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 * Misc:: Everything else.
57 @node Target Structure
58 @section The Global @code{targetm} Variable
60 @cindex target functions
62 @deftypevar {struct gcc_target} targetm
63 The target @file{.c} file must define the global @code{targetm} variable
64 which contains pointers to functions and data relating to the target
65 machine. The variable is declared in @file{target.h};
66 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
67 used to initialize the variable, and macros for the default initializers
68 for elements of the structure. The @file{.c} file should override those
69 macros for which the default definition is inappropriate. For example:
72 #include "target-def.h"
74 /* @r{Initialize the GCC target structure.} */
76 #undef TARGET_COMP_TYPE_ATTRIBUTES
77 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
79 struct gcc_target targetm = TARGET_INITIALIZER;
83 Where a macro should be defined in the @file{.c} file in this manner to
84 form part of the @code{targetm} structure, it is documented below as a
85 ``Target Hook'' with a prototype. Many macros will change in future
86 from being defined in the @file{.h} file to being part of the
87 @code{targetm} structure.
90 @section Controlling the Compilation Driver, @file{gcc}
92 @cindex controlling the compilation driver
94 @c prevent bad page break with this line
95 You can control the compilation driver.
97 @defmac SWITCH_TAKES_ARG (@var{char})
98 A C expression which determines whether the option @option{-@var{char}}
99 takes arguments. The value should be the number of arguments that
100 option takes--zero, for many options.
102 By default, this macro is defined as
103 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
104 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
105 wish to add additional options which take arguments. Any redefinition
106 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
110 @defmac WORD_SWITCH_TAKES_ARG (@var{name})
111 A C expression which determines whether the option @option{-@var{name}}
112 takes arguments. The value should be the number of arguments that
113 option takes--zero, for many options. This macro rather than
114 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
116 By default, this macro is defined as
117 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
118 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
119 wish to add additional options which take arguments. Any redefinition
120 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
124 @defmac SWITCH_CURTAILS_COMPILATION (@var{char})
125 A C expression which determines whether the option @option{-@var{char}}
126 stops compilation before the generation of an executable. The value is
127 boolean, nonzero if the option does stop an executable from being
128 generated, zero otherwise.
130 By default, this macro is defined as
131 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
132 options properly. You need not define
133 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
134 options which affect the generation of an executable. Any redefinition
135 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
136 for additional options.
139 @defmac SWITCHES_NEED_SPACES
140 A string-valued C expression which enumerates the options for which
141 the linker needs a space between the option and its argument.
143 If this macro is not defined, the default value is @code{""}.
146 @defmac TARGET_OPTION_TRANSLATE_TABLE
147 If defined, a list of pairs of strings, the first of which is a
148 potential command line target to the @file{gcc} driver program, and the
149 second of which is a space-separated (tabs and other whitespace are not
150 supported) list of options with which to replace the first option. The
151 target defining this list is responsible for assuring that the results
152 are valid. Replacement options may not be the @code{--opt} style, they
153 must be the @code{-opt} style. It is the intention of this macro to
154 provide a mechanism for substitution that affects the multilibs chosen,
155 such as one option that enables many options, some of which select
156 multilibs. Example nonsensical definition, where @code{-malt-abi},
157 @code{-EB}, and @code{-mspoo} cause different multilibs to be chosen:
160 #define TARGET_OPTION_TRANSLATE_TABLE \
161 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
162 @{ "-compat", "-EB -malign=4 -mspoo" @}
166 @defmac DRIVER_SELF_SPECS
167 A list of specs for the driver itself. It should be a suitable
168 initializer for an array of strings, with no surrounding braces.
170 The driver applies these specs to its own command line between loading
171 default @file{specs} files (but not command-line specified ones) and
172 choosing the multilib directory or running any subcommands. It
173 applies them in the order given, so each spec can depend on the
174 options added by earlier ones. It is also possible to remove options
175 using @samp{%<@var{option}} in the usual way.
177 This macro can be useful when a port has several interdependent target
178 options. It provides a way of standardizing the command line so
179 that the other specs are easier to write.
181 Do not define this macro if it does not need to do anything.
184 @defmac OPTION_DEFAULT_SPECS
185 A list of specs used to support configure-time default options (i.e.@:
186 @option{--with} options) in the driver. It should be a suitable initializer
187 for an array of structures, each containing two strings, without the
188 outermost pair of surrounding braces.
190 The first item in the pair is the name of the default. This must match
191 the code in @file{config.gcc} for the target. The second item is a spec
192 to apply if a default with this name was specified. The string
193 @samp{%(VALUE)} in the spec will be replaced by the value of the default
194 everywhere it occurs.
196 The driver will apply these specs to its own command line between loading
197 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
198 the same mechanism as @code{DRIVER_SELF_SPECS}.
200 Do not define this macro if it does not need to do anything.
204 A C string constant that tells the GCC driver program options to
205 pass to CPP@. It can also specify how to translate options you
206 give to GCC into options for GCC to pass to the CPP@.
208 Do not define this macro if it does not need to do anything.
211 @defmac CPLUSPLUS_CPP_SPEC
212 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
213 than C@. If you do not define this macro, then the value of
214 @code{CPP_SPEC} (if any) will be used instead.
218 A C string constant that tells the GCC driver program options to
219 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
221 It can also specify how to translate options you give to GCC into options
222 for GCC to pass to front ends.
224 Do not define this macro if it does not need to do anything.
228 A C string constant that tells the GCC driver program options to
229 pass to @code{cc1plus}. It can also specify how to translate options you
230 give to GCC into options for GCC to pass to the @code{cc1plus}.
232 Do not define this macro if it does not need to do anything.
233 Note that everything defined in CC1_SPEC is already passed to
234 @code{cc1plus} so there is no need to duplicate the contents of
235 CC1_SPEC in CC1PLUS_SPEC@.
239 A C string constant that tells the GCC driver program options to
240 pass to the assembler. It can also specify how to translate options
241 you give to GCC into options for GCC to pass to the assembler.
242 See the file @file{sun3.h} for an example of this.
244 Do not define this macro if it does not need to do anything.
247 @defmac ASM_FINAL_SPEC
248 A C string constant that tells the GCC driver program how to
249 run any programs which cleanup after the normal assembler.
250 Normally, this is not needed. See the file @file{mips.h} for
253 Do not define this macro if it does not need to do anything.
256 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
257 Define this macro, with no value, if the driver should give the assembler
258 an argument consisting of a single dash, @option{-}, to instruct it to
259 read from its standard input (which will be a pipe connected to the
260 output of the compiler proper). This argument is given after any
261 @option{-o} option specifying the name of the output file.
263 If you do not define this macro, the assembler is assumed to read its
264 standard input if given no non-option arguments. If your assembler
265 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
266 see @file{mips.h} for instance.
270 A C string constant that tells the GCC driver program options to
271 pass to the linker. It can also specify how to translate options you
272 give to GCC into options for GCC to pass to the linker.
274 Do not define this macro if it does not need to do anything.
278 Another C string constant used much like @code{LINK_SPEC}. The difference
279 between the two is that @code{LIB_SPEC} is used at the end of the
280 command given to the linker.
282 If this macro is not defined, a default is provided that
283 loads the standard C library from the usual place. See @file{gcc.c}.
287 Another C string constant that tells the GCC driver program
288 how and when to place a reference to @file{libgcc.a} into the
289 linker command line. This constant is placed both before and after
290 the value of @code{LIB_SPEC}.
292 If this macro is not defined, the GCC driver provides a default that
293 passes the string @option{-lgcc} to the linker.
296 @defmac REAL_LIBGCC_SPEC
297 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
298 @code{LIBGCC_SPEC} is not directly used by the driver program but is
299 instead modified to refer to different versions of @file{libgcc.a}
300 depending on the values of the command line flags @code{-static},
301 @code{-shared}, @code{-static-libgcc}, and @code{-shared-libgcc}. On
302 targets where these modifications are inappropriate, define
303 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
304 driver how to place a reference to @file{libgcc} on the link command
305 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
308 @defmac STARTFILE_SPEC
309 Another C string constant used much like @code{LINK_SPEC}. The
310 difference between the two is that @code{STARTFILE_SPEC} is used at
311 the very beginning of the command given to the linker.
313 If this macro is not defined, a default is provided that loads the
314 standard C startup file from the usual place. See @file{gcc.c}.
318 Another C string constant used much like @code{LINK_SPEC}. The
319 difference between the two is that @code{ENDFILE_SPEC} is used at
320 the very end of the command given to the linker.
322 Do not define this macro if it does not need to do anything.
325 @defmac THREAD_MODEL_SPEC
326 GCC @code{-v} will print the thread model GCC was configured to use.
327 However, this doesn't work on platforms that are multilibbed on thread
328 models, such as AIX 4.3. On such platforms, define
329 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
330 blanks that names one of the recognized thread models. @code{%*}, the
331 default value of this macro, will expand to the value of
332 @code{thread_file} set in @file{config.gcc}.
335 @defmac SYSROOT_SUFFIX_SPEC
336 Define this macro to add a suffix to the target sysroot when GCC is
337 configured with a sysroot. This will cause GCC to search for usr/lib,
338 et al, within sysroot+suffix.
341 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
342 Define this macro to add a headers_suffix to the target sysroot when
343 GCC is configured with a sysroot. This will cause GCC to pass the
344 updated sysroot+headers_suffix to CPP, causing it to search for
345 usr/include, et al, within sysroot+headers_suffix.
349 Define this macro to provide additional specifications to put in the
350 @file{specs} file that can be used in various specifications like
353 The definition should be an initializer for an array of structures,
354 containing a string constant, that defines the specification name, and a
355 string constant that provides the specification.
357 Do not define this macro if it does not need to do anything.
359 @code{EXTRA_SPECS} is useful when an architecture contains several
360 related targets, which have various @code{@dots{}_SPECS} which are similar
361 to each other, and the maintainer would like one central place to keep
364 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
365 define either @code{_CALL_SYSV} when the System V calling sequence is
366 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
369 The @file{config/rs6000/rs6000.h} target file defines:
372 #define EXTRA_SPECS \
373 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
375 #define CPP_SYS_DEFAULT ""
378 The @file{config/rs6000/sysv.h} target file defines:
382 "%@{posix: -D_POSIX_SOURCE @} \
383 %@{mcall-sysv: -D_CALL_SYSV @} \
384 %@{!mcall-sysv: %(cpp_sysv_default) @} \
385 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
387 #undef CPP_SYSV_DEFAULT
388 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
391 while the @file{config/rs6000/eabiaix.h} target file defines
392 @code{CPP_SYSV_DEFAULT} as:
395 #undef CPP_SYSV_DEFAULT
396 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
400 @defmac LINK_LIBGCC_SPECIAL
401 Define this macro if the driver program should find the library
402 @file{libgcc.a} itself and should not pass @option{-L} options to the
403 linker. If you do not define this macro, the driver program will pass
404 the argument @option{-lgcc} to tell the linker to do the search and will
405 pass @option{-L} options to it.
408 @defmac LINK_LIBGCC_SPECIAL_1
409 Define this macro if the driver program should find the library
410 @file{libgcc.a}. If you do not define this macro, the driver program will pass
411 the argument @option{-lgcc} to tell the linker to do the search.
412 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
413 not affect @option{-L} options.
416 @defmac LINK_GCC_C_SEQUENCE_SPEC
417 The sequence in which libgcc and libc are specified to the linker.
418 By default this is @code{%G %L %G}.
421 @defmac LINK_COMMAND_SPEC
422 A C string constant giving the complete command line need to execute the
423 linker. When you do this, you will need to update your port each time a
424 change is made to the link command line within @file{gcc.c}. Therefore,
425 define this macro only if you need to completely redefine the command
426 line for invoking the linker and there is no other way to accomplish
427 the effect you need. Overriding this macro may be avoidable by overriding
428 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
431 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
432 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
433 directories from linking commands. Do not give it a nonzero value if
434 removing duplicate search directories changes the linker's semantics.
437 @defmac MULTILIB_DEFAULTS
438 Define this macro as a C expression for the initializer of an array of
439 string to tell the driver program which options are defaults for this
440 target and thus do not need to be handled specially when using
441 @code{MULTILIB_OPTIONS}.
443 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
444 the target makefile fragment or if none of the options listed in
445 @code{MULTILIB_OPTIONS} are set by default.
446 @xref{Target Fragment}.
449 @defmac RELATIVE_PREFIX_NOT_LINKDIR
450 Define this macro to tell @command{gcc} that it should only translate
451 a @option{-B} prefix into a @option{-L} linker option if the prefix
452 indicates an absolute file name.
455 @defmac MD_EXEC_PREFIX
456 If defined, this macro is an additional prefix to try after
457 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
458 when the @option{-b} option is used, or the compiler is built as a cross
459 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
460 to the list of directories used to find the assembler in @file{configure.in}.
463 @defmac STANDARD_STARTFILE_PREFIX
464 Define this macro as a C string constant if you wish to override the
465 standard choice of @code{libdir} as the default prefix to
466 try when searching for startup files such as @file{crt0.o}.
467 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
468 is built as a cross compiler.
471 @defmac MD_STARTFILE_PREFIX
472 If defined, this macro supplies an additional prefix to try after the
473 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
474 @option{-b} option is used, or when the compiler is built as a cross
478 @defmac MD_STARTFILE_PREFIX_1
479 If defined, this macro supplies yet another prefix to try after the
480 standard prefixes. It is not searched when the @option{-b} option is
481 used, or when the compiler is built as a cross compiler.
484 @defmac INIT_ENVIRONMENT
485 Define this macro as a C string constant if you wish to set environment
486 variables for programs called by the driver, such as the assembler and
487 loader. The driver passes the value of this macro to @code{putenv} to
488 initialize the necessary environment variables.
491 @defmac LOCAL_INCLUDE_DIR
492 Define this macro as a C string constant if you wish to override the
493 standard choice of @file{/usr/local/include} as the default prefix to
494 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
495 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
497 Cross compilers do not search either @file{/usr/local/include} or its
501 @defmac MODIFY_TARGET_NAME
502 Define this macro if you wish to define command-line switches that
503 modify the default target name.
505 For each switch, you can include a string to be appended to the first
506 part of the configuration name or a string to be deleted from the
507 configuration name, if present. The definition should be an initializer
508 for an array of structures. Each array element should have three
509 elements: the switch name (a string constant, including the initial
510 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
511 indicate whether the string should be inserted or deleted, and the string
512 to be inserted or deleted (a string constant).
514 For example, on a machine where @samp{64} at the end of the
515 configuration name denotes a 64-bit target and you want the @option{-32}
516 and @option{-64} switches to select between 32- and 64-bit targets, you would
520 #define MODIFY_TARGET_NAME \
521 @{ @{ "-32", DELETE, "64"@}, \
522 @{"-64", ADD, "64"@}@}
526 @defmac SYSTEM_INCLUDE_DIR
527 Define this macro as a C string constant if you wish to specify a
528 system-specific directory to search for header files before the standard
529 directory. @code{SYSTEM_INCLUDE_DIR} comes before
530 @code{STANDARD_INCLUDE_DIR} in the search order.
532 Cross compilers do not use this macro and do not search the directory
536 @defmac STANDARD_INCLUDE_DIR
537 Define this macro as a C string constant if you wish to override the
538 standard choice of @file{/usr/include} as the default prefix to
539 try when searching for header files.
541 Cross compilers ignore this macro and do not search either
542 @file{/usr/include} or its replacement.
545 @defmac STANDARD_INCLUDE_COMPONENT
546 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
547 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
548 If you do not define this macro, no component is used.
551 @defmac INCLUDE_DEFAULTS
552 Define this macro if you wish to override the entire default search path
553 for include files. For a native compiler, the default search path
554 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
555 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
556 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
557 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
558 and specify private search areas for GCC@. The directory
559 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
561 The definition should be an initializer for an array of structures.
562 Each array element should have four elements: the directory name (a
563 string constant), the component name (also a string constant), a flag
564 for C++-only directories,
565 and a flag showing that the includes in the directory don't need to be
566 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
567 the array with a null element.
569 The component name denotes what GNU package the include file is part of,
570 if any, in all uppercase letters. For example, it might be @samp{GCC}
571 or @samp{BINUTILS}. If the package is part of a vendor-supplied
572 operating system, code the component name as @samp{0}.
574 For example, here is the definition used for VAX/VMS:
577 #define INCLUDE_DEFAULTS \
579 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
580 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
581 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
588 Here is the order of prefixes tried for exec files:
592 Any prefixes specified by the user with @option{-B}.
595 The environment variable @code{GCC_EXEC_PREFIX}, if any.
598 The directories specified by the environment variable @code{COMPILER_PATH}.
601 The macro @code{STANDARD_EXEC_PREFIX}.
604 @file{/usr/lib/gcc/}.
607 The macro @code{MD_EXEC_PREFIX}, if any.
610 Here is the order of prefixes tried for startfiles:
614 Any prefixes specified by the user with @option{-B}.
617 The environment variable @code{GCC_EXEC_PREFIX}, if any.
620 The directories specified by the environment variable @code{LIBRARY_PATH}
621 (or port-specific name; native only, cross compilers do not use this).
624 The macro @code{STANDARD_EXEC_PREFIX}.
627 @file{/usr/lib/gcc/}.
630 The macro @code{MD_EXEC_PREFIX}, if any.
633 The macro @code{MD_STARTFILE_PREFIX}, if any.
636 The macro @code{STANDARD_STARTFILE_PREFIX}.
645 @node Run-time Target
646 @section Run-time Target Specification
647 @cindex run-time target specification
648 @cindex predefined macros
649 @cindex target specifications
651 @c prevent bad page break with this line
652 Here are run-time target specifications.
654 @defmac TARGET_CPU_CPP_BUILTINS ()
655 This function-like macro expands to a block of code that defines
656 built-in preprocessor macros and assertions for the target cpu, using
657 the functions @code{builtin_define}, @code{builtin_define_std} and
658 @code{builtin_assert}. When the front end
659 calls this macro it provides a trailing semicolon, and since it has
660 finished command line option processing your code can use those
663 @code{builtin_assert} takes a string in the form you pass to the
664 command-line option @option{-A}, such as @code{cpu=mips}, and creates
665 the assertion. @code{builtin_define} takes a string in the form
666 accepted by option @option{-D} and unconditionally defines the macro.
668 @code{builtin_define_std} takes a string representing the name of an
669 object-like macro. If it doesn't lie in the user's namespace,
670 @code{builtin_define_std} defines it unconditionally. Otherwise, it
671 defines a version with two leading underscores, and another version
672 with two leading and trailing underscores, and defines the original
673 only if an ISO standard was not requested on the command line. For
674 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
675 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
676 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
677 defines only @code{_ABI64}.
679 You can also test for the C dialect being compiled. The variable
680 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
681 or @code{clk_objective_c}. Note that if we are preprocessing
682 assembler, this variable will be @code{clk_c} but the function-like
683 macro @code{preprocessing_asm_p()} will return true, so you might want
684 to check for that first. If you need to check for strict ANSI, the
685 variable @code{flag_iso} can be used. The function-like macro
686 @code{preprocessing_trad_p()} can be used to check for traditional
690 @defmac TARGET_OS_CPP_BUILTINS ()
691 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
692 and is used for the target operating system instead.
695 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
696 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
697 and is used for the target object format. @file{elfos.h} uses this
698 macro to define @code{__ELF__}, so you probably do not need to define
702 @deftypevar {extern int} target_flags
703 This declaration should be present.
706 @cindex optional hardware or system features
707 @cindex features, optional, in system conventions
709 @defmac TARGET_@var{featurename}
710 This series of macros is to allow compiler command arguments to
711 enable or disable the use of optional features of the target machine.
712 For example, one machine description serves both the 68000 and
713 the 68020; a command argument tells the compiler whether it should
714 use 68020-only instructions or not. This command argument works
715 by means of a macro @code{TARGET_68020} that tests a bit in
718 Define a macro @code{TARGET_@var{featurename}} for each such option.
719 Its definition should test a bit in @code{target_flags}. It is
720 recommended that a helper macro @code{MASK_@var{featurename}}
721 is defined for each bit-value to test, and used in
722 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
726 #define TARGET_MASK_68020 1
727 #define TARGET_68020 (target_flags & MASK_68020)
730 One place where these macros are used is in the condition-expressions
731 of instruction patterns. Note how @code{TARGET_68020} appears
732 frequently in the 68000 machine description file, @file{m68k.md}.
733 Another place they are used is in the definitions of the other
734 macros in the @file{@var{machine}.h} file.
737 @defmac TARGET_SWITCHES
738 This macro defines names of command options to set and clear
739 bits in @code{target_flags}. Its definition is an initializer
740 with a subgrouping for each command option.
742 Each subgrouping contains a string constant, that defines the option
743 name, a number, which contains the bits to set in
744 @code{target_flags}, and a second string which is the description
745 displayed by @option{--help}. If the number is negative then the bits specified
746 by the number are cleared instead of being set. If the description
747 string is present but empty, then no help information will be displayed
748 for that option, but it will not count as an undocumented option. The
749 actual option name is made by appending @samp{-m} to the specified name.
750 Non-empty description strings should be marked with @code{N_(@dots{})} for
751 @command{xgettext}. Please do not mark empty strings because the empty
752 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
753 of the message catalog with meta information, not the empty string.
755 In addition to the description for @option{--help},
756 more detailed documentation for each option should be added to
759 One of the subgroupings should have a null string. The number in
760 this grouping is the default value for @code{target_flags}. Any
761 target options act starting with that value.
763 Here is an example which defines @option{-m68000} and @option{-m68020}
764 with opposite meanings, and picks the latter as the default:
767 #define TARGET_SWITCHES \
768 @{ @{ "68020", MASK_68020, "" @}, \
769 @{ "68000", -MASK_68020, \
770 N_("Compile for the 68000") @}, \
771 @{ "", MASK_68020, "" @}, \
776 @defmac TARGET_OPTIONS
777 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
778 options that have values. Its definition is an initializer with a
779 subgrouping for each command option.
781 Each subgrouping contains a string constant, that defines the option
782 name, the address of a variable, a description string, and a value.
783 Non-empty description strings should be marked with @code{N_(@dots{})}
784 for @command{xgettext}. Please do not mark empty strings because the
785 empty string is reserved by GNU gettext. @code{gettext("")} returns the
786 header entry of the message catalog with meta information, not the empty
789 If the value listed in the table is @code{NULL}, then the variable, type
790 @code{char *}, is set to the variable part of the given option if the
791 fixed part matches. In other words, if the first part of the option
792 matches what's in the table, the variable will be set to point to the
793 rest of the option. This allows the user to specify a value for that
794 option. The actual option name is made by appending @samp{-m} to the
795 specified name. Again, each option should also be documented in
798 If the value listed in the table is non-@code{NULL}, then the option
799 must match the option in the table exactly (with @samp{-m}), and the
800 variable is set to point to the value listed in the table.
802 Here is an example which defines @option{-mshort-data-@var{number}}. If the
803 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
804 will be set to the string @code{"512"}.
807 extern char *m88k_short_data;
808 #define TARGET_OPTIONS \
809 @{ @{ "short-data-", &m88k_short_data, \
810 N_("Specify the size of the short data section"), 0 @} @}
813 Here is a variant of the above that allows the user to also specify
814 just @option{-mshort-data} where a default of @code{"64"} is used.
817 extern char *m88k_short_data;
818 #define TARGET_OPTIONS \
819 @{ @{ "short-data-", &m88k_short_data, \
820 N_("Specify the size of the short data section"), 0 @} \
821 @{ "short-data", &m88k_short_data, "", "64" @},
825 Here is an example which defines @option{-mno-alu}, @option{-malu1}, and
826 @option{-malu2} as a three-state switch, along with suitable macros for
827 checking the state of the option (documentation is elided for brevity).
831 char *chip_alu = ""; /* Specify default here. */
834 extern char *chip_alu;
835 #define TARGET_OPTIONS \
836 @{ @{ "no-alu", &chip_alu, "", "" @}, \
837 @{ "alu1", &chip_alu, "", "1" @}, \
838 @{ "alu2", &chip_alu, "", "2" @}, @}
839 #define TARGET_ALU (chip_alu[0] != '\0')
840 #define TARGET_ALU1 (chip_alu[0] == '1')
841 #define TARGET_ALU2 (chip_alu[0] == '2')
845 @defmac TARGET_VERSION
846 This macro is a C statement to print on @code{stderr} a string
847 describing the particular machine description choice. Every machine
848 description should define @code{TARGET_VERSION}. For example:
852 #define TARGET_VERSION \
853 fprintf (stderr, " (68k, Motorola syntax)");
855 #define TARGET_VERSION \
856 fprintf (stderr, " (68k, MIT syntax)");
861 @defmac OVERRIDE_OPTIONS
862 Sometimes certain combinations of command options do not make sense on
863 a particular target machine. You can define a macro
864 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
865 defined, is executed once just after all the command options have been
868 Don't use this macro to turn on various extra optimizations for
869 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
872 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
873 Some machines may desire to change what optimizations are performed for
874 various optimization levels. This macro, if defined, is executed once
875 just after the optimization level is determined and before the remainder
876 of the command options have been parsed. Values set in this macro are
877 used as the default values for the other command line options.
879 @var{level} is the optimization level specified; 2 if @option{-O2} is
880 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
882 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
884 You should not use this macro to change options that are not
885 machine-specific. These should uniformly selected by the same
886 optimization level on all supported machines. Use this macro to enable
887 machine-specific optimizations.
889 @strong{Do not examine @code{write_symbols} in
890 this macro!} The debugging options are not supposed to alter the
894 @defmac CAN_DEBUG_WITHOUT_FP
895 Define this macro if debugging can be performed even without a frame
896 pointer. If this macro is defined, GCC will turn on the
897 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
900 @node Per-Function Data
901 @section Defining data structures for per-function information.
902 @cindex per-function data
903 @cindex data structures
905 If the target needs to store information on a per-function basis, GCC
906 provides a macro and a couple of variables to allow this. Note, just
907 using statics to store the information is a bad idea, since GCC supports
908 nested functions, so you can be halfway through encoding one function
909 when another one comes along.
911 GCC defines a data structure called @code{struct function} which
912 contains all of the data specific to an individual function. This
913 structure contains a field called @code{machine} whose type is
914 @code{struct machine_function *}, which can be used by targets to point
915 to their own specific data.
917 If a target needs per-function specific data it should define the type
918 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
919 This macro should be used to initialize the function pointer
920 @code{init_machine_status}. This pointer is explained below.
922 One typical use of per-function, target specific data is to create an
923 RTX to hold the register containing the function's return address. This
924 RTX can then be used to implement the @code{__builtin_return_address}
925 function, for level 0.
927 Note---earlier implementations of GCC used a single data area to hold
928 all of the per-function information. Thus when processing of a nested
929 function began the old per-function data had to be pushed onto a
930 stack, and when the processing was finished, it had to be popped off the
931 stack. GCC used to provide function pointers called
932 @code{save_machine_status} and @code{restore_machine_status} to handle
933 the saving and restoring of the target specific information. Since the
934 single data area approach is no longer used, these pointers are no
937 @defmac INIT_EXPANDERS
938 Macro called to initialize any target specific information. This macro
939 is called once per function, before generation of any RTL has begun.
940 The intention of this macro is to allow the initialization of the
941 function pointer @code{init_machine_status}.
944 @deftypevar {void (*)(struct function *)} init_machine_status
945 If this function pointer is non-@code{NULL} it will be called once per
946 function, before function compilation starts, in order to allow the
947 target to perform any target specific initialization of the
948 @code{struct function} structure. It is intended that this would be
949 used to initialize the @code{machine} of that structure.
951 @code{struct machine_function} structures are expected to be freed by GC.
952 Generally, any memory that they reference must be allocated by using
953 @code{ggc_alloc}, including the structure itself.
957 @section Storage Layout
958 @cindex storage layout
960 Note that the definitions of the macros in this table which are sizes or
961 alignments measured in bits do not need to be constant. They can be C
962 expressions that refer to static variables, such as the @code{target_flags}.
963 @xref{Run-time Target}.
965 @defmac BITS_BIG_ENDIAN
966 Define this macro to have the value 1 if the most significant bit in a
967 byte has the lowest number; otherwise define it to have the value zero.
968 This means that bit-field instructions count from the most significant
969 bit. If the machine has no bit-field instructions, then this must still
970 be defined, but it doesn't matter which value it is defined to. This
971 macro need not be a constant.
973 This macro does not affect the way structure fields are packed into
974 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
977 @defmac BYTES_BIG_ENDIAN
978 Define this macro to have the value 1 if the most significant byte in a
979 word has the lowest number. This macro need not be a constant.
982 @defmac WORDS_BIG_ENDIAN
983 Define this macro to have the value 1 if, in a multiword object, the
984 most significant word has the lowest number. This applies to both
985 memory locations and registers; GCC fundamentally assumes that the
986 order of words in memory is the same as the order in registers. This
987 macro need not be a constant.
990 @defmac LIBGCC2_WORDS_BIG_ENDIAN
991 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
992 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
993 used only when compiling @file{libgcc2.c}. Typically the value will be set
994 based on preprocessor defines.
997 @defmac FLOAT_WORDS_BIG_ENDIAN
998 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
999 @code{TFmode} floating point numbers are stored in memory with the word
1000 containing the sign bit at the lowest address; otherwise define it to
1001 have the value 0. This macro need not be a constant.
1003 You need not define this macro if the ordering is the same as for
1004 multi-word integers.
1007 @defmac BITS_PER_UNIT
1008 Define this macro to be the number of bits in an addressable storage
1009 unit (byte). If you do not define this macro the default is 8.
1012 @defmac BITS_PER_WORD
1013 Number of bits in a word. If you do not define this macro, the default
1014 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
1017 @defmac MAX_BITS_PER_WORD
1018 Maximum number of bits in a word. If this is undefined, the default is
1019 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
1020 largest value that @code{BITS_PER_WORD} can have at run-time.
1023 @defmac UNITS_PER_WORD
1024 Number of storage units in a word; normally 4.
1027 @defmac MIN_UNITS_PER_WORD
1028 Minimum number of units in a word. If this is undefined, the default is
1029 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
1030 smallest value that @code{UNITS_PER_WORD} can have at run-time.
1033 @defmac POINTER_SIZE
1034 Width of a pointer, in bits. You must specify a value no wider than the
1035 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1036 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1037 a value the default is @code{BITS_PER_WORD}.
1040 @defmac POINTERS_EXTEND_UNSIGNED
1041 A C expression whose value is greater than zero if pointers that need to be
1042 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
1043 be zero-extended and zero if they are to be sign-extended. If the value
1044 is less then zero then there must be an "ptr_extend" instruction that
1045 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
1047 You need not define this macro if the @code{POINTER_SIZE} is equal
1048 to the width of @code{Pmode}.
1051 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1052 A macro to update @var{m} and @var{unsignedp} when an object whose type
1053 is @var{type} and which has the specified mode and signedness is to be
1054 stored in a register. This macro is only called when @var{type} is a
1057 On most RISC machines, which only have operations that operate on a full
1058 register, define this macro to set @var{m} to @code{word_mode} if
1059 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1060 cases, only integer modes should be widened because wider-precision
1061 floating-point operations are usually more expensive than their narrower
1064 For most machines, the macro definition does not change @var{unsignedp}.
1065 However, some machines, have instructions that preferentially handle
1066 either signed or unsigned quantities of certain modes. For example, on
1067 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1068 sign-extend the result to 64 bits. On such machines, set
1069 @var{unsignedp} according to which kind of extension is more efficient.
1071 Do not define this macro if it would never modify @var{m}.
1074 @defmac PROMOTE_FUNCTION_MODE
1075 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
1076 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1077 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1079 The default is @code{PROMOTE_MODE}.
1082 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1083 This target hook should return @code{true} if the promotion described by
1084 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1088 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1089 This target hook should return @code{true} if the promotion described by
1090 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1093 If this target hook returns @code{true}, @code{FUNCTION_VALUE} must
1094 perform the same promotions done by @code{PROMOTE_FUNCTON_MODE}.
1097 @defmac PARM_BOUNDARY
1098 Normal alignment required for function parameters on the stack, in
1099 bits. All stack parameters receive at least this much alignment
1100 regardless of data type. On most machines, this is the same as the
1104 @defmac STACK_BOUNDARY
1105 Define this macro to the minimum alignment enforced by hardware for the
1106 stack pointer on this machine. The definition is a C expression for the
1107 desired alignment (measured in bits). This value is used as a default
1108 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1109 this should be the same as @code{PARM_BOUNDARY}.
1112 @defmac PREFERRED_STACK_BOUNDARY
1113 Define this macro if you wish to preserve a certain alignment for the
1114 stack pointer, greater than what the hardware enforces. The definition
1115 is a C expression for the desired alignment (measured in bits). This
1116 macro must evaluate to a value equal to or larger than
1117 @code{STACK_BOUNDARY}.
1120 @defmac FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1121 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1122 not guaranteed by the runtime and we should emit code to align the stack
1123 at the beginning of @code{main}.
1125 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1126 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1127 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1128 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1129 be momentarily unaligned while pushing arguments.
1132 @defmac FUNCTION_BOUNDARY
1133 Alignment required for a function entry point, in bits.
1136 @defmac BIGGEST_ALIGNMENT
1137 Biggest alignment that any data type can require on this machine, in bits.
1140 @defmac MINIMUM_ATOMIC_ALIGNMENT
1141 If defined, the smallest alignment, in bits, that can be given to an
1142 object that can be referenced in one operation, without disturbing any
1143 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1144 on machines that don't have byte or half-word store operations.
1147 @defmac BIGGEST_FIELD_ALIGNMENT
1148 Biggest alignment that any structure or union field can require on this
1149 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1150 structure and union fields only, unless the field alignment has been set
1151 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1154 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1155 An expression for the alignment of a structure field @var{field} if the
1156 alignment computed in the usual way (including applying of
1157 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1158 alignment) is @var{computed}. It overrides alignment only if the
1159 field alignment has not been set by the
1160 @code{__attribute__ ((aligned (@var{n})))} construct.
1163 @defmac MAX_OFILE_ALIGNMENT
1164 Biggest alignment supported by the object file format of this machine.
1165 Use this macro to limit the alignment which can be specified using the
1166 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1167 the default value is @code{BIGGEST_ALIGNMENT}.
1170 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1171 If defined, a C expression to compute the alignment for a variable in
1172 the static store. @var{type} is the data type, and @var{basic-align} is
1173 the alignment that the object would ordinarily have. The value of this
1174 macro is used instead of that alignment to align the object.
1176 If this macro is not defined, then @var{basic-align} is used.
1179 One use of this macro is to increase alignment of medium-size data to
1180 make it all fit in fewer cache lines. Another is to cause character
1181 arrays to be word-aligned so that @code{strcpy} calls that copy
1182 constants to character arrays can be done inline.
1185 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1186 If defined, a C expression to compute the alignment given to a constant
1187 that is being placed in memory. @var{constant} is the constant and
1188 @var{basic-align} is the alignment that the object would ordinarily
1189 have. The value of this macro is used instead of that alignment to
1192 If this macro is not defined, then @var{basic-align} is used.
1194 The typical use of this macro is to increase alignment for string
1195 constants to be word aligned so that @code{strcpy} calls that copy
1196 constants can be done inline.
1199 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1200 If defined, a C expression to compute the alignment for a variable in
1201 the local store. @var{type} is the data type, and @var{basic-align} is
1202 the alignment that the object would ordinarily have. The value of this
1203 macro is used instead of that alignment to align the object.
1205 If this macro is not defined, then @var{basic-align} is used.
1207 One use of this macro is to increase alignment of medium-size data to
1208 make it all fit in fewer cache lines.
1211 @defmac EMPTY_FIELD_BOUNDARY
1212 Alignment in bits to be given to a structure bit-field that follows an
1213 empty field such as @code{int : 0;}.
1215 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1218 @defmac STRUCTURE_SIZE_BOUNDARY
1219 Number of bits which any structure or union's size must be a multiple of.
1220 Each structure or union's size is rounded up to a multiple of this.
1222 If you do not define this macro, the default is the same as
1223 @code{BITS_PER_UNIT}.
1226 @defmac STRICT_ALIGNMENT
1227 Define this macro to be the value 1 if instructions will fail to work
1228 if given data not on the nominal alignment. If instructions will merely
1229 go slower in that case, define this macro as 0.
1232 @defmac PCC_BITFIELD_TYPE_MATTERS
1233 Define this if you wish to imitate the way many other C compilers handle
1234 alignment of bit-fields and the structures that contain them.
