands must not have side effects.
8150 @defmac REAL_VALUE_TYPE
8151 The C data type to be used to hold a floating point value in the target
8152 machine's format. Typically this is a @code{struct} containing an
8153 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8157 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8158 Compares for equality the two values, @var{x} and @var{y}. If the target
8159 floating point format supports negative zeroes and/or NaNs,
8160 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8161 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8164 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8165 Tests whether @var{x} is less than @var{y}.
8168 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8169 Truncates @var{x} to a signed integer, rounding toward zero.
8172 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8173 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8174 @var{x} is negative, returns zero.
8177 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8178 Converts @var{string} into a floating point number in the target machine's
8179 representation for mode @var{mode}. This routine can handle both
8180 decimal and hexadecimal floating point constants, using the syntax
8181 defined by the C language for both.
8184 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8185 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8188 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8189 Determines whether @var{x} represents infinity (positive or negative).
8192 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8193 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8196 @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})
8197 Calculates an arithmetic operation on the two floating point values
8198 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8201 The operation to be performed is specified by @var{code}. Only the
8202 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8203 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8205 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8206 target's floating point format cannot represent infinity, it will call
8207 @code{abort}. Callers should check for this situation first, using
8208 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8211 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8212 Returns the negative of the floating point value @var{x}.
8215 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8216 Returns the absolute value of @var{x}.
8219 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8220 Truncates the floating point value @var{x} to fit in @var{mode}. The
8221 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8222 appropriate bit pattern to be output asa floating constant whose
8223 precision accords with mode @var{mode}.
8226 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8227 Converts a floating point value @var{x} into a double-precision integer
8228 which is then stored into @var{low} and @var{high}. If the value is not
8229 integral, it is truncated.
8232 @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})
8233 Converts a double-precision integer found in @var{low} and @var{high},
8234 into a floating point value which is then stored into @var{x}. The
8235 value is truncated to fit in mode @var{mode}.
8238 @node Mode Switching
8239 @section Mode Switching Instructions
8240 @cindex mode switching
8241 The following macros control mode switching optimizations:
8243 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8244 Define this macro if the port needs extra instructions inserted for mode
8245 switching in an optimizing compilation.
8247 For an example, the SH4 can perform both single and double precision
8248 floating point operations, but to perform a single precision operation,
8249 the FPSCR PR bit has to be cleared, while for a double precision
8250 operation, this bit has to be set. Changing the PR bit requires a general
8251 purpose register as a scratch register, hence these FPSCR sets have to
8252 be inserted before reload, i.e.@: you can't put this into instruction emitting
8253 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
8255 You can have multiple entities that are mode-switched, and select at run time
8256 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8257 return nonzero for any @var{entity} that needs mode-switching.
8258 If you define this macro, you also have to define
8259 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8260 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8261 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
8265 @defmac NUM_MODES_FOR_MODE_SWITCHING
8266 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8267 initializer for an array of integers. Each initializer element
8268 N refers to an entity that needs mode switching, and specifies the number
8269 of different modes that might need to be set for this entity.
8270 The position of the initializer in the initializer - starting counting at
8271 zero - determines the integer that is used to refer to the mode-switched
8273 In macros that take mode arguments / yield a mode result, modes are
8274 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8275 switch is needed / supplied.
8278 @defmac MODE_NEEDED (@var{entity}, @var{insn})
8279 @var{entity} is an integer specifying a mode-switched entity. If
8280 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8281 return an integer value not larger than the corresponding element in
8282 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8283 be switched into prior to the execution of @var{insn}.
8286 @defmac MODE_AFTER (@var{mode}, @var{insn})
8287 If this macro is defined, it is evaluated for every @var{insn} during
8288 mode switching. It determines the mode that an insn results in (if
8289 different from the incoming mode).
8292 @defmac MODE_ENTRY (@var{entity})
8293 If this macro is defined, it is evaluated for every @var{entity} that needs
8294 mode switching. It should evaluate to an integer, which is a mode that
8295 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
8296 is defined then @code{MODE_EXIT} must be defined.
8299 @defmac MODE_EXIT (@var{entity})
8300 If this macro is defined, it is evaluated for every @var{entity} that needs
8301 mode switching. It should evaluate to an integer, which is a mode that
8302 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
8303 is defined then @code{MODE_ENTRY} must be defined.
8306 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8307 This macro specifies the order in which modes for @var{entity} are processed.
8308 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8309 lowest. The value of the macro should be an integer designating a mode
8310 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8311 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8312 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8315 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8316 Generate one or more insns to set @var{entity} to @var{mode}.
