1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Target Description Macros and Functions
8 @cindex machine description macros
9 @cindex target description macros
10 @cindex macros, target description
11 @cindex @file{tm.h} macros
13 In addition to the file @file{@var{machine}.md}, a machine description
14 includes a C header file conventionally given the name
15 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
16 The header file defines numerous macros that convey the information
17 about the target machine that does not fit into the scheme of the
18 @file{.md} file. The file @file{tm.h} should be a link to
19 @file{@var{machine}.h}. The header file @file{config.h} includes
20 @file{tm.h} and most compiler source files include @file{config.h}. The
21 source file defines a variable @code{targetm}, which is a structure
22 containing pointers to functions and data relating to the target
23 machine. @file{@var{machine}.c} should also contain their definitions,
24 if they are not defined elsewhere in GCC, and other functions called
25 through the macros defined in the @file{.h} file.
28 * Target Structure:: The @code{targetm} variable.
29 * Driver:: Controlling how the driver runs the compilation passes.
30 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
31 * Per-Function Data:: Defining data structures for per-function information.
32 * Storage Layout:: Defining sizes and alignments of data.
33 * Type Layout:: Defining sizes and properties of basic user data types.
34 * Escape Sequences:: Defining the value of target character escape sequences
35 * Registers:: Naming and describing the hardware registers.
36 * Register Classes:: Defining the classes of hardware registers.
37 * Stack and Calling:: Defining which way the stack grows and by how much.
38 * Varargs:: Defining the varargs macros.
39 * Trampolines:: Code set up at run time to enter a nested function.
40 * Library Calls:: Controlling how library routines are implicitly called.
41 * Addressing Modes:: Defining addressing modes valid for memory operands.
42 * Condition Code:: Defining how insns update the condition code.
43 * Costs:: Defining relative costs of different operations.
44 * Scheduling:: Adjusting the behavior of the instruction scheduler.
45 * Sections:: Dividing storage into text, data, and other sections.
46 * PIC:: Macros for position independent code.
47 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
48 * Debugging Info:: Defining the format of debugging output.
49 * Floating Point:: Handling floating point for cross-compilers.
50 * Mode Switching:: Insertion of mode-switching instructions.
51 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
52 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
53 * Misc:: Everything else.
56 @node Target Structure
57 @section The Global @code{targetm} Variable
59 @cindex target functions
61 @deftypevar {struct gcc_target} targetm
62 The target @file{.c} file must define the global @code{targetm} variable
63 which contains pointers to functions and data relating to the target
64 machine. The variable is declared in @file{target.h};
65 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
66 used to initialize the variable, and macros for the default initializers
67 for elements of the structure. The @file{.c} file should override those
68 macros for which the default definition is inappropriate. For example:
71 #include "target-def.h"
73 /* @r{Initialize the GCC target structure.} */
75 #undef TARGET_COMP_TYPE_ATTRIBUTES
76 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
78 struct gcc_target targetm = TARGET_INITIALIZER;
82 Where a macro should be defined in the @file{.c} file in this manner to
83 form part of the @code{targetm} structure, it is documented below as a
84 ``Target Hook'' with a prototype. Many macros will change in future
85 from being defined in the @file{.h} file to being part of the
86 @code{targetm} structure.
89 @section Controlling the Compilation Driver, @file{gcc}
91 @cindex controlling the compilation driver
93 @c prevent bad page break with this line
94 You can control the compilation driver.
96 @defmac SWITCH_TAKES_ARG (@var{char})
97 A C expression which determines whether the option @option{-@var{char}}
98 takes arguments. The value should be the number of arguments that
99 option takes--zero, for many options.
101 By default, this macro is defined as
102 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
103 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
104 wish to add additional options which take arguments. Any redefinition
105 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
109 @defmac WORD_SWITCH_TAKES_ARG (@var{name})
110 A C expression which determines whether the option @option{-@var{name}}
111 takes arguments. The value should be the number of arguments that
112 option takes--zero, for many options. This macro rather than
113 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
115 By default, this macro is defined as
116 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
117 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
118 wish to add additional options which take arguments. Any redefinition
119 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
123 @defmac SWITCH_CURTAILS_COMPILATION (@var{char})
124 A C expression which determines whether the option @option{-@var{char}}
125 stops compilation before the generation of an executable. The value is
126 boolean, nonzero if the option does stop an executable from being
127 generated, zero otherwise.