1236 The behavior is that the type written for a named bit-field (@code{int},
1237 @code{short}, or other integer type) imposes an alignment for the entire
1238 structure, as if the structure really did contain an ordinary field of
1239 that type. In addition, the bit-field is placed within the structure so
1240 that it would fit within such a field, not crossing a boundary for it.
1242 Thus, on most machines, a named bit-field whose type is written as
1243 @code{int} would not cross a four-byte boundary, and would force
1244 four-byte alignment for the whole structure. (The alignment used may
1245 not be four bytes; it is controlled by the other alignment parameters.)
1247 An unnamed bit-field will not affect the alignment of the containing
1250 If the macro is defined, its definition should be a C expression;
1251 a nonzero value for the expression enables this behavior.
1253 Note that if this macro is not defined, or its value is zero, some
1254 bit-fields may cross more than one alignment boundary. The compiler can
1255 support such references if there are @samp{insv}, @samp{extv}, and
1256 @samp{extzv} insns that can directly reference memory.
1258 The other known way of making bit-fields work is to define
1259 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1260 Then every structure can be accessed with fullwords.
1262 Unless the machine has bit-field instructions or you define
1263 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1264 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1266 If your aim is to make GCC use the same conventions for laying out
1267 bit-fields as are used by another compiler, here is how to investigate
1268 what the other compiler does. Compile and run this program:
1287 printf ("Size of foo1 is %d\n",
1288 sizeof (struct foo1));
1289 printf ("Size of foo2 is %d\n",
1290 sizeof (struct foo2));
1295 If this prints 2 and 5, then the compiler's behavior is what you would
1296 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1299 @defmac BITFIELD_NBYTES_LIMITED
1300 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1301 to aligning a bit-field within the structure.
1304 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1305 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1306 whether unnamed bitfields affect the alignment of the containing
1307 structure. The hook should return true if the structure should inherit
1308 the alignment requirements of an unnamed bitfield's type.
1311 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1312 Return 1 if a structure or array containing @var{field} should be accessed using
1315 If @var{field} is the only field in the structure, @var{mode} is its
1316 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1317 case where structures of one field would require the structure's mode to
1318 retain the field's mode.
1320 Normally, this is not needed. See the file @file{c4x.h} for an example
1321 of how to use this macro to prevent a structure having a floating point
1322 field from being accessed in an integer mode.
1325 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1326 Define this macro as an expression for the alignment of a type (given
1327 by @var{type} as a tree node) if the alignment computed in the usual
1328 way is @var{computed} and the alignment explicitly specified was
1331 The default is to use @var{specified} if it is larger; otherwise, use
1332 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1335 @defmac MAX_FIXED_MODE_SIZE
1336 An integer expression for the size in bits of the largest integer
1337 machine mode that should actually be used. All integer machine modes of
1338 this size or smaller can be used for structures and unions with the
1339 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1340 (DImode)} is assumed.
1343 @defmac VECTOR_MODE_SUPPORTED_P (@var{mode})
1344 Define this macro to be nonzero if the port is prepared to handle insns
1345 involving vector mode @var{mode}. At the very least, it must have move
1346 patterns for this mode.
1349 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1350 If defined, an expression of type @code{enum machine_mode} that
1351 specifies the mode of the save area operand of a
1352 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1353 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1354 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1355 having its mode specified.
1357 You need not define this macro if it always returns @code{Pmode}. You
1358 would most commonly define this macro if the
1359 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1363 @defmac STACK_SIZE_MODE
1364 If defined, an expression of type @code{enum machine_mode} that
1365 specifies the mode of the size increment operand of an
1366 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1368 You need not define this macro if it always returns @code{word_mode}.
1369 You would most commonly define this macro if the @code{allocate_stack}
1370 pattern needs to support both a 32- and a 64-bit mode.
1373 @defmac TARGET_FLOAT_FORMAT
1374 A code distinguishing the floating point format of the target machine.
1375 There are four defined values:
1378 @item IEEE_FLOAT_FORMAT
1379 This code indicates IEEE floating point. It is the default; there is no
1380 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1382 @item VAX_FLOAT_FORMAT
1383 This code indicates the ``F float'' (for @code{float}) and ``D float''
1384 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1386 @item IBM_FLOAT_FORMAT
1387 This code indicates the format used on the IBM System/370.
1389 @item C4X_FLOAT_FORMAT
1390 This code indicates the format used on the TMS320C3x/C4x.
1393 If your target uses a floating point format other than these, you must
1394 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1395 it to @file{real.c}.
1397 The ordering of the component words of floating point values stored in
1398 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1401 @defmac MODE_HAS_NANS (@var{mode})
1402 When defined, this macro should be true if @var{mode} has a NaN
1403 representation. The compiler assumes that NaNs are not equal to
1404 anything (including themselves) and that addition, subtraction,
1405 multiplication and division all return NaNs when one operand is
1408 By default, this macro is true if @var{mode} is a floating-point
1409 mode and the target floating-point format is IEEE@.
1412 @defmac MODE_HAS_INFINITIES (@var{mode})
1413 This macro should be true if @var{mode} can represent infinity. At
1414 present, the compiler uses this macro to decide whether @samp{x - x}
1415 is always defined. By default, the macro is true when @var{mode}
1416 is a floating-point mode and the target format is IEEE@.
1419 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1420 True if @var{mode} distinguishes between positive and negative zero.
1421 The rules are expected to follow the IEEE standard:
1425 @samp{x + x} has the same sign as @samp{x}.
1428 If the sum of two values with opposite sign is zero, the result is
1429 positive for all rounding modes expect towards @minus{}infinity, for
1430 which it is negative.
1433 The sign of a product or quotient is negative when exactly one
1434 of the operands is negative.
1437 The default definition is true if @var{mode} is a floating-point
1438 mode and the target format is IEEE@.
1441 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1442 If defined, this macro should be true for @var{mode} if it has at
1443 least one rounding mode in which @samp{x} and @samp{-x} can be
1444 rounded to numbers of different magnitude. Two such modes are
1445 towards @minus{}infinity and towards +infinity.
1447 The default definition of this macro is true if @var{mode} is
1448 a floating-point mode and the target format is IEEE@.
1451 @defmac ROUND_TOWARDS_ZERO
1452 If defined, this macro should be true if the prevailing rounding
1453 mode is towards zero. A true value has the following effects:
1457 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1460 @file{libgcc.a}'s floating-point emulator will round towards zero
1461 rather than towards nearest.
1464 The compiler's floating-point emulator will round towards zero after
1465 doing arithmetic, and when converting from the internal float format to
1469 The macro does not affect the parsing of string literals. When the
1470 primary rounding mode is towards zero, library functions like
1471 @code{strtod} might still round towards nearest, and the compiler's
1472 parser should behave like the target's @code{strtod} where possible.
1474 Not defining this macro is equivalent to returning zero.
1477 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1478 This macro should return true if floats with @var{size}
1479 bits do not have a NaN or infinity representation, but use the largest
1480 exponent for normal numbers instead.
1482 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1483 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1484 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1485 floating-point arithmetic.
1487 The default definition of this macro returns false for all sizes.
1490 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1491 This target hook should return @code{true} a vector is opaque. That
1492 is, if no cast is needed when copying a vector value of type
1493 @var{type} into another vector lvalue of the same size. Vector opaque
1494 types cannot be initialized. The default is that there are no such
1498 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1499 This target hook returns @code{true} if bit-fields in the given
1500 @var{record_type} are to be laid out following the rules of Microsoft
1501 Visual C/C++, namely: (i) a bit-field won't share the same storage
1502 unit with the previous bit-field if their underlying types have
1503 different sizes, and the bit-field will be aligned to the highest
1504 alignment of the underlying types of itself and of the previous
1505 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1506 the whole enclosing structure, even if it is unnamed; except that
1507 (iii) a zero-sized bit-field will be disregarded unless it follows
1508 another bit-field of nonzero size. If this hook returns @code{true},
1509 other macros that control bit-field layout are ignored.
1511 When a bit-field is inserted into a packed record, the whole size
1512 of the underlying type is used by one or more same-size adjacent
1513 bit-fields (that is, if its long:3, 32 bits is used in the record,
1514 and any additional adjacent long bit-fields are packed into the same
1515 chunk of 32 bits. However, if the size changes, a new field of that
1516 size is allocated). In an unpacked record, this is the same as using
1517 alignment, but not equivalent when packing.
1519 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1520 the latter will take precedence. If @samp{__attribute__((packed))} is
1521 used on a single field when MS bit-fields are in use, it will take
1522 precedence for that field, but the alignment of the rest of the structure
1523 may affect its placement.
1526 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_FUNDAMENTAL_TYPE (tree @var{type})
1527 If your target defines any fundamental types, define this hook to
1528 return the appropriate encoding for these types as part of a C++
1529 mangled name. The @var{type} argument is the tree structure
1530 representing the type to be mangled. The hook may be applied to trees
1531 which are not target-specific fundamental types; it should return
1532 @code{NULL} for all such types, as well as arguments it does not
1533 recognize. If the return value is not @code{NULL}, it must point to
1534 a statically-allocated string constant.
1536 Target-specific fundamental types might be new fundamental types or
1537 qualified versions of ordinary fundamental types. Encode new
1538 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1539 is the name used for the type in source code, and @var{n} is the
1540 length of @var{name} in decimal. Encode qualified versions of
1541 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1542 @var{name} is the name used for the type qualifier in source code,
1543 @var{n} is the length of @var{name} as above, and @var{code} is the
1544 code used to represent the unqualified version of this type. (See
1545 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1546 codes.) In both cases the spaces are for clarity; do not include any
1547 spaces in your string.
1549 The default version of this hook always returns @code{NULL}, which is
1550 appropriate for a target that does not define any new fundamental
1555 @section Layout of Source Language Data Types
1557 These macros define the sizes and other characteristics of the standard
1558 basic data types used in programs being compiled. Unlike the macros in
1559 the previous section, these apply to specific features of C and related
1560 languages, rather than to fundamental aspects of storage layout.
1562 @defmac INT_TYPE_SIZE
1563 A C expression for the size in bits of the type @code{int} on the
1564 target machine. If you don't define this, the default is one word.
1567 @defmac SHORT_TYPE_SIZE
1568 A C expression for the size in bits of the type @code{short} on the
1569 target machine. If you don't define this, the default is half a word.
1570 (If this would be less than one storage unit, it is rounded up to one
1574 @defmac LONG_TYPE_SIZE
1575 A C expression for the size in bits of the type @code{long} on the
1576 target machine. If you don't define this, the default is one word.
1579 @defmac ADA_LONG_TYPE_SIZE
1580 On some machines, the size used for the Ada equivalent of the type
1581 @code{long} by a native Ada compiler differs from that used by C. In
1582 that situation, define this macro to be a C expression to be used for
1583 the size of that type. If you don't define this, the default is the
1584 value of @code{LONG_TYPE_SIZE}.
1587 @defmac LONG_LONG_TYPE_SIZE
1588 A C expression for the size in bits of the type @code{long long} on the
1589 target machine. If you don't define this, the default is two
1590 words. If you want to support GNU Ada on your machine, the value of this
1591 macro must be at least 64.
1594 @defmac CHAR_TYPE_SIZE
1595 A C expression for the size in bits of the type @code{char} on the
1596 target machine. If you don't define this, the default is
1597 @code{BITS_PER_UNIT}.
1600 @defmac BOOL_TYPE_SIZE
1601 A C expression for the size in bits of the C++ type @code{bool} and
1602 C99 type @code{_Bool} on the target machine. If you don't define
1603 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1606 @defmac FLOAT_TYPE_SIZE
1607 A C expression for the size in bits of the type @code{float} on the
1608 target machine. If you don't define this, the default is one word.
1611 @defmac DOUBLE_TYPE_SIZE
1612 A C expression for the size in bits of the type @code{double} on the
1613 target machine. If you don't define this, the default is two
1617 @defmac LONG_DOUBLE_TYPE_SIZE
1618 A C expression for the size in bits of the type @code{long double} on
1619 the target machine. If you don't define this, the default is two
1623 @defmac TARGET_FLT_EVAL_METHOD
1624 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1625 assuming, if applicable, that the floating-point control word is in its
1626 default state. If you do not define this macro the value of
1627 @code{FLT_EVAL_METHOD} will be zero.
1630 @defmac WIDEST_HARDWARE_FP_SIZE
1631 A C expression for the size in bits of the widest floating-point format
1632 supported by the hardware. If you define this macro, you must specify a
1633 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1634 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1638 @defmac DEFAULT_SIGNED_CHAR
1639 An expression whose value is 1 or 0, according to whether the type
1640 @code{char} should be signed or unsigned by default. The user can
1641 always override this default with the options @option{-fsigned-char}
1642 and @option{-funsigned-char}.
1645 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1646 This target hook should return true if the compiler should give an
1647 @code{enum} type only as many bytes as it takes to represent the range
1648 of possible values of that type. It should return false if all
1649 @code{enum} types should be allocated like @code{int}.
1651 The default is to return false.
1655 A C expression for a string describing the name of the data type to use
1656 for size values. The typedef name @code{size_t} is defined using the
1657 contents of the string.
1659 The string can contain more than one keyword. If so, separate them with
1660 spaces, and write first any length keyword, then @code{unsigned} if
1661 appropriate, and finally @code{int}. The string must exactly match one
1662 of the data type names defined in the function
1663 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1664 omit @code{int} or change the order---that would cause the compiler to
1667 If you don't define this macro, the default is @code{"long unsigned
1671 @defmac PTRDIFF_TYPE
1672 A C expression for a string describing the name of the data type to use
1673 for the result of subtracting two pointers. The typedef name
1674 @code{ptrdiff_t} is defined using the contents of the string. See
1675 @code{SIZE_TYPE} above for more information.
1677 If you don't define this macro, the default is @code{"long int"}.
1681 A C expression for a string describing the name of the data type to use
1682 for wide characters. The typedef name @code{wchar_t} is defined using
1683 the contents of the string. See @code{SIZE_TYPE} above for more
1686 If you don't define this macro, the default is @code{"int"}.
1689 @defmac WCHAR_TYPE_SIZE
1690 A C expression for the size in bits of the data type for wide
1691 characters. This is used in @code{cpp}, which cannot make use of
1696 A C expression for a string describing the name of the data type to
1697 use for wide characters passed to @code{printf} and returned from
1698 @code{getwc}. The typedef name @code{wint_t} is defined using the
1699 contents of the string. See @code{SIZE_TYPE} above for more
1702 If you don't define this macro, the default is @code{"unsigned int"}.
1706 A C expression for a string describing the name of the data type that
1707 can represent any value of any standard or extended signed integer type.
1708 The typedef name @code{intmax_t} is defined using the contents of the
1709 string. See @code{SIZE_TYPE} above for more information.
1711 If you don't define this macro, the default is the first of
1712 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1713 much precision as @code{long long int}.
1716 @defmac UINTMAX_TYPE
1717 A C expression for a string describing the name of the data type that
1718 can represent any value of any standard or extended unsigned integer
1719 type. The typedef name @code{uintmax_t} is defined using the contents
1720 of the string. See @code{SIZE_TYPE} above for more information.
1722 If you don't define this macro, the default is the first of
1723 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1724 unsigned int"} that has as much precision as @code{long long unsigned
1728 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1729 The C++ compiler represents a pointer-to-member-function with a struct
1736 ptrdiff_t vtable_index;
1743 The C++ compiler must use one bit to indicate whether the function that
1744 will be called through a pointer-to-member-function is virtual.
1745 Normally, we assume that the low-order bit of a function pointer must
1746 always be zero. Then, by ensuring that the vtable_index is odd, we can
1747 distinguish which variant of the union is in use. But, on some
1748 platforms function pointers can be odd, and so this doesn't work. In
1749 that case, we use the low-order bit of the @code{delta} field, and shift
1750 the remainder of the @code{delta} field to the left.
1752 GCC will automatically make the right selection about where to store
1753 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1754 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1755 set such that functions always start at even addresses, but the lowest
1756 bit of pointers to functions indicate whether the function at that
1757 address is in ARM or Thumb mode. If this is the case of your
1758 architecture, you should define this macro to
1759 @code{ptrmemfunc_vbit_in_delta}.
1761 In general, you should not have to define this macro. On architectures
1762 in which function addresses are always even, according to
1763 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1764 @code{ptrmemfunc_vbit_in_pfn}.
1767 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1768 Normally, the C++ compiler uses function pointers in vtables. This
1769 macro allows the target to change to use ``function descriptors''
1770 instead. Function descriptors are found on targets for whom a
1771 function pointer is actually a small data structure. Normally the
1772 data structure consists of the actual code address plus a data
1773 pointer to which the function's data is relative.
1775 If vtables are used, the value of this macro should be the number
1776 of words that the function descriptor occupies.
1779 @defmac TARGET_VTABLE_ENTRY_ALIGN
1780 By default, the vtable entries are void pointers, the so the alignment
1781 is the same as pointer alignment. The value of this macro specifies
1782 the alignment of the vtable entry in bits. It should be defined only
1783 when special alignment is necessary. */
1786 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1787 There are a few non-descriptor entries in the vtable at offsets below
1788 zero. If these entries must be padded (say, to preserve the alignment
1789 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1790 of words in each data entry.
1793 @node Escape Sequences
1794 @section Target Character Escape Sequences
1795 @cindex escape sequences
1797 By default, GCC assumes that the C character escape sequences and other
1798 characters take on their ASCII values for the target. If this is not
1799 correct, you must explicitly define all of the macros below. All of
1800 them must evaluate to constants; they are used in @code{case}
1806 @findex TARGET_DIGIT0
1809 @findex TARGET_NEWLINE
1812 @multitable {@code{TARGET_NEWLINE}} {Escape} {ASCII character}
1813 @item Macro @tab Escape @tab ASCII character
1814 @item @code{TARGET_BELL} @tab @kbd{\a} @tab @code{07}, @code{BEL}
1815 @item @code{TARGET_BS} @tab @kbd{\b} @tab @code{08}, @code{BS}
1816 @item @code{TARGET_CR} @tab @kbd{\r} @tab @code{0D}, @code{CR}
1817 @item @code{TARGET_DIGIT0} @tab @kbd{0} @tab @code{30}, @code{ZERO}
1818 @item @code{TARGET_ESC} @tab @kbd{\e}, @kbd{\E} @tab @code{1B}, @code{ESC}
1819 @item @code{TARGET_FF} @tab @kbd{\f} @tab @code{0C}, @code{FF}
1820 @item @code{TARGET_NEWLINE} @tab @kbd{\n} @tab @code{0A}, @code{LF}
1821 @item @code{TARGET_TAB} @tab @kbd{\t} @tab @code{09}, @code{HT}
1822 @item @code{TARGET_VT} @tab @kbd{\v} @tab @code{0B}, @code{VT}
1826 Note that the @kbd{\e} and @kbd{\E} escapes are GNU extensions, not
1827 part of the C standard.
1830 @section Register Usage
1831 @cindex register usage
1833 This section explains how to describe what registers the target machine
1834 has, and how (in general) they can be used.
1836 The description of which registers a specific instruction can use is
1837 done with register classes; see @ref{Register Classes}. For information
1838 on using registers to access a stack frame, see @ref{Frame Registers}.
1839 For passing values in registers, see @ref{Register Arguments}.
1840 For returning values in registers, see @ref{Scalar Return}.
1843 * Register Basics:: Number and kinds of registers.
1844 * Allocation Order:: Order in which registers are allocated.
1845 * Values in Registers:: What kinds of values each reg can hold.
1846 * Leaf Functions:: Renumbering registers for leaf functions.
1847 * Stack Registers:: Handling a register stack such as 80387.
1850 @node Register Basics
1851 @subsection Basic Characteristics of Registers
1853 @c prevent bad page break with this line
1854 Registers have various characteristics.
1856 @defmac FIRST_PSEUDO_REGISTER
1857 Number of hardware registers known to the compiler. They receive
1858 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1859 pseudo register's number really is assigned the number
1860 @code{FIRST_PSEUDO_REGISTER}.
1863 @defmac FIXED_REGISTERS
1864 @cindex fixed register
1865 An initializer that says which registers are used for fixed purposes
1866 all throughout the compiled code and are therefore not available for
1867 general allocation. These would include the stack pointer, the frame
1868 pointer (except on machines where that can be used as a general
1869 register when no frame pointer is needed), the program counter on
1870 machines where that is considered one of the addressable registers,
1871 and any other numbered register with a standard use.
1873 This information is expressed as a sequence of numbers, separated by
1874 commas and surrounded by braces. The @var{n}th number is 1 if
1875 register @var{n} is fixed, 0 otherwise.
1877 The table initialized from this macro, and the table initialized by
1878 the following one, may be overridden at run time either automatically,
1879 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1880 the user with the command options @option{-ffixed-@var{reg}},
1881 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1884 @defmac CALL_USED_REGISTERS
1885 @cindex call-used register
1886 @cindex call-clobbered register
1887 @cindex call-saved register
1888 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1889 clobbered (in general) by function calls as well as for fixed
1890 registers. This macro therefore identifies the registers that are not
1891 available for general allocation of values that must live across
1894 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1895 automatically saves it on function entry and restores it on function
1896 exit, if the register is used within the function.
1899 @defmac CALL_REALLY_USED_REGISTERS
1900 @cindex call-used register
1901 @cindex call-clobbered register
1902 @cindex call-saved register
1903 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1904 that the entire set of @code{FIXED_REGISTERS} be included.
1905 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1906 This macro is optional. If not specified, it defaults to the value
1907 of @code{CALL_USED_REGISTERS}.
1910 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1911 @cindex call-used register
1912 @cindex call-clobbered register
1913 @cindex call-saved register
1914 A C expression that is nonzero if it is not permissible to store a
1915 value of mode @var{mode} in hard register number @var{regno} across a
1916 call without some part of it being clobbered. For most machines this
1917 macro need not be defined. It is only required for machines that do not
1918 preserve the entire contents of a register across a call.
1922 @findex call_used_regs
1925 @findex reg_class_contents
1926 @defmac CONDITIONAL_REGISTER_USAGE
1927 Zero or more C statements that may conditionally modify five variables
1928 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1929 @code{reg_names}, and @code{reg_class_contents}, to take into account
1930 any dependence of these register sets on target flags. The first three
1931 of these are of type @code{char []} (interpreted as Boolean vectors).
1932 @code{global_regs} is a @code{const char *[]}, and
1933 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1934 called, @code{fixed_regs}, @code{call_used_regs},
1935 @code{reg_class_contents}, and @code{reg_names} have been initialized
1936 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1937 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1938 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1939 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1940 command options have been applied.
1942 You need not define this macro if it has no work to do.
1944 @cindex disabling certain registers
1945 @cindex controlling register usage
1946 If the usage of an entire class of registers depends on the target
1947 flags, you may indicate this to GCC by using this macro to modify
1948 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1949 registers in the classes which should not be used by GCC@. Also define
1950 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1951 to return @code{NO_REGS} if it
1952 is called with a letter for a class that shouldn't be used.
1954 (However, if this class is not included in @code{GENERAL_REGS} and all
1955 of the insn patterns whose constraints permit this class are
1956 controlled by target switches, then GCC will automatically avoid using
1957 these registers when the target switches are opposed to them.)
1960 @defmac NON_SAVING_SETJMP
1961 If this macro is defined and has a nonzero value, it means that
1962 @code{setjmp} and related functions fail to save the registers, or that
1963 @code{longjmp} fails to restore them. To compensate, the compiler
1964 avoids putting variables in registers in functions that use
1968 @defmac INCOMING_REGNO (@var{out})
1969 Define this macro if the target machine has register windows. This C
1970 expression returns the register number as seen by the called function
1971 corresponding to the register number @var{out} as seen by the calling
1972 function. Return @var{out} if register number @var{out} is not an
1976 @defmac OUTGOING_REGNO (@var{in})
1977 Define this macro if the target machine has register windows. This C
1978 expression returns the register number as seen by the calling function
1979 corresponding to the register number @var{in} as seen by the called
1980 function. Return @var{in} if register number @var{in} is not an inbound
1984 @defmac LOCAL_REGNO (@var{regno})
1985 Define this macro if the target machine has register windows. This C
1986 expression returns true if the register is call-saved but is in the
1987 register window. Unlike most call-saved registers, such registers
1988 need not be explicitly restored on function exit or during non-local
1993 If the program counter has a register number, define this as that
1994 register number. Otherwise, do not define it.
1997 @node Allocation Order
1998 @subsection Order of Allocation of Registers
1999 @cindex order of register allocation
2000 @cindex register allocation order
2002 @c prevent bad page break with this line
2003 Registers are allocated in order.
2005 @defmac REG_ALLOC_ORDER
2006 If defined, an initializer for a vector of integers, containing the
2007 numbers of hard registers in the order in which GCC should prefer
2008 to use them (from most preferred to least).
2010 If this macro is not defined, registers are used lowest numbered first
2011 (all else being equal).
2013 One use of this macro is on machines where the highest numbered
2014 registers must always be saved and the save-multiple-registers
2015 instruction supports only sequences of consecutive registers. On such
2016 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2017 the highest numbered allocable register first.
2020 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
2021 A C statement (sans semicolon) to choose the order in which to allocate
2022 hard registers for pseudo-registers local to a basic block.
2024 Store the desired register order in the array @code{reg_alloc_order}.
2025 Element 0 should be the register to allocate first; element 1, the next
2026 register; and so on.
2028 The macro body should not assume anything about the contents of
2029 @code{reg_alloc_order} before execution of the macro.
2031 On most machines, it is not necessary to define this macro.
2034 @node Values in Registers
2035 @subsection How Values Fit in Registers
2037 This section discusses the macros that describe which kinds of values
2038 (specifically, which machine modes) each register can hold, and how many
2039 consecutive registers are needed for a given mode.
2041 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2042 A C expression for the number of consecutive hard registers, starting
2043 at register number @var{regno}, required to hold a value of mode
2046 On a machine where all registers are exactly one word, a suitable
2047 definition of this macro is
2050 #define HARD_REGNO_NREGS(REGNO, MODE) \
2051 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2056 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2057 Define this macro if the natural size of registers that hold values
2058 of mode @var{mode} is not the word size. It is a C expression that
2059 should give the natural size in bytes for the specified mode. It is
2060 used by the register allocator to try to optimize its results. This
2061 happens for example on SPARC 64-bit where the natural size of
2062 floating-point registers is still 32-bit.
2065 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2066 A C expression that is nonzero if it is permissible to store a value
2067 of mode @var{mode} in hard register number @var{regno} (or in several
2068 registers starting with that one). For a machine where all registers
2069 are equivalent, a suitable definition is
2072 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2075 You need not include code to check for the numbers of fixed registers,
2076 because the allocation mechanism considers them to be always occupied.
2078 @cindex register pairs
2079 On some machines, double-precision values must be kept in even/odd
2080 register pairs. You can implement that by defining this macro to reject
2081 odd register numbers for such modes.
2083 The minimum requirement for a mode to be OK in a register is that the
2084 @samp{mov@var{mode}} instruction pattern support moves between the
2085 register and other hard register in the same class and that moving a
2086 value into the register and back out not alter it.
2088 Since the same instruction used to move @code{word_mode} will work for
2089 all narrower integer modes, it is not necessary on any machine for
2090 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2091 you define patterns @samp{movhi}, etc., to take advantage of this. This
2092 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2093 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2096 Many machines have special registers for floating point arithmetic.
2097 Often people assume that floating point machine modes are allowed only
2098 in floating point registers. This is not true. Any registers that
2099 can hold integers can safely @emph{hold} a floating point machine
2100 mode, whether or not floating arithmetic can be done on it in those
2101 registers. Integer move instructions can be used to move the values.
2103 On some machines, though, the converse is true: fixed-point machine
2104 modes may not go in floating registers. This is true if the floating
2105 registers normalize any value stored in them, because storing a
2106 non-floating value there would garble it. In this case,
2107 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2108 floating registers. But if the floating registers do not automatically
2109 normalize, if you can store any bit pattern in one and retrieve it
2110 unchanged without a trap, then any machine mode may go in a floating
2111 register, so you can define this macro to say so.
2113 The primary significance of special floating registers is rather that
2114 they are the registers acceptable in floating point arithmetic
2115 instructions. However, this is of no concern to
2116 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2117 constraints for those instructions.
2119 On some machines, the floating registers are especially slow to access,
2120 so that it is better to store a value in a stack frame than in such a
2121 register if floating point arithmetic is not being done. As long as the
2122 floating registers are not in class @code{GENERAL_REGS}, they will not
2123 be used unless some pattern's constraint asks for one.
2126 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2127 A C expression that is nonzero if it is OK to rename a hard register
2128 @var{from} to another hard register @var{to}.
2130 One common use of this macro is to prevent renaming of a register to
2131 another register that is not saved by a prologue in an interrupt
2134 The default is always nonzero.
2137 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2138 A C expression that is nonzero if a value of mode
2139 @var{mode1} is accessible in mode @var{mode2} without copying.
2141 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2142 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2143 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2144 should be nonzero. If they differ for any @var{r}, you should define
2145 this macro to return zero unless some other mechanism ensures the
2146 accessibility of the value in a narrower mode.
2148 You should define this macro to return nonzero in as many cases as
2149 possible since doing so will allow GCC to perform better register
2153 @defmac AVOID_CCMODE_COPIES
2154 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2155 registers. You should only define this macro if support for copying to/from
2156 @code{CCmode} is incomplete.
2159 @node Leaf Functions
2160 @subsection Handling Leaf Functions
2162 @cindex leaf functions
2163 @cindex functions, leaf
2164 On some machines, a leaf function (i.e., one which makes no calls) can run
2165 more efficiently if it does not make its own register window. Often this
2166 means it is required to receive its arguments in the registers where they
2167 are passed by the caller, instead of the registers where they would
2170 The special treatment for leaf functions generally applies only when
2171 other conditions are met; for example, often they may use only those
2172 registers for its own variables and temporaries. We use the term ``leaf
2173 function'' to mean a function that is suitable for this special
2174 handling, so that functions with no calls are not necessarily ``leaf
2177 GCC assigns register numbers before it knows whether the function is
2178 suitable for leaf function treatment. So it needs to renumber the
2179 registers in order to output a leaf function. The following macros
2182 @defmac LEAF_REGISTERS
2183 Name of a char vector, indexed by hard register number, which
2184 contains 1 for a register that is allowable in a candidate for leaf
2187 If leaf function treatment involves renumbering the registers, then the
2188 registers marked here should be the ones before renumbering---those that
2189 GCC would ordinarily allocate. The registers which will actually be
2190 used in the assembler code, after renumbering, should not be marked with 1
2193 Define this macro only if the target machine offers a way to optimize
2194 the treatment of leaf functions.
2197 @defmac LEAF_REG_REMAP (@var{regno})
2198 A C expression whose value is the register number to which @var{regno}
2199 should be renumbered, when a function is treated as a leaf function.
2201 If @var{regno} is a register number which should not appear in a leaf
2202 function before renumbering, then the expression should yield @minus{}1, which
2203 will cause the compiler to abort.
2205 Define this macro only if the target machine offers a way to optimize the
2206 treatment of leaf functions, and registers need to be renumbered to do
2210 @findex current_function_is_leaf
2211 @findex current_function_uses_only_leaf_regs
2212 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2213 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2214 specially. They can test the C variable @code{current_function_is_leaf}
2215 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2216 set prior to local register allocation and is valid for the remaining
2217 compiler passes. They can also test the C variable
2218 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2219 functions which only use leaf registers.
2220 @code{current_function_uses_only_leaf_regs} is valid after reload and is
2221 only useful if @code{LEAF_REGISTERS} is defined.
2222 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2223 @c of the next paragraph?! --mew 2feb93
2225 @node Stack Registers
2226 @subsection Registers That Form a Stack
2228 There are special features to handle computers where some of the
2229 ``registers'' form a stack. Stack registers are normally written by
2230 pushing onto the stack, and are numbered relative to the top of the
2233 Currently, GCC can only handle one group of stack-like registers, and
2234 they must be consecutively numbered. Furthermore, the existing
2235 support for stack-like registers is specific to the 80387 floating
2236 point coprocessor. If you have a new architecture that uses
2237 stack-like registers, you will need to do substantial work on
2238 @file{reg-stack.c} and write your machine description to cooperate
2239 with it, as well as defining these macros.
2242 Define this if the machine has any stack-like registers.
2245 @defmac FIRST_STACK_REG
2246 The number of the first stack-like register. This one is the top
2250 @defmac LAST_STACK_REG
2251 The number of the last stack-like register. This one is the bottom of
2255 @node Register Classes
2256 @section Register Classes
2257 @cindex register class definitions
2258 @cindex class definitions, register
2260 On many machines, the numbered registers are not all equivalent.
2261 For example, certain registers may not be allowed for indexed addressing;
2262 certain registers may not be allowed in some instructions. These machine
2263 restrictions are described to the compiler using @dfn{register classes}.
2265 You define a number of register classes, giving each one a name and saying
2266 which of the registers belong to it. Then you can specify register classes
2267 that are allowed as operands to particular instruction patterns.
2271 In general, each register will belong to several classes. In fact, one
2272 class must be named @code{ALL_REGS} and contain all the registers. Another
2273 class must be named @code{NO_REGS} and contain no registers. Often the
2274 union of two classes will be another class; however, this is not required.
2276 @findex GENERAL_REGS
2277 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2278 terribly special about the name, but the operand constraint letters
2279 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2280 the same as @code{ALL_REGS}, just define it as a macro which expands
2283 Order the classes so that if class @var{x} is contained in class @var{y}
2284 then @var{x} has a lower class number than @var{y}.
2286 The way classes other than @code{GENERAL_REGS} are specified in operand
2287 constraints is through machine-dependent operand constraint letters.
2288 You can define such letters to correspond to various classes, then use
2289 them in operand constraints.
2291 You should define a class for the union of two classes whenever some
2292 instruction allows both classes. For example, if an instruction allows
2293 either a floating point (coprocessor) register or a general register for a
2294 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2295 which includes both of them. Otherwise you will get suboptimal code.
2297 You must also specify certain redundant information about the register
2298 classes: for each class, which classes contain it and which ones are
2299 contained in it; for each pair of classes, the largest class contained
2302 When a value occupying several consecutive registers is expected in a
2303 certain class, all the registers used must belong to that class.
2304 Therefore, register classes cannot be used to enforce a requirement for
2305 a register pair to start with an even-numbered register. The way to
2306 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2308 Register classes used for input-operands of bitwise-and or shift
2309 instructions have a special requirement: each such class must have, for
2310 each fixed-point machine mode, a subclass whose registers can transfer that
2311 mode to or from memory. For example, on some machines, the operations for
2312 single-byte values (@code{QImode}) are limited to certain registers. When
2313 this is so, each register class that is used in a bitwise-and or shift
2314 instruction must have a subclass consisting of registers from which
2315 single-byte values can be loaded or stored. This is so that
2316 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2318 @deftp {Data type} {enum reg_class}
2319 An enumeral type that must be defined with all the register class names
2320 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
2321 must be the last register class, followed by one more enumeral value,
2322 @code{LIM_REG_CLASSES}, which is not a register class but rather
2323 tells how many classes there are.
2325 Each register class has a number, which is the value of casting
2326 the class name to type @code{int}. The number serves as an index
2327 in many of the tables described below.
2330 @defmac N_REG_CLASSES
2331 The number of distinct register classes, defined as follows:
2334 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2338 @defmac REG_CLASS_NAMES
2339 An initializer containing the names of the register classes as C string
2340 constants. These names are used in writing some of the debugging dumps.
2343 @defmac REG_CLASS_CONTENTS
2344 An initializer containing the contents of the register classes, as integers
2345 which are bit masks. The @var{n}th integer specifies the contents of class
2346 @var{n}. The way the integer @var{mask} is interpreted is that
2347 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2349 When the machine has more than 32 registers, an integer does not suffice.
2350 Then the integers are replaced by sub-initializers, braced groupings containing
2351 several integers. Each sub-initializer must be suitable as an initializer
2352 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2353 In this situation, the first integer in each sub-initializer corresponds to
2354 registers 0 through 31, the second integer to registers 32 through 63, and
2358 @defmac REGNO_REG_CLASS (@var{regno})
2359 A C expression whose value is a register class containing hard register
2360 @var{regno}. In general there is more than one such class; choose a class
2361 which is @dfn{minimal}, meaning that no smaller class also contains the
2365 @defmac BASE_REG_CLASS
2366 A macro whose definition is the name of the class to which a valid
2367 base register must belong. A base register is one used in an address
2368 which is the register value plus a displacement.
2371 @defmac MODE_BASE_REG_CLASS (@var{mode})
2372 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2373 the selection of a base register in a mode dependent manner. If
2374 @var{mode} is VOIDmode then it should return the same value as
2375 @code{BASE_REG_CLASS}.