8317 @var{hard_reg_live} is the set of hard registers live at the point where
8318 the insn(s) are to be inserted.
8321 @node Target Attributes
8322 @section Defining target-specific uses of @code{__attribute__}
8323 @cindex target attributes
8324 @cindex machine attributes
8325 @cindex attributes, target-specific
8327 Target-specific attributes may be defined for functions, data and types.
8328 These are described using the following target hooks; they also need to
8329 be documented in @file{extend.texi}.
8331 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8332 If defined, this target hook points to an array of @samp{struct
8333 attribute_spec} (defined in @file{tree.h}) specifying the machine
8334 specific attributes for this target and some of the restrictions on the
8335 entities to which these attributes are applied and the arguments they
8339 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8340 If defined, this target hook is a function which returns zero if the attributes on
8341 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8342 and two if they are nearly compatible (which causes a warning to be
8343 generated). If this is not defined, machine-specific attributes are
8344 supposed always to be compatible.
8347 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8348 If defined, this target hook is a function which assigns default attributes to
8349 newly defined @var{type}.
8352 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8353 Define this target hook if the merging of type attributes needs special
8354 handling. If defined, the result is a list of the combined
8355 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8356 that @code{comptypes} has already been called and returned 1. This
8357 function may call @code{merge_attributes} to handle machine-independent
8361 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8362 Define this target hook if the merging of decl attributes needs special
8363 handling. If defined, the result is a list of the combined
8364 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8365 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8366 when this is needed are when one attribute overrides another, or when an
8367 attribute is nullified by a subsequent definition. This function may
8368 call @code{merge_attributes} to handle machine-independent merging.
8370 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8371 If the only target-specific handling you require is @samp{dllimport} for
8372 Microsoft Windows targets, you should define the macro
8373 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function
8374 called @code{merge_dllimport_decl_attributes} which can then be defined
8375 as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done
8376 in @file{i386/cygwin.h} and @file{i386/i386.c}, for example.
8379 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8380 Define this target hook if you want to be able to add attributes to a decl
8381 when it is being created. This is normally useful for back ends which
8382 wish to implement a pragma by using the attributes which correspond to
8383 the pragma's effect. The @var{node} argument is the decl which is being
8384 created. The @var{attr_ptr} argument is a pointer to the attribute list
8385 for this decl. The list itself should not be modified, since it may be
8386 shared with other decls, but attributes may be chained on the head of
8387 the list and @code{*@var{attr_ptr}} modified to point to the new
8388 attributes, or a copy of the list may be made if further changes are
8392 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8394 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8395 into the current function, despite its having target-specific
8396 attributes, @code{false} otherwise. By default, if a function has a
8397 target specific attribute attached to it, it will not be inlined.
8400 @node MIPS Coprocessors
8401 @section Defining coprocessor specifics for MIPS targets.
8402 @cindex MIPS coprocessor-definition macros
8404 The MIPS specification allows MIPS implementations to have as many as 4
8405 coprocessors, each with as many as 32 private registers. GCC supports
8406 accessing these registers and transferring values between the registers
8407 and memory using asm-ized variables. For example:
8410 register unsigned int cp0count asm ("c0r1");
8416 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8417 names may be added as described below, or the default names may be
8418 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8420 Coprocessor registers are assumed to be epilogue-used; sets to them will
8421 be preserved even if it does not appear that the register is used again
8422 later in the function.
8424 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8425 the FPU. One accesses COP1 registers through standard mips
8426 floating-point support; they are not included in this mechanism.
8428 There is one macro used in defining the MIPS coprocessor interface which
8429 you may want to override in subtargets; it is described below.
8431 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
8432 A comma-separated list (with leading comma) of pairs describing the
8433 alternate names of coprocessor registers. The format of each entry should be
8435 @{ @var{alternatename}, @var{register_number}@}
8441 @section Parameters for Precompiled Header Validity Checking
8442 @cindex parameters, precompiled headers
8444 @deftypefn {Target Hook} void * TARGET_GET_PCH_VALIDITY (size_t * @var{sz})
8445 Define this hook if your target needs to check a different collection
8446 of flags than the default, which is every flag defined by
8447 @code{TARGET_SWITCHES} and @code{TARGET_OPTIONS}. It should return
8448 some data which will be saved in the PCH file and presented to
8449 @code{TARGET_PCH_VALID_P} later; it should set @code{SZ} to the size
8453 @deftypefn {Target Hook} const char * TARGET_PCH_VALID_P (const void * @var{data}, size_t @var{sz})
8454 Define this hook if your target needs to check a different collection of
8455 flags than the default, which is every flag defined by @code{TARGET_SWITCHES}