129 By default, this macro is defined as
130 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
131 options properly. You need not define
132 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
133 options which affect the generation of an executable. Any redefinition
134 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
135 for additional options.
138 @defmac SWITCHES_NEED_SPACES
139 A string-valued C expression which enumerates the options for which
140 the linker needs a space between the option and its argument.
142 If this macro is not defined, the default value is @code{""}.
145 @defmac TARGET_OPTION_TRANSLATE_TABLE
146 If defined, a list of pairs of strings, the first of which is a
147 potential command line target to the @file{gcc} driver program, and the
148 second of which is a space-separated (tabs and other whitespace are not
149 supported) list of options with which to replace the first option. The
150 target defining this list is responsible for assuring that the results
151 are valid. Replacement options may not be the @code{--opt} style, they
152 must be the @code{-opt} style. It is the intention of this macro to
153 provide a mechanism for substitution that affects the multilibs chosen,
154 such as one option that enables many options, some of which select
155 multilibs. Example nonsensical definition, where @code{-malt-abi},
156 @code{-EB}, and @code{-mspoo} cause different multilibs to be chosen:
159 #define TARGET_OPTION_TRANSLATE_TABLE \
160 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
161 @{ "-compat", "-EB -malign=4 -mspoo" @}
165 @defmac DRIVER_SELF_SPECS
166 A list of specs for the driver itself. It should be a suitable
167 initializer for an array of strings, with no surrounding braces.
169 The driver applies these specs to its own command line between loading
170 default @file{specs} files (but not command-line specified ones) and
171 choosing the multilib directory or running any subcommands. It
172 applies them in the order given, so each spec can depend on the
173 options added by earlier ones. It is also possible to remove options
174 using @samp{%<@var{option}} in the usual way.
176 This macro can be useful when a port has several interdependent target
177 options. It provides a way of standardizing the command line so
178 that the other specs are easier to write.
180 Do not define this macro if it does not need to do anything.
183 @defmac OPTION_DEFAULT_SPECS
184 A list of specs used to support configure-time default options (i.e.@:
185 @option{--with} options) in the driver. It should be a suitable initializer
186 for an array of structures, each containing two strings, without the
187 outermost pair of surrounding braces.
189 The first item in the pair is the name of the default. This must match
190 the code in @file{config.gcc} for the target. The second item is a spec
191 to apply if a default with this name was specified. The string
192 @samp{%(VALUE)} in the spec will be replaced by the value of the default
193 everywhere it occurs.
195 The driver will apply these specs to its own command line between loading
196 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
197 the same mechanism as @code{DRIVER_SELF_SPECS}.
199 Do not define this macro if it does not need to do anything.
203 A C string constant that tells the GCC driver program options to
204 pass to CPP@. It can also specify how to translate options you
205 give to GCC into options for GCC to pass to the CPP@.
207 Do not define this macro if it does not need to do anything.
210 @defmac CPLUSPLUS_CPP_SPEC
211 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
212 than C@. If you do not define this macro, then the value of
213 @code{CPP_SPEC} (if any) will be used instead.
217 A C string constant that tells the GCC driver program options to
218 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
220 It can also specify how to translate options you give to GCC into options
221 for GCC to pass to front ends.
223 Do not define this macro if it does not need to do anything.
227 A C string constant that tells the GCC driver program options to
228 pass to @code{cc1plus}. It can also specify how to translate options you
229 give to GCC into options for GCC to pass to the @code{cc1plus}.
231 Do not define this macro if it does not need to do anything.
232 Note that everything defined in CC1_SPEC is already passed to
233 @code{cc1plus} so there is no need to duplicate the contents of
234 CC1_SPEC in CC1PLUS_SPEC@.
238 A C string constant that tells the GCC driver program options to
239 pass to the assembler. It can also specify how to translate options
240 you give to GCC into options for GCC to pass to the assembler.
241 See the file @file{sun3.h} for an example of this.
243 Do not define this macro if it does not need to do anything.
246 @defmac ASM_FINAL_SPEC
247 A C string constant that tells the GCC driver program how to
248 run any programs which cleanup after the normal assembler.