2378 @defmac INDEX_REG_CLASS
2379 A macro whose definition is the name of the class to which a valid
2380 index register must belong. An index register is one used in an
2381 address where its value is either multiplied by a scale factor or
2382 added to another register (as well as added to a displacement).
2385 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2386 For the constraint at the start of @var{str}, which starts with the letter
2387 @var{c}, return the length. This allows you to have register class /
2388 constant / extra constraints that are longer than a single letter;
2389 you don't need to define this macro if you can do with single-letter
2390 constraints only. The definition of this macro should use
2391 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2392 to handle specially.
2393 There are some sanity checks in genoutput.c that check the constraint lengths
2394 for the md file, so you can also use this macro to help you while you are
2395 transitioning from a byzantine single-letter-constraint scheme: when you
2396 return a negative length for a constraint you want to re-use, genoutput
2397 will complain about every instance where it is used in the md file.
2400 @defmac REG_CLASS_FROM_LETTER (@var{char})
2401 A C expression which defines the machine-dependent operand constraint
2402 letters for register classes. If @var{char} is such a letter, the
2403 value should be the register class corresponding to it. Otherwise,
2404 the value should be @code{NO_REGS}. The register letter @samp{r},
2405 corresponding to class @code{GENERAL_REGS}, will not be passed
2406 to this macro; you do not need to handle it.
2409 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2410 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2411 passed in @var{str}, so that you can use suffixes to distinguish between
2415 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2416 A C expression which is nonzero if register number @var{num} is
2417 suitable for use as a base register in operand addresses. It may be
2418 either a suitable hard register or a pseudo register that has been
2419 allocated such a hard register.
2422 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2423 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2424 that expression may examine the mode of the memory reference in
2425 @var{mode}. You should define this macro if the mode of the memory
2426 reference affects whether a register may be used as a base register. If
2427 you define this macro, the compiler will use it instead of
2428 @code{REGNO_OK_FOR_BASE_P}.
2431 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2432 A C expression which is nonzero if register number @var{num} is
2433 suitable for use as an index register in operand addresses. It may be
2434 either a suitable hard register or a pseudo register that has been
2435 allocated such a hard register.
2437 The difference between an index register and a base register is that
2438 the index register may be scaled. If an address involves the sum of
2439 two registers, neither one of them scaled, then either one may be
2440 labeled the ``base'' and the other the ``index''; but whichever
2441 labeling is used must fit the machine's constraints of which registers
2442 may serve in each capacity. The compiler will try both labelings,
2443 looking for one that is valid, and will reload one or both registers
2444 only if neither labeling works.
2447 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2448 A C expression that places additional restrictions on the register class
2449 to use when it is necessary to copy value @var{x} into a register in class
2450 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2451 another, smaller class. On many machines, the following definition is
2455 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2458 Sometimes returning a more restrictive class makes better code. For
2459 example, on the 68000, when @var{x} is an integer constant that is in range
2460 for a @samp{moveq} instruction, the value of this macro is always
2461 @code{DATA_REGS} as long as @var{class} includes the data registers.
2462 Requiring a data register guarantees that a @samp{moveq} will be used.
2464 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2465 @var{class} is if @var{x} is a legitimate constant which cannot be
2466 loaded into some register class. By returning @code{NO_REGS} you can
2467 force @var{x} into a memory location. For example, rs6000 can load
2468 immediate values into general-purpose registers, but does not have an
2469 instruction for loading an immediate value into a floating-point
2470 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2471 @var{x} is a floating-point constant. If the constant can't be loaded
2472 into any kind of register, code generation will be better if
2473 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2474 of using @code{PREFERRED_RELOAD_CLASS}.
2477 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2478 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2479 input reloads. If you don't define this macro, the default is to use
2480 @var{class}, unchanged.
2483 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2484 A C expression that places additional restrictions on the register class
2485 to use when it is necessary to be able to hold a value of mode
2486 @var{mode} in a reload register for which class @var{class} would
2489 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2490 there are certain modes that simply can't go in certain reload classes.
2492 The value is a register class; perhaps @var{class}, or perhaps another,
2495 Don't define this macro unless the target machine has limitations which
2496 require the macro to do something nontrivial.
2499 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2500 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2501 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2502 Many machines have some registers that cannot be copied directly to or
2503 from memory or even from other types of registers. An example is the
2504 @samp{MQ} register, which on most machines, can only be copied to or
2505 from general registers, but not memory. Some machines allow copying all
2506 registers to and from memory, but require a scratch register for stores
2507 to some memory locations (e.g., those with symbolic address on the RT,
2508 and those with certain symbolic address on the SPARC when compiling
2509 PIC)@. In some cases, both an intermediate and a scratch register are
2512 You should define these macros to indicate to the reload phase that it may
2513 need to allocate at least one register for a reload in addition to the
2514 register to contain the data. Specifically, if copying @var{x} to a
2515 register @var{class} in @var{mode} requires an intermediate register,
2516 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2517 largest register class all of whose registers can be used as
2518 intermediate registers or scratch registers.
2520 If copying a register @var{class} in @var{mode} to @var{x} requires an
2521 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2522 should be defined to return the largest register class required. If the
2523 requirements for input and output reloads are the same, the macro
2524 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2527 The values returned by these macros are often @code{GENERAL_REGS}.
2528 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2529 can be directly copied to or from a register of @var{class} in
2530 @var{mode} without requiring a scratch register. Do not define this
2531 macro if it would always return @code{NO_REGS}.
2533 If a scratch register is required (either with or without an
2534 intermediate register), you should define patterns for
2535 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2536 (@pxref{Standard Names}. These patterns, which will normally be
2537 implemented with a @code{define_expand}, should be similar to the
2538 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2541 Define constraints for the reload register and scratch register that
2542 contain a single register class. If the original reload register (whose
2543 class is @var{class}) can meet the constraint given in the pattern, the
2544 value returned by these macros is used for the class of the scratch
2545 register. Otherwise, two additional reload registers are required.
2546 Their classes are obtained from the constraints in the insn pattern.
2548 @var{x} might be a pseudo-register or a @code{subreg} of a
2549 pseudo-register, which could either be in a hard register or in memory.
2550 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2551 in memory and the hard register number if it is in a register.
2553 These macros should not be used in the case where a particular class of
2554 registers can only be copied to memory and not to another class of
2555 registers. In that case, secondary reload registers are not needed and
2556 would not be helpful. Instead, a stack location must be used to perform
2557 the copy and the @code{mov@var{m}} pattern should use memory as an
2558 intermediate storage. This case often occurs between floating-point and
2562 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2563 Certain machines have the property that some registers cannot be copied
2564 to some other registers without using memory. Define this macro on
2565 those machines to be a C expression that is nonzero if objects of mode
2566 @var{m} in registers of @var{class1} can only be copied to registers of
2567 class @var{class2} by storing a register of @var{class1} into memory
2568 and loading that memory location into a register of @var{class2}.
2570 Do not define this macro if its value would always be zero.
2573 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2574 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2575 allocates a stack slot for a memory location needed for register copies.
2576 If this macro is defined, the compiler instead uses the memory location
2577 defined by this macro.
2579 Do not define this macro if you do not define
2580 @code{SECONDARY_MEMORY_NEEDED}.
2583 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2584 When the compiler needs a secondary memory location to copy between two
2585 registers of mode @var{mode}, it normally allocates sufficient memory to
2586 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2587 load operations in a mode that many bits wide and whose class is the
2588 same as that of @var{mode}.
2590 This is right thing to do on most machines because it ensures that all
2591 bits of the register are copied and prevents accesses to the registers
2592 in a narrower mode, which some machines prohibit for floating-point
2595 However, this default behavior is not correct on some machines, such as
2596 the DEC Alpha, that store short integers in floating-point registers
2597 differently than in integer registers. On those machines, the default
2598 widening will not work correctly and you must define this macro to
2599 suppress that widening in some cases. See the file @file{alpha.h} for
2602 Do not define this macro if you do not define
2603 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2604 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2607 @defmac SMALL_REGISTER_CLASSES
2608 On some machines, it is risky to let hard registers live across arbitrary
2609 insns. Typically, these machines have instructions that require values
2610 to be in specific registers (like an accumulator), and reload will fail
2611 if the required hard register is used for another purpose across such an
2614 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2615 value on these machines. When this macro has a nonzero value, the
2616 compiler will try to minimize the lifetime of hard registers.
2618 It is always safe to define this macro with a nonzero value, but if you
2619 unnecessarily define it, you will reduce the amount of optimizations
2620 that can be performed in some cases. If you do not define this macro
2621 with a nonzero value when it is required, the compiler will run out of
2622 spill registers and print a fatal error message. For most machines, you
2623 should not define this macro at all.
2626 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2627 A C expression whose value is nonzero if pseudos that have been assigned
2628 to registers of class @var{class} would likely be spilled because
2629 registers of @var{class} are needed for spill registers.
2631 The default value of this macro returns 1 if @var{class} has exactly one
2632 register and zero otherwise. On most machines, this default should be
2633 used. Only define this macro to some other expression if pseudos
2634 allocated by @file{local-alloc.c} end up in memory because their hard
2635 registers were needed for spill registers. If this macro returns nonzero
2636 for those classes, those pseudos will only be allocated by
2637 @file{global.c}, which knows how to reallocate the pseudo to another
2638 register. If there would not be another register available for
2639 reallocation, you should not change the definition of this macro since
2640 the only effect of such a definition would be to slow down register
2644 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2645 A C expression for the maximum number of consecutive registers
2646 of class @var{class} needed to hold a value of mode @var{mode}.
2648 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2649 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2650 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2651 @var{mode})} for all @var{regno} values in the class @var{class}.
2653 This macro helps control the handling of multiple-word values
2657 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2658 If defined, a C expression that returns nonzero for a @var{class} for which
2659 a change from mode @var{from} to mode @var{to} is invalid.
2661 For the example, loading 32-bit integer or floating-point objects into
2662 floating-point registers on the Alpha extends them to 64 bits.
2663 Therefore loading a 64-bit object and then storing it as a 32-bit object
2664 does not store the low-order 32 bits, as would be the case for a normal
2665 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2669 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2670 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2671 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2675 Three other special macros describe which operands fit which constraint
2678 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2679 A C expression that defines the machine-dependent operand constraint
2680 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2681 particular ranges of integer values. If @var{c} is one of those
2682 letters, the expression should check that @var{value}, an integer, is in
2683 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2684 not one of those letters, the value should be 0 regardless of
2688 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2689 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2690 string passed in @var{str}, so that you can use suffixes to distinguish
2691 between different variants.
2694 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2695 A C expression that defines the machine-dependent operand constraint
2696 letters that specify particular ranges of @code{const_double} values
2697 (@samp{G} or @samp{H}).
2699 If @var{c} is one of those letters, the expression should check that
2700 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2701 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2702 letters, the value should be 0 regardless of @var{value}.
2704 @code{const_double} is used for all floating-point constants and for
2705 @code{DImode} fixed-point constants. A given letter can accept either
2706 or both kinds of values. It can use @code{GET_MODE} to distinguish
2707 between these kinds.
2710 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2711 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2712 string passed in @var{str}, so that you can use suffixes to distinguish
2713 between different variants.
2716 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2717 A C expression that defines the optional machine-dependent constraint
2718 letters that can be used to segregate specific types of operands, usually
2719 memory references, for the target machine. Any letter that is not
2720 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2721 @code{REG_CLASS_FROM_CONSTRAINT}
2722 may be used. Normally this macro will not be defined.
2724 If it is required for a particular target machine, it should return 1
2725 if @var{value} corresponds to the operand type represented by the
2726 constraint letter @var{c}. If @var{c} is not defined as an extra
2727 constraint, the value returned should be 0 regardless of @var{value}.
2729 For example, on the ROMP, load instructions cannot have their output
2730 in r0 if the memory reference contains a symbolic address. Constraint
2731 letter @samp{Q} is defined as representing a memory address that does
2732 @emph{not} contain a symbolic address. An alternative is specified with
2733 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2734 alternative specifies @samp{m} on the input and a register class that
2735 does not include r0 on the output.
2738 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2739 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2740 in @var{str}, so that you can use suffixes to distinguish between different
2744 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2745 A C expression that defines the optional machine-dependent constraint
2746 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2747 be treated like memory constraints by the reload pass.
2749 It should return 1 if the operand type represented by the constraint
2750 at the start of @var{str}, the first letter of which is the letter @var{c},
2751 comprises a subset of all memory references including
2752 all those whose address is simply a base register. This allows the reload
2753 pass to reload an operand, if it does not directly correspond to the operand
2754 type of @var{c}, by copying its address into a base register.
2756 For example, on the S/390, some instructions do not accept arbitrary
2757 memory references, but only those that do not make use of an index
2758 register. The constraint letter @samp{Q} is defined via
2759 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2760 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2761 a @samp{Q} constraint can handle any memory operand, because the
2762 reload pass knows it can be reloaded by copying the memory address
2763 into a base register if required. This is analogous to the way
2764 a @samp{o} constraint can handle any memory operand.
2767 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2768 A C expression that defines the optional machine-dependent constraint
2769 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2770 @code{EXTRA_CONSTRAINT_STR}, that should
2771 be treated like address constraints by the reload pass.
2773 It should return 1 if the operand type represented by the constraint
2774 at the start of @var{str}, which starts with the letter @var{c}, comprises
2775 a subset of all memory addresses including
2776 all those that consist of just a base register. This allows the reload
2777 pass to reload an operand, if it does not directly correspond to the operand
2778 type of @var{str}, by copying it into a base register.
2780 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2781 be used with the @code{address_operand} predicate. It is treated
2782 analogously to the @samp{p} constraint.
2785 @node Stack and Calling
2786 @section Stack Layout and Calling Conventions
2787 @cindex calling conventions
2789 @c prevent bad page break with this line
2790 This describes the stack layout and calling conventions.
2794 * Exception Handling::
2799 * Register Arguments::
2801 * Aggregate Return::
2809 @subsection Basic Stack Layout
2810 @cindex stack frame layout
2811 @cindex frame layout
2813 @c prevent bad page break with this line
2814 Here is the basic stack layout.
2816 @defmac STACK_GROWS_DOWNWARD
2817 Define this macro if pushing a word onto the stack moves the stack
2818 pointer to a smaller address.
2820 When we say, ``define this macro if @dots{},'' it means that the
2821 compiler checks this macro only with @code{#ifdef} so the precise
2822 definition used does not matter.
2825 @defmac STACK_PUSH_CODE
2826 This macro defines the operation used when something is pushed
2827 on the stack. In RTL, a push operation will be
2828 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2830 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2831 and @code{POST_INC}. Which of these is correct depends on
2832 the stack direction and on whether the stack pointer points
2833 to the last item on the stack or whether it points to the
2834 space for the next item on the stack.
2836 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2837 defined, which is almost always right, and @code{PRE_INC} otherwise,
2838 which is often wrong.
2841 @defmac FRAME_GROWS_DOWNWARD
2842 Define this macro if the addresses of local variable slots are at negative
2843 offsets from the frame pointer.
2846 @defmac ARGS_GROW_DOWNWARD
2847 Define this macro if successive arguments to a function occupy decreasing
2848 addresses on the stack.
2851 @defmac STARTING_FRAME_OFFSET
2852 Offset from the frame pointer to the first local variable slot to be allocated.
2854 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2855 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2856 Otherwise, it is found by adding the length of the first slot to the
2857 value @code{STARTING_FRAME_OFFSET}.
2858 @c i'm not sure if the above is still correct.. had to change it to get
2859 @c rid of an overfull. --mew 2feb93
2862 @defmac STACK_ALIGNMENT_NEEDED
2863 Define to zero to disable final alignment of the stack during reload.
2864 The nonzero default for this macro is suitable for most ports.
2866 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2867 is a register save block following the local block that doesn't require
2868 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2869 stack alignment and do it in the backend.
2872 @defmac STACK_POINTER_OFFSET
2873 Offset from the stack pointer register to the first location at which
2874 outgoing arguments are placed. If not specified, the default value of
2875 zero is used. This is the proper value for most machines.
2877 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2878 the first location at which outgoing arguments are placed.
2881 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2882 Offset from the argument pointer register to the first argument's
2883 address. On some machines it may depend on the data type of the
2886 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2887 the first argument's address.
2890 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2891 Offset from the stack pointer register to an item dynamically allocated
2892 on the stack, e.g., by @code{alloca}.
2894 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2895 length of the outgoing arguments. The default is correct for most
2896 machines. See @file{function.c} for details.
2899 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2900 A C expression whose value is RTL representing the address in a stack
2901 frame where the pointer to the caller's frame is stored. Assume that
2902 @var{frameaddr} is an RTL expression for the address of the stack frame
2905 If you don't define this macro, the default is to return the value
2906 of @var{frameaddr}---that is, the stack frame address is also the
2907 address of the stack word that points to the previous frame.
2910 @defmac SETUP_FRAME_ADDRESSES
2911 If defined, a C expression that produces the machine-specific code to
2912 setup the stack so that arbitrary frames can be accessed. For example,
2913 on the SPARC, we must flush all of the register windows to the stack
2914 before we can access arbitrary stack frames. You will seldom need to
2918 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
2919 This target hook should return an rtx that is used to store
2920 the address of the current frame into the built in @code{setjmp} buffer.
2921 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2922 machines. One reason you may need to define this target hook is if
2923 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2926 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2927 A C expression whose value is RTL representing the value of the return
2928 address for the frame @var{count} steps up from the current frame, after
2929 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2930 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2931 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2933 The value of the expression must always be the correct address when
2934 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2935 determine the return address of other frames.
2938 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2939 Define this if the return address of a particular stack frame is accessed
2940 from the frame pointer of the previous stack frame.
2943 @defmac INCOMING_RETURN_ADDR_RTX
2944 A C expression whose value is RTL representing the location of the
2945 incoming return address at the beginning of any function, before the
2946 prologue. This RTL is either a @code{REG}, indicating that the return
2947 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2950 You only need to define this macro if you want to support call frame
2951 debugging information like that provided by DWARF 2.
2953 If this RTL is a @code{REG}, you should also define
2954 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2957 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2958 A C expression whose value is an integer giving a DWARF 2 column
2959 number that may be used as an alternate return column. This should
2960 be defined only if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2961 general register, but an alternate column needs to be used for
2965 @defmac INCOMING_FRAME_SP_OFFSET
2966 A C expression whose value is an integer giving the offset, in bytes,
2967 from the value of the stack pointer register to the top of the stack
2968 frame at the beginning of any function, before the prologue. The top of
2969 the frame is defined to be the value of the stack pointer in the
2970 previous frame, just before the call instruction.
2972 You only need to define this macro if you want to support call frame
2973 debugging information like that provided by DWARF 2.
2976 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2977 A C expression whose value is an integer giving the offset, in bytes,
2978 from the argument pointer to the canonical frame address (cfa). The
2979 final value should coincide with that calculated by
2980 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2981 during virtual register instantiation.
2983 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2984 which is correct for most machines; in general, the arguments are found
2985 immediately before the stack frame. Note that this is not the case on
2986 some targets that save registers into the caller's frame, such as SPARC
2987 and rs6000, and so such targets need to define this macro.
2989 You only need to define this macro if the default is incorrect, and you
2990 want to support call frame debugging information like that provided by
2994 @node Exception Handling
2995 @subsection Exception Handling Support
2996 @cindex exception handling
2998 @defmac EH_RETURN_DATA_REGNO (@var{N})
2999 A C expression whose value is the @var{N}th register number used for
3000 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3001 @var{N} registers are usable.
3003 The exception handling library routines communicate with the exception
3004 handlers via a set of agreed upon registers. Ideally these registers
3005 should be call-clobbered; it is possible to use call-saved registers,
3006 but may negatively impact code size. The target must support at least
3007 2 data registers, but should define 4 if there are enough free registers.
3009 You must define this macro if you want to support call frame exception
3010 handling like that provided by DWARF 2.
3013 @defmac EH_RETURN_STACKADJ_RTX
3014 A C expression whose value is RTL representing a location in which
3015 to store a stack adjustment to be applied before function return.
3016 This is used to unwind the stack to an exception handler's call frame.
3017 It will be assigned zero on code paths that return normally.
3019 Typically this is a call-clobbered hard register that is otherwise
3020 untouched by the epilogue, but could also be a stack slot.
3022 Do not define this macro if the stack pointer is saved and restored
3023 by the regular prolog and epilog code in the call frame itself; in
3024 this case, the exception handling library routines will update the
3025 stack location to be restored in place. Otherwise, you must define
3026 this macro if you want to support call frame exception handling like
3027 that provided by DWARF 2.
3030 @defmac EH_RETURN_HANDLER_RTX
3031 A C expression whose value is RTL representing a location in which
3032 to store the address of an exception handler to which we should
3033 return. It will not be assigned on code paths that return normally.
3035 Typically this is the location in the call frame at which the normal
3036 return address is stored. For targets that return by popping an
3037 address off the stack, this might be a memory address just below
3038 the @emph{target} call frame rather than inside the current call
3039 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3040 been assigned, so it may be used to calculate the location of the
3043 Some targets have more complex requirements than storing to an
3044 address calculable during initial code generation. In that case
3045 the @code{eh_return} instruction pattern should be used instead.
3047 If you want to support call frame exception handling, you must
3048 define either this macro or the @code{eh_return} instruction pattern.
3051 @defmac RETURN_ADDR_OFFSET
3052 If defined, an integer-valued C expression for which rtl will be generated
3053 to add it to the exception handler address before it is searched in the
3054 exception handling tables, and to subtract it again from the address before
3055 using it to return to the exception handler.
3058 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3059 This macro chooses the encoding of pointers embedded in the exception
3060 handling sections. If at all possible, this should be defined such
3061 that the exception handling section will not require dynamic relocations,
3062 and so may be read-only.
3064 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3065 @var{global} is true if the symbol may be affected by dynamic relocations.
3066 The macro should return a combination of the @code{DW_EH_PE_*} defines
3067 as found in @file{dwarf2.h}.
3069 If this macro is not defined, pointers will not be encoded but
3070 represented directly.
3073 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3074 This macro allows the target to emit whatever special magic is required
3075 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3076 Generic code takes care of pc-relative and indirect encodings; this must
3077 be defined if the target uses text-relative or data-relative encodings.
3079 This is a C statement that branches to @var{done} if the format was
3080 handled. @var{encoding} is the format chosen, @var{size} is the number
3081 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3085 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs}, @var{success})
3086 This macro allows the target to add cpu and operating system specific
3087 code to the call-frame unwinder for use when there is no unwind data
3088 available. The most common reason to implement this macro is to unwind
3089 through signal frames.
3091 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3092 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3093 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3094 for the address of the code being executed and @code{context->cfa} for
3095 the stack pointer value. If the frame can be decoded, the register save
3096 addresses should be updated in @var{fs} and the macro should branch to
3097 @var{success}. If the frame cannot be decoded, the macro should do
3100 For proper signal handling in Java this macro is accompanied by
3101 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3104 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3105 This macro allows the target to add operating system specific code to the
3106 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3107 usually used for signal or interrupt frames.
3109 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3110 @var{context} is an @code{_Unwind_Context};
3111 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3112 for the abi and context in the @code{.unwabi} directive. If the
3113 @code{.unwabi} directive can be handled, the register save addresses should
3114 be updated in @var{fs}.
3117 @defmac TARGET_USES_WEAK_UNWIND_INFO
3118 A C expression that evaluates to true if the target requires unwind
3119 info to be given comdat linkage. Define it to be @code{1} if comdat
3120 linkage is necessary. The default is @code{0}.
3123 @node Stack Checking
3124 @subsection Specifying How Stack Checking is Done
3126 GCC will check that stack references are within the boundaries of
3127 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3131 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3132 will assume that you have arranged for stack checking to be done at
3133 appropriate places in the configuration files, e.g., in
3134 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3138 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3139 called @code{check_stack} in your @file{md} file, GCC will call that
3140 pattern with one argument which is the address to compare the stack
3141 value against. You must arrange for this pattern to report an error if
3142 the stack pointer is out of range.
3145 If neither of the above are true, GCC will generate code to periodically
3146 ``probe'' the stack pointer using the values of the macros defined below.
3149 Normally, you will use the default values of these macros, so GCC
3150 will use the third approach.
3152 @defmac STACK_CHECK_BUILTIN
3153 A nonzero value if stack checking is done by the configuration files in a
3154 machine-dependent manner. You should define this macro if stack checking
3155 is require by the ABI of your machine or if you would like to have to stack
3156 checking in some more efficient way than GCC's portable approach.
3157 The default value of this macro is zero.
3160 @defmac STACK_CHECK_PROBE_INTERVAL
3161 An integer representing the interval at which GCC must generate stack
3162 probe instructions. You will normally define this macro to be no larger
3163 than the size of the ``guard pages'' at the end of a stack area. The
3164 default value of 4096 is suitable for most systems.
3167 @defmac STACK_CHECK_PROBE_LOAD
3168 A integer which is nonzero if GCC should perform the stack probe
3169 as a load instruction and zero if GCC should use a store instruction.
3170 The default is zero, which is the most efficient choice on most systems.
3173 @defmac STACK_CHECK_PROTECT
3174 The number of bytes of stack needed to recover from a stack overflow,
3175 for languages where such a recovery is supported. The default value of
3176 75 words should be adequate for most machines.
3179 @defmac STACK_CHECK_MAX_FRAME_SIZE
3180 The maximum size of a stack frame, in bytes. GCC will generate probe
3181 instructions in non-leaf functions to ensure at least this many bytes of
3182 stack are available. If a stack frame is larger than this size, stack
3183 checking will not be reliable and GCC will issue a warning. The
3184 default is chosen so that GCC only generates one instruction on most
3185 systems. You should normally not change the default value of this macro.
3188 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3189 GCC uses this value to generate the above warning message. It
3190 represents the amount of fixed frame used by a function, not including
3191 space for any callee-saved registers, temporaries and user variables.
3192 You need only specify an upper bound for this amount and will normally
3193 use the default of four words.
3196 @defmac STACK_CHECK_MAX_VAR_SIZE
3197 The maximum size, in bytes, of an object that GCC will place in the
3198 fixed area of the stack frame when the user specifies
3199 @option{-fstack-check}.
3200 GCC computed the default from the values of the above macros and you will
3201 normally not need to override that default.
3205 @node Frame Registers
3206 @subsection Registers That Address the Stack Frame
3208 @c prevent bad page break with this line
3209 This discusses registers that address the stack frame.
3211 @defmac STACK_POINTER_REGNUM
3212 The register number of the stack pointer register, which must also be a
3213 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3214 the hardware determines which register this is.
3217 @defmac FRAME_POINTER_REGNUM
3218 The register number of the frame pointer register, which is used to
3219 access automatic variables in the stack frame. On some machines, the
3220 hardware determines which register this is. On other machines, you can
3221 choose any register you wish for this purpose.
3224 @defmac HARD_FRAME_POINTER_REGNUM
3225 On some machines the offset between the frame pointer and starting
3226 offset of the automatic variables is not known until after register
3227 allocation has been done (for example, because the saved registers are
3228 between these two locations). On those machines, define
3229 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3230 be used internally until the offset is known, and define
3231 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3232 used for the frame pointer.
3234 You should define this macro only in the very rare circumstances when it
3235 is not possible to calculate the offset between the frame pointer and
3236 the automatic variables until after register allocation has been
3237 completed. When this macro is defined, you must also indicate in your
3238 definition of @code{ELIMINABLE_REGS} how to eliminate
3239 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3240 or @code{STACK_POINTER_REGNUM}.
3242 Do not define this macro if it would be the same as
3243 @code{FRAME_POINTER_REGNUM}.
3246 @defmac ARG_POINTER_REGNUM
3247 The register number of the arg pointer register, which is used to access
3248 the function's argument list. On some machines, this is the same as the
3249 frame pointer register. On some machines, the hardware determines which
3250 register this is. On other machines, you can choose any register you
3251 wish for this purpose. If this is not the same register as the frame
3252 pointer register, then you must mark it as a fixed register according to
3253 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3254 (@pxref{Elimination}).
3257 @defmac RETURN_ADDRESS_POINTER_REGNUM
3258 The register number of the return address pointer register, which is used to
3259 access the current function's return address from the stack. On some
3260 machines, the return address is not at a fixed offset from the frame
3261 pointer or stack pointer or argument pointer. This register can be defined
3262 to point to the return address on the stack, and then be converted by
3263 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3265 Do not define this macro unless there is no other way to get the return
3266 address from the stack.
3269 @defmac STATIC_CHAIN_REGNUM
3270 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3271 Register numbers used for passing a function's static chain pointer. If
3272 register windows are used, the register number as seen by the called
3273 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3274 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3275 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3278 The static chain register need not be a fixed register.
3280 If the static chain is passed in memory, these macros should not be
3281 defined; instead, the next two macros should be defined.
3284 @defmac STATIC_CHAIN
3285 @defmacx STATIC_CHAIN_INCOMING
3286 If the static chain is passed in memory, these macros provide rtx giving
3287 @code{mem} expressions that denote where they are stored.
3288 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3289 as seen by the calling and called functions, respectively. Often the former
3290 will be at an offset from the stack pointer and the latter at an offset from
3293 @findex stack_pointer_rtx
3294 @findex frame_pointer_rtx
3295 @findex arg_pointer_rtx
3296 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3297 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3298 macros and should be used to refer to those items.
3300 If the static chain is passed in a register, the two previous macros should
3304 @defmac DWARF_FRAME_REGISTERS
3305 This macro specifies the maximum number of hard registers that can be
3306 saved in a call frame. This is used to size data structures used in
3307 DWARF2 exception handling.
3309 Prior to GCC 3.0, this macro was needed in order to establish a stable
3310 exception handling ABI in the face of adding new hard registers for ISA
3311 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3312 in the number of hard registers. Nevertheless, this macro can still be
3313 used to reduce the runtime memory requirements of the exception handling
3314 routines, which can be substantial if the ISA contains a lot of
3315 registers that are not call-saved.
3317 If this macro is not defined, it defaults to
3318 @code{FIRST_PSEUDO_REGISTER}.
3321 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3323 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3324 for backward compatibility in pre GCC 3.0 compiled code.
3326 If this macro is not defined, it defaults to
3327 @code{DWARF_FRAME_REGISTERS}.
3330 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3332 Define this macro if the target's representation for dwarf registers
3333 is different than the internal representation for unwind column.
3334 Given a dwarf register, this macro should return the internal unwind
3335 column number to use instead.
3337 See the PowerPC's SPE target for an example.
3340 @defmac DWARF_FRAME_REGNUM (@var{regno})
3342 Define this macro if the target's representation for dwarf registers
3343 used in .eh_frame or .debug_frame is different from that used in other
3344 debug info sections. Given a GCC hard register number, this macro
3345 should return the .eh_frame register number. The default is
3346 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3350 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3352 Define this macro to map register numbers held in the call frame info
3353 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3354 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3355 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3356 return @code{@var{regno}}.
3361 @subsection Eliminating Frame Pointer and Arg Pointer
3363 @c prevent bad page break with this line
3364 This is about eliminating the frame pointer and arg pointer.
3366 @defmac FRAME_POINTER_REQUIRED
3367 A C expression which is nonzero if a function must have and use a frame
3368 pointer. This expression is evaluated in the reload pass. If its value is
3369 nonzero the function will have a frame pointer.
3371 The expression can in principle examine the current function and decide
3372 according to the facts, but on most machines the constant 0 or the
3373 constant 1 suffices. Use 0 when the machine allows code to be generated
3374 with no frame pointer, and doing so saves some time or space. Use 1
3375 when there is no possible advantage to avoiding a frame pointer.
3377 In certain cases, the compiler does not know how to produce valid code
3378 without a frame pointer. The compiler recognizes those cases and
3379 automatically gives the function a frame pointer regardless of what
3380 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3383 In a function that does not require a frame pointer, the frame pointer
3384 register can be allocated for ordinary usage, unless you mark it as a
3385 fixed register. See @code{FIXED_REGISTERS} for more information.
3388 @findex get_frame_size
3389 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3390 A C statement to store in the variable @var{depth-var} the difference
3391 between the frame pointer and the stack pointer values immediately after
3392 the function prologue. The value would be computed from information
3393 such as the result of @code{get_frame_size ()} and the tables of
3394 registers @code{regs_ever_live} and @code{call_used_regs}.
3396 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3397 need not be defined. Otherwise, it must be defined even if
3398 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3399 case, you may set @var{depth-var} to anything.
3402 @defmac ELIMINABLE_REGS
3403 If defined, this macro specifies a table of register pairs used to
3404 eliminate unneeded registers that point into the stack frame. If it is not
3405 defined, the only elimination attempted by the compiler is to replace
3406 references to the frame pointer with references to the stack pointer.
3408 The definition of this macro is a list of structure initializations, each
3409 of which specifies an original and replacement register.
3411 On some machines, the position of the argument pointer is not known until
3412 the compilation is completed. In such a case, a separate hard register
3413 must be used for the argument pointer. This register can be eliminated by
3414 replacing it with either the frame pointer or the argument pointer,
3415 depending on whether or not the frame pointer has been eliminated.
3417 In this case, you might specify:
3419 #define ELIMINABLE_REGS \
3420 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3421 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3422 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3425 Note that the elimination of the argument pointer with the stack pointer is
3426 specified first since that is the preferred elimination.
3429 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3430 A C expression that returns nonzero if the compiler is allowed to try
3431 to replace register number @var{from-reg} with register number
3432 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3433 is defined, and will usually be the constant 1, since most of the cases
3434 preventing register elimination are things that the compiler already
3438 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3439 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3440 specifies the initial difference between the specified pair of
3441 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3445 @node Stack Arguments
3446 @subsection Passing Function Arguments on the Stack
3447 @cindex arguments on stack
3448 @cindex stack arguments
3450 The macros in this section control how arguments are passed
3451 on the stack. See the following section for other macros that
3452 control passing certain arguments in registers.
3454 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3455 This target hook returns @code{true} if an argument declared in a
3456 prototype as an integral type smaller than @code{int} should actually be
3457 passed as an @code{int}. In addition to avoiding errors in certain
3458 cases of mismatch, it also makes for better code on certain machines.
3459 The default is to not promote prototypes.
3463 A C expression. If nonzero, push insns will be used to pass
3465 If the target machine does not have a push instruction, set it to zero.
3466 That directs GCC to use an alternate strategy: to
3467 allocate the entire argument block and then store the arguments into
3468 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3471 @defmac PUSH_ARGS_REVERSED
3472 A C expression. If nonzero, function arguments will be evaluated from
3473 last to first, rather than from first to last. If this macro is not
3474 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3475 and args grow in opposite directions, and 0 otherwise.
3478 @defmac PUSH_ROUNDING (@var{npushed})
3479 A C expression that is the number of bytes actually pushed onto the
3480 stack when an instruction attempts to push @var{npushed} bytes.
3482 On some machines, the definition
3485 #define PUSH_ROUNDING(BYTES) (BYTES)
3489 will suffice. But on other machines, instructions that appear
3490 to push one byte actually push two bytes in an attempt to maintain
3491 alignment. Then the definition should be
3494 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3498 @findex current_function_outgoing_args_size
3499 @defmac ACCUMULATE_OUTGOING_ARGS
3500 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3501 will be computed and placed into the variable
3502 @code{current_function_outgoing_args_size}. No space will be pushed
3503 onto the stack for each call; instead, the function prologue should
3504 increase the stack frame size by this amount.
3506 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3510 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3511 Define this macro if functions should assume that stack space has been
3512 allocated for arguments even when their values are passed in
3515 The value of this macro is the size, in bytes, of the area reserved for
3516 arguments passed in registers for the function represented by @var{fndecl},
3517 which can be zero if GCC is calling a library function.
3519 This space can be allocated by the caller, or be a part of the
3520 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3523 @c above is overfull. not sure what to do. --mew 5feb93 did
3524 @c something, not sure if it looks good. --mew 10feb93
3526 @defmac OUTGOING_REG_PARM_STACK_SPACE
3527 Define this if it is the responsibility of the caller to allocate the area
3528 reserved for arguments passed in registers.
3530 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3531 whether the space for these arguments counts in the value of
3532 @code{current_function_outgoing_args_size}.
3535 @defmac STACK_PARMS_IN_REG_PARM_AREA
3536 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3537 stack parameters don't skip the area specified by it.
3538 @c i changed this, makes more sens and it should have taken care of the
3539 @c overfull.. not as specific, tho. --mew 5feb93
3541 Normally, when a parameter is not passed in registers, it is placed on the
3542 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3543 suppresses this behavior and causes the parameter to be passed on the
3544 stack in its natural location.