8456 and @code{TARGET_OPTIONS}. It is given data which came from
8457 @code{TARGET_GET_PCH_VALIDITY} (in this version of this compiler, so there
8458 is no need for extensive validity checking). It returns @code{NULL} if
8459 it is safe to load a PCH file with this data, or a suitable error message
8460 if not. The error message will be presented to the user, so it should
8465 @section C++ ABI parameters
8466 @cindex parameters, c++ abi
8468 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
8469 Define this hook to override the integer type used for guard variables.
8470 These are used to implement one-time construction of static objects. The
8471 default is long_long_integer_type_node.
8474 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
8475 This hook determines how guard variables are used. It should return
8476 @code{false} (the default) if first byte should be used. A return value of
8477 @code{true} indicates the least significant bit should be used.
8481 @section Miscellaneous Parameters
8482 @cindex parameters, miscellaneous
8484 @c prevent bad page break with this line
8485 Here are several miscellaneous parameters.
8487 @defmac PREDICATE_CODES
8488 Define this if you have defined special-purpose predicates in the file
8489 @file{@var{machine}.c}. This macro is called within an initializer of an
8490 array of structures. The first field in the structure is the name of a
8491 predicate and the second field is an array of rtl codes. For each
8492 predicate, list all rtl codes that can be in expressions matched by the
8493 predicate. The list should have a trailing comma. Here is an example
8494 of two entries in the list for a typical RISC machine:
8497 #define PREDICATE_CODES \
8498 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8499 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8502 Defining this macro does not affect the generated code (however,
8503 incorrect definitions that omit an rtl code that may be matched by the
8504 predicate can cause the compiler to malfunction). Instead, it allows
8505 the table built by @file{genrecog} to be more compact and efficient,
8506 thus speeding up the compiler. The most important predicates to include
8507 in the list specified by this macro are those used in the most insn
8510 For each predicate function named in @code{PREDICATE_CODES}, a
8511 declaration will be generated in @file{insn-codes.h}.
8514 @defmac HAS_LONG_COND_BRANCH
8515 Define this boolean macro to indicate whether or not your architecture
8516 has conditional branches that can span all of memory. It is used in
8517 conjunction with an optimization that partitions hot and cold basic
8518 blocks into separate sections of the executable. If this macro is
8519 set to false, gcc will convert any conditional branches that attempt
8520 to cross between sections into unconditional branches or indirect jumps.
8523 @defmac HAS_LONG_UNCOND_BRANCH
8524 Define this boolean macro to indicate whether or not your architecture
8525 has unconditional branches that can span all of memory. It is used in
8526 conjunction with an optimization that partitions hot and cold basic
8527 blocks into separate sections of the executable. If this macro is
8528 set to false, gcc will convert any unconditional branches that attempt
8529 to cross between sections into indirect jumps.
8532 @defmac SPECIAL_MODE_PREDICATES
8533 Define this if you have special predicates that know special things
8534 about modes. Genrecog will warn about certain forms of
8535 @code{match_operand} without a mode; if the operand predicate is
8536 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8539 Here is an example from the IA-32 port (@code{ext_register_operand}
8540 specially checks for @code{HImode} or @code{SImode} in preparation
8541 for a byte extraction from @code{%ah} etc.).
8544 #define SPECIAL_MODE_PREDICATES \
8545 "ext_register_operand",
8549 @defmac CASE_VECTOR_MODE
8550 An alias for a machine mode name. This is the machine mode that
8551 elements of a jump-table should have.
8554 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8555 Optional: return the preferred mode for an @code{addr_diff_vec}
8556 when the minimum and maximum offset are known. If you define this,
8557 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8558 To make this work, you also have to define @code{INSN_ALIGN} and
8559 make the alignment for @code{addr_diff_vec} explicit.
8560 The @var{body} argument is provided so that the offset_unsigned and scale
8561 flags can be updated.
8564 @defmac CASE_VECTOR_PC_RELATIVE
8565 Define this macro to be a C expression to indicate when jump-tables
8566 should contain relative addresses. You need not define this macro if
8567 jump-tables never contain relative addresses, or jump-tables should
8568 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
8572 @defmac CASE_DROPS_THROUGH
8573 Define this if control falls through a @code{case} insn when the index
8574 value is out of range. This means the specified default-label is
8575 actually ignored by the @code{case} insn proper.
8578 @defmac CASE_VALUES_THRESHOLD
8579 Define this to be the smallest number of different values for which it
8580 is best to use a jump-table instead of a tree of conditional branches.