249 Normally, this is not needed. See the file @file{mips.h} for
252 Do not define this macro if it does not need to do anything.
255 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
256 Define this macro, with no value, if the driver should give the assembler
257 an argument consisting of a single dash, @option{-}, to instruct it to
258 read from its standard input (which will be a pipe connected to the
259 output of the compiler proper). This argument is given after any
260 @option{-o} option specifying the name of the output file.
262 If you do not define this macro, the assembler is assumed to read its
263 standard input if given no non-option arguments. If your assembler
264 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
265 see @file{mips.h} for instance.
269 A C string constant that tells the GCC driver program options to
270 pass to the linker. It can also specify how to translate options you
271 give to GCC into options for GCC to pass to the linker.
273 Do not define this macro if it does not need to do anything.
277 Another C string constant used much like @code{LINK_SPEC}. The difference
278 between the two is that @code{LIB_SPEC} is used at the end of the
279 command given to the linker.
281 If this macro is not defined, a default is provided that
282 loads the standard C library from the usual place. See @file{gcc.c}.
286 Another C string constant that tells the GCC driver program
287 how and when to place a reference to @file{libgcc.a} into the
288 linker command line. This constant is placed both before and after
289 the value of @code{LIB_SPEC}.
291 If this macro is not defined, the GCC driver provides a default that
292 passes the string @option{-lgcc} to the linker.
295 @defmac STARTFILE_SPEC
296 Another C string constant used much like @code{LINK_SPEC}. The
297 difference between the two is that @code{STARTFILE_SPEC} is used at
298 the very beginning of the command given to the linker.
300 If this macro is not defined, a default is provided that loads the
301 standard C startup file from the usual place. See @file{gcc.c}.
305 Another C string constant used much like @code{LINK_SPEC}. The
306 difference between the two is that @code{ENDFILE_SPEC} is used at
307 the very end of the command given to the linker.
309 Do not define this macro if it does not need to do anything.
312 @defmac THREAD_MODEL_SPEC
313 GCC @code{-v} will print the thread model GCC was configured to use.
314 However, this doesn't work on platforms that are multilibbed on thread
315 models, such as AIX 4.3. On such platforms, define
316 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
317 blanks that names one of the recognized thread models. @code{%*}, the
318 default value of this macro, will expand to the value of
319 @code{thread_file} set in @file{config.gcc}.
322 @defmac SYSROOT_SUFFIX_SPEC
323 Define this macro to add a suffix to the target sysroot when GCC is
324 configured with a sysroot. This will cause GCC to search for usr/lib,
325 et al, within sysroot+suffix.
328 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
329 Define this macro to add a headers_suffix to the target sysroot when
330 GCC is configured with a sysroot. This will cause GCC to pass the
331 updated sysroot+headers_suffix to CPP@, causing it to search for
332 usr/include, et al, within sysroot+headers_suffix.
336 Define this macro to provide additional specifications to put in the
337 @file{specs} file that can be used in various specifications like
340 The definition should be an initializer for an array of structures,
341 containing a string constant, that defines the specification name, and a
342 string constant that provides the specification.
344 Do not define this macro if it does not need to do anything.
346 @code{EXTRA_SPECS} is useful when an architecture contains several
347 related targets, which have various @code{@dots{}_SPECS} which are similar
348 to each other, and the maintainer would like one central place to keep
351 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
352 define either @code{_CALL_SYSV} when the System V calling sequence is
353 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
356 The @file{config/rs6000/rs6000.h} target file defines:
359 #define EXTRA_SPECS \
360 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
362 #define CPP_SYS_DEFAULT ""
365 The @file{config/rs6000/sysv.h} target file defines:
369 "%@{posix: -D_POSIX_SOURCE @} \
370 %@{mcall-sysv: -D_CALL_SYSV @} \
371 %@{!mcall-sysv: %(cpp_sysv_default) @} \
372 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
374 #undef CPP_SYSV_DEFAULT
375 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
378 while the @file{config/rs6000/eabiaix.h} target file defines
379 @code{CPP_SYSV_DEFAULT} as:
382 #undef CPP_SYSV_DEFAULT
383 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
387 @defmac LINK_LIBGCC_SPECIAL
388 Define this macro if the driver program should find the library
389 @file{libgcc.a} itself and should not pass @option{-L} options to the
390 linker. If you do not define this macro, the driver program will pass
391 the argument @option{-lgcc} to tell the linker to do the search and will
392 pass @option{-L} options to it.