3547 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3548 A C expression that should indicate the number of bytes of its own
3549 arguments that a function pops on returning, or 0 if the
3550 function pops no arguments and the caller must therefore pop them all
3551 after the function returns.
3553 @var{fundecl} is a C variable whose value is a tree node that describes
3554 the function in question. Normally it is a node of type
3555 @code{FUNCTION_DECL} that describes the declaration of the function.
3556 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3558 @var{funtype} is a C variable whose value is a tree node that
3559 describes the function in question. Normally it is a node of type
3560 @code{FUNCTION_TYPE} that describes the data type of the function.
3561 From this it is possible to obtain the data types of the value and
3562 arguments (if known).
3564 When a call to a library function is being considered, @var{fundecl}
3565 will contain an identifier node for the library function. Thus, if
3566 you need to distinguish among various library functions, you can do so
3567 by their names. Note that ``library function'' in this context means
3568 a function used to perform arithmetic, whose name is known specially
3569 in the compiler and was not mentioned in the C code being compiled.
3571 @var{stack-size} is the number of bytes of arguments passed on the
3572 stack. If a variable number of bytes is passed, it is zero, and
3573 argument popping will always be the responsibility of the calling function.
3575 On the VAX, all functions always pop their arguments, so the definition
3576 of this macro is @var{stack-size}. On the 68000, using the standard
3577 calling convention, no functions pop their arguments, so the value of
3578 the macro is always 0 in this case. But an alternative calling
3579 convention is available in which functions that take a fixed number of
3580 arguments pop them but other functions (such as @code{printf}) pop
3581 nothing (the caller pops all). When this convention is in use,
3582 @var{funtype} is examined to determine whether a function takes a fixed
3583 number of arguments.
3586 @defmac CALL_POPS_ARGS (@var{cum})
3587 A C expression that should indicate the number of bytes a call sequence
3588 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3589 when compiling a function call.
3591 @var{cum} is the variable in which all arguments to the called function
3592 have been accumulated.
3594 On certain architectures, such as the SH5, a call trampoline is used
3595 that pops certain registers off the stack, depending on the arguments
3596 that have been passed to the function. Since this is a property of the
3597 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3601 @node Register Arguments
3602 @subsection Passing Arguments in Registers
3603 @cindex arguments in registers
3604 @cindex registers arguments
3606 This section describes the macros which let you control how various
3607 types of arguments are passed in registers or how they are arranged in
3610 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3611 A C expression that controls whether a function argument is passed
3612 in a register, and which register.
3614 The arguments are @var{cum}, which summarizes all the previous
3615 arguments; @var{mode}, the machine mode of the argument; @var{type},
3616 the data type of the argument as a tree node or 0 if that is not known
3617 (which happens for C support library functions); and @var{named},
3618 which is 1 for an ordinary argument and 0 for nameless arguments that
3619 correspond to @samp{@dots{}} in the called function's prototype.
3620 @var{type} can be an incomplete type if a syntax error has previously
3623 The value of the expression is usually either a @code{reg} RTX for the
3624 hard register in which to pass the argument, or zero to pass the
3625 argument on the stack.
3627 For machines like the VAX and 68000, where normally all arguments are
3628 pushed, zero suffices as a definition.
3630 The value of the expression can also be a @code{parallel} RTX@. This is
3631 used when an argument is passed in multiple locations. The mode of the
3632 @code{parallel} should be the mode of the entire argument. The
3633 @code{parallel} holds any number of @code{expr_list} pairs; each one
3634 describes where part of the argument is passed. In each
3635 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3636 register in which to pass this part of the argument, and the mode of the
3637 register RTX indicates how large this part of the argument is. The
3638 second operand of the @code{expr_list} is a @code{const_int} which gives
3639 the offset in bytes into the entire argument of where this part starts.
3640 As a special exception the first @code{expr_list} in the @code{parallel}
3641 RTX may have a first operand of zero. This indicates that the entire
3642 argument is also stored on the stack.
3644 The last time this macro is called, it is called with @code{MODE ==
3645 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3646 pattern as operands 2 and 3 respectively.
3648 @cindex @file{stdarg.h} and register arguments
3649 The usual way to make the ISO library @file{stdarg.h} work on a machine
3650 where some arguments are usually passed in registers, is to cause
3651 nameless arguments to be passed on the stack instead. This is done
3652 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3654 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3655 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3656 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3657 in the definition of this macro to determine if this argument is of a
3658 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3659 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3660 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3661 defined, the argument will be computed in the stack and then loaded into
3665 @defmac MUST_PASS_IN_STACK (@var{mode}, @var{type})
3666 Define as a C expression that evaluates to nonzero if we do not know how
3667 to pass TYPE solely in registers. The file @file{expr.h} defines a
3668 definition that is usually appropriate, refer to @file{expr.h} for additional
3672 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3673 Define this macro if the target machine has ``register windows'', so
3674 that the register in which a function sees an arguments is not
3675 necessarily the same as the one in which the caller passed the
3678 For such machines, @code{FUNCTION_ARG} computes the register in which
3679 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3680 be defined in a similar fashion to tell the function being called
3681 where the arguments will arrive.
3683 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3684 serves both purposes.
3687 @defmac FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3688 A C expression for the number of words, at the beginning of an
3689 argument, that must be put in registers. The value must be zero for
3690 arguments that are passed entirely in registers or that are entirely
3691 pushed on the stack.
3693 On some machines, certain arguments must be passed partially in
3694 registers and partially in memory. On these machines, typically the
3695 first @var{n} words of arguments are passed in registers, and the rest
3696 on the stack. If a multi-word argument (a @code{double} or a
3697 structure) crosses that boundary, its first few words must be passed
3698 in registers and the rest must be pushed. This macro tells the
3699 compiler when this occurs, and how many of the words should go in
3702 @code{FUNCTION_ARG} for these arguments should return the first
3703 register to be used by the caller for this argument; likewise
3704 @code{FUNCTION_INCOMING_ARG}, for the called function.
3707 @defmac FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3708 A C expression that indicates when an argument must be passed by reference.
3709 If nonzero for an argument, a copy of that argument is made in memory and a
3710 pointer to the argument is passed instead of the argument itself.
3711 The pointer is passed in whatever way is appropriate for passing a pointer
3714 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3715 definition of this macro might be
3717 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3718 (CUM, MODE, TYPE, NAMED) \
3719 MUST_PASS_IN_STACK (MODE, TYPE)
3721 @c this is *still* too long. --mew 5feb93
3724 @defmac FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3725 If defined, a C expression that indicates when it is the called function's
3726 responsibility to make a copy of arguments passed by invisible reference.
3727 Normally, the caller makes a copy and passes the address of the copy to the
3728 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3729 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3730 ``live'' value. The called function must not modify this value. If it can be
3731 determined that the value won't be modified, it need not make a copy;
3732 otherwise a copy must be made.
3735 @defmac CUMULATIVE_ARGS
3736 A C type for declaring a variable that is used as the first argument of
3737 @code{FUNCTION_ARG} and other related values. For some target machines,
3738 the type @code{int} suffices and can hold the number of bytes of
3741 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3742 arguments that have been passed on the stack. The compiler has other
3743 variables to keep track of that. For target machines on which all
3744 arguments are passed on the stack, there is no need to store anything in
3745 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3746 should not be empty, so use @code{int}.
3749 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3750 A C statement (sans semicolon) for initializing the variable
3751 @var{cum} for the state at the beginning of the argument list. The
3752 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3753 is the tree node for the data type of the function which will receive
3754 the args, or 0 if the args are to a compiler support library function.
3755 For direct calls that are not libcalls, @var{fndecl} contain the
3756 declaration node of the function. @var{fndecl} is also set when
3757 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3758 being compiled. @var{n_named_args} is set to the number of named
3759 arguments, including a structure return address if it is passed as a
3760 parameter, when making a call. When processing incoming arguments,
3761 @var{n_named_args} is set to -1.
3763 When processing a call to a compiler support library function,
3764 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3765 contains the name of the function, as a string. @var{libname} is 0 when
3766 an ordinary C function call is being processed. Thus, each time this
3767 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3768 never both of them at once.
3771 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3772 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3773 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3774 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3775 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3776 0)} is used instead.
3779 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3780 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3781 finding the arguments for the function being compiled. If this macro is
3782 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3784 The value passed for @var{libname} is always 0, since library routines
3785 with special calling conventions are never compiled with GCC@. The
3786 argument @var{libname} exists for symmetry with
3787 @code{INIT_CUMULATIVE_ARGS}.
3788 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3789 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3792 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3793 A C statement (sans semicolon) to update the summarizer variable
3794 @var{cum} to advance past an argument in the argument list. The
3795 values @var{mode}, @var{type} and @var{named} describe that argument.
3796 Once this is done, the variable @var{cum} is suitable for analyzing
3797 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3799 This macro need not do anything if the argument in question was passed
3800 on the stack. The compiler knows how to track the amount of stack space
3801 used for arguments without any special help.
3804 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3805 If defined, a C expression which determines whether, and in which direction,
3806 to pad out an argument with extra space. The value should be of type
3807 @code{enum direction}: either @code{upward} to pad above the argument,
3808 @code{downward} to pad below, or @code{none} to inhibit padding.
3810 The @emph{amount} of padding is always just enough to reach the next
3811 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3814 This macro has a default definition which is right for most systems.
3815 For little-endian machines, the default is to pad upward. For
3816 big-endian machines, the default is to pad downward for an argument of
3817 constant size shorter than an @code{int}, and upward otherwise.
3820 @defmac PAD_VARARGS_DOWN
3821 If defined, a C expression which determines whether the default
3822 implementation of va_arg will attempt to pad down before reading the
3823 next argument, if that argument is smaller than its aligned space as
3824 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3825 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3828 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3829 Specify padding for the last element of a block move between registers and
3830 memory. @var{first} is nonzero if this is the only element. Defining this
3831 macro allows better control of register function parameters on big-endian
3832 machines, without using @code{PARALLEL} rtl. In particular,
3833 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3834 registers, as there is no longer a "wrong" part of a register; For example,
3835 a three byte aggregate may be passed in the high part of a register if so
3839 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3840 If defined, a C expression that gives the alignment boundary, in bits,
3841 of an argument with the specified mode and type. If it is not defined,
3842 @code{PARM_BOUNDARY} is used for all arguments.
3845 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3846 A C expression that is nonzero if @var{regno} is the number of a hard
3847 register in which function arguments are sometimes passed. This does
3848 @emph{not} include implicit arguments such as the static chain and
3849 the structure-value address. On many machines, no registers can be
3850 used for this purpose since all function arguments are pushed on the
3854 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
3855 This hook should return true if parameter of type @var{type} are passed
3856 as two scalar parameters. By default, GCC will attempt to pack complex
3857 arguments into the target's word size. Some ABIs require complex arguments
3858 to be split and treated as their individual components. For example, on
3859 AIX64, complex floats should be passed in a pair of floating point
3860 registers, even though a complex float would fit in one 64-bit floating
3863 The default value of this hook is @code{NULL}, which is treated as always
3868 @subsection How Scalar Function Values Are Returned
3869 @cindex return values in registers
3870 @cindex values, returned by functions
3871 @cindex scalars, returned as values
3873 This section discusses the macros that control returning scalars as
3874 values---values that can fit in registers.
3876 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3877 A C expression to create an RTX representing the place where a
3878 function returns a value of data type @var{valtype}. @var{valtype} is
3879 a tree node representing a data type. Write @code{TYPE_MODE
3880 (@var{valtype})} to get the machine mode used to represent that type.
3881 On many machines, only the mode is relevant. (Actually, on most
3882 machines, scalar values are returned in the same place regardless of
3885 The value of the expression is usually a @code{reg} RTX for the hard
3886 register where the return value is stored. The value can also be a
3887 @code{parallel} RTX, if the return value is in multiple places. See
3888 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3890 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply the same
3891 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3894 If the precise function being called is known, @var{func} is a tree
3895 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3896 pointer. This makes it possible to use a different value-returning
3897 convention for specific functions when all their calls are
3900 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3901 types, because these are returned in another way. See
3902 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
3905 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3906 Define this macro if the target machine has ``register windows''
3907 so that the register in which a function returns its value is not
3908 the same as the one in which the caller sees the value.
3910 For such machines, @code{FUNCTION_VALUE} computes the register in which
3911 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3912 defined in a similar fashion to tell the function where to put the
3915 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3916 @code{FUNCTION_VALUE} serves both purposes.
3918 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3919 aggregate data types, because these are returned in another way. See
3920 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
3923 @defmac LIBCALL_VALUE (@var{mode})
3924 A C expression to create an RTX representing the place where a library
3925 function returns a value of mode @var{mode}. If the precise function
3926 being called is known, @var{func} is a tree node
3927 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3928 pointer. This makes it possible to use a different value-returning
3929 convention for specific functions when all their calls are
3932 Note that ``library function'' in this context means a compiler
3933 support routine, used to perform arithmetic, whose name is known
3934 specially by the compiler and was not mentioned in the C code being
3937 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3938 data types, because none of the library functions returns such types.
3941 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3942 A C expression that is nonzero if @var{regno} is the number of a hard
3943 register in which the values of called function may come back.
3945 A register whose use for returning values is limited to serving as the
3946 second of a pair (for a value of type @code{double}, say) need not be
3947 recognized by this macro. So for most machines, this definition
3951 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3954 If the machine has register windows, so that the caller and the called
3955 function use different registers for the return value, this macro
3956 should recognize only the caller's register numbers.
3959 @defmac APPLY_RESULT_SIZE
3960 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3961 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3962 saving and restoring an arbitrary return value.
3965 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
3966 This hook should return true if values of type @var{type} are returned
3967 at the most significant end of a register (in other words, if they are
3968 padded at the least significant end). You can assume that @var{type}
3969 is returned in a register; the caller is required to check this.
3971 Note that the register provided by @code{FUNCTION_VALUE} must be able
3972 to hold the complete return value. For example, if a 1-, 2- or 3-byte
3973 structure is returned at the most significant end of a 4-byte register,
3974 @code{FUNCTION_VALUE} should provide an @code{SImode} rtx.
3977 @node Aggregate Return
3978 @subsection How Large Values Are Returned
3979 @cindex aggregates as return values
3980 @cindex large return values
3981 @cindex returning aggregate values
3982 @cindex structure value address
3984 When a function value's mode is @code{BLKmode} (and in some other
3985 cases), the value is not returned according to @code{FUNCTION_VALUE}
3986 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3987 block of memory in which the value should be stored. This address
3988 is called the @dfn{structure value address}.
3990 This section describes how to control returning structure values in
3993 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
3994 This target hook should return a nonzero value to say to return the
3995 function value in memory, just as large structures are always returned.
3996 Here @var{type} will be the data type of the value, and @var{fntype}
3997 will be the type of the function doing the returning, or @code{NULL} for
4000 Note that values of mode @code{BLKmode} must be explicitly handled
4001 by this function. Also, the option @option{-fpcc-struct-return}
4002 takes effect regardless of this macro. On most systems, it is
4003 possible to leave the hook undefined; this causes a default
4004 definition to be used, whose value is the constant 1 for @code{BLKmode}
4005 values, and 0 otherwise.
4007 Do not use this hook to indicate that structures and unions should always
4008 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4012 @defmac DEFAULT_PCC_STRUCT_RETURN
4013 Define this macro to be 1 if all structure and union return values must be
4014 in memory. Since this results in slower code, this should be defined
4015 only if needed for compatibility with other compilers or with an ABI@.
4016 If you define this macro to be 0, then the conventions used for structure
4017 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4020 If not defined, this defaults to the value 1.
4023 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4024 This target hook should return the location of the structure value
4025 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4026 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4027 be @code{NULL}, for libcalls. You do not need to define this target
4028 hook if the address is always passed as an ``invisible'' first
4031 On some architectures the place where the structure value address
4032 is found by the called function is not the same place that the
4033 caller put it. This can be due to register windows, or it could
4034 be because the function prologue moves it to a different place.
4035 @var{incoming} is @code{true} when the location is needed in
4036 the context of the called function, and @code{false} in the context of
4039 If @var{incoming} is @code{true} and the address is to be found on the
4040 stack, return a @code{mem} which refers to the frame pointer.
4043 @defmac PCC_STATIC_STRUCT_RETURN
4044 Define this macro if the usual system convention on the target machine
4045 for returning structures and unions is for the called function to return
4046 the address of a static variable containing the value.
4048 Do not define this if the usual system convention is for the caller to
4049 pass an address to the subroutine.
4051 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4052 nothing when you use @option{-freg-struct-return} mode.
4056 @subsection Caller-Saves Register Allocation
4058 If you enable it, GCC can save registers around function calls. This
4059 makes it possible to use call-clobbered registers to hold variables that
4060 must live across calls.
4062 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4063 A C expression to determine whether it is worthwhile to consider placing
4064 a pseudo-register in a call-clobbered hard register and saving and
4065 restoring it around each function call. The expression should be 1 when
4066 this is worth doing, and 0 otherwise.
4068 If you don't define this macro, a default is used which is good on most
4069 machines: @code{4 * @var{calls} < @var{refs}}.
4072 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4073 A C expression specifying which mode is required for saving @var{nregs}
4074 of a pseudo-register in call-clobbered hard register @var{regno}. If
4075 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4076 returned. For most machines this macro need not be defined since GCC
4077 will select the smallest suitable mode.
4080 @node Function Entry
4081 @subsection Function Entry and Exit
4082 @cindex function entry and exit
4086 This section describes the macros that output function entry
4087 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4089 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4090 If defined, a function that outputs the assembler code for entry to a
4091 function. The prologue is responsible for setting up the stack frame,
4092 initializing the frame pointer register, saving registers that must be
4093 saved, and allocating @var{size} additional bytes of storage for the
4094 local variables. @var{size} is an integer. @var{file} is a stdio
4095 stream to which the assembler code should be output.
4097 The label for the beginning of the function need not be output by this
4098 macro. That has already been done when the macro is run.
4100 @findex regs_ever_live
4101 To determine which registers to save, the macro can refer to the array
4102 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4103 @var{r} is used anywhere within the function. This implies the function
4104 prologue should save register @var{r}, provided it is not one of the
4105 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4106 @code{regs_ever_live}.)
4108 On machines that have ``register windows'', the function entry code does
4109 not save on the stack the registers that are in the windows, even if
4110 they are supposed to be preserved by function calls; instead it takes
4111 appropriate steps to ``push'' the register stack, if any non-call-used
4112 registers are used in the function.
4114 @findex frame_pointer_needed
4115 On machines where functions may or may not have frame-pointers, the
4116 function entry code must vary accordingly; it must set up the frame
4117 pointer if one is wanted, and not otherwise. To determine whether a
4118 frame pointer is in wanted, the macro can refer to the variable
4119 @code{frame_pointer_needed}. The variable's value will be 1 at run
4120 time in a function that needs a frame pointer. @xref{Elimination}.
4122 The function entry code is responsible for allocating any stack space
4123 required for the function. This stack space consists of the regions
4124 listed below. In most cases, these regions are allocated in the
4125 order listed, with the last listed region closest to the top of the
4126 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4127 the highest address if it is not defined). You can use a different order
4128 for a machine if doing so is more convenient or required for
4129 compatibility reasons. Except in cases where required by standard
4130 or by a debugger, there is no reason why the stack layout used by GCC
4131 need agree with that used by other compilers for a machine.
4134 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4135 If defined, a function that outputs assembler code at the end of a
4136 prologue. This should be used when the function prologue is being
4137 emitted as RTL, and you have some extra assembler that needs to be
4138 emitted. @xref{prologue instruction pattern}.
4141 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4142 If defined, a function that outputs assembler code at the start of an
4143 epilogue. This should be used when the function epilogue is being
4144 emitted as RTL, and you have some extra assembler that needs to be
4145 emitted. @xref{epilogue instruction pattern}.
4148 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4149 If defined, a function that outputs the assembler code for exit from a
4150 function. The epilogue is responsible for restoring the saved
4151 registers and stack pointer to their values when the function was
4152 called, and returning control to the caller. This macro takes the
4153 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4154 registers to restore are determined from @code{regs_ever_live} and
4155 @code{CALL_USED_REGISTERS} in the same way.
4157 On some machines, there is a single instruction that does all the work
4158 of returning from the function. On these machines, give that
4159 instruction the name @samp{return} and do not define the macro
4160 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4162 Do not define a pattern named @samp{return} if you want the
4163 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4164 switches to control whether return instructions or epilogues are used,
4165 define a @samp{return} pattern with a validity condition that tests the
4166 target switches appropriately. If the @samp{return} pattern's validity
4167 condition is false, epilogues will be used.
4169 On machines where functions may or may not have frame-pointers, the
4170 function exit code must vary accordingly. Sometimes the code for these
4171 two cases is completely different. To determine whether a frame pointer
4172 is wanted, the macro can refer to the variable
4173 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4174 a function that needs a frame pointer.
4176 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4177 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4178 The C variable @code{current_function_is_leaf} is nonzero for such a
4179 function. @xref{Leaf Functions}.
4181 On some machines, some functions pop their arguments on exit while
4182 others leave that for the caller to do. For example, the 68020 when
4183 given @option{-mrtd} pops arguments in functions that take a fixed
4184 number of arguments.
4186 @findex current_function_pops_args
4187 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4188 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4189 needs to know what was decided. The variable that is called
4190 @code{current_function_pops_args} is the number of bytes of its
4191 arguments that a function should pop. @xref{Scalar Return}.
4192 @c what is the "its arguments" in the above sentence referring to, pray
4193 @c tell? --mew 5feb93
4198 @findex current_function_pretend_args_size
4199 A region of @code{current_function_pretend_args_size} bytes of
4200 uninitialized space just underneath the first argument arriving on the
4201 stack. (This may not be at the very start of the allocated stack region
4202 if the calling sequence has pushed anything else since pushing the stack
4203 arguments. But usually, on such machines, nothing else has been pushed
4204 yet, because the function prologue itself does all the pushing.) This
4205 region is used on machines where an argument may be passed partly in
4206 registers and partly in memory, and, in some cases to support the
4207 features in @code{<stdarg.h>}.
4210 An area of memory used to save certain registers used by the function.
4211 The size of this area, which may also include space for such things as
4212 the return address and pointers to previous stack frames, is
4213 machine-specific and usually depends on which registers have been used
4214 in the function. Machines with register windows often do not require
4218 A region of at least @var{size} bytes, possibly rounded up to an allocation
4219 boundary, to contain the local variables of the function. On some machines,
4220 this region and the save area may occur in the opposite order, with the
4221 save area closer to the top of the stack.
4224 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4225 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4226 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4227 argument lists of the function. @xref{Stack Arguments}.
4230 Normally, it is necessary for the macros
4231 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4232 @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially.
4233 The C variable @code{current_function_is_leaf} is nonzero for such a
4236 @defmac EXIT_IGNORE_STACK
4237 Define this macro as a C expression that is nonzero if the return
4238 instruction or the function epilogue ignores the value of the stack
4239 pointer; in other words, if it is safe to delete an instruction to
4240 adjust the stack pointer before a return from the function. The
4243 Note that this macro's value is relevant only for functions for which
4244 frame pointers are maintained. It is never safe to delete a final
4245 stack adjustment in a function that has no frame pointer, and the
4246 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4249 @defmac EPILOGUE_USES (@var{regno})
4250 Define this macro as a C expression that is nonzero for registers that are
4251 used by the epilogue or the @samp{return} pattern. The stack and frame
4252 pointer registers are already be assumed to be used as needed.
4255 @defmac EH_USES (@var{regno})
4256 Define this macro as a C expression that is nonzero for registers that are
4257 used by the exception handling mechanism, and so should be considered live
4258 on entry to an exception edge.
4261 @defmac DELAY_SLOTS_FOR_EPILOGUE
4262 Define this macro if the function epilogue contains delay slots to which
4263 instructions from the rest of the function can be ``moved''. The
4264 definition should be a C expression whose value is an integer
4265 representing the number of delay slots there.
4268 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4269 A C expression that returns 1 if @var{insn} can be placed in delay
4270 slot number @var{n} of the epilogue.
4272 The argument @var{n} is an integer which identifies the delay slot now
4273 being considered (since different slots may have different rules of
4274 eligibility). It is never negative and is always less than the number
4275 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4276 If you reject a particular insn for a given delay slot, in principle, it
4277 may be reconsidered for a subsequent delay slot. Also, other insns may
4278 (at least in principle) be considered for the so far unfilled delay
4281 @findex current_function_epilogue_delay_list
4282 @findex final_scan_insn
4283 The insns accepted to fill the epilogue delay slots are put in an RTL
4284 list made with @code{insn_list} objects, stored in the variable
4285 @code{current_function_epilogue_delay_list}. The insn for the first
4286 delay slot comes first in the list. Your definition of the macro
4287 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4288 outputting the insns in this list, usually by calling
4289 @code{final_scan_insn}.
4291 You need not define this macro if you did not define
4292 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4295 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, tree @var{function})
4296 A function that outputs the assembler code for a thunk
4297 function, used to implement C++ virtual function calls with multiple
4298 inheritance. The thunk acts as a wrapper around a virtual function,
4299 adjusting the implicit object parameter before handing control off to
4302 First, emit code to add the integer @var{delta} to the location that
4303 contains the incoming first argument. Assume that this argument
4304 contains a pointer, and is the one used to pass the @code{this} pointer
4305 in C++. This is the incoming argument @emph{before} the function prologue,
4306 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4307 all other incoming arguments.
4309 After the addition, emit code to jump to @var{function}, which is a
4310 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4311 not touch the return address. Hence returning from @var{FUNCTION} will
4312 return to whoever called the current @samp{thunk}.
4314 The effect must be as if @var{function} had been called directly with
4315 the adjusted first argument. This macro is responsible for emitting all
4316 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4317 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4319 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4320 have already been extracted from it.) It might possibly be useful on
4321 some targets, but probably not.
4323 If you do not define this macro, the target-independent code in the C++
4324 front end will generate a less efficient heavyweight thunk that calls
4325 @var{function} instead of jumping to it. The generic approach does
4326 not support varargs.
4329 @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})
4330 A function like @code{TARGET_ASM_OUTPUT_MI_THUNK}, except that if
4331 @var{vcall_offset} is nonzero, an additional adjustment should be made
4332 after adding @code{delta}. In particular, if @var{p} is the
4333 adjusted pointer, the following adjustment should be made:
4336 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4340 If this function is defined, it will always be used in place of
4341 @code{TARGET_ASM_OUTPUT_MI_THUNK}.
4345 @subsection Generating Code for Profiling
4346 @cindex profiling, code generation
4348 These macros will help you generate code for profiling.
4350 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4351 A C statement or compound statement to output to @var{file} some
4352 assembler code to call the profiling subroutine @code{mcount}.
4355 The details of how @code{mcount} expects to be called are determined by
4356 your operating system environment, not by GCC@. To figure them out,
4357 compile a small program for profiling using the system's installed C
4358 compiler and look at the assembler code that results.
4360 Older implementations of @code{mcount} expect the address of a counter
4361 variable to be loaded into some register. The name of this variable is
4362 @samp{LP} followed by the number @var{labelno}, so you would generate
4363 the name using @samp{LP%d} in a @code{fprintf}.
4366 @defmac PROFILE_HOOK
4367 A C statement or compound statement to output to @var{file} some assembly
4368 code to call the profiling subroutine @code{mcount} even the target does
4369 not support profiling.
4372 @defmac NO_PROFILE_COUNTERS
4373 Define this macro if the @code{mcount} subroutine on your system does
4374 not need a counter variable allocated for each function. This is true
4375 for almost all modern implementations. If you define this macro, you
4376 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4379 @defmac PROFILE_BEFORE_PROLOGUE
4380 Define this macro if the code for function profiling should come before
4381 the function prologue. Normally, the profiling code comes after.
4385 @subsection Permitting tail calls
4388 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4389 True if it is ok to do sibling call optimization for the specified
4390 call expression @var{exp}. @var{decl} will be the called function,
4391 or @code{NULL} if this is an indirect call.
4393 It is not uncommon for limitations of calling conventions to prevent
4394 tail calls to functions outside the current unit of translation, or
4395 during PIC compilation. The hook is used to enforce these restrictions,
4396 as the @code{sibcall} md pattern can not fail, or fall over to a
4397 ``normal'' call. The criteria for successful sibling call optimization
4398 may vary greatly between different architectures.
4402 @section Implementing the Varargs Macros
4403 @cindex varargs implementation
4405 GCC comes with an implementation of @code{<varargs.h>} and
4406 @code{<stdarg.h>} that work without change on machines that pass arguments
4407 on the stack. Other machines require their own implementations of
4408 varargs, and the two machine independent header files must have
4409 conditionals to include it.
4411 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4412 the calling convention for @code{va_start}. The traditional
4413 implementation takes just one argument, which is the variable in which
4414 to store the argument pointer. The ISO implementation of
4415 @code{va_start} takes an additional second argument. The user is
4416 supposed to write the last named argument of the function here.
4418 However, @code{va_start} should not use this argument. The way to find
4419 the end of the named arguments is with the built-in functions described
4422 @defmac __builtin_saveregs ()
4423 Use this built-in function to save the argument registers in memory so
4424 that the varargs mechanism can access them. Both ISO and traditional
4425 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4426 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4428 On some machines, @code{__builtin_saveregs} is open-coded under the
4429 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4430 other machines, it calls a routine written in assembler language,
4431 found in @file{libgcc2.c}.
4433 Code generated for the call to @code{__builtin_saveregs} appears at the
4434 beginning of the function, as opposed to where the call to
4435 @code{__builtin_saveregs} is written, regardless of what the code is.
4436 This is because the registers must be saved before the function starts
4437 to use them for its own purposes.
4438 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4442 @defmac __builtin_args_info (@var{category})
4443 Use this built-in function to find the first anonymous arguments in
4446 In general, a machine may have several categories of registers used for
4447 arguments, each for a particular category of data types. (For example,
4448 on some machines, floating-point registers are used for floating-point
4449 arguments while other arguments are passed in the general registers.)
4450 To make non-varargs functions use the proper calling convention, you
4451 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4452 registers in each category have been used so far
4454 @code{__builtin_args_info} accesses the same data structure of type
4455 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4456 with it, with @var{category} specifying which word to access. Thus, the
4457 value indicates the first unused register in a given category.
4459 Normally, you would use @code{__builtin_args_info} in the implementation
4460 of @code{va_start}, accessing each category just once and storing the
4461 value in the @code{va_list} object. This is because @code{va_list} will
4462 have to update the values, and there is no way to alter the
4463 values accessed by @code{__builtin_args_info}.
4466 @defmac __builtin_next_arg (@var{lastarg})
4467 This is the equivalent of @code{__builtin_args_info}, for stack
4468 arguments. It returns the address of the first anonymous stack
4469 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4470 returns the address of the location above the first anonymous stack
4471 argument. Use it in @code{va_start} to initialize the pointer for
4472 fetching arguments from the stack. Also use it in @code{va_start} to
4473 verify that the second parameter @var{lastarg} is the last named argument
4474 of the current function.
4477 @defmac __builtin_classify_type (@var{object})
4478 Since each machine has its own conventions for which data types are
4479 passed in which kind of register, your implementation of @code{va_arg}
4480 has to embody these conventions. The easiest way to categorize the
4481 specified data type is to use @code{__builtin_classify_type} together
4482 with @code{sizeof} and @code{__alignof__}.
4484 @code{__builtin_classify_type} ignores the value of @var{object},
4485 considering only its data type. It returns an integer describing what
4486 kind of type that is---integer, floating, pointer, structure, and so on.
4488 The file @file{typeclass.h} defines an enumeration that you can use to
4489 interpret the values of @code{__builtin_classify_type}.
4492 These machine description macros help implement varargs:
4494 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4495 If defined, this hook produces the machine-specific code for a call to
4496 @code{__builtin_saveregs}. This code will be moved to the very
4497 beginning of the function, before any parameter access are made. The
4498 return value of this function should be an RTX that contains the value
4499 to use as the return of @code{__builtin_saveregs}.
4502 @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})
4503 This target hook offers an alternative to using
4504 @code{__builtin_saveregs} and defining the hook
4505 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4506 register arguments into the stack so that all the arguments appear to
4507 have been passed consecutively on the stack. Once this is done, you can
4508 use the standard implementation of varargs that works for machines that
4509 pass all their arguments on the stack.
4511 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4512 structure, containing the values that are obtained after processing the
4513 named arguments. The arguments @var{mode} and @var{type} describe the
4514 last named argument---its machine mode and its data type as a tree node.
4516 The target hook should do two things: first, push onto the stack all the
4517 argument registers @emph{not} used for the named arguments, and second,
4518 store the size of the data thus pushed into the @code{int}-valued
4519 variable pointed to by @var{pretend_args_size}. The value that you
4520 store here will serve as additional offset for setting up the stack
4523 Because you must generate code to push the anonymous arguments at
4524 compile time without knowing their data types,
4525 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4526 have just a single category of argument register and use it uniformly
4529 If the argument @var{second_time} is nonzero, it means that the
4530 arguments of the function are being analyzed for the second time. This
4531 happens for an inline function, which is not actually compiled until the
4532 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4533 not generate any instructions in this case.
4536 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4537 Define this hook to return @code{true} if the location where a function
4538 argument is passed depends on whether or not it is a named argument.
4540 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4541 is set for varargs and stdarg functions. If this hook returns
4542 @code{true}, the @var{named} argument is always true for named
4543 arguments, and false for unnamed arguments. If it returns @code{false},
4544 but @code{TARGET_PRETEND_OUTOGOING_VARARGS_NAMED} returns @code{true},
4545 then all arguments are treated as named. Otherwise, all named arguments
4546 except the last are treated as named.
4548 You need not define this hook if it always returns zero.
4551 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4552 If you need to conditionally change ABIs so that one works with
4553 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4554 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4555 defined, then define this hook to return @code{true} if
4556 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4557 Otherwise, you should not define this hook.
4561 @section Trampolines for Nested Functions
4562 @cindex trampolines for nested functions
4563 @cindex nested functions, trampolines for
4565 A @dfn{trampoline} is a small piece of code that is created at run time
4566 when the address of a nested function is taken. It normally resides on
4567 the stack, in the stack frame of the containing function. These macros
4568 tell GCC how to generate code to allocate and initialize a
4571 The instructions in the trampoline must do two things: load a constant
4572 address into the static chain register, and jump to the real address of
4573 the nested function. On CISC machines such as the m68k, this requires
4574 two instructions, a move immediate and a jump. Then the two addresses
4575 exist in the trampoline as word-long immediate operands. On RISC
4576 machines, it is often necessary to load each address into a register in
4577 two parts. Then pieces of each address form separate immediate
4580 The code generated to initialize the trampoline must store the variable
4581 parts---the static chain value and the function address---into the
4582 immediate operands of the instructions. On a CISC machine, this is
4583 simply a matter of copying each address to a memory reference at the
4584 proper offset from the start of the trampoline. On a RISC machine, it
4585 may be necessary to take out pieces of the address and store them
4588 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4589 A C statement to output, on the stream @var{file}, assembler code for a
4590 block of data that contains the constant parts of a trampoline. This
4591 code should not include a label---the label is taken care of
4594 If you do not define this macro, it means no template is needed
4595 for the target. Do not define this macro on systems where the block move
4596 code to copy the trampoline into place would be larger than the code
4597 to generate it on the spot.
4600 @defmac TRAMPOLINE_SECTION
4601 The name of a subroutine to switch to the section in which the
4602 trampoline template is to be placed (@pxref{Sections}). The default is
4603 a value of @samp{readonly_data_section}, which places the trampoline in
4604 the section containing read-only data.
4607 @defmac TRAMPOLINE_SIZE
4608 A C expression for the size in bytes of the trampoline, as an integer.
4611 @defmac TRAMPOLINE_ALIGNMENT
4612 Alignment required for trampolines, in bits.
4614 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4615 is used for aligning trampolines.
4618 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4619 A C statement to initialize the variable parts of a trampoline.
4620 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4621 an RTX for the address of the nested function; @var{static_chain} is an
4622 RTX for the static chain value that should be passed to the function
4626 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4627 A C statement that should perform any machine-specific adjustment in
4628 the address of the trampoline. Its argument contains the address that
4629 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4630 used for a function call should be different from the address in which
4631 the template was stored, the different address should be assigned to
4632 @var{addr}. If this macro is not defined, @var{addr} will be used for
4635 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4636 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4637 If this macro is not defined, by default the trampoline is allocated as
4638 a stack slot. This default is right for most machines. The exceptions
4639 are machines where it is impossible to execute instructions in the stack
4640 area. On such machines, you may have to implement a separate stack,
4641 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4642 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4644 @var{fp} points to a data structure, a @code{struct function}, which
4645 describes the compilation status of the immediate containing function of
4646 the function which the trampoline is for. The stack slot for the
4647 trampoline is in the stack frame of this containing function. Other
4648 allocation strategies probably must do something analogous with this
4652 Implementing trampolines is difficult on many machines because they have
4653 separate instruction and data caches. Writing into a stack location
4654 fails to clear the memory in the instruction cache, so when the program
4655 jumps to that location, it executes the old contents.