8581 The default is four for machines with a @code{casesi} instruction and
8582 five otherwise. This is best for most machines.
8585 @defmac CASE_USE_BIT_TESTS
8586 Define this macro to be a C expression to indicate whether C switch
8587 statements may be implemented by a sequence of bit tests. This is
8588 advantageous on processors that can efficiently implement left shift
8589 of 1 by the number of bits held in a register, but inappropriate on
8590 targets that would require a loop. By default, this macro returns
8591 @code{true} if the target defines an @code{ashlsi3} pattern, and
8592 @code{false} otherwise.
8595 @defmac WORD_REGISTER_OPERATIONS
8596 Define this macro if operations between registers with integral mode
8597 smaller than a word are always performed on the entire register.
8598 Most RISC machines have this property and most CISC machines do not.
8601 @defmac LOAD_EXTEND_OP (@var{mem_mode})
8602 Define this macro to be a C expression indicating when insns that read
8603 memory in @var{mem_mode}, an integral mode narrower than a word, set the
8604 bits outside of @var{mem_mode} to be either the sign-extension or the
8605 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8606 of @var{mem_mode} for which the
8607 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8608 @code{NIL} for other modes.
8610 This macro is not called with @var{mem_mode} non-integral or with a width
8611 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8612 value in this case. Do not define this macro if it would always return
8613 @code{NIL}. On machines where this macro is defined, you will normally
8614 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8616 You may return a non-@code{NIL} value even if for some hard registers
8617 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
8618 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
8619 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
8620 integral mode larger than this but not larger than @code{word_mode}.
8622 You must return @code{NIL} if for some hard registers that allow this
8623 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
8624 @code{word_mode}, but that they can change to another integral mode that
8625 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
8628 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
8629 Define this macro if loading short immediate values into registers sign
8633 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
8634 Define this macro if the same instructions that convert a floating
8635 point number to a signed fixed point number also convert validly to an
8640 The maximum number of bytes that a single instruction can move quickly
8641 between memory and registers or between two memory locations.
8644 @defmac MAX_MOVE_MAX
8645 The maximum number of bytes that a single instruction can move quickly
8646 between memory and registers or between two memory locations. If this
8647 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8648 constant value that is the largest value that @code{MOVE_MAX} can have
8652 @defmac SHIFT_COUNT_TRUNCATED
8653 A C expression that is nonzero if on this machine the number of bits
8654 actually used for the count of a shift operation is equal to the number
8655 of bits needed to represent the size of the object being shifted. When
8656 this macro is nonzero, the compiler will assume that it is safe to omit
8657 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8658 truncates the count of a shift operation. On machines that have
8659 instructions that act on bit-fields at variable positions, which may
8660 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8661 also enables deletion of truncations of the values that serve as
8662 arguments to bit-field instructions.
8664 If both types of instructions truncate the count (for shifts) and
8665 position (for bit-field operations), or if no variable-position bit-field
8666 instructions exist, you should define this macro.
8668 However, on some machines, such as the 80386 and the 680x0, truncation
8669 only applies to shift operations and not the (real or pretended)
8670 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8671 such machines. Instead, add patterns to the @file{md} file that include
8672 the implied truncation of the shift instructions.
8674 You need not define this macro if it would always have the value of zero.
8677 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8678 A C expression which is nonzero if on this machine it is safe to
8679 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8680 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8681 operating on it as if it had only @var{outprec} bits.
8683 On many machines, this expression can be 1.
8685 @c rearranged this, removed the phrase "it is reported that". this was
8686 @c to fix an overfull hbox. --mew 10feb93
8687 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8688 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8689 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8690 such cases may improve things.
8693 @defmac STORE_FLAG_VALUE
8694 A C expression describing the value returned by a comparison operator
8695 with an integral mode and stored by a store-flag instruction
8696 (@samp{s@var{cond}}) when the condition is true. This description must
8697 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8698 comparison operators whose results have a @code{MODE_INT} mode.
8700 A value of 1 or @minus{}1 means that the instruction implementing the
8701 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8702 and 0 when the comparison is false. Otherwise, the value indicates
8703 which bits of the result are guaranteed to be 1 when the comparison is
8704 true. This value is interpreted in the mode of the comparison
8705 operation, which is given by the mode of the first operand in the
8706 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8707 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8710 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8711 generate code that depends only on the specified bits. It can also
8712 replace comparison operators with equivalent operations if they cause
8713 the required bits to be set, even if the remaining bits are undefined.