395 @defmac LINK_LIBGCC_SPECIAL_1
396 Define this macro if the driver program should find the library
397 @file{libgcc.a}. If you do not define this macro, the driver program will pass
398 the argument @option{-lgcc} to tell the linker to do the search.
399 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
400 not affect @option{-L} options.
403 @defmac LINK_GCC_C_SEQUENCE_SPEC
404 The sequence in which libgcc and libc are specified to the linker.
405 By default this is @code{%G %L %G}.
408 @defmac LINK_COMMAND_SPEC
409 A C string constant giving the complete command line need to execute the
410 linker. When you do this, you will need to update your port each time a
411 change is made to the link command line within @file{gcc.c}. Therefore,
412 define this macro only if you need to completely redefine the command
413 line for invoking the linker and there is no other way to accomplish
414 the effect you need. Overriding this macro may be avoidable by overriding
415 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
418 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
419 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
420 directories from linking commands. Do not give it a nonzero value if
421 removing duplicate search directories changes the linker's semantics.
424 @defmac MULTILIB_DEFAULTS
425 Define this macro as a C expression for the initializer of an array of
426 string to tell the driver program which options are defaults for this
427 target and thus do not need to be handled specially when using
428 @code{MULTILIB_OPTIONS}.
430 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
431 the target makefile fragment or if none of the options listed in
432 @code{MULTILIB_OPTIONS} are set by default.
433 @xref{Target Fragment}.
436 @defmac RELATIVE_PREFIX_NOT_LINKDIR
437 Define this macro to tell @command{gcc} that it should only translate
438 a @option{-B} prefix into a @option{-L} linker option if the prefix
439 indicates an absolute file name.
442 @defmac MD_EXEC_PREFIX
443 If defined, this macro is an additional prefix to try after
444 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
445 when the @option{-b} option is used, or the compiler is built as a cross
446 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
447 to the list of directories used to find the assembler in @file{configure.in}.
450 @defmac STANDARD_STARTFILE_PREFIX
451 Define this macro as a C string constant if you wish to override the
452 standard choice of @file{/usr/local/lib/} as the default prefix to
453 try when searching for startup files such as @file{crt0.o}.
456 @defmac MD_STARTFILE_PREFIX
457 If defined, this macro supplies an additional prefix to try after the
458 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
459 @option{-b} option is used, or when the compiler is built as a cross
463 @defmac MD_STARTFILE_PREFIX_1
464 If defined, this macro supplies yet another prefix to try after the
465 standard prefixes. It is not searched when the @option{-b} option is
466 used, or when the compiler is built as a cross compiler.
469 @defmac INIT_ENVIRONMENT
470 Define this macro as a C string constant if you wish to set environment
471 variables for programs called by the driver, such as the assembler and
472 loader. The driver passes the value of this macro to @code{putenv} to
473 initialize the necessary environment variables.
476 @defmac LOCAL_INCLUDE_DIR
477 Define this macro as a C string constant if you wish to override the
478 standard choice of @file{/usr/local/include} as the default prefix to
479 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
480 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
482 Cross compilers do not search either @file{/usr/local/include} or its
486 @defmac MODIFY_TARGET_NAME
487 Define this macro if you wish to define command-line switches that
488 modify the default target name.
490 For each switch, you can include a string to be appended to the first
491 part of the configuration name or a string to be deleted from the
492 configuration name, if present. The definition should be an initializer
493 for an array of structures. Each array element should have three
494 elements: the switch name (a string constant, including the initial
495 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
496 indicate whether the string should be inserted or deleted, and the string
497 to be inserted or deleted (a string constant).
499 For example, on a machine where @samp{64} at the end of the
500 configuration name denotes a 64-bit target and you want the @option{-32}
501 and @option{-64} switches to select between 32- and 64-bit targets, you would
505 #define MODIFY_TARGET_NAME \
506 @{ @{ "-32", DELETE, "64"@}, \
507 @{"-64", ADD, "64"@}@}
511 @defmac SYSTEM_INCLUDE_DIR
512 Define this macro as a C string constant if you wish to specify a
513 system-specific directory to search for header files before the standard
514 directory. @code{SYSTEM_INCLUDE_DIR} comes before
515 @code{STANDARD_INCLUDE_DIR} in the search order.