4657 Here are two possible solutions. One is to clear the relevant parts of
4658 the instruction cache whenever a trampoline is set up. The other is to
4659 make all trampolines identical, by having them jump to a standard
4660 subroutine. The former technique makes trampoline execution faster; the
4661 latter makes initialization faster.
4663 To clear the instruction cache when a trampoline is initialized, define
4664 the following macro.
4666 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4667 If defined, expands to a C expression clearing the @emph{instruction
4668 cache} in the specified interval. The definition of this macro would
4669 typically be a series of @code{asm} statements. Both @var{beg} and
4670 @var{end} are both pointer expressions.
4673 To use a standard subroutine, define the following macro. In addition,
4674 you must make sure that the instructions in a trampoline fill an entire
4675 cache line with identical instructions, or else ensure that the
4676 beginning of the trampoline code is always aligned at the same point in
4677 its cache line. Look in @file{m68k.h} as a guide.
4679 @defmac TRANSFER_FROM_TRAMPOLINE
4680 Define this macro if trampolines need a special subroutine to do their
4681 work. The macro should expand to a series of @code{asm} statements
4682 which will be compiled with GCC@. They go in a library function named
4683 @code{__transfer_from_trampoline}.
4685 If you need to avoid executing the ordinary prologue code of a compiled
4686 C function when you jump to the subroutine, you can do so by placing a
4687 special label of your own in the assembler code. Use one @code{asm}
4688 statement to generate an assembler label, and another to make the label
4689 global. Then trampolines can use that label to jump directly to your
4690 special assembler code.
4694 @section Implicit Calls to Library Routines
4695 @cindex library subroutine names
4696 @cindex @file{libgcc.a}
4698 @c prevent bad page break with this line
4699 Here is an explanation of implicit calls to library routines.
4701 @defmac DECLARE_LIBRARY_RENAMES
4702 This macro, if defined, should expand to a piece of C code that will get
4703 expanded when compiling functions for libgcc.a. It can be used to
4704 provide alternate names for GCC's internal library functions if there
4705 are ABI-mandated names that the compiler should provide.
4708 @findex init_one_libfunc
4709 @findex set_optab_libfunc
4710 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
4711 This hook should declare additional library routines or rename
4712 existing ones, using the functions @code{set_optab_libfunc} and
4713 @code{init_one_libfunc} defined in @file{optabs.c}.
4714 @code{init_optabs} calls this macro after initializing all the normal
4717 The default is to do nothing. Most ports don't need to define this hook.
4720 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4721 This macro should return @code{true} if the library routine that
4722 implements the floating point comparison operator @var{comparison} in
4723 mode @var{mode} will return a boolean, and @var{false} if it will
4726 GCC's own floating point libraries return tristates from the
4727 comparison operators, so the default returns false always. Most ports
4728 don't need to define this macro.
4731 @cindex US Software GOFAST, floating point emulation library
4732 @cindex floating point emulation library, US Software GOFAST
4733 @cindex GOFAST, floating point emulation library
4734 @findex gofast_maybe_init_libfuncs
4735 @defmac US_SOFTWARE_GOFAST
4736 Define this macro if your system C library uses the US Software GOFAST
4737 library to provide floating point emulation.
4739 In addition to defining this macro, your architecture must set
4740 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
4741 else call that function from its version of that hook. It is defined
4742 in @file{config/gofast.h}, which must be included by your
4743 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
4746 If this macro is defined, the
4747 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
4748 false for @code{SFmode} and @code{DFmode} comparisons.
4751 @cindex @code{EDOM}, implicit usage
4754 The value of @code{EDOM} on the target machine, as a C integer constant
4755 expression. If you don't define this macro, GCC does not attempt to
4756 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4757 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4760 If you do not define @code{TARGET_EDOM}, then compiled code reports
4761 domain errors by calling the library function and letting it report the
4762 error. If mathematical functions on your system use @code{matherr} when
4763 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4764 that @code{matherr} is used normally.
4767 @cindex @code{errno}, implicit usage
4768 @defmac GEN_ERRNO_RTX
4769 Define this macro as a C expression to create an rtl expression that
4770 refers to the global ``variable'' @code{errno}. (On certain systems,
4771 @code{errno} may not actually be a variable.) If you don't define this
4772 macro, a reasonable default is used.
4775 @cindex @code{bcopy}, implicit usage
4776 @cindex @code{memcpy}, implicit usage
4777 @cindex @code{memmove}, implicit usage
4778 @cindex @code{bzero}, implicit usage
4779 @cindex @code{memset}, implicit usage
4780 @defmac TARGET_MEM_FUNCTIONS
4781 Define this macro if GCC should generate calls to the ISO C
4782 (and System V) library functions @code{memcpy}, @code{memmove} and
4783 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4786 @cindex C99 math functions, implicit usage
4787 @defmac TARGET_C99_FUNCTIONS
4788 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4789 @code{sinf} and similarly for other functions defined by C99 standard. The
4790 default is nonzero that should be proper value for most modern systems, however
4791 number of existing systems lacks support for these functions in the runtime so
4792 they needs this macro to be redefined to 0.
4795 @defmac NEXT_OBJC_RUNTIME
4796 Define this macro to generate code for Objective-C message sending using
4797 the calling convention of the NeXT system. This calling convention
4798 involves passing the object, the selector and the method arguments all
4799 at once to the method-lookup library function.
4801 The default calling convention passes just the object and the selector
4802 to the lookup function, which returns a pointer to the method.
4805 @node Addressing Modes
4806 @section Addressing Modes
4807 @cindex addressing modes
4809 @c prevent bad page break with this line
4810 This is about addressing modes.
4812 @defmac HAVE_PRE_INCREMENT
4813 @defmacx HAVE_PRE_DECREMENT
4814 @defmacx HAVE_POST_INCREMENT
4815 @defmacx HAVE_POST_DECREMENT
4816 A C expression that is nonzero if the machine supports pre-increment,
4817 pre-decrement, post-increment, or post-decrement addressing respectively.
4820 @defmac HAVE_PRE_MODIFY_DISP
4821 @defmacx HAVE_POST_MODIFY_DISP
4822 A C expression that is nonzero if the machine supports pre- or
4823 post-address side-effect generation involving constants other than
4824 the size of the memory operand.
4827 @defmac HAVE_PRE_MODIFY_REG
4828 @defmacx HAVE_POST_MODIFY_REG
4829 A C expression that is nonzero if the machine supports pre- or
4830 post-address side-effect generation involving a register displacement.
4833 @defmac CONSTANT_ADDRESS_P (@var{x})
4834 A C expression that is 1 if the RTX @var{x} is a constant which
4835 is a valid address. On most machines, this can be defined as
4836 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4837 in which constant addresses are supported.
4840 @defmac CONSTANT_P (@var{x})
4841 @code{CONSTANT_P}, which is defined by target-independent code,
4842 accepts integer-values expressions whose values are not explicitly
4843 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4844 expressions and @code{const} arithmetic expressions, in addition to
4845 @code{const_int} and @code{const_double} expressions.
4848 @defmac MAX_REGS_PER_ADDRESS
4849 A number, the maximum number of registers that can appear in a valid
4850 memory address. Note that it is up to you to specify a value equal to
4851 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4855 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4856 A C compound statement with a conditional @code{goto @var{label};}
4857 executed if @var{x} (an RTX) is a legitimate memory address on the
4858 target machine for a memory operand of mode @var{mode}.
4860 It usually pays to define several simpler macros to serve as
4861 subroutines for this one. Otherwise it may be too complicated to
4864 This macro must exist in two variants: a strict variant and a
4865 non-strict one. The strict variant is used in the reload pass. It
4866 must be defined so that any pseudo-register that has not been
4867 allocated a hard register is considered a memory reference. In
4868 contexts where some kind of register is required, a pseudo-register
4869 with no hard register must be rejected.
4871 The non-strict variant is used in other passes. It must be defined to
4872 accept all pseudo-registers in every context where some kind of
4873 register is required.
4875 @findex REG_OK_STRICT
4876 Compiler source files that want to use the strict variant of this
4877 macro define the macro @code{REG_OK_STRICT}. You should use an
4878 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4879 in that case and the non-strict variant otherwise.
4881 Subroutines to check for acceptable registers for various purposes (one
4882 for base registers, one for index registers, and so on) are typically
4883 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4884 Then only these subroutine macros need have two variants; the higher
4885 levels of macros may be the same whether strict or not.
4887 Normally, constant addresses which are the sum of a @code{symbol_ref}
4888 and an integer are stored inside a @code{const} RTX to mark them as
4889 constant. Therefore, there is no need to recognize such sums
4890 specifically as legitimate addresses. Normally you would simply
4891 recognize any @code{const} as legitimate.
4893 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4894 sums that are not marked with @code{const}. It assumes that a naked
4895 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4896 naked constant sums as illegitimate addresses, so that none of them will
4897 be given to @code{PRINT_OPERAND_ADDRESS}.
4899 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4900 On some machines, whether a symbolic address is legitimate depends on
4901 the section that the address refers to. On these machines, define the
4902 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4903 into the @code{symbol_ref}, and then check for it here. When you see a
4904 @code{const}, you will have to look inside it to find the
4905 @code{symbol_ref} in order to determine the section. @xref{Assembler
4909 @defmac REG_OK_FOR_BASE_P (@var{x})
4910 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4911 RTX) is valid for use as a base register. For hard registers, it
4912 should always accept those which the hardware permits and reject the
4913 others. Whether the macro accepts or rejects pseudo registers must be
4914 controlled by @code{REG_OK_STRICT} as described above. This usually
4915 requires two variant definitions, of which @code{REG_OK_STRICT}
4916 controls the one actually used.
4919 @defmac REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4920 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4921 that expression may examine the mode of the memory reference in
4922 @var{mode}. You should define this macro if the mode of the memory
4923 reference affects whether a register may be used as a base register. If
4924 you define this macro, the compiler will use it instead of
4925 @code{REG_OK_FOR_BASE_P}.
4928 @defmac REG_OK_FOR_INDEX_P (@var{x})
4929 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4930 RTX) is valid for use as an index register.
4932 The difference between an index register and a base register is that
4933 the index register may be scaled. If an address involves the sum of
4934 two registers, neither one of them scaled, then either one may be
4935 labeled the ``base'' and the other the ``index''; but whichever
4936 labeling is used must fit the machine's constraints of which registers
4937 may serve in each capacity. The compiler will try both labelings,
4938 looking for one that is valid, and will reload one or both registers
4939 only if neither labeling works.
4942 @defmac FIND_BASE_TERM (@var{x})
4943 A C expression to determine the base term of address @var{x}.
4944 This macro is used in only one place: `find_base_term' in alias.c.
4946 It is always safe for this macro to not be defined. It exists so
4947 that alias analysis can understand machine-dependent addresses.
4949 The typical use of this macro is to handle addresses containing
4950 a label_ref or symbol_ref within an UNSPEC@.
4953 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4954 A C compound statement that attempts to replace @var{x} with a valid
4955 memory address for an operand of mode @var{mode}. @var{win} will be a
4956 C statement label elsewhere in the code; the macro definition may use
4959 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4963 to avoid further processing if the address has become legitimate.
4965 @findex break_out_memory_refs
4966 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4967 and @var{oldx} will be the operand that was given to that function to produce
4970 The code generated by this macro should not alter the substructure of
4971 @var{x}. If it transforms @var{x} into a more legitimate form, it
4972 should assign @var{x} (which will always be a C variable) a new value.
4974 It is not necessary for this macro to come up with a legitimate
4975 address. The compiler has standard ways of doing so in all cases. In
4976 fact, it is safe to omit this macro. But often a
4977 machine-dependent strategy can generate better code.
4980 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4981 A C compound statement that attempts to replace @var{x}, which is an address
4982 that needs reloading, with a valid memory address for an operand of mode
4983 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4984 It is not necessary to define this macro, but it might be useful for
4985 performance reasons.
4987 For example, on the i386, it is sometimes possible to use a single
4988 reload register instead of two by reloading a sum of two pseudo
4989 registers into a register. On the other hand, for number of RISC
4990 processors offsets are limited so that often an intermediate address
4991 needs to be generated in order to address a stack slot. By defining
4992 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4993 generated for adjacent some stack slots can be made identical, and thus
4996 @emph{Note}: This macro should be used with caution. It is necessary
4997 to know something of how reload works in order to effectively use this,
4998 and it is quite easy to produce macros that build in too much knowledge
4999 of reload internals.
5001 @emph{Note}: This macro must be able to reload an address created by a
5002 previous invocation of this macro. If it fails to handle such addresses
5003 then the compiler may generate incorrect code or abort.
5006 The macro definition should use @code{push_reload} to indicate parts that
5007 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5008 suitable to be passed unaltered to @code{push_reload}.
5010 The code generated by this macro must not alter the substructure of
5011 @var{x}. If it transforms @var{x} into a more legitimate form, it
5012 should assign @var{x} (which will always be a C variable) a new value.
5013 This also applies to parts that you change indirectly by calling
5016 @findex strict_memory_address_p
5017 The macro definition may use @code{strict_memory_address_p} to test if
5018 the address has become legitimate.
5021 If you want to change only a part of @var{x}, one standard way of doing
5022 this is to use @code{copy_rtx}. Note, however, that is unshares only a
5023 single level of rtl. Thus, if the part to be changed is not at the
5024 top level, you'll need to replace first the top level.
5025 It is not necessary for this macro to come up with a legitimate
5026 address; but often a machine-dependent strategy can generate better code.
5029 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5030 A C statement or compound statement with a conditional @code{goto
5031 @var{label};} executed if memory address @var{x} (an RTX) can have
5032 different meanings depending on the machine mode of the memory
5033 reference it is used for or if the address is valid for some modes
5036 Autoincrement and autodecrement addresses typically have mode-dependent
5037 effects because the amount of the increment or decrement is the size
5038 of the operand being addressed. Some machines have other mode-dependent
5039 addresses. Many RISC machines have no mode-dependent addresses.
5041 You may assume that @var{addr} is a valid address for the machine.
5044 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5045 A C expression that is nonzero if @var{x} is a legitimate constant for
5046 an immediate operand on the target machine. You can assume that
5047 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5048 @samp{1} is a suitable definition for this macro on machines where
5049 anything @code{CONSTANT_P} is valid.
5052 @node Condition Code
5053 @section Condition Code Status
5054 @cindex condition code status
5056 @c prevent bad page break with this line
5057 This describes the condition code status.
5060 The file @file{conditions.h} defines a variable @code{cc_status} to
5061 describe how the condition code was computed (in case the interpretation of
5062 the condition code depends on the instruction that it was set by). This
5063 variable contains the RTL expressions on which the condition code is
5064 currently based, and several standard flags.
5066 Sometimes additional machine-specific flags must be defined in the machine
5067 description header file. It can also add additional machine-specific
5068 information by defining @code{CC_STATUS_MDEP}.
5070 @defmac CC_STATUS_MDEP
5071 C code for a data type which is used for declaring the @code{mdep}
5072 component of @code{cc_status}. It defaults to @code{int}.
5074 This macro is not used on machines that do not use @code{cc0}.
5077 @defmac CC_STATUS_MDEP_INIT
5078 A C expression to initialize the @code{mdep} field to ``empty''.
5079 The default definition does nothing, since most machines don't use
5080 the field anyway. If you want to use the field, you should probably
5081 define this macro to initialize it.
5083 This macro is not used on machines that do not use @code{cc0}.
5086 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5087 A C compound statement to set the components of @code{cc_status}
5088 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5089 this macro's responsibility to recognize insns that set the condition
5090 code as a byproduct of other activity as well as those that explicitly
5093 This macro is not used on machines that do not use @code{cc0}.
5095 If there are insns that do not set the condition code but do alter
5096 other machine registers, this macro must check to see whether they
5097 invalidate the expressions that the condition code is recorded as
5098 reflecting. For example, on the 68000, insns that store in address
5099 registers do not set the condition code, which means that usually
5100 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5101 insns. But suppose that the previous insn set the condition code
5102 based on location @samp{a4@@(102)} and the current insn stores a new
5103 value in @samp{a4}. Although the condition code is not changed by
5104 this, it will no longer be true that it reflects the contents of
5105 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5106 @code{cc_status} in this case to say that nothing is known about the
5107 condition code value.
5109 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5110 with the results of peephole optimization: insns whose patterns are
5111 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5112 constants which are just the operands. The RTL structure of these
5113 insns is not sufficient to indicate what the insns actually do. What
5114 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5115 @code{CC_STATUS_INIT}.
5117 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5118 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5119 @samp{cc}. This avoids having detailed information about patterns in
5120 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5123 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5124 Returns a mode from class @code{MODE_CC} to be used when comparison
5125 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5126 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5127 @pxref{Jump Patterns} for a description of the reason for this
5131 #define SELECT_CC_MODE(OP,X,Y) \
5132 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5133 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5134 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5135 || GET_CODE (X) == NEG) \
5136 ? CC_NOOVmode : CCmode))
5139 You should define this macro if and only if you define extra CC modes
5140 in @file{@var{machine}-modes.def}.
5143 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5144 On some machines not all possible comparisons are defined, but you can
5145 convert an invalid comparison into a valid one. For example, the Alpha
5146 does not have a @code{GT} comparison, but you can use an @code{LT}
5147 comparison instead and swap the order of the operands.
5149 On such machines, define this macro to be a C statement to do any
5150 required conversions. @var{code} is the initial comparison code
5151 and @var{op0} and @var{op1} are the left and right operands of the
5152 comparison, respectively. You should modify @var{code}, @var{op0}, and
5153 @var{op1} as required.
5155 GCC will not assume that the comparison resulting from this macro is
5156 valid but will see if the resulting insn matches a pattern in the
5159 You need not define this macro if it would never change the comparison
5163 @defmac REVERSIBLE_CC_MODE (@var{mode})
5164 A C expression whose value is one if it is always safe to reverse a
5165 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5166 can ever return @var{mode} for a floating-point inequality comparison,
5167 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5169 You need not define this macro if it would always returns zero or if the
5170 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5171 For example, here is the definition used on the SPARC, where floating-point
5172 inequality comparisons are always given @code{CCFPEmode}:
5175 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5179 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5180 A C expression whose value is reversed condition code of the @var{code} for
5181 comparison done in CC_MODE @var{mode}. The macro is used only in case
5182 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5183 machine has some non-standard way how to reverse certain conditionals. For
5184 instance in case all floating point conditions are non-trapping, compiler may
5185 freely convert unordered compares to ordered one. Then definition may look
5189 #define REVERSE_CONDITION(CODE, MODE) \
5190 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5191 : reverse_condition_maybe_unordered (CODE))
5195 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
5196 A C expression that returns true if the conditional execution predicate
5197 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
5198 return 0 if the target has conditional execution predicates that cannot be
5199 reversed safely. If no expansion is specified, this macro is defined as
5203 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5204 ((x) == reverse_condition (y))
5208 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5209 On targets which do not use @code{(cc0)}, and which use a hard
5210 register rather than a pseudo-register to hold condition codes, the
5211 regular CSE passes are often not able to identify cases in which the
5212 hard register is set to a common value. Use this hook to enable a
5213 small pass which optimizes such cases. This hook should return true
5214 to enable this pass, and it should set the integers to which its
5215 arguments point to the hard register numbers used for condition codes.
5216 When there is only one such register, as is true on most systems, the
5217 integer pointed to by the second argument should be set to
5218 @code{INVALID_REGNUM}.
5220 The default version of this hook returns false.
5223 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5224 On targets which use multiple condition code modes in class
5225 @code{MODE_CC}, it is sometimes the case that a comparison can be
5226 validly done in more than one mode. On such a system, define this
5227 target hook to take two mode arguments and to return a mode in which
5228 both comparisons may be validly done. If there is no such mode,
5229 return @code{VOIDmode}.
5231 The default version of this hook checks whether the modes are the
5232 same. If they are, it returns that mode. If they are different, it
5233 returns @code{VOIDmode}.
5237 @section Describing Relative Costs of Operations
5238 @cindex costs of instructions
5239 @cindex relative costs
5240 @cindex speed of instructions
5242 These macros let you describe the relative speed of various operations
5243 on the target machine.
5245 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5246 A C expression for the cost of moving data of mode @var{mode} from a
5247 register in class @var{from} to one in class @var{to}. The classes are
5248 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5249 value of 2 is the default; other values are interpreted relative to
5252 It is not required that the cost always equal 2 when @var{from} is the
5253 same as @var{to}; on some machines it is expensive to move between
5254 registers if they are not general registers.
5256 If reload sees an insn consisting of a single @code{set} between two
5257 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5258 classes returns a value of 2, reload does not check to ensure that the
5259 constraints of the insn are met. Setting a cost of other than 2 will
5260 allow reload to verify that the constraints are met. You should do this
5261 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5264 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5265 A C expression for the cost of moving data of mode @var{mode} between a
5266 register of class @var{class} and memory; @var{in} is zero if the value
5267 is to be written to memory, nonzero if it is to be read in. This cost
5268 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5269 registers and memory is more expensive than between two registers, you
5270 should define this macro to express the relative cost.
5272 If you do not define this macro, GCC uses a default cost of 4 plus
5273 the cost of copying via a secondary reload register, if one is
5274 needed. If your machine requires a secondary reload register to copy
5275 between memory and a register of @var{class} but the reload mechanism is
5276 more complex than copying via an intermediate, define this macro to
5277 reflect the actual cost of the move.
5279 GCC defines the function @code{memory_move_secondary_cost} if
5280 secondary reloads are needed. It computes the costs due to copying via
5281 a secondary register. If your machine copies from memory using a
5282 secondary register in the conventional way but the default base value of
5283 4 is not correct for your machine, define this macro to add some other
5284 value to the result of that function. The arguments to that function
5285 are the same as to this macro.
5289 A C expression for the cost of a branch instruction. A value of 1 is
5290 the default; other values are interpreted relative to that.
5293 Here are additional macros which do not specify precise relative costs,
5294 but only that certain actions are more expensive than GCC would
5297 @defmac SLOW_BYTE_ACCESS
5298 Define this macro as a C expression which is nonzero if accessing less
5299 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5300 faster than accessing a word of memory, i.e., if such access
5301 require more than one instruction or if there is no difference in cost
5302 between byte and (aligned) word loads.
5304 When this macro is not defined, the compiler will access a field by
5305 finding the smallest containing object; when it is defined, a fullword
5306 load will be used if alignment permits. Unless bytes accesses are
5307 faster than word accesses, using word accesses is preferable since it
5308 may eliminate subsequent memory access if subsequent accesses occur to
5309 other fields in the same word of the structure, but to different bytes.
5312 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5313 Define this macro to be the value 1 if memory accesses described by the
5314 @var{mode} and @var{alignment} parameters have a cost many times greater
5315 than aligned accesses, for example if they are emulated in a trap
5318 When this macro is nonzero, the compiler will act as if
5319 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5320 moves. This can cause significantly more instructions to be produced.
5321 Therefore, do not set this macro nonzero if unaligned accesses only add a
5322 cycle or two to the time for a memory access.
5324 If the value of this macro is always zero, it need not be defined. If
5325 this macro is defined, it should produce a nonzero value when
5326 @code{STRICT_ALIGNMENT} is nonzero.
5330 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5331 which a sequence of insns should be generated instead of a
5332 string move insn or a library call. Increasing the value will always
5333 make code faster, but eventually incurs high cost in increased code size.
5335 Note that on machines where the corresponding move insn is a
5336 @code{define_expand} that emits a sequence of insns, this macro counts
5337 the number of such sequences.
5339 If you don't define this, a reasonable default is used.
5342 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5343 A C expression used to determine whether @code{move_by_pieces} will be used to
5344 copy a chunk of memory, or whether some other block move mechanism
5345 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5346 than @code{MOVE_RATIO}.
5349 @defmac MOVE_MAX_PIECES
5350 A C expression used by @code{move_by_pieces} to determine the largest unit
5351 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5355 The threshold of number of scalar move insns, @emph{below} which a sequence
5356 of insns should be generated to clear memory instead of a string clear insn
5357 or a library call. Increasing the value will always make code faster, but
5358 eventually incurs high cost in increased code size.
5360 If you don't define this, a reasonable default is used.
5363 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5364 A C expression used to determine whether @code{clear_by_pieces} will be used
5365 to clear a chunk of memory, or whether some other block clear mechanism
5366 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5367 than @code{CLEAR_RATIO}.
5370 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5371 A C expression used to determine whether @code{store_by_pieces} will be
5372 used to set a chunk of memory to a constant value, or whether some other
5373 mechanism will be used. Used by @code{__builtin_memset} when storing
5374 values other than constant zero and by @code{__builtin_strcpy} when
5375 when called with a constant source string.
5376 Defaults to @code{MOVE_BY_PIECES_P}.
5379 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5380 A C expression used to determine whether a load postincrement is a good
5381 thing to use for a given mode. Defaults to the value of
5382 @code{HAVE_POST_INCREMENT}.
5385 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5386 A C expression used to determine whether a load postdecrement is a good
5387 thing to use for a given mode. Defaults to the value of
5388 @code{HAVE_POST_DECREMENT}.
5391 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5392 A C expression used to determine whether a load preincrement is a good
5393 thing to use for a given mode. Defaults to the value of
5394 @code{HAVE_PRE_INCREMENT}.
5397 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5398 A C expression used to determine whether a load predecrement is a good
5399 thing to use for a given mode. Defaults to the value of
5400 @code{HAVE_PRE_DECREMENT}.
5403 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5404 A C expression used to determine whether a store postincrement is a good
5405 thing to use for a given mode. Defaults to the value of
5406 @code{HAVE_POST_INCREMENT}.
5409 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5410 A C expression used to determine whether a store postdecrement is a good
5411 thing to use for a given mode. Defaults to the value of
5412 @code{HAVE_POST_DECREMENT}.
5415 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5416 This macro is used to determine whether a store preincrement is a good
5417 thing to use for a given mode. Defaults to the value of
5418 @code{HAVE_PRE_INCREMENT}.
5421 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5422 This macro is used to determine whether a store predecrement is a good
5423 thing to use for a given mode. Defaults to the value of
5424 @code{HAVE_PRE_DECREMENT}.
5427 @defmac NO_FUNCTION_CSE
5428 Define this macro if it is as good or better to call a constant
5429 function address than to call an address kept in a register.
5432 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5433 Define this macro if a non-short-circuit operation produced by
5434 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5435 @code{BRANCH_COST} is greater than or equal to the value 2.
5438 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5439 This target hook describes the relative costs of RTL expressions.
5441 The cost may depend on the precise form of the expression, which is
5442 available for examination in @var{x}, and the rtx code of the expression
5443 in which it is contained, found in @var{outer_code}. @var{code} is the
5444 expression code---redundant, since it can be obtained with
5445 @code{GET_CODE (@var{x})}.
5447 In implementing this hook, you can use the construct
5448 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5451 On entry to the hook, @code{*@var{total}} contains a default estimate
5452 for the cost of the expression. The hook should modify this value as
5453 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
5454 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
5455 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
5457 When optimizing for code size, i.e@. when @code{optimize_size} is
5458 non-zero, this target hook should be used to estimate the relative
5459 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
5461 The hook returns true when all subexpressions of @var{x} have been
5462 processed, and false when @code{rtx_cost} should recurse.
5465 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5466 This hook computes the cost of an addressing mode that contains
5467 @var{address}. If not defined, the cost is computed from
5468 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5470 For most CISC machines, the default cost is a good approximation of the
5471 true cost of the addressing mode. However, on RISC machines, all
5472 instructions normally have the same length and execution time. Hence
5473 all addresses will have equal costs.
5475 In cases where more than one form of an address is known, the form with
5476 the lowest cost will be used. If multiple forms have the same, lowest,
5477 cost, the one that is the most complex will be used.
5479 For example, suppose an address that is equal to the sum of a register
5480 and a constant is used twice in the same basic block. When this macro
5481 is not defined, the address will be computed in a register and memory
5482 references will be indirect through that register. On machines where
5483 the cost of the addressing mode containing the sum is no higher than
5484 that of a simple indirect reference, this will produce an additional
5485 instruction and possibly require an additional register. Proper
5486 specification of this macro eliminates this overhead for such machines.
5488 This hook is never called with an invalid address.
5490 On machines where an address involving more than one register is as
5491 cheap as an address computation involving only one register, defining
5492 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5493 be live over a region of code where only one would have been if
5494 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5495 should be considered in the definition of this macro. Equivalent costs
5496 should probably only be given to addresses with different numbers of
5497 registers on machines with lots of registers.
5501 @section Adjusting the Instruction Scheduler
5503 The instruction scheduler may need a fair amount of machine-specific
5504 adjustment in order to produce good code. GCC provides several target
5505 hooks for this purpose. It is usually enough to define just a few of
5506 them: try the first ones in this list first.
5508 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5509 This hook returns the maximum number of instructions that can ever
5510 issue at the same time on the target machine. The default is one.
5511 Although the insn scheduler can define itself the possibility of issue
5512 an insn on the same cycle, the value can serve as an additional
5513 constraint to issue insns on the same simulated processor cycle (see
5514 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5515 This value must be constant over the entire compilation. If you need
5516 it to vary depending on what the instructions are, you must use
5517 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5519 For the automaton based pipeline interface, you could define this hook
5520 to return the value of the macro @code{MAX_DFA_ISSUE_RATE}.
5523 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5524 This hook is executed by the scheduler after it has scheduled an insn
5525 from the ready list. It should return the number of insns which can
5526 still be issued in the current cycle. The default is
5527 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5528 @code{USE}, which normally are not counted against the issue rate.
5529 You should define this hook if some insns take more machine resources
5530 than others, so that fewer insns can follow them in the same cycle.
5531 @var{file} is either a null pointer, or a stdio stream to write any
5532 debug output to. @var{verbose} is the verbose level provided by
5533 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5537 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5538 This function corrects the value of @var{cost} based on the
5539 relationship between @var{insn} and @var{dep_insn} through the
5540 dependence @var{link}. It should return the new value. The default
5541 is to make no adjustment to @var{cost}. This can be used for example
5542 to specify to the scheduler using the traditional pipeline description
5543 that an output- or anti-dependence does not incur the same cost as a
5544 data-dependence. If the scheduler using the automaton based pipeline
5545 description, the cost of anti-dependence is zero and the cost of
5546 output-dependence is maximum of one and the difference of latency
5547 times of the first and the second insns. If these values are not
5548 acceptable, you could use the hook to modify them too. See also
5549 @pxref{Automaton pipeline description}.
5552 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5553 This hook adjusts the integer scheduling priority @var{priority} of
5554 @var{insn}. It should return the new priority. Reduce the priority to
5555 execute @var{insn} earlier, increase the priority to execute @var{insn}
5556 later. Do not define this hook if you do not need to adjust the
5557 scheduling priorities of insns.
5560 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5561 This hook is executed by the scheduler after it has scheduled the ready
5562 list, to allow the machine description to reorder it (for example to
5563 combine two small instructions together on @samp{VLIW} machines).
5564 @var{file} is either a null pointer, or a stdio stream to write any
5565 debug output to. @var{verbose} is the verbose level provided by
5566 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5567 list of instructions that are ready to be scheduled. @var{n_readyp} is
5568 a pointer to the number of elements in the ready list. The scheduler
5569 reads the ready list in reverse order, starting with
5570 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5571 is the timer tick of the scheduler. You may modify the ready list and
5572 the number of ready insns. The return value is the number of insns that
5573 can issue this cycle; normally this is just @code{issue_rate}. See also
5574 @samp{TARGET_SCHED_REORDER2}.
5577 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5578 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5579 function is called whenever the scheduler starts a new cycle. This one
5580 is called once per iteration over a cycle, immediately after
5581 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5582 return the number of insns to be scheduled in the same cycle. Defining
5583 this hook can be useful if there are frequent situations where
5584 scheduling one insn causes other insns to become ready in the same
5585 cycle. These other insns can then be taken into account properly.
5588 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5589 This hook is called after evaluation forward dependencies of insns in
5590 chain given by two parameter values (@var{head} and @var{tail}
5591 correspondingly) but before insns scheduling of the insn chain. For
5592 example, it can be used for better insn classification if it requires
5593 analysis of dependencies. This hook can use backward and forward
5594 dependencies of the insn scheduler because they are already
5598 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5599 This hook is executed by the scheduler at the beginning of each block of
5600 instructions that are to be scheduled. @var{file} is either a null
5601 pointer, or a stdio stream to write any debug output to. @var{verbose}
5602 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5603 @var{max_ready} is the maximum number of insns in the current scheduling
5604 region that can be live at the same time. This can be used to allocate
5605 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5608 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5609 This hook is executed by the scheduler at the end of each block of
5610 instructions that are to be scheduled. It can be used to perform
5611 cleanup of any actions done by the other scheduling hooks. @var{file}
5612 is either a null pointer, or a stdio stream to write any debug output
5613 to. @var{verbose} is the verbose level provided by
5614 @option{-fsched-verbose-@var{n}}.
5617 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
5618 This hook is executed by the scheduler after function level initializations.
5619 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5620 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5621 @var{old_max_uid} is the maximum insn uid when scheduling begins.
5624 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
5625 This is the cleanup hook corresponding to TARGET_SCHED_INIT_GLOBAL.
5626 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5627 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5630 @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void)
5631 This hook is called many times during insn scheduling. If the hook
5632 returns nonzero, the automaton based pipeline description is used for
5633 insn scheduling. Otherwise the traditional pipeline description is
5634 used. The default is usage of the traditional pipeline description.
5636 You should also remember that to simplify the insn scheduler sources
5637 an empty traditional pipeline description interface is generated even
5638 if there is no a traditional pipeline description in the @file{.md}
5639 file. The same is true for the automaton based pipeline description.
5640 That means that you should be accurate in defining the hook.
5643 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5644 The hook returns an RTL insn. The automaton state used in the
5645 pipeline hazard recognizer is changed as if the insn were scheduled
5646 when the new simulated processor cycle starts. Usage of the hook may
5647 simplify the automaton pipeline description for some @acronym{VLIW}
5648 processors. If the hook is defined, it is used only for the automaton
5649 based pipeline description. The default is not to change the state
5650 when the new simulated processor cycle starts.
5653 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5654 The hook can be used to initialize data used by the previous hook.
5657 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5658 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5659 to changed the state as if the insn were scheduled when the new
5660 simulated processor cycle finishes.
5663 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5664 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5665 used to initialize data used by the previous hook.
5668 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5669 This hook controls better choosing an insn from the ready insn queue
5670 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5671 chooses the first insn from the queue. If the hook returns a positive
5672 value, an additional scheduler code tries all permutations of
5673 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5674 subsequent ready insns to choose an insn whose issue will result in
5675 maximal number of issued insns on the same cycle. For the
5676 @acronym{VLIW} processor, the code could actually solve the problem of
5677 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5678 rules of @acronym{VLIW} packing are described in the automaton.
5680 This code also could be used for superscalar @acronym{RISC}
5681 processors. Let us consider a superscalar @acronym{RISC} processor
5682 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5683 @var{B}, some insns can be executed only in pipelines @var{B} or
5684 @var{C}, and one insn can be executed in pipeline @var{B}. The
5685 processor may issue the 1st insn into @var{A} and the 2nd one into
5686 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5687 until the next cycle. If the scheduler issues the 3rd insn the first,
5688 the processor could issue all 3 insns per cycle.
5690 Actually this code demonstrates advantages of the automaton based
5691 pipeline hazard recognizer. We try quickly and easy many insn
5692 schedules to choose the best one.
5694 The default is no multipass scheduling.
5697 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5699 This hook controls what insns from the ready insn queue will be
5700 considered for the multipass insn scheduling. If the hook returns
5701 zero for insn passed as the parameter, the insn will be not chosen to
5704 The default is that any ready insns can be chosen to be issued.
5707 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5709 This hook is called by the insn scheduler before issuing insn passed
5710 as the third parameter on given cycle. If the hook returns nonzero,
5711 the insn is not issued on given processors cycle. Instead of that,
5712 the processor cycle is advanced. If the value passed through the last
5713 parameter is zero, the insn ready queue is not sorted on the new cycle
5714 start as usually. The first parameter passes file for debugging
5715 output. The second one passes the scheduler verbose level of the
5716 debugging output. The forth and the fifth parameter values are
5717 correspondingly processor cycle on which the previous insn has been
5718 issued and the current processor cycle.