8714 For example, on a machine whose comparison operators return an
8715 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8716 @samp{0x80000000}, saying that just the sign bit is relevant, the
8720 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8727 (ashift:SI @var{x} (const_int @var{n}))
8731 where @var{n} is the appropriate shift count to move the bit being
8732 tested into the sign bit.
8734 There is no way to describe a machine that always sets the low-order bit
8735 for a true value, but does not guarantee the value of any other bits,
8736 but we do not know of any machine that has such an instruction. If you
8737 are trying to port GCC to such a machine, include an instruction to
8738 perform a logical-and of the result with 1 in the pattern for the
8739 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8741 Often, a machine will have multiple instructions that obtain a value
8742 from a comparison (or the condition codes). Here are rules to guide the
8743 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8748 Use the shortest sequence that yields a valid definition for
8749 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8750 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8751 comparison operators to do so because there may be opportunities to
8752 combine the normalization with other operations.
8755 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8756 slightly preferred on machines with expensive jumps and 1 preferred on
8760 As a second choice, choose a value of @samp{0x80000001} if instructions
8761 exist that set both the sign and low-order bits but do not define the
8765 Otherwise, use a value of @samp{0x80000000}.
8768 Many machines can produce both the value chosen for
8769 @code{STORE_FLAG_VALUE} and its negation in the same number of
8770 instructions. On those machines, you should also define a pattern for
8771 those cases, e.g., one matching
8774 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8777 Some machines can also perform @code{and} or @code{plus} operations on
8778 condition code values with less instructions than the corresponding
8779 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8780 machines, define the appropriate patterns. Use the names @code{incscc}
8781 and @code{decscc}, respectively, for the patterns which perform
8782 @code{plus} or @code{minus} operations on condition code values. See
8783 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8784 find such instruction sequences on other machines.
8786 If this macro is not defined, the default value, 1, is used. You need
8787 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8788 instructions, or if the value generated by these instructions is 1.
8791 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
8792 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8793 returned when comparison operators with floating-point results are true.
8794 Define this macro on machine that have comparison operations that return
8795 floating-point values. If there are no such operations, do not define
8799 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8800 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8801 A C expression that evaluates to true if the architecture defines a value
8802 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
8803 should be set to this value. If this macro is not defined, the value of
8804 @code{clz} or @code{ctz} is assumed to be undefined.
8806 This macro must be defined if the target's expansion for @code{ffs}
8807 relies on a particular value to get correct results. Otherwise it
8808 is not necessary, though it may be used to optimize some corner cases.
8810 Note that regardless of this macro the ``definedness'' of @code{clz}
8811 and @code{ctz} at zero do @emph{not} extend to the builtin functions
8812 visible to the user. Thus one may be free to adjust the value at will
8813 to match the target expansion of these operations without fear of
8818 An alias for the machine mode for pointers. On most machines, define
8819 this to be the integer mode corresponding to the width of a hardware
8820 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8821 On some machines you must define this to be one of the partial integer
8822 modes, such as @code{PSImode}.
8824 The width of @code{Pmode} must be at least as large as the value of
8825 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8826 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8830 @defmac FUNCTION_MODE
8831 An alias for the machine mode used for memory references to functions
8832 being called, in @code{call} RTL expressions. On most machines this
8833 should be @code{QImode}.
8836 @defmac STDC_0_IN_SYSTEM_HEADERS
8837 In normal operation, the preprocessor expands @code{__STDC__} to the
8838 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8839 hosts, like Solaris, the system compiler uses a different convention,
8840 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8841 strict conformance to the C Standard.
8843 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8844 convention when processing system header files, but when processing user
8845 files @code{__STDC__} will always expand to 1.
8848 @defmac NO_IMPLICIT_EXTERN_C
8849 Define this macro if the system header files support C++ as well as C@.
8850 This macro inhibits the usual method of using system header files in
8851 C++, which is to pretend that the file's contents are enclosed in
8852 @samp{extern "C" @{@dots{}@}}.
8857 @defmac REGISTER_TARGET_PRAGMAS ()
8858 Define this macro if you want to implement any target-specific pragmas.
8859 If defined, it is a C expression which makes a series of calls to
8860 @code{c_register_pragma} for each pragma. The macro may also do any
8861 setup required for the pragmas.
8863 The primary reason to define this macro is to provide compatibility with
8864 other compilers for the same target. In general, we discourage
8865 definition of target-specific pragmas for GCC@.