517 Cross compilers do not use this macro and do not search the directory
521 @defmac STANDARD_INCLUDE_DIR
522 Define this macro as a C string constant if you wish to override the
523 standard choice of @file{/usr/include} as the default prefix to
524 try when searching for header files.
526 Cross compilers ignore this macro and do not search either
527 @file{/usr/include} or its replacement.
530 @defmac STANDARD_INCLUDE_COMPONENT
531 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
532 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
533 If you do not define this macro, no component is used.
536 @defmac INCLUDE_DEFAULTS
537 Define this macro if you wish to override the entire default search path
538 for include files. For a native compiler, the default search path
539 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
540 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
541 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
542 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
543 and specify private search areas for GCC@. The directory
544 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
546 The definition should be an initializer for an array of structures.
547 Each array element should have four elements: the directory name (a
548 string constant), the component name (also a string constant), a flag
549 for C++-only directories,
550 and a flag showing that the includes in the directory don't need to be
551 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
552 the array with a null element.
554 The component name denotes what GNU package the include file is part of,
555 if any, in all uppercase letters. For example, it might be @samp{GCC}
556 or @samp{BINUTILS}. If the package is part of a vendor-supplied
557 operating system, code the component name as @samp{0}.
559 For example, here is the definition used for VAX/VMS:
562 #define INCLUDE_DEFAULTS \
564 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
565 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
566 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
573 Here is the order of prefixes tried for exec files:
577 Any prefixes specified by the user with @option{-B}.
580 The environment variable @code{GCC_EXEC_PREFIX}, if any.
583 The directories specified by the environment variable @code{COMPILER_PATH}.
586 The macro @code{STANDARD_EXEC_PREFIX}.
589 @file{/usr/lib/gcc/}.
592 The macro @code{MD_EXEC_PREFIX}, if any.
595 Here is the order of prefixes tried for startfiles:
599 Any prefixes specified by the user with @option{-B}.
602 The environment variable @code{GCC_EXEC_PREFIX}, if any.
605 The directories specified by the environment variable @code{LIBRARY_PATH}
606 (or port-specific name; native only, cross compilers do not use this).
609 The macro @code{STANDARD_EXEC_PREFIX}.
612 @file{/usr/lib/gcc/}.
615 The macro @code{MD_EXEC_PREFIX}, if any.
618 The macro @code{MD_STARTFILE_PREFIX}, if any.
621 The macro @code{STANDARD_STARTFILE_PREFIX}.
630 @node Run-time Target
631 @section Run-time Target Specification
632 @cindex run-time target specification
633 @cindex predefined macros
634 @cindex target specifications
636 @c prevent bad page break with this line
637 Here are run-time target specifications.
639 @defmac TARGET_CPU_CPP_BUILTINS ()
640 This function-like macro expands to a block of code that defines
641 built-in preprocessor macros and assertions for the target cpu, using
642 the functions @code{builtin_define}, @code{builtin_define_std} and
643 @code{builtin_assert}. When the front end
644 calls this macro it provides a trailing semicolon, and since it has
645 finished command line option processing your code can use those
648 @code{builtin_assert} takes a string in the form you pass to the
649 command-line option @option{-A}, such as @code{cpu=mips}, and creates
650 the assertion. @code{builtin_define} takes a string in the form
651 accepted by option @option{-D} and unconditionally defines the macro.
653 @code{builtin_define_std} takes a string representing the name of an
654 object-like macro. If it doesn't lie in the user's namespace,
655 @code{builtin_define_std} defines it unconditionally. Otherwise, it
656 defines a version with two leading underscores, and another version
657 with two leading and trailing underscores, and defines the original
658 only if an ISO standard was not requested on the command line. For
659 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
660 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
661 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
662 defines only @code{_ABI64}.
664 You can also test for the C dialect being compiled. The variable
665 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
666 or @code{clk_objective_c}. Note that if we are preprocessing
667 assembler, this variable will be @code{clk_c} but the function-like
668 macro @code{preprocessing_asm_p()} will return true, so you might want
669 to check for that first. If you need to check for strict ANSI, the
670 variable @code{flag_iso} can be used. The function-like macro
671 @code{preprocessing_trad_p()} can be used to check for traditional
675 @defmac TARGET_OS_CPP_BUILTINS ()
676 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
677 and is used for the target operating system instead.