5721 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void)
5722 The @acronym{DFA}-based scheduler could take the insertion of nop
5723 operations for better insn scheduling into account. It can be done
5724 only if the multi-pass insn scheduling works (see hook
5725 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}).
5727 Let us consider a @acronym{VLIW} processor insn with 3 slots. Each
5728 insn can be placed only in one of the three slots. We have 3 ready
5729 insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be
5730 placed only in the 1st slot, @var{B} can be placed only in the 3rd
5731 slot. We described the automaton which does not permit empty slot
5732 gaps between insns (usually such description is simpler). Without
5733 this code the scheduler would place each insn in 3 separate
5734 @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd
5735 slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is
5736 the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook
5737 @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or
5738 create the nop insns.
5740 You should remember that the scheduler does not insert the nop insns.
5741 It is not wise because of the following optimizations. The scheduler
5742 only considers such possibility to improve the result schedule. The
5743 nop insns should be inserted lately, e.g. on the final phase.
5746 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index})
5747 This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert
5748 nop operations for better insn scheduling when @acronym{DFA}-based
5749 scheduler makes multipass insn scheduling (see also description of
5750 hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook
5751 returns a nop insn with given @var{index}. The indexes start with
5752 zero. The hook should return @code{NULL} if there are no more nop
5753 insns with indexes greater than given index.
5756 @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})
5757 This hook is used to define which dependences are considered costly by
5758 the target, so costly that it is not advisable to schedule the insns that
5759 are involved in the dependence too close to one another. The parameters
5760 to this hook are as follows: The second parameter @var{insn2} is dependent
5761 upon the first parameter @var{insn1}. The dependence between @var{insn1}
5762 and @var{insn2} is represented by the third parameter @var{dep_link}. The
5763 fourth parameter @var{cost} is the cost of the dependence, and the fifth
5764 parameter @var{distance} is the distance in cycles between the two insns.
5765 The hook returns @code{true} if considering the distance between the two
5766 insns the dependence between them is considered costly by the target,
5767 and @code{false} otherwise.
5769 Defining this hook can be useful in multiple-issue out-of-order machines,
5770 where (a) it's practically hopeless to predict the actual data/resource
5771 delays, however: (b) there's a better chance to predict the actual grouping
5772 that will be formed, and (c) correctly emulating the grouping can be very
5773 important. In such targets one may want to allow issuing dependent insns
5774 closer to one another - i.e, closer than the dependence distance; however,
5775 not in cases of "costly dependences", which this hooks allows to define.
5778 Macros in the following table are generated by the program
5779 @file{genattr} and can be useful for writing the hooks.
5781 @defmac MAX_DFA_ISSUE_RATE
5782 The macro definition is generated in the automaton based pipeline
5783 description interface. Its value is calculated from the automaton
5784 based pipeline description and is equal to maximal number of all insns
5785 described in constructions @samp{define_insn_reservation} which can be
5786 issued on the same processor cycle.
5790 @section Dividing the Output into Sections (Texts, Data, @dots{})
5791 @c the above section title is WAY too long. maybe cut the part between
5792 @c the (...)? --mew 10feb93
5794 An object file is divided into sections containing different types of
5795 data. In the most common case, there are three sections: the @dfn{text
5796 section}, which holds instructions and read-only data; the @dfn{data
5797 section}, which holds initialized writable data; and the @dfn{bss
5798 section}, which holds uninitialized data. Some systems have other kinds
5801 The compiler must tell the assembler when to switch sections. These
5802 macros control what commands to output to tell the assembler this. You
5803 can also define additional sections.
5805 @defmac TEXT_SECTION_ASM_OP
5806 A C expression whose value is a string, including spacing, containing the
5807 assembler operation that should precede instructions and read-only data.
5808 Normally @code{"\t.text"} is right.
5811 @defmac HOT_TEXT_SECTION_NAME
5812 If defined, a C string constant for the name of the section containing most
5813 frequently executed functions of the program. If not defined, GCC will provide
5814 a default definition if the target supports named sections.
5817 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5818 If defined, a C string constant for the name of the section containing unlikely
5819 executed functions in the program.
5822 @defmac DATA_SECTION_ASM_OP
5823 A C expression whose value is a string, including spacing, containing the
5824 assembler operation to identify the following data as writable initialized
5825 data. Normally @code{"\t.data"} is right.
5828 @defmac READONLY_DATA_SECTION_ASM_OP
5829 A C expression whose value is a string, including spacing, containing the
5830 assembler operation to identify the following data as read-only initialized
5834 @defmac READONLY_DATA_SECTION
5835 A macro naming a function to call to switch to the proper section for
5836 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5837 if defined, else fall back to @code{text_section}.
5839 The most common definition will be @code{data_section}, if the target
5840 does not have a special read-only data section, and does not put data
5841 in the text section.
5844 @defmac BSS_SECTION_ASM_OP
5845 If defined, a C expression whose value is a string, including spacing,
5846 containing the assembler operation to identify the following data as
5847 uninitialized global data. If not defined, and neither
5848 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5849 uninitialized global data will be output in the data section if
5850 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5854 @defmac INIT_SECTION_ASM_OP
5855 If defined, a C expression whose value is a string, including spacing,
5856 containing the assembler operation to identify the following data as
5857 initialization code. If not defined, GCC will assume such a section does
5861 @defmac FINI_SECTION_ASM_OP
5862 If defined, a C expression whose value is a string, including spacing,
5863 containing the assembler operation to identify the following data as
5864 finalization code. If not defined, GCC will assume such a section does
5868 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5869 If defined, an ASM statement that switches to a different section
5870 via @var{section_op}, calls @var{function}, and switches back to
5871 the text section. This is used in @file{crtstuff.c} if
5872 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5873 to initialization and finalization functions from the init and fini
5874 sections. By default, this macro uses a simple function call. Some
5875 ports need hand-crafted assembly code to avoid dependencies on
5876 registers initialized in the function prologue or to ensure that
5877 constant pools don't end up too far way in the text section.
5880 @defmac FORCE_CODE_SECTION_ALIGN
5881 If defined, an ASM statement that aligns a code section to some
5882 arbitrary boundary. This is used to force all fragments of the
5883 @code{.init} and @code{.fini} sections to have to same alignment
5884 and thus prevent the linker from having to add any padding.
5889 @defmac EXTRA_SECTIONS
5890 A list of names for sections other than the standard two, which are
5891 @code{in_text} and @code{in_data}. You need not define this macro
5892 on a system with no other sections (that GCC needs to use).
5895 @findex text_section
5896 @findex data_section
5897 @defmac EXTRA_SECTION_FUNCTIONS
5898 One or more functions to be defined in @file{varasm.c}. These
5899 functions should do jobs analogous to those of @code{text_section} and
5900 @code{data_section}, for your additional sections. Do not define this
5901 macro if you do not define @code{EXTRA_SECTIONS}.
5904 @defmac JUMP_TABLES_IN_TEXT_SECTION
5905 Define this macro to be an expression with a nonzero value if jump
5906 tables (for @code{tablejump} insns) should be output in the text
5907 section, along with the assembler instructions. Otherwise, the
5908 readonly data section is used.
5910 This macro is irrelevant if there is no separate readonly data section.
5913 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5914 Switches to the appropriate section for output of @var{exp}. You can
5915 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5916 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5917 requires link-time relocations. Bit 0 is set when variable contains
5918 local relocations only, while bit 1 is set for global relocations.
5919 Select the section by calling @code{data_section} or one of the
5920 alternatives for other sections. @var{align} is the constant alignment
5923 The default version of this function takes care of putting read-only
5924 variables in @code{readonly_data_section}.
5927 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5928 Build up a unique section name, expressed as a @code{STRING_CST} node,
5929 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5930 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5931 the initial value of @var{exp} requires link-time relocations.
5933 The default version of this function appends the symbol name to the
5934 ELF section name that would normally be used for the symbol. For
5935 example, the function @code{foo} would be placed in @code{.text.foo}.
5936 Whatever the actual target object format, this is often good enough.
5939 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
5940 Switches to the appropriate section for output of constant pool entry
5941 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
5942 constant in RTL@. The argument @var{mode} is redundant except in the
5943 case of a @code{const_int} rtx. Select the section by calling
5944 @code{readonly_data_section} or one of the alternatives for other
5945 sections. @var{align} is the constant alignment in bits.
5947 The default version of this function takes care of putting symbolic
5948 constants in @code{flag_pic} mode in @code{data_section} and everything
5949 else in @code{readonly_data_section}.
5952 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
5953 Define this hook if references to a symbol or a constant must be
5954 treated differently depending on something about the variable or
5955 function named by the symbol (such as what section it is in).
5957 The hook is executed immediately after rtl has been created for
5958 @var{decl}, which may be a variable or function declaration or
5959 an entry in the constant pool. In either case, @var{rtl} is the
5960 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
5961 in this hook; that field may not have been initialized yet.
5963 In the case of a constant, it is safe to assume that the rtl is
5964 a @code{mem} whose address is a @code{symbol_ref}. Most decls
5965 will also have this form, but that is not guaranteed. Global
5966 register variables, for instance, will have a @code{reg} for their
5967 rtl. (Normally the right thing to do with such unusual rtl is
5970 The @var{new_decl_p} argument will be true if this is the first time
5971 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
5972 be false for subsequent invocations, which will happen for duplicate
5973 declarations. Whether or not anything must be done for the duplicate
5974 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
5975 @var{new_decl_p} is always true when the hook is called for a constant.
5977 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
5978 The usual thing for this hook to do is to record flags in the
5979 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
5980 Historically, the name string was modified if it was necessary to
5981 encode more than one bit of information, but this practice is now
5982 discouraged; use @code{SYMBOL_REF_FLAGS}.
5984 The default definition of this hook, @code{default_encode_section_info}
5985 in @file{varasm.c}, sets a number of commonly-useful bits in
5986 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
5987 before overriding it.
5990 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
5991 Decode @var{name} and return the real name part, sans
5992 the characters that @code{TARGET_ENCODE_SECTION_INFO}
5996 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
5997 Returns true if @var{exp} should be placed into a ``small data'' section.
5998 The default version of this hook always returns false.
6001 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6002 Contains the value true if the target places read-only
6003 ``small data'' into a separate section. The default value is false.
6006 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6007 Returns true if @var{exp} names an object for which name resolution
6008 rules must resolve to the current ``module'' (dynamic shared library
6009 or executable image).
6011 The default version of this hook implements the name resolution rules
6012 for ELF, which has a looser model of global name binding than other
6013 currently supported object file formats.
6016 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6017 Contains the value true if the target supports thread-local storage.
6018 The default value is false.
6023 @section Position Independent Code
6024 @cindex position independent code
6027 This section describes macros that help implement generation of position
6028 independent code. Simply defining these macros is not enough to
6029 generate valid PIC; you must also add support to the macros
6030 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6031 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6032 @samp{movsi} to do something appropriate when the source operand
6033 contains a symbolic address. You may also need to alter the handling of
6034 switch statements so that they use relative addresses.
6035 @c i rearranged the order of the macros above to try to force one of
6036 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6038 @defmac PIC_OFFSET_TABLE_REGNUM
6039 The register number of the register used to address a table of static
6040 data addresses in memory. In some cases this register is defined by a
6041 processor's ``application binary interface'' (ABI)@. When this macro
6042 is defined, RTL is generated for this register once, as with the stack
6043 pointer and frame pointer registers. If this macro is not defined, it
6044 is up to the machine-dependent files to allocate such a register (if
6045 necessary). Note that this register must be fixed when in use (e.g.@:
6046 when @code{flag_pic} is true).
6049 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6050 Define this macro if the register defined by
6051 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6052 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6055 @defmac FINALIZE_PIC
6056 By generating position-independent code, when two different programs (A
6057 and B) share a common library (libC.a), the text of the library can be
6058 shared whether or not the library is linked at the same address for both
6059 programs. In some of these environments, position-independent code
6060 requires not only the use of different addressing modes, but also
6061 special code to enable the use of these addressing modes.
6063 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
6064 codes once the function is being compiled into assembly code, but not
6065 before. (It is not done before, because in the case of compiling an
6066 inline function, it would lead to multiple PIC prologues being
6067 included in functions which used inline functions and were compiled to
6071 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6072 A C expression that is nonzero if @var{x} is a legitimate immediate
6073 operand on the target machine when generating position independent code.
6074 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6075 check this. You can also assume @var{flag_pic} is true, so you need not
6076 check it either. You need not define this macro if all constants
6077 (including @code{SYMBOL_REF}) can be immediate operands when generating
6078 position independent code.
6081 @node Assembler Format
6082 @section Defining the Output Assembler Language
6084 This section describes macros whose principal purpose is to describe how
6085 to write instructions in assembler language---rather than what the
6089 * File Framework:: Structural information for the assembler file.
6090 * Data Output:: Output of constants (numbers, strings, addresses).
6091 * Uninitialized Data:: Output of uninitialized variables.
6092 * Label Output:: Output and generation of labels.
6093 * Initialization:: General principles of initialization
6094 and termination routines.
6095 * Macros for Initialization::
6096 Specific macros that control the handling of
6097 initialization and termination routines.
6098 * Instruction Output:: Output of actual instructions.
6099 * Dispatch Tables:: Output of jump tables.
6100 * Exception Region Output:: Output of exception region code.
6101 * Alignment Output:: Pseudo ops for alignment and skipping data.
6104 @node File Framework
6105 @subsection The Overall Framework of an Assembler File
6106 @cindex assembler format
6107 @cindex output of assembler code
6109 @c prevent bad page break with this line
6110 This describes the overall framework of an assembly file.
6112 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6113 @findex default_file_start
6114 Output to @code{asm_out_file} any text which the assembler expects to
6115 find at the beginning of a file. The default behavior is controlled
6116 by two flags, documented below. Unless your target's assembler is
6117 quite unusual, if you override the default, you should call
6118 @code{default_file_start} at some point in your target hook. This
6119 lets other target files rely on these variables.
6122 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6123 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6124 printed as the very first line in the assembly file, unless
6125 @option{-fverbose-asm} is in effect. (If that macro has been defined
6126 to the empty string, this variable has no effect.) With the normal
6127 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6128 assembler that it need not bother stripping comments or extra
6129 whitespace from its input. This allows it to work a bit faster.
6131 The default is false. You should not set it to true unless you have
6132 verified that your port does not generate any extra whitespace or
6133 comments that will cause GAS to issue errors in NO_APP mode.
6136 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6137 If this flag is true, @code{output_file_directive} will be called
6138 for the primary source file, immediately after printing
6139 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6140 this to be done. The default is false.
6143 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6144 Output to @code{asm_out_file} any text which the assembler expects
6145 to find at the end of a file. The default is to output nothing.
6148 @deftypefun void file_end_indicate_exec_stack ()
6149 Some systems use a common convention, the @samp{.note.GNU-stack}
6150 special section, to indicate whether or not an object file relies on
6151 the stack being executable. If your system uses this convention, you
6152 should define @code{TARGET_ASM_FILE_END} to this function. If you
6153 need to do other things in that hook, have your hook function call
6157 @defmac ASM_COMMENT_START
6158 A C string constant describing how to begin a comment in the target
6159 assembler language. The compiler assumes that the comment will end at
6160 the end of the line.
6164 A C string constant for text to be output before each @code{asm}
6165 statement or group of consecutive ones. Normally this is
6166 @code{"#APP"}, which is a comment that has no effect on most
6167 assemblers but tells the GNU assembler that it must check the lines
6168 that follow for all valid assembler constructs.
6172 A C string constant for text to be output after each @code{asm}
6173 statement or group of consecutive ones. Normally this is
6174 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6175 time-saving assumptions that are valid for ordinary compiler output.
6178 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6179 A C statement to output COFF information or DWARF debugging information
6180 which indicates that filename @var{name} is the current source file to
6181 the stdio stream @var{stream}.
6183 This macro need not be defined if the standard form of output
6184 for the file format in use is appropriate.
6187 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6188 A C statement to output the string @var{string} to the stdio stream
6189 @var{stream}. If you do not call the function @code{output_quoted_string}
6190 in your config files, GCC will only call it to output filenames to
6191 the assembler source. So you can use it to canonicalize the format
6192 of the filename using this macro.
6195 @defmac ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
6196 A C statement to output DBX or SDB debugging information before code
6197 for line number @var{line} of the current source file to the
6198 stdio stream @var{stream}. @var{counter} is the number of time the
6199 macro was invoked, including the current invocation; it is intended
6200 to generate unique labels in the assembly output.
6202 This macro need not be defined if the standard form of debugging
6203 information for the debugger in use is appropriate.
6206 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6207 A C statement to output something to the assembler file to handle a
6208 @samp{#ident} directive containing the text @var{string}. If this
6209 macro is not defined, nothing is output for a @samp{#ident} directive.
6212 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6213 Output assembly directives to switch to section @var{name}. The section
6214 should have attributes as specified by @var{flags}, which is a bit mask
6215 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6216 is nonzero, it contains an alignment in bytes to be used for the section,
6217 otherwise some target default should be used. Only targets that must
6218 specify an alignment within the section directive need pay attention to
6219 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6222 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6223 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6226 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6227 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6228 based on a variable or function decl, a section name, and whether or not the
6229 declaration's initializer may contain runtime relocations. @var{decl} may be
6230 null, in which case read-write data should be assumed.
6232 The default version if this function handles choosing code vs data,
6233 read-only vs read-write data, and @code{flag_pic}. You should only
6234 need to override this if your target has special flags that might be
6235 set via @code{__attribute__}.
6240 @subsection Output of Data
6243 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6244 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6245 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6246 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6247 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6248 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6249 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6250 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6251 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6252 These hooks specify assembly directives for creating certain kinds
6253 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6254 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6255 aligned two-byte object, and so on. Any of the hooks may be
6256 @code{NULL}, indicating that no suitable directive is available.
6258 The compiler will print these strings at the start of a new line,
6259 followed immediately by the object's initial value. In most cases,
6260 the string should contain a tab, a pseudo-op, and then another tab.
6263 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6264 The @code{assemble_integer} function uses this hook to output an
6265 integer object. @var{x} is the object's value, @var{size} is its size
6266 in bytes and @var{aligned_p} indicates whether it is aligned. The
6267 function should return @code{true} if it was able to output the
6268 object. If it returns false, @code{assemble_integer} will try to
6269 split the object into smaller parts.
6271 The default implementation of this hook will use the
6272 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6273 when the relevant string is @code{NULL}.
6276 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6277 A C statement to recognize @var{rtx} patterns that
6278 @code{output_addr_const} can't deal with, and output assembly code to
6279 @var{stream} corresponding to the pattern @var{x}. This may be used to
6280 allow machine-dependent @code{UNSPEC}s to appear within constants.
6282 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6283 @code{goto fail}, so that a standard error message is printed. If it
6284 prints an error message itself, by calling, for example,
6285 @code{output_operand_lossage}, it may just complete normally.
6288 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6289 A C statement to output to the stdio stream @var{stream} an assembler
6290 instruction to assemble a string constant containing the @var{len}
6291 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6292 @code{char *} and @var{len} a C expression of type @code{int}.
6294 If the assembler has a @code{.ascii} pseudo-op as found in the
6295 Berkeley Unix assembler, do not define the macro
6296 @code{ASM_OUTPUT_ASCII}.
6299 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6300 A C statement to output word @var{n} of a function descriptor for
6301 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6302 is defined, and is otherwise unused.
6305 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6306 You may define this macro as a C expression. You should define the
6307 expression to have a nonzero value if GCC should output the constant
6308 pool for a function before the code for the function, or a zero value if
6309 GCC should output the constant pool after the function. If you do
6310 not define this macro, the usual case, GCC will output the constant
6311 pool before the function.
6314 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6315 A C statement to output assembler commands to define the start of the
6316 constant pool for a function. @var{funname} is a string giving
6317 the name of the function. Should the return type of the function
6318 be required, it can be obtained via @var{fundecl}. @var{size}
6319 is the size, in bytes, of the constant pool that will be written
6320 immediately after this call.
6322 If no constant-pool prefix is required, the usual case, this macro need
6326 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6327 A C statement (with or without semicolon) to output a constant in the
6328 constant pool, if it needs special treatment. (This macro need not do
6329 anything for RTL expressions that can be output normally.)
6331 The argument @var{file} is the standard I/O stream to output the
6332 assembler code on. @var{x} is the RTL expression for the constant to
6333 output, and @var{mode} is the machine mode (in case @var{x} is a
6334 @samp{const_int}). @var{align} is the required alignment for the value
6335 @var{x}; you should output an assembler directive to force this much
6338 The argument @var{labelno} is a number to use in an internal label for
6339 the address of this pool entry. The definition of this macro is
6340 responsible for outputting the label definition at the proper place.
6341 Here is how to do this:
6344 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6347 When you output a pool entry specially, you should end with a
6348 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6349 entry from being output a second time in the usual manner.
6351 You need not define this macro if it would do nothing.
6354 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6355 A C statement to output assembler commands to at the end of the constant
6356 pool for a function. @var{funname} is a string giving the name of the
6357 function. Should the return type of the function be required, you can
6358 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6359 constant pool that GCC wrote immediately before this call.
6361 If no constant-pool epilogue is required, the usual case, you need not
6365 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6366 Define this macro as a C expression which is nonzero if @var{C} is
6367 used as a logical line separator by the assembler.
6369 If you do not define this macro, the default is that only
6370 the character @samp{;} is treated as a logical line separator.
6373 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6374 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6375 These target hooks are C string constants, describing the syntax in the
6376 assembler for grouping arithmetic expressions. If not overridden, they
6377 default to normal parentheses, which is correct for most assemblers.
6380 These macros are provided by @file{real.h} for writing the definitions
6381 of @code{ASM_OUTPUT_DOUBLE} and the like:
6383 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6384 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6385 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6386 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6387 floating point representation, and store its bit pattern in the variable
6388 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6389 be a simple @code{long int}. For the others, it should be an array of
6390 @code{long int}. The number of elements in this array is determined by
6391 the size of the desired target floating point data type: 32 bits of it
6392 go in each @code{long int} array element. Each array element holds 32
6393 bits of the result, even if @code{long int} is wider than 32 bits on the
6396 The array element values are designed so that you can print them out
6397 using @code{fprintf} in the order they should appear in the target
6401 @node Uninitialized Data
6402 @subsection Output of Uninitialized Variables
6404 Each of the macros in this section is used to do the whole job of
6405 outputting a single uninitialized variable.
6407 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6408 A C statement (sans semicolon) to output to the stdio stream
6409 @var{stream} the assembler definition of a common-label named
6410 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6411 is the size rounded up to whatever alignment the caller wants.
6413 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6414 output the name itself; before and after that, output the additional
6415 assembler syntax for defining the name, and a newline.
6417 This macro controls how the assembler definitions of uninitialized
6418 common global variables are output.
6421 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6422 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6423 separate, explicit argument. If you define this macro, it is used in
6424 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6425 handling the required alignment of the variable. The alignment is specified
6426 as the number of bits.
6429 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6430 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6431 variable to be output, if there is one, or @code{NULL_TREE} if there
6432 is no corresponding variable. If you define this macro, GCC will use it
6433 in place of both @code{ASM_OUTPUT_COMMON} and
6434 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6435 the variable's decl in order to chose what to output.
6438 @defmac ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6439 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6440 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6444 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6445 A C statement (sans semicolon) to output to the stdio stream
6446 @var{stream} the assembler definition of uninitialized global @var{decl} named
6447 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6448 is the size rounded up to whatever alignment the caller wants.
6450 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6451 defining this macro. If unable, use the expression
6452 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6453 before and after that, output the additional assembler syntax for defining
6454 the name, and a newline.
6456 This macro controls how the assembler definitions of uninitialized global
6457 variables are output. This macro exists to properly support languages like
6458 C++ which do not have @code{common} data. However, this macro currently
6459 is not defined for all targets. If this macro and
6460 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6461 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6462 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6465 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6466 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6467 separate, explicit argument. If you define this macro, it is used in
6468 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6469 handling the required alignment of the variable. The alignment is specified
6470 as the number of bits.
6472 Try to use function @code{asm_output_aligned_bss} defined in file
6473 @file{varasm.c} when defining this macro.
6476 @defmac ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6477 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6478 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6482 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6483 A C statement (sans semicolon) to output to the stdio stream
6484 @var{stream} the assembler definition of a local-common-label named
6485 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6486 is the size rounded up to whatever alignment the caller wants.
6488 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6489 output the name itself; before and after that, output the additional
6490 assembler syntax for defining the name, and a newline.
6492 This macro controls how the assembler definitions of uninitialized
6493 static variables are output.
6496 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6497 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6498 separate, explicit argument. If you define this macro, it is used in
6499 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6500 handling the required alignment of the variable. The alignment is specified
6501 as the number of bits.
6504 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6505 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6506 variable to be output, if there is one, or @code{NULL_TREE} if there
6507 is no corresponding variable. If you define this macro, GCC will use it
6508 in place of both @code{ASM_OUTPUT_DECL} and
6509 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6510 the variable's decl in order to chose what to output.
6513 @defmac ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6514 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6515 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6520 @subsection Output and Generation of Labels
6522 @c prevent bad page break with this line
6523 This is about outputting labels.
6525 @findex assemble_name
6526 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6527 A C statement (sans semicolon) to output to the stdio stream
6528 @var{stream} the assembler definition of a label named @var{name}.
6529 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6530 output the name itself; before and after that, output the additional
6531 assembler syntax for defining the name, and a newline. A default
6532 definition of this macro is provided which is correct for most systems.
6536 A C string containing the appropriate assembler directive to specify the
6537 size of a symbol, without any arguments. On systems that use ELF, the
6538 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6539 systems, the default is not to define this macro.
6541 Define this macro only if it is correct to use the default definitions
6542 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6543 for your system. If you need your own custom definitions of those
6544 macros, or if you do not need explicit symbol sizes at all, do not
6548 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6549 A C statement (sans semicolon) to output to the stdio stream
6550 @var{stream} a directive telling the assembler that the size of the
6551 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6552 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6556 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6557 A C statement (sans semicolon) to output to the stdio stream
6558 @var{stream} a directive telling the assembler to calculate the size of
6559 the symbol @var{name} by subtracting its address from the current
6562 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6563 provided. The default assumes that the assembler recognizes a special
6564 @samp{.} symbol as referring to the current address, and can calculate
6565 the difference between this and another symbol. If your assembler does
6566 not recognize @samp{.} or cannot do calculations with it, you will need
6567 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6571 A C string containing the appropriate assembler directive to specify the
6572 type of a symbol, without any arguments. On systems that use ELF, the
6573 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6574 systems, the default is not to define this macro.
6576 Define this macro only if it is correct to use the default definition of
6577 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6578 custom definition of this macro, or if you do not need explicit symbol
6579 types at all, do not define this macro.
6582 @defmac TYPE_OPERAND_FMT
6583 A C string which specifies (using @code{printf} syntax) the format of
6584 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6585 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6586 the default is not to define this macro.
6588 Define this macro only if it is correct to use the default definition of
6589 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6590 custom definition of this macro, or if you do not need explicit symbol
6591 types at all, do not define this macro.
6594 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6595 A C statement (sans semicolon) to output to the stdio stream
6596 @var{stream} a directive telling the assembler that the type of the
6597 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6598 that string is always either @samp{"function"} or @samp{"object"}, but
6599 you should not count on this.
6601 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6602 definition of this macro is provided.
6605 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6606 A C statement (sans semicolon) to output to the stdio stream
6607 @var{stream} any text necessary for declaring the name @var{name} of a
6608 function which is being defined. This macro is responsible for
6609 outputting the label definition (perhaps using
6610 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6611 @code{FUNCTION_DECL} tree node representing the function.
6613 If this macro is not defined, then the function name is defined in the
6614 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6616 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6620 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6621 A C statement (sans semicolon) to output to the stdio stream
6622 @var{stream} any text necessary for declaring the size of a function
6623 which is being defined. The argument @var{name} is the name of the
6624 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6625 representing the function.
6627 If this macro is not defined, then the function size is not defined.
6629 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6633 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6634 A C statement (sans semicolon) to output to the stdio stream
6635 @var{stream} any text necessary for declaring the name @var{name} of an
6636 initialized variable which is being defined. This macro must output the
6637 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6638 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6640 If this macro is not defined, then the variable name is defined in the
6641 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6643 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6644 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6647 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
6648 A C statement (sans semicolon) to output to the stdio stream
6649 @var{stream} any text necessary for declaring the name @var{name} of a
6650 constant which is being defined. This macro is responsible for
6651 outputting the label definition (perhaps using
6652 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
6653 value of the constant, and @var{size} is the size of the constant
6654 in bytes. @var{name} will be an internal label.
6656 If this macro is not defined, then the @var{name} is defined in the
6657 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6659 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6663 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6664 A C statement (sans semicolon) to output to the stdio stream
6665 @var{stream} any text necessary for claiming a register @var{regno}
6666 for a global variable @var{decl} with name @var{name}.
6668 If you don't define this macro, that is equivalent to defining it to do
6672 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6673 A C statement (sans semicolon) to finish up declaring a variable name
6674 once the compiler has processed its initializer fully and thus has had a
6675 chance to determine the size of an array when controlled by an
6676 initializer. This is used on systems where it's necessary to declare
6677 something about the size of the object.
6679 If you don't define this macro, that is equivalent to defining it to do
6682 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6683 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6686 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6687 This target hook is a function to output to the stdio stream
6688 @var{stream} some commands that will make the label @var{name} global;
6689 that is, available for reference from other files.
6691 The default implementation relies on a proper definition of
6692 @code{GLOBAL_ASM_OP}.
6695 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6696 A C statement (sans semicolon) to output to the stdio stream
6697 @var{stream} some commands that will make the label @var{name} weak;
6698 that is, available for reference from other files but only used if
6699 no other definition is available. Use the expression
6700 @code{assemble_name (@var{stream}, @var{name})} to output the name
6701 itself; before and after that, output the additional assembler syntax
6702 for making that name weak, and a newline.
6704 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6705 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6709 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6710 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6711 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6712 or variable decl. If @var{value} is not @code{NULL}, this C statement
6713 should output to the stdio stream @var{stream} assembler code which
6714 defines (equates) the weak symbol @var{name} to have the value
6715 @var{value}. If @var{value} is @code{NULL}, it should output commands
6716 to make @var{name} weak.
6719 @defmac SUPPORTS_WEAK
6720 A C expression which evaluates to true if the target supports weak symbols.
6722 If you don't define this macro, @file{defaults.h} provides a default
6723 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6724 is defined, the default definition is @samp{1}; otherwise, it is
6725 @samp{0}. Define this macro if you want to control weak symbol support
6726 with a compiler flag such as @option{-melf}.
6729 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
6730 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6731 public symbol such that extra copies in multiple translation units will
6732 be discarded by the linker. Define this macro if your object file
6733 format provides support for this concept, such as the @samp{COMDAT}
6734 section flags in the Microsoft Windows PE/COFF format, and this support
6735 requires changes to @var{decl}, such as putting it in a separate section.
6738 @defmac SUPPORTS_ONE_ONLY
6739 A C expression which evaluates to true if the target supports one-only
6742 If you don't define this macro, @file{varasm.c} provides a default
6743 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6744 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6745 you want to control one-only symbol support with a compiler flag, or if
6746 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6747 be emitted as one-only.
6750 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6751 This target hook is a function to output to @var{asm_out_file} some
6752 commands that will make the symbol(s) associated with @var{decl} have
6753 hidden, protected or internal visibility as specified by @var{visibility}.
6756 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
6757 A C expression that evaluates to true if the target's linker expects
6758 that weak symbols do not appear in a static archive's table of contents.
6759 The default is @code{0}.
6761 Leaving weak symbols out of an archive's table of contents means that,
6762 if a symbol will only have a definition in one translation unit and
6763 will have undefined references from other translation units, that
6764 symbol should not be weak. Defining this macro to be nonzero will
6765 thus have the effect that certain symbols that would normally be weak
6766 (explicit template instantiations, and vtables for polymorphic classes
6767 with noninline key methods) will instead be nonweak.
6769 The C++ ABI requires this macro to be zero. Define this macro for
6770 targets where full C++ ABI compliance is impossible and where linker
6771 restrictions require weak symbols to be left out of a static archive's
6775 @defmac TARGET_SUPPORTS_HIDDEN
6776 A C expression that evaluates to true if the target supports hidden
6777 visibility. By default this expression is true if and only if
6778 @code{HAS_GAS_HIDDEN} is defined. Set this macro if the
6779 @code{HAS_GAS_HIDDEN} macro gives the wrong answer for this
6780 target. (For example, if the target's mechanism for supporting
6781 hidden visibility is not the same as GAS's.)
6784 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6785 A C statement (sans semicolon) to output to the stdio stream
6786 @var{stream} any text necessary for declaring the name of an external
6787 symbol named @var{name} which is referenced in this compilation but
6788 not defined. The value of @var{decl} is the tree node for the
6791 This macro need not be defined if it does not need to output anything.
6792 The GNU assembler and most Unix assemblers don't require anything.
6795 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
6796 This target hook is a function to output to @var{asm_out_file} an assembler
6797 pseudo-op to declare a library function name external. The name of the
6798 library function is given by @var{symref}, which is a @code{symbol_ref}.
6801 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6802 A C statement (sans semicolon) to output to the stdio stream
6803 @var{stream} a reference in assembler syntax to a label named
6804 @var{name}. This should add @samp{_} to the front of the name, if that
6805 is customary on your operating system, as it is in most Berkeley Unix
6806 systems. This macro is used in @code{assemble_name}.
6809 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6810 A C statement (sans semicolon) to output a reference to
6811 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6812 will be used to output the name of the symbol. This macro may be used
6813 to modify the way a symbol is referenced depending on information
6814 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6817 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6818 A C statement (sans semicolon) to output a reference to @var{buf}, the
6819 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6820 @code{assemble_name} will be used to output the name of the symbol.
6821 This macro is not used by @code{output_asm_label}, or the @code{%l}
6822 specifier that calls it; the intention is that this macro should be set
6823 when it is necessary to output a label differently when its address is
6827 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6828 A function to output to the stdio stream @var{stream} a label whose
6829 name is made from the string @var{prefix} and the number @var{labelno}.
6831 It is absolutely essential that these labels be distinct from the labels
6832 used for user-level functions and variables. Otherwise, certain programs
6833 will have name conflicts with internal labels.
6835 It is desirable to exclude internal labels from the symbol table of the
6836 object file. Most assemblers have a naming convention for labels that
6837 should be excluded; on many systems, the letter @samp{L} at the
6838 beginning of a label has this effect. You should find out what
6839 convention your system uses, and follow it.
6841 The default version of this function utilizes ASM_GENERATE_INTERNAL_LABEL.
6844 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6845 A C statement to output to the stdio stream @var{stream} a debug info
6846 label whose name is made from the string @var{prefix} and the number
6847 @var{num}. This is useful for VLIW targets, where debug info labels
6848 may need to be treated differently than branch target labels. On some
6849 systems, branch target labels must be at the beginning of instruction
6850 bundles, but debug info labels can occur in the middle of instruction
6853 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
6857 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6858 A C statement to store into the string @var{string} a label whose name
6859 is made from the string @var{prefix} and the number @var{num}.
6861 This string, when output subsequently by @code{assemble_name}, should
6862 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
6863 with the same @var{prefix} and @var{num}.
6865 If the string begins with @samp{*}, then @code{assemble_name} will
6866 output the rest of the string unchanged. It is often convenient for
6867 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6868 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6869 to output the string, and may change it. (Of course,
6870 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6871 you should know what it does on your machine.)
6874 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6875 A C expression to assign to @var{outvar} (which is a variable of type
6876 @code{char *}) a newly allocated string made from the string
6877 @var{name} and the number @var{number}, with some suitable punctuation
6878 added. Use @code{alloca} to get space for the string.
6880 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6881 produce an assembler label for an internal static variable whose name is
6882 @var{name}. Therefore, the string must be such as to result in valid
6883 assembler code. The argument @var{number} is different each time this
6884 macro is executed; it prevents conflicts between similarly-named
6885 internal static variables in different scopes.
6887 Ideally this string should not be a valid C identifier, to prevent any
6888 conflict with the user's own symbols. Most assemblers allow periods
6889 or percent signs in assembler symbols; putting at least one of these
6890 between the name and the number will suffice.
6892 If this macro is not defined, a default definition will be provided
6893 which is correct for most systems.
6896 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6897 A C statement to output to the stdio stream @var{stream} assembler code
6898 which defines (equates) the symbol @var{name} to have the value @var{value}.
6901 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6902 correct for most systems.