8867 If the pragma can be implemented by attributes then you should consider
8868 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8870 Preprocessor macros that appear on pragma lines are not expanded. All
8871 @samp{#pragma} directives that do not match any registered pragma are
8872 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8875 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
8877 Each call to @code{c_register_pragma} establishes one pragma. The
8878 @var{callback} routine will be called when the preprocessor encounters a
8882 #pragma [@var{space}] @var{name} @dots{}
8885 @var{space} is the case-sensitive namespace of the pragma, or
8886 @code{NULL} to put the pragma in the global namespace. The callback
8887 routine receives @var{pfile} as its first argument, which can be passed
8888 on to cpplib's functions if necessary. You can lex tokens after the
8889 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8890 callback will be silently ignored. The end of the line is indicated by
8891 a token of type @code{CPP_EOF}
8893 For an example use of this routine, see @file{c4x.h} and the callback
8894 routines defined in @file{c4x-c.c}.
8896 Note that the use of @code{c_lex} is specific to the C and C++
8897 compilers. It will not work in the Java or Fortran compilers, or any
8898 other language compilers for that matter. Thus if @code{c_lex} is going
8899 to be called from target-specific code, it must only be done so when
8900 building the C and C++ compilers. This can be done by defining the
8901 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8902 target entry in the @file{config.gcc} file. These variables should name
8903 the target-specific, language-specific object file which contains the
8904 code that uses @code{c_lex}. Note it will also be necessary to add a
8905 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8906 how to build this object file.
8911 @defmac HANDLE_SYSV_PRAGMA
8912 Define this macro (to a value of 1) if you want the System V style
8913 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8914 [=<value>]} to be supported by gcc.
8916 The pack pragma specifies the maximum alignment (in bytes) of fields
8917 within a structure, in much the same way as the @samp{__aligned__} and
8918 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8919 the behavior to the default.
8921 A subtlety for Microsoft Visual C/C++ style bit-field packing
8922 (e.g. -mms-bitfields) for targets that support it:
8923 When a bit-field is inserted into a packed record, the whole size
8924 of the underlying type is used by one or more same-size adjacent
8925 bit-fields (that is, if its long:3, 32 bits is used in the record,
8926 and any additional adjacent long bit-fields are packed into the same
8927 chunk of 32 bits. However, if the size changes, a new field of that
8930 If both MS bit-fields and @samp{__attribute__((packed))} are used,
8931 the latter will take precedence. If @samp{__attribute__((packed))} is
8932 used on a single field when MS bit-fields are in use, it will take
8933 precedence for that field, but the alignment of the rest of the structure
8934 may affect its placement.
8936 The weak pragma only works if @code{SUPPORTS_WEAK} and
8937 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8938 of specifically named weak labels, optionally with a value.
8943 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
8944 Define this macro (to a value of 1) if you want to support the Win32
8945 style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma
8946 pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment
8947 (in bytes) of fields within a structure, in much the same way as the
8948 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8949 pack value of zero resets the behavior to the default. Successive
8950 invocations of this pragma cause the previous values to be stacked, so
8951 that invocations of @samp{#pragma pack(pop)} will return to the previous
8955 @defmac DOLLARS_IN_IDENTIFIERS
8956 Define this macro to control use of the character @samp{$} in
8957 identifier names for the C family of languages. 0 means @samp{$} is
8958 not allowed by default; 1 means it is allowed. 1 is the default;
8959 there is no need to define this macro in that case.
8962 @defmac NO_DOLLAR_IN_LABEL
8963 Define this macro if the assembler does not accept the character
8964 @samp{$} in label names. By default constructors and destructors in
8965 G++ have @samp{$} in the identifiers. If this macro is defined,
8966 @samp{.} is used instead.
8969 @defmac NO_DOT_IN_LABEL
8970 Define this macro if the assembler does not accept the character
8971 @samp{.} in label names. By default constructors and destructors in G++
8972 have names that use @samp{.}. If this macro is defined, these names
8973 are rewritten to avoid @samp{.}.
8976 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
8977 Define this macro as a C expression that is nonzero if it is safe for the
8978 delay slot scheduler to place instructions in the delay slot of @var{insn},
8979 even if they appear to use a resource set or clobbered in @var{insn}.
8980 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8981 every @code{call_insn} has this behavior. On machines where some @code{insn}
8982 or @code{jump_insn} is really a function call and hence has this behavior,
8983 you should define this macro.
8985 You need not define this macro if it would always return zero.
8988 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
8989 Define this macro as a C expression that is nonzero if it is safe for the
8990 delay slot scheduler to place instructions in the delay slot of @var{insn},
8991 even if they appear to set or clobber a resource referenced in @var{insn}.