6905 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6906 A C statement to output to the stdio stream @var{stream} assembler code
6907 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6908 to have the value of the tree node @var{decl_of_value}. This macro will
6909 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6910 the tree nodes are available.
6913 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6914 correct for most systems.
6917 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6918 A C statement to output to the stdio stream @var{stream} assembler code
6919 which defines (equates) the weak symbol @var{name} to have the value
6920 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6921 an undefined weak symbol.
6923 Define this macro if the target only supports weak aliases; define
6924 @code{ASM_OUTPUT_DEF} instead if possible.
6927 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6928 Define this macro to override the default assembler names used for
6929 Objective-C methods.
6931 The default name is a unique method number followed by the name of the
6932 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6933 the category is also included in the assembler name (e.g.@:
6936 These names are safe on most systems, but make debugging difficult since
6937 the method's selector is not present in the name. Therefore, particular
6938 systems define other ways of computing names.
6940 @var{buf} is an expression of type @code{char *} which gives you a
6941 buffer in which to store the name; its length is as long as
6942 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6943 50 characters extra.
6945 The argument @var{is_inst} specifies whether the method is an instance
6946 method or a class method; @var{class_name} is the name of the class;
6947 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6948 in a category); and @var{sel_name} is the name of the selector.
6950 On systems where the assembler can handle quoted names, you can use this
6951 macro to provide more human-readable names.
6954 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
6955 A C statement (sans semicolon) to output to the stdio stream
6956 @var{stream} commands to declare that the label @var{name} is an
6957 Objective-C class reference. This is only needed for targets whose
6958 linkers have special support for NeXT-style runtimes.
6961 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6962 A C statement (sans semicolon) to output to the stdio stream
6963 @var{stream} commands to declare that the label @var{name} is an
6964 unresolved Objective-C class reference. This is only needed for targets
6965 whose linkers have special support for NeXT-style runtimes.
6968 @node Initialization
6969 @subsection How Initialization Functions Are Handled
6970 @cindex initialization routines
6971 @cindex termination routines
6972 @cindex constructors, output of
6973 @cindex destructors, output of
6975 The compiled code for certain languages includes @dfn{constructors}
6976 (also called @dfn{initialization routines})---functions to initialize
6977 data in the program when the program is started. These functions need
6978 to be called before the program is ``started''---that is to say, before
6979 @code{main} is called.
6981 Compiling some languages generates @dfn{destructors} (also called
6982 @dfn{termination routines}) that should be called when the program
6985 To make the initialization and termination functions work, the compiler
6986 must output something in the assembler code to cause those functions to
6987 be called at the appropriate time. When you port the compiler to a new
6988 system, you need to specify how to do this.
6990 There are two major ways that GCC currently supports the execution of
6991 initialization and termination functions. Each way has two variants.
6992 Much of the structure is common to all four variations.
6994 @findex __CTOR_LIST__
6995 @findex __DTOR_LIST__
6996 The linker must build two lists of these functions---a list of
6997 initialization functions, called @code{__CTOR_LIST__}, and a list of
6998 termination functions, called @code{__DTOR_LIST__}.
7000 Each list always begins with an ignored function pointer (which may hold
7001 0, @minus{}1, or a count of the function pointers after it, depending on
7002 the environment). This is followed by a series of zero or more function
7003 pointers to constructors (or destructors), followed by a function
7004 pointer containing zero.
7006 Depending on the operating system and its executable file format, either
7007 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7008 time and exit time. Constructors are called in reverse order of the
7009 list; destructors in forward order.
7011 The best way to handle static constructors works only for object file
7012 formats which provide arbitrarily-named sections. A section is set
7013 aside for a list of constructors, and another for a list of destructors.
7014 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7015 object file that defines an initialization function also puts a word in
7016 the constructor section to point to that function. The linker
7017 accumulates all these words into one contiguous @samp{.ctors} section.
7018 Termination functions are handled similarly.
7020 This method will be chosen as the default by @file{target-def.h} if
7021 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7022 support arbitrary sections, but does support special designated
7023 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7024 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7026 When arbitrary sections are available, there are two variants, depending
7027 upon how the code in @file{crtstuff.c} is called. On systems that
7028 support a @dfn{.init} section which is executed at program startup,
7029 parts of @file{crtstuff.c} are compiled into that section. The
7030 program is linked by the @command{gcc} driver like this:
7033 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7036 The prologue of a function (@code{__init}) appears in the @code{.init}
7037 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7038 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7039 files are provided by the operating system or by the GNU C library, but
7040 are provided by GCC for a few targets.
7042 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7043 compiled from @file{crtstuff.c}. They contain, among other things, code
7044 fragments within the @code{.init} and @code{.fini} sections that branch
7045 to routines in the @code{.text} section. The linker will pull all parts
7046 of a section together, which results in a complete @code{__init} function
7047 that invokes the routines we need at startup.
7049 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7052 If no init section is available, when GCC compiles any function called
7053 @code{main} (or more accurately, any function designated as a program
7054 entry point by the language front end calling @code{expand_main_function}),
7055 it inserts a procedure call to @code{__main} as the first executable code
7056 after the function prologue. The @code{__main} function is defined
7057 in @file{libgcc2.c} and runs the global constructors.
7059 In file formats that don't support arbitrary sections, there are again
7060 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7061 and an `a.out' format must be used. In this case,
7062 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7063 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7064 and with the address of the void function containing the initialization
7065 code as its value. The GNU linker recognizes this as a request to add
7066 the value to a @dfn{set}; the values are accumulated, and are eventually
7067 placed in the executable as a vector in the format described above, with
7068 a leading (ignored) count and a trailing zero element.
7069 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7070 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7071 the compilation of @code{main} to call @code{__main} as above, starting
7072 the initialization process.
7074 The last variant uses neither arbitrary sections nor the GNU linker.
7075 This is preferable when you want to do dynamic linking and when using
7076 file formats which the GNU linker does not support, such as `ECOFF'@. In
7077 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7078 termination functions are recognized simply by their names. This requires
7079 an extra program in the linkage step, called @command{collect2}. This program
7080 pretends to be the linker, for use with GCC; it does its job by running
7081 the ordinary linker, but also arranges to include the vectors of
7082 initialization and termination functions. These functions are called
7083 via @code{__main} as described above. In order to use this method,
7084 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7087 The following section describes the specific macros that control and
7088 customize the handling of initialization and termination functions.
7091 @node Macros for Initialization
7092 @subsection Macros Controlling Initialization Routines
7094 Here are the macros that control how the compiler handles initialization
7095 and termination functions:
7097 @defmac INIT_SECTION_ASM_OP
7098 If defined, a C string constant, including spacing, for the assembler
7099 operation to identify the following data as initialization code. If not
7100 defined, GCC will assume such a section does not exist. When you are
7101 using special sections for initialization and termination functions, this
7102 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7103 run the initialization functions.
7106 @defmac HAS_INIT_SECTION
7107 If defined, @code{main} will not call @code{__main} as described above.
7108 This macro should be defined for systems that control start-up code
7109 on a symbol-by-symbol basis, such as OSF/1, and should not
7110 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7113 @defmac LD_INIT_SWITCH
7114 If defined, a C string constant for a switch that tells the linker that
7115 the following symbol is an initialization routine.
7118 @defmac LD_FINI_SWITCH
7119 If defined, a C string constant for a switch that tells the linker that
7120 the following symbol is a finalization routine.
7123 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7124 If defined, a C statement that will write a function that can be
7125 automatically called when a shared library is loaded. The function
7126 should call @var{func}, which takes no arguments. If not defined, and
7127 the object format requires an explicit initialization function, then a
7128 function called @code{_GLOBAL__DI} will be generated.
7130 This function and the following one are used by collect2 when linking a
7131 shared library that needs constructors or destructors, or has DWARF2
7132 exception tables embedded in the code.
7135 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7136 If defined, a C statement that will write a function that can be
7137 automatically called when a shared library is unloaded. The function
7138 should call @var{func}, which takes no arguments. If not defined, and
7139 the object format requires an explicit finalization function, then a
7140 function called @code{_GLOBAL__DD} will be generated.
7143 @defmac INVOKE__main
7144 If defined, @code{main} will call @code{__main} despite the presence of
7145 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7146 where the init section is not actually run automatically, but is still
7147 useful for collecting the lists of constructors and destructors.
7150 @defmac SUPPORTS_INIT_PRIORITY
7151 If nonzero, the C++ @code{init_priority} attribute is supported and the
7152 compiler should emit instructions to control the order of initialization
7153 of objects. If zero, the compiler will issue an error message upon
7154 encountering an @code{init_priority} attribute.
7157 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7158 This value is true if the target supports some ``native'' method of
7159 collecting constructors and destructors to be run at startup and exit.
7160 It is false if we must use @command{collect2}.
7163 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7164 If defined, a function that outputs assembler code to arrange to call
7165 the function referenced by @var{symbol} at initialization time.
7167 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7168 no arguments and with no return value. If the target supports initialization
7169 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7170 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7172 If this macro is not defined by the target, a suitable default will
7173 be chosen if (1) the target supports arbitrary section names, (2) the
7174 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7178 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7179 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7180 functions rather than initialization functions.
7183 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7184 generated for the generated object file will have static linkage.
7186 If your system uses @command{collect2} as the means of processing
7187 constructors, then that program normally uses @command{nm} to scan
7188 an object file for constructor functions to be called.
7190 On certain kinds of systems, you can define this macro to make
7191 @command{collect2} work faster (and, in some cases, make it work at all):
7193 @defmac OBJECT_FORMAT_COFF
7194 Define this macro if the system uses COFF (Common Object File Format)
7195 object files, so that @command{collect2} can assume this format and scan
7196 object files directly for dynamic constructor/destructor functions.
7198 This macro is effective only in a native compiler; @command{collect2} as
7199 part of a cross compiler always uses @command{nm} for the target machine.
7202 @defmac COLLECT_PARSE_FLAG (@var{flag})
7203 Define this macro to be C code that examines @command{collect2} command
7204 line option @var{flag} and performs special actions if
7205 @command{collect2} needs to behave differently depending on @var{flag}.
7208 @defmac REAL_NM_FILE_NAME
7209 Define this macro as a C string constant containing the file name to use
7210 to execute @command{nm}. The default is to search the path normally for
7213 If your system supports shared libraries and has a program to list the
7214 dynamic dependencies of a given library or executable, you can define
7215 these macros to enable support for running initialization and
7216 termination functions in shared libraries:
7220 Define this macro to a C string constant containing the name of the program
7221 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7224 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7225 Define this macro to be C code that extracts filenames from the output
7226 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7227 of type @code{char *} that points to the beginning of a line of output
7228 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7229 code must advance @var{ptr} to the beginning of the filename on that
7230 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7233 @node Instruction Output
7234 @subsection Output of Assembler Instructions
7236 @c prevent bad page break with this line
7237 This describes assembler instruction output.
7239 @defmac REGISTER_NAMES
7240 A C initializer containing the assembler's names for the machine
7241 registers, each one as a C string constant. This is what translates
7242 register numbers in the compiler into assembler language.
7245 @defmac ADDITIONAL_REGISTER_NAMES
7246 If defined, a C initializer for an array of structures containing a name
7247 and a register number. This macro defines additional names for hard
7248 registers, thus allowing the @code{asm} option in declarations to refer
7249 to registers using alternate names.
7252 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7253 Define this macro if you are using an unusual assembler that
7254 requires different names for the machine instructions.
7256 The definition is a C statement or statements which output an
7257 assembler instruction opcode to the stdio stream @var{stream}. The
7258 macro-operand @var{ptr} is a variable of type @code{char *} which
7259 points to the opcode name in its ``internal'' form---the form that is
7260 written in the machine description. The definition should output the
7261 opcode name to @var{stream}, performing any translation you desire, and
7262 increment the variable @var{ptr} to point at the end of the opcode
7263 so that it will not be output twice.
7265 In fact, your macro definition may process less than the entire opcode
7266 name, or more than the opcode name; but if you want to process text
7267 that includes @samp{%}-sequences to substitute operands, you must take
7268 care of the substitution yourself. Just be sure to increment
7269 @var{ptr} over whatever text should not be output normally.
7271 @findex recog_data.operand
7272 If you need to look at the operand values, they can be found as the
7273 elements of @code{recog_data.operand}.
7275 If the macro definition does nothing, the instruction is output
7279 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7280 If defined, a C statement to be executed just prior to the output of
7281 assembler code for @var{insn}, to modify the extracted operands so
7282 they will be output differently.
7284 Here the argument @var{opvec} is the vector containing the operands
7285 extracted from @var{insn}, and @var{noperands} is the number of
7286 elements of the vector which contain meaningful data for this insn.
7287 The contents of this vector are what will be used to convert the insn
7288 template into assembler code, so you can change the assembler output
7289 by changing the contents of the vector.
7291 This macro is useful when various assembler syntaxes share a single
7292 file of instruction patterns; by defining this macro differently, you
7293 can cause a large class of instructions to be output differently (such
7294 as with rearranged operands). Naturally, variations in assembler
7295 syntax affecting individual insn patterns ought to be handled by
7296 writing conditional output routines in those patterns.
7298 If this macro is not defined, it is equivalent to a null statement.
7301 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7302 A C compound statement to output to stdio stream @var{stream} the
7303 assembler syntax for an instruction operand @var{x}. @var{x} is an
7306 @var{code} is a value that can be used to specify one of several ways
7307 of printing the operand. It is used when identical operands must be
7308 printed differently depending on the context. @var{code} comes from
7309 the @samp{%} specification that was used to request printing of the
7310 operand. If the specification was just @samp{%@var{digit}} then
7311 @var{code} is 0; if the specification was @samp{%@var{ltr}
7312 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7315 If @var{x} is a register, this macro should print the register's name.
7316 The names can be found in an array @code{reg_names} whose type is
7317 @code{char *[]}. @code{reg_names} is initialized from
7318 @code{REGISTER_NAMES}.
7320 When the machine description has a specification @samp{%@var{punct}}
7321 (a @samp{%} followed by a punctuation character), this macro is called
7322 with a null pointer for @var{x} and the punctuation character for
7326 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7327 A C expression which evaluates to true if @var{code} is a valid
7328 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7329 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7330 punctuation characters (except for the standard one, @samp{%}) are used
7334 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7335 A C compound statement to output to stdio stream @var{stream} the
7336 assembler syntax for an instruction operand that is a memory reference
7337 whose address is @var{x}. @var{x} is an RTL expression.
7339 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7340 On some machines, the syntax for a symbolic address depends on the
7341 section that the address refers to. On these machines, define the hook
7342 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7343 @code{symbol_ref}, and then check for it here. @xref{Assembler
7347 @findex dbr_sequence_length
7348 @defmac DBR_OUTPUT_SEQEND (@var{file})
7349 A C statement, to be executed after all slot-filler instructions have
7350 been output. If necessary, call @code{dbr_sequence_length} to
7351 determine the number of slots filled in a sequence (zero if not
7352 currently outputting a sequence), to decide how many no-ops to output,
7355 Don't define this macro if it has nothing to do, but it is helpful in
7356 reading assembly output if the extent of the delay sequence is made
7357 explicit (e.g.@: with white space).
7360 @findex final_sequence
7361 Note that output routines for instructions with delay slots must be
7362 prepared to deal with not being output as part of a sequence
7363 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7364 found.) The variable @code{final_sequence} is null when not
7365 processing a sequence, otherwise it contains the @code{sequence} rtx
7369 @defmac REGISTER_PREFIX
7370 @defmacx LOCAL_LABEL_PREFIX
7371 @defmacx USER_LABEL_PREFIX
7372 @defmacx IMMEDIATE_PREFIX
7373 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7374 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7375 @file{final.c}). These are useful when a single @file{md} file must
7376 support multiple assembler formats. In that case, the various @file{tm.h}
7377 files can define these macros differently.
7380 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
7381 If defined this macro should expand to a series of @code{case}
7382 statements which will be parsed inside the @code{switch} statement of
7383 the @code{asm_fprintf} function. This allows targets to define extra
7384 printf formats which may useful when generating their assembler
7385 statements. Note that uppercase letters are reserved for future
7386 generic extensions to asm_fprintf, and so are not available to target
7387 specific code. The output file is given by the parameter @var{file}.
7388 The varargs input pointer is @var{argptr} and the rest of the format
7389 string, starting the character after the one that is being switched
7390 upon, is pointed to by @var{format}.
7393 @defmac ASSEMBLER_DIALECT
7394 If your target supports multiple dialects of assembler language (such as
7395 different opcodes), define this macro as a C expression that gives the
7396 numeric index of the assembler language dialect to use, with zero as the
7399 If this macro is defined, you may use constructs of the form
7401 @samp{@{option0|option1|option2@dots{}@}}
7404 in the output templates of patterns (@pxref{Output Template}) or in the
7405 first argument of @code{asm_fprintf}. This construct outputs
7406 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7407 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7408 within these strings retain their usual meaning. If there are fewer
7409 alternatives within the braces than the value of
7410 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7412 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7413 @samp{@}} do not have any special meaning when used in templates or
7414 operands to @code{asm_fprintf}.
7416 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7417 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7418 the variations in assembler language syntax with that mechanism. Define
7419 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7420 if the syntax variant are larger and involve such things as different
7421 opcodes or operand order.
7424 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7425 A C expression to output to @var{stream} some assembler code
7426 which will push hard register number @var{regno} onto the stack.
7427 The code need not be optimal, since this macro is used only when
7431 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7432 A C expression to output to @var{stream} some assembler code
7433 which will pop hard register number @var{regno} off of the stack.
7434 The code need not be optimal, since this macro is used only when
7438 @node Dispatch Tables
7439 @subsection Output of Dispatch Tables
7441 @c prevent bad page break with this line
7442 This concerns dispatch tables.
7444 @cindex dispatch table
7445 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7446 A C statement to output to the stdio stream @var{stream} an assembler
7447 pseudo-instruction to generate a difference between two labels.
7448 @var{value} and @var{rel} are the numbers of two internal labels. The
7449 definitions of these labels are output using
7450 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7451 way here. For example,
7454 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7455 @var{value}, @var{rel})
7458 You must provide this macro on machines where the addresses in a
7459 dispatch table are relative to the table's own address. If defined, GCC
7460 will also use this macro on all machines when producing PIC@.
7461 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7462 mode and flags can be read.
7465 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7466 This macro should be provided on machines where the addresses
7467 in a dispatch table are absolute.
7469 The definition should be a C statement to output to the stdio stream
7470 @var{stream} an assembler pseudo-instruction to generate a reference to
7471 a label. @var{value} is the number of an internal label whose
7472 definition is output using @code{(*targetm.asm_out.internal_label)}.
7476 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7480 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7481 Define this if the label before a jump-table needs to be output
7482 specially. The first three arguments are the same as for
7483 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7484 jump-table which follows (a @code{jump_insn} containing an
7485 @code{addr_vec} or @code{addr_diff_vec}).
7487 This feature is used on system V to output a @code{swbeg} statement
7490 If this macro is not defined, these labels are output with
7491 @code{(*targetm.asm_out.internal_label)}.
7494 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7495 Define this if something special must be output at the end of a
7496 jump-table. The definition should be a C statement to be executed
7497 after the assembler code for the table is written. It should write
7498 the appropriate code to stdio stream @var{stream}. The argument
7499 @var{table} is the jump-table insn, and @var{num} is the label-number
7500 of the preceding label.
7502 If this macro is not defined, nothing special is output at the end of
7506 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
7507 This target hook emits a label at the beginning of each FDE. It
7508 should be defined on targets where FDEs need special labels, and it
7509 should write the appropriate label, for the FDE associated with the
7510 function declaration @var{decl}, to the stdio stream @var{stream}.
7511 The third argument, @var{for_eh}, is a boolean: true if this is for an
7512 exception table. The fourth argument, @var{empty}, is a boolean:
7513 true if this is a placeholder label for an omitted FDE.
7515 The default is that FDEs are not given nonlocal labels.
7518 @node Exception Region Output
7519 @subsection Assembler Commands for Exception Regions
7521 @c prevent bad page break with this line
7523 This describes commands marking the start and the end of an exception
7526 @defmac EH_FRAME_SECTION_NAME
7527 If defined, a C string constant for the name of the section containing
7528 exception handling frame unwind information. If not defined, GCC will
7529 provide a default definition if the target supports named sections.
7530 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7532 You should define this symbol if your target supports DWARF 2 frame
7533 unwind information and the default definition does not work.
7536 @defmac EH_FRAME_IN_DATA_SECTION
7537 If defined, DWARF 2 frame unwind information will be placed in the
7538 data section even though the target supports named sections. This
7539 might be necessary, for instance, if the system linker does garbage
7540 collection and sections cannot be marked as not to be collected.
7542 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7546 @defmac MASK_RETURN_ADDR
7547 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7548 that it does not contain any extraneous set bits in it.
7551 @defmac DWARF2_UNWIND_INFO
7552 Define this macro to 0 if your target supports DWARF 2 frame unwind
7553 information, but it does not yet work with exception handling.
7554 Otherwise, if your target supports this information (if it defines
7555 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7556 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7559 If this macro is defined to 1, the DWARF 2 unwinder will be the default
7560 exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by
7563 If this macro is defined to anything, the DWARF 2 unwinder will be used
7564 instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case.
7567 @defmac MUST_USE_SJLJ_EXCEPTIONS
7568 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
7569 runtime-variable. In that case, @file{except.h} cannot correctly
7570 determine the corresponding definition of
7571 @code{MUST_USE_SJLJ_EXCEPTIONS}, so the target must provide it directly.
7574 @defmac DWARF_CIE_DATA_ALIGNMENT
7575 This macro need only be defined if the target might save registers in the
7576 function prologue at an offset to the stack pointer that is not aligned to
7577 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7578 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7579 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7580 the target supports DWARF 2 frame unwind information.
7583 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7584 If defined, a function that switches to the section in which the main
7585 exception table is to be placed (@pxref{Sections}). The default is a
7586 function that switches to a section named @code{.gcc_except_table} on
7587 machines that support named sections via
7588 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7589 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7590 @code{readonly_data_section}.
7593 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7594 If defined, a function that switches to the section in which the DWARF 2
7595 frame unwind information to be placed (@pxref{Sections}). The default
7596 is a function that outputs a standard GAS section directive, if
7597 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7598 directive followed by a synthetic label.
7601 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7602 Contains the value true if the target should add a zero word onto the
7603 end of a Dwarf-2 frame info section when used for exception handling.
7604 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7608 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
7609 Given a register, this hook should return a parallel of registers to
7610 represent where to find the register pieces. Define this hook if the
7611 register and its mode are represented in Dwarf in non-contiguous
7612 locations, or if the register should be represented in more than one
7613 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
7614 If not defined, the default is to return @code{NULL_RTX}.
7617 @node Alignment Output
7618 @subsection Assembler Commands for Alignment
7620 @c prevent bad page break with this line
7621 This describes commands for alignment.
7623 @defmac JUMP_ALIGN (@var{label})
7624 The alignment (log base 2) to put in front of @var{label}, which is
7625 a common destination of jumps and has no fallthru incoming edge.
7627 This macro need not be defined if you don't want any special alignment
7628 to be done at such a time. Most machine descriptions do not currently
7631 Unless it's necessary to inspect the @var{label} parameter, it is better
7632 to set the variable @var{align_jumps} in the target's
7633 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7634 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7637 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
7638 The alignment (log base 2) to put in front of @var{label}, which follows
7641 This macro need not be defined if you don't want any special alignment
7642 to be done at such a time. Most machine descriptions do not currently
7646 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7647 The maximum number of bytes to skip when applying
7648 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7649 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7652 @defmac LOOP_ALIGN (@var{label})
7653 The alignment (log base 2) to put in front of @var{label}, which follows
7654 a @code{NOTE_INSN_LOOP_BEG} note.
7656 This macro need not be defined if you don't want any special alignment
7657 to be done at such a time. Most machine descriptions do not currently
7660 Unless it's necessary to inspect the @var{label} parameter, it is better
7661 to set the variable @code{align_loops} in the target's
7662 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7663 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7666 @defmac LOOP_ALIGN_MAX_SKIP
7667 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7668 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7671 @defmac LABEL_ALIGN (@var{label})
7672 The alignment (log base 2) to put in front of @var{label}.
7673 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7674 the maximum of the specified values is used.
7676 Unless it's necessary to inspect the @var{label} parameter, it is better
7677 to set the variable @code{align_labels} in the target's
7678 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7679 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7682 @defmac LABEL_ALIGN_MAX_SKIP
7683 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7684 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7687 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7688 A C statement to output to the stdio stream @var{stream} an assembler
7689 instruction to advance the location counter by @var{nbytes} bytes.
7690 Those bytes should be zero when loaded. @var{nbytes} will be a C
7691 expression of type @code{int}.
7694 @defmac ASM_NO_SKIP_IN_TEXT
7695 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7696 text section because it fails to put zeros in the bytes that are skipped.
7697 This is true on many Unix systems, where the pseudo--op to skip bytes
7698 produces no-op instructions rather than zeros when used in the text
7702 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7703 A C statement to output to the stdio stream @var{stream} an assembler
7704 command to advance the location counter to a multiple of 2 to the
7705 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7708 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7709 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7710 for padding, if necessary.
7713 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7714 A C statement to output to the stdio stream @var{stream} an assembler
7715 command to advance the location counter to a multiple of 2 to the
7716 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7717 satisfy the alignment request. @var{power} and @var{max_skip} will be
7718 a C expression of type @code{int}.
7722 @node Debugging Info
7723 @section Controlling Debugging Information Format
7725 @c prevent bad page break with this line
7726 This describes how to specify debugging information.
7729 * All Debuggers:: Macros that affect all debugging formats uniformly.
7730 * DBX Options:: Macros enabling specific options in DBX format.
7731 * DBX Hooks:: Hook macros for varying DBX format.
7732 * File Names and DBX:: Macros controlling output of file names in DBX format.
7733 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7734 * VMS Debug:: Macros for VMS debug format.
7738 @subsection Macros Affecting All Debugging Formats
7740 @c prevent bad page break with this line
7741 These macros affect all debugging formats.
7743 @defmac DBX_REGISTER_NUMBER (@var{regno})
7744 A C expression that returns the DBX register number for the compiler
7745 register number @var{regno}. In the default macro provided, the value
7746 of this expression will be @var{regno} itself. But sometimes there are
7747 some registers that the compiler knows about and DBX does not, or vice
7748 versa. In such cases, some register may need to have one number in the
7749 compiler and another for DBX@.
7751 If two registers have consecutive numbers inside GCC, and they can be
7752 used as a pair to hold a multiword value, then they @emph{must} have
7753 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7754 Otherwise, debuggers will be unable to access such a pair, because they
7755 expect register pairs to be consecutive in their own numbering scheme.
7757 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7758 does not preserve register pairs, then what you must do instead is
7759 redefine the actual register numbering scheme.
7762 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
7763 A C expression that returns the integer offset value for an automatic
7764 variable having address @var{x} (an RTL expression). The default
7765 computation assumes that @var{x} is based on the frame-pointer and
7766 gives the offset from the frame-pointer. This is required for targets
7767 that produce debugging output for DBX or COFF-style debugging output
7768 for SDB and allow the frame-pointer to be eliminated when the
7769 @option{-g} options is used.
7772 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7773 A C expression that returns the integer offset value for an argument
7774 having address @var{x} (an RTL expression). The nominal offset is
7778 @defmac PREFERRED_DEBUGGING_TYPE
7779 A C expression that returns the type of debugging output GCC should
7780 produce when the user specifies just @option{-g}. Define
7781 this if you have arranged for GCC to support more than one format of
7782 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7783 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7784 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7786 When the user specifies @option{-ggdb}, GCC normally also uses the
7787 value of this macro to select the debugging output format, but with two
7788 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
7789 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7790 defined, GCC uses @code{DBX_DEBUG}.
7792 The value of this macro only affects the default debugging output; the
7793 user can always get a specific type of output by using @option{-gstabs},
7794 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
7798 @subsection Specific Options for DBX Output
7800 @c prevent bad page break with this line
7801 These are specific options for DBX output.
7803 @defmac DBX_DEBUGGING_INFO
7804 Define this macro if GCC should produce debugging output for DBX
7805 in response to the @option{-g} option.
7808 @defmac XCOFF_DEBUGGING_INFO
7809 Define this macro if GCC should produce XCOFF format debugging output
7810 in response to the @option{-g} option. This is a variant of DBX format.
7813 @defmac DEFAULT_GDB_EXTENSIONS
7814 Define this macro to control whether GCC should by default generate
7815 GDB's extended version of DBX debugging information (assuming DBX-format
7816 debugging information is enabled at all). If you don't define the
7817 macro, the default is 1: always generate the extended information
7818 if there is any occasion to.
7821 @defmac DEBUG_SYMS_TEXT
7822 Define this macro if all @code{.stabs} commands should be output while
7823 in the text section.
7826 @defmac ASM_STABS_OP
7827 A C string constant, including spacing, naming the assembler pseudo op to
7828 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7829 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7830 applies only to DBX debugging information format.
7833 @defmac ASM_STABD_OP
7834 A C string constant, including spacing, naming the assembler pseudo op to
7835 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7836 value is the current location. If you don't define this macro,
7837 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7841 @defmac ASM_STABN_OP
7842 A C string constant, including spacing, naming the assembler pseudo op to
7843 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7844 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7845 macro applies only to DBX debugging information format.
7848 @defmac DBX_NO_XREFS
7849 Define this macro if DBX on your system does not support the construct
7850 @samp{xs@var{tagname}}. On some systems, this construct is used to
7851 describe a forward reference to a structure named @var{tagname}.
7852 On other systems, this construct is not supported at all.
7855 @defmac DBX_CONTIN_LENGTH
7856 A symbol name in DBX-format debugging information is normally
7857 continued (split into two separate @code{.stabs} directives) when it
7858 exceeds a certain length (by default, 80 characters). On some
7859 operating systems, DBX requires this splitting; on others, splitting
7860 must not be done. You can inhibit splitting by defining this macro
7861 with the value zero. You can override the default splitting-length by
7862 defining this macro as an expression for the length you desire.
7865 @defmac DBX_CONTIN_CHAR
7866 Normally continuation is indicated by adding a @samp{\} character to
7867 the end of a @code{.stabs} string when a continuation follows. To use
7868 a different character instead, define this macro as a character
7869 constant for the character you want to use. Do not define this macro
7870 if backslash is correct for your system.
7873 @defmac DBX_STATIC_STAB_DATA_SECTION
7874 Define this macro if it is necessary to go to the data section before
7875 outputting the @samp{.stabs} pseudo-op for a non-global static
7879 @defmac DBX_TYPE_DECL_STABS_CODE
7880 The value to use in the ``code'' field of the @code{.stabs} directive
7881 for a typedef. The default is @code{N_LSYM}.
7884 @defmac DBX_STATIC_CONST_VAR_CODE
7885 The value to use in the ``code'' field of the @code{.stabs} directive
7886 for a static variable located in the text section. DBX format does not
7887 provide any ``right'' way to do this. The default is @code{N_FUN}.
7890 @defmac DBX_REGPARM_STABS_CODE
7891 The value to use in the ``code'' field of the @code{.stabs} directive
7892 for a parameter passed in registers. DBX format does not provide any
7893 ``right'' way to do this. The default is @code{N_RSYM}.
7896 @defmac DBX_REGPARM_STABS_LETTER
7897 The letter to use in DBX symbol data to identify a symbol as a parameter
7898 passed in registers. DBX format does not customarily provide any way to
7899 do this. The default is @code{'P'}.
7902 @defmac DBX_MEMPARM_STABS_LETTER
7903 The letter to use in DBX symbol data to identify a symbol as a stack
7904 parameter. The default is @code{'p'}.
7907 @defmac DBX_FUNCTION_FIRST
7908 Define this macro if the DBX information for a function and its
7909 arguments should precede the assembler code for the function. Normally,
7910 in DBX format, the debugging information entirely follows the assembler
7914 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
7915 Define this macro if the value of a symbol describing the scope of a
7916 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7917 of the enclosing function. Normally, GCC uses an absolute address.
7920 @defmac DBX_USE_BINCL
7921 Define this macro if GCC should generate @code{N_BINCL} and
7922 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7923 macro also directs GCC to output a type number as a pair of a file
7924 number and a type number within the file. Normally, GCC does not
7925 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7926 number for a type number.
7930 @subsection Open-Ended Hooks for DBX Format
7932 @c prevent bad page break with this line
7933 These are hooks for DBX format.
7935 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7936 Define this macro to say how to output to @var{stream} the debugging
7937 information for the start of a scope level for variable names. The
7938 argument @var{name} is the name of an assembler symbol (for use with
7939 @code{assemble_name}) whose value is the address where the scope begins.
7942 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7943 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7946 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
7947 Define this macro if the target machine requires special handling to
7948 output an @code{N_FUN} entry for the function @var{decl}.
7951 @defmac DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7952 Define this macro if the target machine requires special output at the
7953 end of the debugging information for a function. The definition should
7954 be a C statement (sans semicolon) to output the appropriate information
7955 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7959 @defmac NO_DBX_FUNCTION_END
7960 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7961 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
7962 On those machines, define this macro to turn this feature off without
7963 disturbing the rest of the gdb extensions.
7966 @node File Names and DBX
7967 @subsection File Names in DBX Format
7969 @c prevent bad page break with this line
7970 This describes file names in DBX format.
7972 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7973 A C statement to output DBX debugging information to the stdio stream
7974 @var{stream} which indicates that file @var{name} is the main source
7975 file---the file specified as the input file for compilation.
7976 This macro is called only once, at the beginning of compilation.
7978 This macro need not be defined if the standard form of output
7979 for DBX debugging information is appropriate.
7982 @defmac DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
7983 A C statement to output DBX debugging information to the stdio stream
7984 @var{stream} which indicates that the current directory during
7985 compilation is named @var{name}.
7987 This macro need not be defined if the standard form of output
7988 for DBX debugging information is appropriate.
7991 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7992 A C statement to output DBX debugging information at the end of
7993 compilation of the main source file @var{name}.
7995 If you don't define this macro, nothing special is output at the end
7996 of compilation, which is correct for most machines.
8001 @subsection Macros for SDB and DWARF Output
8003 @c prevent bad page break with this line
8004 Here are macros for SDB and DWARF output.
8006 @defmac SDB_DEBUGGING_INFO
8007 Define this macro if GCC should produce COFF-style debugging output
8008 for SDB in response to the @option{-g} option.
8011 @defmac DWARF2_DEBUGGING_INFO
8012 Define this macro if GCC should produce dwarf version 2 format
8013 debugging output in response to the @option{-g} option.
8015 To support optional call frame debugging information, you must also
8016 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8017 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8018 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8019 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8022 @defmac DWARF2_FRAME_INFO
8023 Define this macro to a nonzero value if GCC should always output
8024 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8025 (@pxref{Exception Region Output} is nonzero, GCC will output this
8026 information not matter how you define @code{DWARF2_FRAME_INFO}.
8029 @defmac DWARF2_GENERATE_TEXT_SECTION_LABEL
8030 By default, the Dwarf 2 debugging information generator will generate a
8031 label to mark the beginning of the text section. If it is better simply
8032 to use the name of the text section itself, rather than an explicit label,
8033 to indicate the beginning of the text section, define this macro to zero.
8036 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8037 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8038 line debug info sections. This will result in much more compact line number
8039 tables, and hence is desirable if it works.
8042 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8043 A C statement to issue assembly directives that create a difference
8044 between the two given labels, using an integer of the given size.
8047 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8048 A C statement to issue assembly directives that create a
8049 section-relative reference to the given label, using an integer of the
8053 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8054 A C statement to issue assembly directives that create a self-relative
8055 reference to the given label, using an integer of the given size.
8058 @defmac PUT_SDB_@dots{}
8059 Define these macros to override the assembler syntax for the special
8060 SDB assembler directives. See @file{sdbout.c} for a list of these
8061 macros and their arguments. If the standard syntax is used, you need
8062 not define them yourself.
8066 Some assemblers do not support a semicolon as a delimiter, even between
8067 SDB assembler directives. In that case, define this macro to be the
8068 delimiter to use (usually @samp{\n}). It is not necessary to define
8069 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8073 @defmac SDB_GENERATE_FAKE
8074 Define this macro to override the usual method of constructing a dummy
8075 name for anonymous structure and union types. See @file{sdbout.c} for
8079 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8080 Define this macro to allow references to unknown structure,
8081 union, or enumeration tags to be emitted. Standard COFF does not
8082 allow handling of unknown references, MIPS ECOFF has support for
8086 @defmac SDB_ALLOW_FORWARD_REFERENCES
8087 Define this macro to allow references to structure, union, or
8088 enumeration tags that have not yet been seen to be handled. Some
8089 assemblers choke if forward tags are used, while some require it.
8094 @subsection Macros for VMS Debug Format
8096 @c prevent bad page break with this line
8097 Here are macros for VMS debug format.
8099 @defmac VMS_DEBUGGING_INFO
8100 Define this macro if GCC should produce debugging output for VMS
8101 in response to the @option{-g} option. The default behavior for VMS
8102 is to generate minimal debug info for a traceback in the absence of
8103 @option{-g} unless explicitly overridden with @option{-g0}. This
8104 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8105 @code{OVERRIDE_OPTIONS}.
8108 @node Floating Point
8109 @section Cross Compilation and Floating Point
8110 @cindex cross compilation and floating point
8111 @cindex floating point and cross compilation
8113 While all modern machines use twos-complement representation for integers,
8114 there are a variety of representations for floating point numbers. This
8115 means that in a cross-compiler the representation of floating point numbers
8116 in the compiled program may be different from that used in the machine
8117 doing the compilation.
8119 Because different representation systems may offer different amounts of
8120 range and precision, all floating point constants must be represented in
8121 the target machine's format. Therefore, the cross compiler cannot
8122 safely use the host machine's floating point arithmetic; it must emulate
8123 the target's arithmetic. To ensure consistency, GCC always uses
8124 emulation to work with floating point values, even when the host and
8125 target floating point formats are identical.
8127 The following macros are provided by @file{real.h} for the compiler to
8128 use. All parts of the compiler which generate or optimize
8129 floating-point calculations must use these macros. They may evaluate
8130 their operands more than once, so operands must not have side effects.
8132 @defmac REAL_VALUE_TYPE
8133 The C data type to be used to hold a floating point value in the target
8134 machine's format. Typically this is a @code{struct} containing an
8135 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8139 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8140 Compares for equality the two values, @var{x} and @var{y}. If the target
8141 floating point format supports negative zeroes and/or NaNs,
8142 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8143 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8146 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8147 Tests whether @var{x} is less than @var{y}.
8150 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8151 Truncates @var{x} to a signed integer, rounding toward zero.
8154 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8155 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8156 @var{x} is negative, returns zero.
8159 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8160 Converts @var{string} into a floating point number in the target machine's
8161 representation for mode @var{mode}. This routine can handle both
8162 decimal and hexadecimal floating point constants, using the syntax
8163 defined by the C language for both.
8166 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8167 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8170 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8171 Determines whether @var{x} represents infinity (positive or negative).
8174 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8175 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8178 @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})
8179 Calculates an arithmetic operation on the two floating point values
8180 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8183 The operation to be performed is specified by @var{code}. Only the
8184 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8185 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8187 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8188 target's floating point format cannot represent infinity, it will call
8189 @code{abort}. Callers should check for this situation first, using
8190 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8193 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8194 Returns the negative of the floating point value @var{x}.
8197 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8198 Returns the absolute value of @var{x}.
8201 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8202 Truncates the floating point value @var{x} to fit in @var{mode}. The
8203 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8204 appropriate bit pattern to be output asa floating constant whose
8205 precision accords with mode @var{mode}.
8208 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8209 Converts a floating point value @var{x} into a double-precision integer
8210 which is then stored into @var{low} and @var{high}. If the value is not
8211 integral, it is truncated.
8214 @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})
8215 Converts a double-precision integer found in @var{low} and @var{high},
8216 into a floating point value which is then stored into @var{x}. The
8217 value is truncated to fit in mode @var{mode}.
8220 @node Mode Switching
8221 @section Mode Switching Instructions
8222 @cindex mode switching
8223 The following macros control mode switching optimizations:
8225 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8226 Define this macro if the port needs extra instructions inserted for mode
8227 switching in an optimizing compilation.
8229 For an example, the SH4 can perform both single and double precision
8230 floating point operations, but to perform a single precision operation,
8231 the FPSCR PR bit has to be cleared, while for a double precision
8232 operation, this bit has to be set. Changing the PR bit requires a general
8233 purpose register as a scratch register, hence these FPSCR sets have to
8234 be inserted before reload, i.e.@: you can't put this into instruction emitting
8235 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
8237 You can have multiple entities that are mode-switched, and select at run time
8238 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8239 return nonzero for any @var{entity} that needs mode-switching.
8240 If you define this macro, you also have to define
8241 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8242 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8243 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
8247 @defmac NUM_MODES_FOR_MODE_SWITCHING
8248 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8249 initializer for an array of integers. Each initializer element
8250 N refers to an entity that needs mode switching, and specifies the number
8251 of different modes that might need to be set for this entity.
8252 The position of the initializer in the initializer - starting counting at
8253 zero - determines the integer that is used to refer to the mode-switched
8255 In macros that take mode arguments / yield a mode result, modes are
8256 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8257 switch is needed / supplied.
8260 @defmac MODE_NEEDED (@var{entity}, @var{insn})
8261 @var{entity} is an integer specifying a mode-switched entity. If
8262 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8263 return an integer value not larger than the corresponding element in
8264 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8265 be switched into prior to the execution of @var{insn}.
8268 @defmac MODE_AFTER (@var{mode}, @var{insn})
8269 If this macro is defined, it is evaluated for every @var{insn} during
8270 mode switching. It determines the mode that an insn results in (if
8271 different from the incoming mode).
8274 @defmac MODE_ENTRY (@var{entity})
8275 If this macro is defined, it is evaluated for every @var{entity} that needs
8276 mode switching. It should evaluate to an integer, which is a mode that
8277 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
8278 is defined then @code{MODE_EXIT} must be defined.
8281 @defmac MODE_EXIT (@var{entity})
8282 If this macro is defined, it is evaluated for every @var{entity} that needs
8283 mode switching. It should evaluate to an integer, which is a mode that
8284 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
8285 is defined then @code{MODE_ENTRY} must be defined.
8288 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8289 This macro specifies the order in which modes for @var{entity} are processed.
8290 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8291 lowest. The value of the macro should be an integer designating a mode
8292 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8293 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8294 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8297 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8298 Generate one or more insns to set @var{entity} to @var{mode}.
8299 @var{hard_reg_live} is the set of hard registers live at the point where
8300 the insn(s) are to be inserted.
8303 @node Target Attributes
8304 @section Defining target-specific uses of @code{__attribute__}
8305 @cindex target attributes
8306 @cindex machine attributes
8307 @cindex attributes, target-specific
8309 Target-specific attributes may be defined for functions, data and types.
8310 These are described using the following target hooks; they also need to
8311 be documented in @file{extend.texi}.
8313 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8314 If defined, this target hook points to an array of @samp{struct
8315 attribute_spec} (defined in @file{tree.h}) specifying the machine
8316 specific attributes for this target and some of the restrictions on the
8317 entities to which these attributes are applied and the arguments they
8321 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8322 If defined, this target hook is a function which returns zero if the attributes on
8323 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8324 and two if they are nearly compatible (which causes a warning to be
8325 generated). If this is not defined, machine-specific attributes are
8326 supposed always to be compatible.
8329 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8330 If defined, this target hook is a function which assigns default attributes to
8331 newly defined @var{type}.
8334 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8335 Define this target hook if the merging of type attributes needs special
8336 handling. If defined, the result is a list of the combined
8337 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8338 that @code{comptypes} has already been called and returned 1. This
8339 function may call @code{merge_attributes} to handle machine-independent
8343 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8344 Define this target hook if the merging of decl attributes needs special
8345 handling. If defined, the result is a list of the combined
8346 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8347 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8348 when this is needed are when one attribute overrides another, or when an
8349 attribute is nullified by a subsequent definition. This function may
8350 call @code{merge_attributes} to handle machine-independent merging.
8352 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8353 If the only target-specific handling you require is @samp{dllimport} for
8354 Microsoft Windows targets, you should define the macro
8355 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function
8356 called @code{merge_dllimport_decl_attributes} which can then be defined
8357 as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done
8358 in @file{i386/cygwin.h} and @file{i386/i386.c}, for example.
8361 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8362 Define this target hook if you want to be able to add attributes to a decl
8363 when it is being created. This is normally useful for back ends which
8364 wish to implement a pragma by using the attributes which correspond to
8365 the pragma's effect. The @var{node} argument is the decl which is being
8366 created. The @var{attr_ptr} argument is a pointer to the attribute list
8367 for this decl. The list itself should not be modified, since it may be
8368 shared with other decls, but attributes may be chained on the head of
8369 the list and @code{*@var{attr_ptr}} modified to point to the new
8370 attributes, or a copy of the list may be made if further changes are
8374 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8376 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8377 into the current function, despite its having target-specific
8378 attributes, @code{false} otherwise. By default, if a function has a
8379 target specific attribute attached to it, it will not be inlined.
8382 @node MIPS Coprocessors
8383 @section Defining coprocessor specifics for MIPS targets.
8384 @cindex MIPS coprocessor-definition macros
8386 The MIPS specification allows MIPS implementations to have as many as 4
8387 coprocessors, each with as many as 32 private registers. GCC supports
8388 accessing these registers and transferring values between the registers
8389 and memory using asm-ized variables. For example:
8392 register unsigned int cp0count asm ("c0r1");
8398 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8399 names may be added as described below, or the default names may be
8400 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8402 Coprocessor registers are assumed to be epilogue-used; sets to them will
8403 be preserved even if it does not appear that the register is used again
8404 later in the function.
8406 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8407 the FPU. One accesses COP1 registers through standard mips
8408 floating-point support; they are not included in this mechanism.
8410 There is one macro used in defining the MIPS coprocessor interface which
8411 you may want to override in subtargets; it is described below.
8413 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
8414 A comma-separated list (with leading comma) of pairs describing the
8415 alternate names of coprocessor registers. The format of each entry should be
8417 @{ @var{alternatename}, @var{register_number}@}
8423 @section Parameters for Precompiled Header Validity Checking
8424 @cindex parameters, precompiled headers
8426 @deftypefn {Target Hook} void * TARGET_GET_PCH_VALIDITY (size_t * @var{sz})
8427 Define this hook if your target needs to check a different collection
8428 of flags than the default, which is every flag defined by
8429 @code{TARGET_SWITCHES} and @code{TARGET_OPTIONS}. It should return
8430 some data which will be saved in the PCH file and presented to
8431 @code{TARGET_PCH_VALID_P} later; it should set @code{SZ} to the size
8435 @deftypefn {Target Hook} const char * TARGET_PCH_VALID_P (const void * @var{data}, size_t @var{sz})
8436 Define this hook if your target needs to check a different collection of
8437 flags than the default, which is every flag defined by @code{TARGET_SWITCHES}
8438 and @code{TARGET_OPTIONS}. It is given data which came from
8439 @code{TARGET_GET_PCH_VALIDITY} (in this version of this compiler, so there
8440 is no need for extensive validity checking). It returns @code{NULL} if
8441 it is safe to load a PCH file with this data, or a suitable error message
8442 if not. The error message will be presented to the user, so it should
8447 @section Miscellaneous Parameters
8448 @cindex parameters, miscellaneous
8450 @c prevent bad page break with this line
8451 Here are several miscellaneous parameters.
8453 @defmac PREDICATE_CODES
8454 Define this if you have defined special-purpose predicates in the file
8455 @file{@var{machine}.c}. This macro is called within an initializer of an
8456 array of structures. The first field in the structure is the name of a
8457 predicate and the second field is an array of rtl codes. For each
8458 predicate, list all rtl codes that can be in expressions matched by the
8459 predicate. The list should have a trailing comma. Here is an example
8460 of two entries in the list for a typical RISC machine:
8463 #define PREDICATE_CODES \
8464 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8465 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8468 Defining this macro does not affect the generated code (however,
8469 incorrect definitions that omit an rtl code that may be matched by the
8470 predicate can cause the compiler to malfunction). Instead, it allows
8471 the table built by @file{genrecog} to be more compact and efficient,
8472 thus speeding up the compiler. The most important predicates to include
8473 in the list specified by this macro are those used in the most insn
8476 For each predicate function named in @code{PREDICATE_CODES}, a
8477 declaration will be generated in @file{insn-codes.h}.
8480 @defmac HAS_LONG_COND_BRANCH
8481 Define this boolean macro to indicate whether or not your architecture
8482 has conditional branches that can span all of memory. It is used in
8483 conjunction with an optimization that partitions hot and cold basic
8484 blocks into separate sections of the executable. If this macro is
8485 set to false, gcc will convert any conditional branches that attempt
8486 to cross between sections into unconditional branches or indirect jumps.
8489 @defmac HAS_LONG_UNCOND_BRANCH
8490 Define this boolean macro to indicate whether or not your architecture
8491 has unconditional branches that can span all of memory. It is used in
8492 conjunction with an optimization that partitions hot and cold basic
8493 blocks into separate sections of the executable. If this macro is
8494 set to false, gcc will convert any unconditional branches that attempt
8495 to cross between sections into indirect jumps.
8498 @defmac SPECIAL_MODE_PREDICATES
8499 Define this if you have special predicates that know special things
8500 about modes. Genrecog will warn about certain forms of
8501 @code{match_operand} without a mode; if the operand predicate is
8502 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8505 Here is an example from the IA-32 port (@code{ext_register_operand}
8506 specially checks for @code{HImode} or @code{SImode} in preparation
8507 for a byte extraction from @code{%ah} etc.).
8510 #define SPECIAL_MODE_PREDICATES \
8511 "ext_register_operand",
8515 @defmac CASE_VECTOR_MODE
8516 An alias for a machine mode name. This is the machine mode that
8517 elements of a jump-table should have.
8520 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8521 Optional: return the preferred mode for an @code{addr_diff_vec}
8522 when the minimum and maximum offset are known. If you define this,
8523 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8524 To make this work, you also have to define @code{INSN_ALIGN} and
8525 make the alignment for @code{addr_diff_vec} explicit.
8526 The @var{body} argument is provided so that the offset_unsigned and scale
8527 flags can be updated.
8530 @defmac CASE_VECTOR_PC_RELATIVE
8531 Define this macro to be a C expression to indicate when jump-tables
8532 should contain relative addresses. You need not define this macro if
8533 jump-tables never contain relative addresses, or jump-tables should
8534 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
8538 @defmac CASE_DROPS_THROUGH
8539 Define this if control falls through a @code{case} insn when the index
8540 value is out of range. This means the specified default-label is
8541 actually ignored by the @code{case} insn proper.
8544 @defmac CASE_VALUES_THRESHOLD
8545 Define this to be the smallest number of different values for which it
8546 is best to use a jump-table instead of a tree of conditional branches.
8547 The default is four for machines with a @code{casesi} instruction and
8548 five otherwise. This is best for most machines.
8551 @defmac CASE_USE_BIT_TESTS
8552 Define this macro to be a C expression to indicate whether C switch
8553 statements may be implemented by a sequence of bit tests. This is
8554 advantageous on processors that can efficiently implement left shift
8555 of 1 by the number of bits held in a register, but inappropriate on
8556 targets that would require a loop. By default, this macro returns
8557 @code{true} if the target defines an @code{ashlsi3} pattern, and
8558 @code{false} otherwise.
8561 @defmac WORD_REGISTER_OPERATIONS
8562 Define this macro if operations between registers with integral mode
8563 smaller than a word are always performed on the entire register.
8564 Most RISC machines have this property and most CISC machines do not.
8567 @defmac LOAD_EXTEND_OP (@var{mem_mode})
8568 Define this macro to be a C expression indicating when insns that read
8569 memory in @var{mem_mode}, an integral mode narrower than a word, set the
8570 bits outside of @var{mem_mode} to be either the sign-extension or the
8571 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8572 of @var{mem_mode} for which the
8573 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8574 @code{NIL} for other modes.
8576 This macro is not called with @var{mem_mode} non-integral or with a width
8577 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8578 value in this case. Do not define this macro if it would always return
8579 @code{NIL}. On machines where this macro is defined, you will normally
8580 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8582 You may return a non-@code{NIL} value even if for some hard registers
8583 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
8584 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
8585 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
8586 integral mode larger than this but not larger than @code{word_mode}.
8588 You must return @code{NIL} if for some hard registers that allow this
8589 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
8590 @code{word_mode}, but that they can change to another integral mode that
8591 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
8594 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
8595 Define this macro if loading short immediate values into registers sign
8599 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
8600 Define this macro if the same instructions that convert a floating
8601 point number to a signed fixed point number also convert validly to an
8606 The maximum number of bytes that a single instruction can move quickly
8607 between memory and registers or between two memory locations.
8610 @defmac MAX_MOVE_MAX
8611 The maximum number of bytes that a single instruction can move quickly
8612 between memory and registers or between two memory locations. If this
8613 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8614 constant value that is the largest value that @code{MOVE_MAX} can have
8618 @defmac SHIFT_COUNT_TRUNCATED
8619 A C expression that is nonzero if on this machine the number of bits
8620 actually used for the count of a shift operation is equal to the number
8621 of bits needed to represent the size of the object being shifted. When
8622 this macro is nonzero, the compiler will assume that it is safe to omit
8623 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8624 truncates the count of a shift operation. On machines that have
8625 instructions that act on bit-fields at variable positions, which may
8626 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8627 also enables deletion of truncations of the values that serve as
8628 arguments to bit-field instructions.
8630 If both types of instructions truncate the count (for shifts) and
8631 position (for bit-field operations), or if no variable-position bit-field
8632 instructions exist, you should define this macro.
8634 However, on some machines, such as the 80386 and the 680x0, truncation
8635 only applies to shift operations and not the (real or pretended)
8636 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8637 such machines. Instead, add patterns to the @file{md} file that include
8638 the implied truncation of the shift instructions.
8640 You need not define this macro if it would always have the value of zero.
8643 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8644 A C expression which is nonzero if on this machine it is safe to
8645 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8646 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8647 operating on it as if it had only @var{outprec} bits.
8649 On many machines, this expression can be 1.
8651 @c rearranged this, removed the phrase "it is reported that". this was
8652 @c to fix an overfull hbox. --mew 10feb93
8653 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8654 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8655 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8656 such cases may improve things.
8659 @defmac STORE_FLAG_VALUE
8660 A C expression describing the value returned by a comparison operator
8661 with an integral mode and stored by a store-flag instruction
8662 (@samp{s@var{cond}}) when the condition is true. This description must
8663 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8664 comparison operators whose results have a @code{MODE_INT} mode.
8666 A value of 1 or @minus{}1 means that the instruction implementing the
8667 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8668 and 0 when the comparison is false. Otherwise, the value indicates
8669 which bits of the result are guaranteed to be 1 when the comparison is
8670 true. This value is interpreted in the mode of the comparison
8671 operation, which is given by the mode of the first operand in the
8672 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8673 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8676 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8677 generate code that depends only on the specified bits. It can also
8678 replace comparison operators with equivalent operations if they cause
8679 the required bits to be set, even if the remaining bits are undefined.
8680 For example, on a machine whose comparison operators return an
8681 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8682 @samp{0x80000000}, saying that just the sign bit is relevant, the
8686 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8693 (ashift:SI @var{x} (const_int @var{n}))
8697 where @var{n} is the appropriate shift count to move the bit being
8698 tested into the sign bit.
8700 There is no way to describe a machine that always sets the low-order bit
8701 for a true value, but does not guarantee the value of any other bits,
8702 but we do not know of any machine that has such an instruction. If you
8703 are trying to port GCC to such a machine, include an instruction to
8704 perform a logical-and of the result with 1 in the pattern for the
8705 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8707 Often, a machine will have multiple instructions that obtain a value
8708 from a comparison (or the condition codes). Here are rules to guide the
8709 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8714 Use the shortest sequence that yields a valid definition for
8715 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8716 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8717 comparison operators to do so because there may be opportunities to
8718 combine the normalization with other operations.
8721 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8722 slightly preferred on machines with expensive jumps and 1 preferred on
8726 As a second choice, choose a value of @samp{0x80000001} if instructions
8727 exist that set both the sign and low-order bits but do not define the
8731 Otherwise, use a value of @samp{0x80000000}.
8734 Many machines can produce both the value chosen for
8735 @code{STORE_FLAG_VALUE} and its negation in the same number of
8736 instructions. On those machines, you should also define a pattern for
8737 those cases, e.g., one matching
8740 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8743 Some machines can also perform @code{and} or @code{plus} operations on
8744 condition code values with less instructions than the corresponding
8745 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8746 machines, define the appropriate patterns. Use the names @code{incscc}
8747 and @code{decscc}, respectively, for the patterns which perform
8748 @code{plus} or @code{minus} operations on condition code values. See
8749 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8750 find such instruction sequences on other machines.
8752 If this macro is not defined, the default value, 1, is used. You need
8753 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8754 instructions, or if the value generated by these instructions is 1.
8757 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
8758 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8759 returned when comparison operators with floating-point results are true.
8760 Define this macro on machine that have comparison operations that return
8761 floating-point values. If there are no such operations, do not define
8765 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8766 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8767 A C expression that evaluates to true if the architecture defines a value
8768 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
8769 should be set to this value. If this macro is not defined, the value of
8770 @code{clz} or @code{ctz} is assumed to be undefined.
8772 This macro must be defined if the target's expansion for @code{ffs}
8773 relies on a particular value to get correct results. Otherwise it
8774 is not necessary, though it may be used to optimize some corner cases.
8776 Note that regardless of this macro the ``definedness'' of @code{clz}
8777 and @code{ctz} at zero do @emph{not} extend to the builtin functions
8778 visible to the user. Thus one may be free to adjust the value at will
8779 to match the target expansion of these operations without fear of
8784 An alias for the machine mode for pointers. On most machines, define
8785 this to be the integer mode corresponding to the width of a hardware
8786 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8787 On some machines you must define this to be one of the partial integer
8788 modes, such as @code{PSImode}.
8790 The width of @code{Pmode} must be at least as large as the value of
8791 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8792 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8796 @defmac FUNCTION_MODE
8797 An alias for the machine mode used for memory references to functions
8798 being called, in @code{call} RTL expressions. On most machines this
8799 should be @code{QImode}.
8802 @defmac STDC_0_IN_SYSTEM_HEADERS
8803 In normal operation, the preprocessor expands @code{__STDC__} to the
8804 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8805 hosts, like Solaris, the system compiler uses a different convention,
8806 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8807 strict conformance to the C Standard.
8809 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8810 convention when processing system header files, but when processing user
8811 files @code{__STDC__} will always expand to 1.
8814 @defmac NO_IMPLICIT_EXTERN_C
8815 Define this macro if the system header files support C++ as well as C@.
8816 This macro inhibits the usual method of using system header files in
8817 C++, which is to pretend that the file's contents are enclosed in
8818 @samp{extern "C" @{@dots{}@}}.
8823 @defmac REGISTER_TARGET_PRAGMAS ()
8824 Define this macro if you want to implement any target-specific pragmas.
8825 If defined, it is a C expression which makes a series of calls to
8826 @code{c_register_pragma} for each pragma. The macro may also do any
8827 setup required for the pragmas.
8829 The primary reason to define this macro is to provide compatibility with
8830 other compilers for the same target. In general, we discourage
8831 definition of target-specific pragmas for GCC@.
8833 If the pragma can be implemented by attributes then you should consider
8834 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8836 Preprocessor macros that appear on pragma lines are not expanded. All
8837 @samp{#pragma} directives that do not match any registered pragma are
8838 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8841 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
8843 Each call to @code{c_register_pragma} establishes one pragma. The
8844 @var{callback} routine will be called when the preprocessor encounters a
8848 #pragma [@var{space}] @var{name} @dots{}
8851 @var{space} is the case-sensitive namespace of the pragma, or
8852 @code{NULL} to put the pragma in the global namespace. The callback
8853 routine receives @var{pfile} as its first argument, which can be passed
8854 on to cpplib's functions if necessary. You can lex tokens after the
8855 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8856 callback will be silently ignored. The end of the line is indicated by
8857 a token of type @code{CPP_EOF}
8859 For an example use of this routine, see @file{c4x.h} and the callback
8860 routines defined in @file{c4x-c.c}.
8862 Note that the use of @code{c_lex} is specific to the C and C++
8863 compilers. It will not work in the Java or Fortran compilers, or any
8864 other language compilers for that matter. Thus if @code{c_lex} is going
8865 to be called from target-specific code, it must only be done so when
8866 building the C and C++ compilers. This can be done by defining the
8867 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8868 target entry in the @file{config.gcc} file. These variables should name
8869 the target-specific, language-specific object file which contains the
8870 code that uses @code{c_lex}. Note it will also be necessary to add a
8871 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8872 how to build this object file.
8877 @defmac HANDLE_SYSV_PRAGMA
8878 Define this macro (to a value of 1) if you want the System V style
8879 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8880 [=<value>]} to be supported by gcc.
8882 The pack pragma specifies the maximum alignment (in bytes) of fields
8883 within a structure, in much the same way as the @samp{__aligned__} and
8884 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8885 the behavior to the default.
8887 A subtlety for Microsoft Visual C/C++ style bit-field packing
8888 (e.g. -mms-bitfields) for targets that support it:
8889 When a bit-field is inserted into a packed record, the whole size
8890 of the underlying type is used by one or more same-size adjacent
8891 bit-fields (that is, if its long:3, 32 bits is used in the record,
8892 and any additional adjacent long bit-fields are packed into the same
8893 chunk of 32 bits. However, if the size changes, a new field of that
8896 If both MS bit-fields and @samp{__attribute__((packed))} are used,
8897 the latter will take precedence. If @samp{__attribute__((packed))} is
8898 used on a single field when MS bit-fields are in use, it will take
8899 precedence for that field, but the alignment of the rest of the structure
8900 may affect its placement.
8902 The weak pragma only works if @code{SUPPORTS_WEAK} and
8903 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8904 of specifically named weak labels, optionally with a value.
8909 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
8910 Define this macro (to a value of 1) if you want to support the Win32
8911 style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma
8912 pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment
8913 (in bytes) of fields within a structure, in much the same way as the
8914 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8915 pack value of zero resets the behavior to the default. Successive
8916 invocations of this pragma cause the previous values to be stacked, so
8917 that invocations of @samp{#pragma pack(pop)} will return to the previous
8921 @defmac DOLLARS_IN_IDENTIFIERS
8922 Define this macro to control use of the character @samp{$} in
8923 identifier names for the C family of languages. 0 means @samp{$} is
8924 not allowed by default; 1 means it is allowed. 1 is the default;
8925 there is no need to define this macro in that case.
8928 @defmac NO_DOLLAR_IN_LABEL
8929 Define this macro if the assembler does not accept the character
8930 @samp{$} in label names. By default constructors and destructors in
8931 G++ have @samp{$} in the identifiers. If this macro is defined,
8932 @samp{.} is used instead.
8935 @defmac NO_DOT_IN_LABEL
8936 Define this macro if the assembler does not accept the character
8937 @samp{.} in label names. By default constructors and destructors in G++
8938 have names that use @samp{.}. If this macro is defined, these names
8939 are rewritten to avoid @samp{.}.
8942 @defmac DEFAULT_MAIN_RETURN
8943 Define this macro if the target system expects every program's @code{main}
8944 function to return a standard ``success'' value by default (if no other
8945 value is explicitly returned).
8947 The definition should be a C statement (sans semicolon) to generate the
8948 appropriate rtl instructions. It is used only when compiling the end of
8952 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
8953 Define this macro as a C expression that is nonzero if it is safe for the
8954 delay slot scheduler to place instructions in the delay slot of @var{insn},
8955 even if they appear to use a resource set or clobbered in @var{insn}.
8956 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8957 every @code{call_insn} has this behavior. On machines where some @code{insn}
8958 or @code{jump_insn} is really a function call and hence has this behavior,
8959 you should define this macro.
8961 You need not define this macro if it would always return zero.
8964 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
8965 Define this macro as a C expression that is nonzero if it is safe for the
8966 delay slot scheduler to place instructions in the delay slot of @var{insn},
8967 even if they appear to set or clobber a resource referenced in @var{insn}.
8968 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8969 some @code{insn} or @code{jump_insn} is really a function call and its operands
8970 are registers whose use is actually in the subroutine it calls, you should
8971 define this macro. Doing so allows the delay slot scheduler to move
8972 instructions which copy arguments into the argument registers into the delay
8975 You need not define this macro if it would always return zero.
8978 @defmac MULTIPLE_SYMBOL_SPACES
8979 Define this macro if in some cases global symbols from one translation
8980 unit may not be bound to undefined symbols in another translation unit
8981 without user intervention. For instance, under Microsoft Windows
8982 symbols must be explicitly imported from shared libraries (DLLs).
8985 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{clobbers})
8986 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
8987 any hard regs the port wishes to automatically clobber for all asms.
8988 It should return the result of the last @code{tree_cons} used to add a
8992 @defmac MATH_LIBRARY
8993 Define this macro as a C string constant for the linker argument to link
8994 in the system math library, or @samp{""} if the target does not have a
8995 separate math library.
8997 You need only define this macro if the default of @samp{"-lm"} is wrong.
9000 @defmac LIBRARY_PATH_ENV
9001 Define this macro as a C string constant for the environment variable that
9002 specifies where the linker should look for libraries.
9004 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9008 @defmac TARGET_HAS_F_SETLKW
9009 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
9010 Note that this functionality is part of POSIX@.
9011 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
9012 to use file locking when exiting a program, which avoids race conditions
9013 if the program has forked.
9016 @defmac MAX_CONDITIONAL_EXECUTE
9018 A C expression for the maximum number of instructions to execute via
9019 conditional execution instructions instead of a branch. A value of
9020 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9021 1 if it does use cc0.
9024 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
9025 Used if the target needs to perform machine-dependent modifications on the
9026 conditionals used for turning basic blocks into conditionally executed code.
9027 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9028 contains information about the currently processed blocks. @var{true_expr}
9029 and @var{false_expr} are the tests that are used for converting the
9030 then-block and the else-block, respectively. Set either @var{true_expr} or
9031 @var{false_expr} to a null pointer if the tests cannot be converted.
9034 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9035 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9036 if-statements into conditions combined by @code{and} and @code{or} operations.
9037 @var{bb} contains the basic block that contains the test that is currently
9038 being processed and about to be turned into a condition.
9041 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
9042 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9043 be converted to conditional execution format. @var{ce_info} points to
9044 a data structure, @code{struct ce_if_block}, which contains information
9045 about the currently processed blocks.
9048 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
9049 A C expression to perform any final machine dependent modifications in
9050 converting code to conditional execution. The involved basic blocks
9051 can be found in the @code{struct ce_if_block} structure that is pointed
9052 to by @var{ce_info}.
9055 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
9056 A C expression to cancel any machine dependent modifications in
9057 converting code to conditional execution. The involved basic blocks
9058 can be found in the @code{struct ce_if_block} structure that is pointed
9059 to by @var{ce_info}.
9062 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
9063 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9064 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9067 @defmac IFCVT_EXTRA_FIELDS
9068 If defined, it should expand to a set of field declarations that will be
9069 added to the @code{struct ce_if_block} structure. These should be initialized
9070 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9073 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
9074 If non-null, this hook performs a target-specific pass over the
9075 instruction stream. The compiler will run it at all optimization levels,
9076 just before the point at which it normally does delayed-branch scheduling.
9078 The exact purpose of the hook varies from target to target. Some use
9079 it to do transformations that are necessary for correctness, such as
9080 laying out in-function constant pools or avoiding hardware hazards.
9081 Others use it as an opportunity to do some machine-dependent optimizations.
9083 You need not implement the hook if it has nothing to do. The default
9087 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9088 Define this hook if you have any machine-specific built-in functions
9089 that need to be defined. It should be a function that performs the
9092 Machine specific built-in functions can be useful to expand special machine
9093 instructions that would otherwise not normally be generated because
9094 they have no equivalent in the source language (for example, SIMD vector
9095 instructions or prefetch instructions).
9097 To create a built-in function, call the function @code{builtin_function}
9098 which is defined by the language front end. You can use any type nodes set
9099 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9100 only language front ends that use those two functions will call
9101 @samp{TARGET_INIT_BUILTINS}.
9104 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9106 Expand a call to a machine specific built-in function that was set up by
9107 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9108 function call; the result should go to @var{target} if that is
9109 convenient, and have mode @var{mode} if that is convenient.
9110 @var{subtarget} may be used as the target for computing one of
9111 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9112 ignored. This function should return the result of the call to the
9116 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
9118 Take a branch insn in @var{branch1} and another in @var{branch2}.
9119 Return true if redirecting @var{branch1} to the destination of
9120 @var{branch2} is possible.
9122 On some targets, branches may have a limited range. Optimizing the
9123 filling of delay slots can result in branches being redirected, and this
9124 may in turn cause a branch offset to overflow.
9127 @defmac ALLOCATE_INITIAL_VALUE (@var{hard_reg})
9129 When the initial value of a hard register has been copied in a pseudo
9130 register, it is often not necessary to actually allocate another register
9131 to this pseudo register, because the original hard register or a stack slot
9132 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9133 defined, is called at the start of register allocation once for each
9134 hard register that had its initial value copied by using
9135 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9136 Possible values are @code{NULL_RTX}, if you don't want
9137 to do any special allocation, a @code{REG} rtx---that would typically be
9138 the hard register itself, if it is known not to be clobbered---or a
9140 If you are returning a @code{MEM}, this is only a hint for the allocator;
9141 it might decide to use another register anyways.
9142 You may use @code{current_function_leaf_function} in the definition of the
9143 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9144 register in question will not be clobbered.
9147 @defmac TARGET_OBJECT_SUFFIX
9148 Define this macro to be a C string representing the suffix for object
9149 files on your target machine. If you do not define this macro, GCC will
9150 use @samp{.o} as the suffix for object files.
9153 @defmac TARGET_EXECUTABLE_SUFFIX
9154 Define this macro to be a C string representing the suffix to be
9155 automatically added to executable files on your target machine. If you
9156 do not define this macro, GCC will use the null string as the suffix for
9160 @defmac COLLECT_EXPORT_LIST
9161 If defined, @code{collect2} will scan the individual object files
9162 specified on its command line and create an export list for the linker.
9163 Define this macro for systems like AIX, where the linker discards
9164 object files that are not referenced from @code{main} and uses export
9168 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
9169 Define this macro to a C expression representing a variant of the
9170 method call @var{mdecl}, if Java Native Interface (JNI) methods
9171 must be invoked differently from other methods on your target.
9172 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
9173 the @code{stdcall} calling convention and this macro is then
9174 defined as this expression:
9177 build_type_attribute_variant (@var{mdecl},
9179 (get_identifier ("stdcall"),
9184 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9185 This target hook returns @code{true} past the point in which new jump
9186 instructions could be created. On machines that require a register for
9187 every jump such as the SHmedia ISA of SH5, this point would typically be
9188 reload, so this target hook should be defined to a function such as:
9192 cannot_modify_jumps_past_reload_p ()
9194 return (reload_completed || reload_in_progress);
9199 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
9200 This target hook returns a register class for which branch target register
9201 optimizations should be applied. All registers in this class should be
9202 usable interchangeably. After reload, registers in this class will be
9203 re-allocated and loads will be hoisted out of loops and be subjected
9204 to inter-block scheduling.
9207 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
9208 Branch target register optimization will by default exclude callee-saved
9210 that are not already live during the current function; if this target hook
9211 returns true, they will be included. The target code must than make sure
9212 that all target registers in the class returned by
9213 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
9214 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
9215 epilogues have already been generated. Note, even if you only return
9216 true when @var{after_prologue_epilogue_gen} is false, you still are likely
9217 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
9218 to reserve space for caller-saved target registers.
9221 @defmac POWI_MAX_MULTS
9222 If defined, this macro is interpreted as a signed integer C expression
9223 that specifies the maximum number of floating point multiplications
9224 that should be emitted when expanding exponentiation by an integer
9225 constant inline. When this value is defined, exponentiation requiring
9226 more than this number of multiplications is implemented by calling the
9227 system library's @code{pow}, @code{powf} or @code{powl} routines.
9228 The default value places no upper bound on the multiplication count.
9231 @deftypefn Macro void TARGET_EXTRA_INCLUDES (int @var{stdinc})
9232 This target hook should register any extra include files for the
9233 target. The parameter @var{stdinc} indicates if normal include files
9237 @deftypefn Macro void TARGET_OPTF (char *@var{path})
9238 This target hook should register special include paths for the target.
9239 The parameter @var{path} is the include to register. On Darwin
9240 systems, this is used for Framework includes, which have semantics
9241 that are different from @option{-I}.
9244 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
9245 This target hook returns @code{true} if it is safe to use a local alias
9246 for a virtual function @var{fndecl} when constructing thunks,
9247 @code{false} otherwise. By default, the hook returns @code{true} for all
9248 functions, if a target supports aliases (ie. defines
9249 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,