8992 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8993 some @code{insn} or @code{jump_insn} is really a function call and its operands
8994 are registers whose use is actually in the subroutine it calls, you should
8995 define this macro. Doing so allows the delay slot scheduler to move
8996 instructions which copy arguments into the argument registers into the delay
8999 You need not define this macro if it would always return zero.
9002 @defmac MULTIPLE_SYMBOL_SPACES
9003 Define this macro if in some cases global symbols from one translation
9004 unit may not be bound to undefined symbols in another translation unit
9005 without user intervention. For instance, under Microsoft Windows
9006 symbols must be explicitly imported from shared libraries (DLLs).
9009 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{clobbers})
9010 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
9011 any hard regs the port wishes to automatically clobber for all asms.
9012 It should return the result of the last @code{tree_cons} used to add a
9016 @defmac MATH_LIBRARY
9017 Define this macro as a C string constant for the linker argument to link
9018 in the system math library, or @samp{""} if the target does not have a
9019 separate math library.
9021 You need only define this macro if the default of @samp{"-lm"} is wrong.
9024 @defmac LIBRARY_PATH_ENV
9025 Define this macro as a C string constant for the environment variable that
9026 specifies where the linker should look for libraries.
9028 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9032 @defmac TARGET_HAS_F_SETLKW
9033 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
9034 Note that this functionality is part of POSIX@.
9035 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
9036 to use file locking when exiting a program, which avoids race conditions
9037 if the program has forked.
9040 @defmac MAX_CONDITIONAL_EXECUTE
9042 A C expression for the maximum number of instructions to execute via
9043 conditional execution instructions instead of a branch. A value of
9044 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9045 1 if it does use cc0.
9048 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
9049 Used if the target needs to perform machine-dependent modifications on the
9050 conditionals used for turning basic blocks into conditionally executed code.
9051 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9052 contains information about the currently processed blocks. @var{true_expr}
9053 and @var{false_expr} are the tests that are used for converting the
9054 then-block and the else-block, respectively. Set either @var{true_expr} or
9055 @var{false_expr} to a null pointer if the tests cannot be converted.
9058 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9059 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9060 if-statements into conditions combined by @code{and} and @code{or} operations.
9061 @var{bb} contains the basic block that contains the test that is currently
9062 being processed and about to be turned into a condition.
9065 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
9066 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9067 be converted to conditional execution format. @var{ce_info} points to
9068 a data structure, @code{struct ce_if_block}, which contains information
9069 about the currently processed blocks.
9072 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
9073 A C expression to perform any final machine dependent modifications in
9074 converting code to conditional execution. The involved basic blocks
9075 can be found in the @code{struct ce_if_block} structure that is pointed
9076 to by @var{ce_info}.
9079 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
9080 A C expression to cancel any machine dependent modifications in
9081 converting code to conditional execution. The involved basic blocks
9082 can be found in the @code{struct ce_if_block} structure that is pointed
9083 to by @var{ce_info}.
9086 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
9087 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9088 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9091 @defmac IFCVT_EXTRA_FIELDS
9092 If defined, it should expand to a set of field declarations that will be
9093 added to the @code{struct ce_if_block} structure. These should be initialized
9094 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9097 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
9098 If non-null, this hook performs a target-specific pass over the
9099 instruction stream. The compiler will run it at all optimization levels,
9100 just before the point at which it normally does delayed-branch scheduling.
9102 The exact purpose of the hook varies from target to target. Some use
9103 it to do transformations that are necessary for correctness, such as
9104 laying out in-function constant pools or avoiding hardware hazards.
9105 Others use it as an opportunity to do some machine-dependent optimizations.
9107 You need not implement the hook if it has nothing to do. The default
9111 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9112 Define this hook if you have any machine-specific built-in functions
9113 that need to be defined. It should be a function that performs the
9116 Machine specific built-in functions can be useful to expand special machine
9117 instructions that would otherwise not normally be generated because
9118 they have no equivalent in the source language (for example, SIMD vector
9119 instructions or prefetch instructions).
9121 To create a built-in function, call the function @code{builtin_function}
9122 which is defined by the language front end. You can use any type nodes set
9123 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9124 only language front ends that use those two functions will call
9125 @samp{TARGET_INIT_BUILTINS}.
9128 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9130 Expand a call to a machine specific built-in function that was set up by
9131 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9132 function call; the result should go to @var{target} if that is
9133 convenient, and have mode @var{mode} if that is convenient.
9134 @var{subtarget} may be used as the target for computing one of
9135 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9136 ignored. This function should return the result of the call to the
9140 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
9142 Take a branch insn in @var{branch1} and another in @var{branch2}.
9143 Return true if redirecting @var{branch1} to the destination of
9144 @var{branch2} is possible.
9146 On some targets, branches may have a limited range. Optimizing the
9147 filling of delay slots can result in branches being redirected, and this
9148 may in turn cause a branch offset to overflow.
9151 @defmac ALLOCATE_INITIAL_VALUE (@var{hard_reg})
9153 When the initial value of a hard register has been copied in a pseudo
9154 register, it is often not necessary to actually allocate another register
9155 to this pseudo register, because the original hard register or a stack slot
9156 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9157 defined, is called at the start of register allocation once for each
9158 hard register that had its initial value copied by using
9159 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9160 Possible values are @code{NULL_RTX}, if you don't want
9161 to do any special allocation, a @code{REG} rtx---that would typically be
9162 the hard register itself, if it is known not to be clobbered---or a
9164 If you are returning a @code{MEM}, this is only a hint for the allocator;
9165 it might decide to use another register anyways.
9166 You may use @code{current_function_leaf_function} in the definition of the
9167 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9168 register in question will not be clobbered.
9171 @defmac TARGET_OBJECT_SUFFIX
9172 Define this macro to be a C string representing the suffix for object
9173 files on your target machine. If you do not define this macro, GCC will
9174 use @samp{.o} as the suffix for object files.
9177 @defmac TARGET_EXECUTABLE_SUFFIX
9178 Define this macro to be a C string representing the suffix to be
9179 automatically added to executable files on your target machine. If you
9180 do not define this macro, GCC will use the null string as the suffix for
9184 @defmac COLLECT_EXPORT_LIST
9185 If defined, @code{collect2} will scan the individual object files
9186 specified on its command line and create an export list for the linker.
9187 Define this macro for systems like AIX, where the linker discards
9188 object files that are not referenced from @code{main} and uses export
9192 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
9193 Define this macro to a C expression representing a variant of the
9194 method call @var{mdecl}, if Java Native Interface (JNI) methods
9195 must be invoked differently from other methods on your target.
9196 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
9197 the @code{stdcall} calling convention and this macro is then
9198 defined as this expression:
9201 build_type_attribute_variant (@var{mdecl},
9203 (get_identifier ("stdcall"),
9208 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9209 This target hook returns @code{true} past the point in which new jump
9210 instructions could be created. On machines that require a register for
9211 every jump such as the SHmedia ISA of SH5, this point would typically be
9212 reload, so this target hook should be defined to a function such as:
9216 cannot_modify_jumps_past_reload_p ()
9218 return (reload_completed || reload_in_progress);
9223 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
9224 This target hook returns a register class for which branch target register
9225 optimizations should be applied. All registers in this class should be
9226 usable interchangeably. After reload, registers in this class will be
9227 re-allocated and loads will be hoisted out of loops and be subjected
9228 to inter-block scheduling.
9231 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
9232 Branch target register optimization will by default exclude callee-saved
9234 that are not already live during the current function; if this target hook
9235 returns true, they will be included. The target code must than make sure
9236 that all target registers in the class returned by
9237 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
9238 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
9239 epilogues have already been generated. Note, even if you only return
9240 true when @var{after_prologue_epilogue_gen} is false, you still are likely
9241 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
9242 to reserve space for caller-saved target registers.
9245 @defmac POWI_MAX_MULTS
9246 If defined, this macro is interpreted as a signed integer C expression
9247 that specifies the maximum number of floating point multiplications
9248 that should be emitted when expanding exponentiation by an integer
9249 constant inline. When this value is defined, exponentiation requiring
9250 more than this number of multiplications is implemented by calling the
9251 system library's @code{pow}, @code{powf} or @code{powl} routines.
9252 The default value places no upper bound on the multiplication count.
9255 @deftypefn Macro void TARGET_EXTRA_INCLUDES (int @var{stdinc})
9256 This target hook should register any extra include files for the
9257 target. The parameter @var{stdinc} indicates if normal include files
9261 @deftypefn Macro void TARGET_OPTF (char *@var{path})
9262 This target hook should register special include paths for the target.
9263 The parameter @var{path} is the include to register. On Darwin
9264 systems, this is used for Framework includes, which have semantics
9265 that are different from @option{-I}.
9268 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
9269 This target hook returns @code{true} if it is safe to use a local alias
9270 for a virtual function @var{fndecl} when constructing thunks,
9271 @code{false} otherwise. By default, the hook returns @code{true} for all
9272 functions, if a target supports aliases (ie. defines
9273 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,