680 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
681 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
682 and is used for the target object format. @file{elfos.h} uses this
683 macro to define @code{__ELF__}, so you probably do not need to define
687 @deftypevar {extern int} target_flags
688 This declaration should be present.
691 @cindex optional hardware or system features
692 @cindex features, optional, in system conventions
694 @defmac TARGET_@var{featurename}
695 This series of macros is to allow compiler command arguments to
696 enable or disable the use of optional features of the target machine.
697 For example, one machine description serves both the 68000 and
698 the 68020; a command argument tells the compiler whether it should
699 use 68020-only instructions or not. This command argument works
700 by means of a macro @code{TARGET_68020} that tests a bit in
703 Define a macro @code{TARGET_@var{featurename}} for each such option.
704 Its definition should test a bit in @code{target_flags}. It is
705 recommended that a helper macro @code{MASK_@var{featurename}}
706 is defined for each bit-value to test, and used in
707 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
711 #define TARGET_MASK_68020 1
712 #define TARGET_68020 (target_flags & MASK_68020)
715 One place where these macros are used is in the condition-expressions
716 of instruction patterns. Note how @code{TARGET_68020} appears
717 frequently in the 68000 machine description file, @file{m68k.md}.
718 Another place they are used is in the definitions of the other
719 macros in the @file{@var{machine}.h} file.
722 @defmac TARGET_SWITCHES
723 This macro defines names of command options to set and clear
724 bits in @code{target_flags}. Its definition is an initializer
725 with a subgrouping for each command option.
727 Each subgrouping contains a string constant, that defines the option
728 name, a number, which contains the bits to set in
729 @code{target_flags}, and a second string which is the description
730 displayed by @option{--help}. If the number is negative then the bits specified
731 by the number are cleared instead of being set. If the description
732 string is present but empty, then no help information will be displayed
733 for that option, but it will not count as an undocumented option. The
734 actual option name is made by appending @samp{-m} to the specified name.
735 Non-empty description strings should be marked with @code{N_(@dots{})} for
736 @command{xgettext}. Please do not mark empty strings because the empty
737 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
738 of the message catalog with meta information, not the empty string.
740 In addition to the description for @option{--help},
741 more detailed documentation for each option should be added to
744 One of the subgroupings should have a null string. The number in
745 this grouping is the default value for @code{target_flags}. Any
746 target options act starting with that value.
748 Here is an example which defines @option{-m68000} and @option{-m68020}
749 with opposite meanings, and picks the latter as the default:
752 #define TARGET_SWITCHES \
753 @{ @{ "68020", MASK_68020, "" @}, \
754 @{ "68000", -MASK_68020, \
755 N_("Compile for the 68000") @}, \
756 @{ "", MASK_68020, "" @}, \
761 @defmac TARGET_OPTIONS
762 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
763 options that have values. Its definition is an initializer with a
764 subgrouping for each command option.
766 Each subgrouping contains a string constant, that defines the option
767 name, the address of a variable, a description string, and a value.
768 Non-empty description strings should be marked with @code{N_(@dots{})}
769 for @command{xgettext}. Please do not mark empty strings because the
770 empty string is reserved by GNU gettext. @code{gettext("")} returns the
771 header entry of the message catalog with meta information, not the empty
774 If the value listed in the table is @code{NULL}, then the variable, type
775 @code{char *}, is set to the variable part of the given option if the
776 fixed part matches. In other words, if the first part of the option
777 matches what's in the table, the variable will be set to point to the
778 rest of the option. This allows the user to specify a value for that
779 option. The actual option name is made by appending @samp{-m} to the
780 specified name. Again, each option should also be documented in
783 If the value listed in the table is non-@code{NULL}, then the option
784 must match the option in the table exactly (with @samp{-m}), and the
785 variable is set to point to the value listed in the table.
787 Here is an example which defines @option{-mshort-data-@var{number}}. If the
788 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
789 will be set to the string @code{"512"}.
792 extern char *m88k_short_data;
793 #define TARGET_OPTIONS \
794 @{ @{ "short-data-", &m88k_short_data, \
795 N_("Specify the size of the short data section"), 0 @} @}
798 Here is a variant of the above that allows the user to also specify
799 just @option{-mshort-data} where a default of @code{"64"} is used.
802 extern char *m88k_short_data;
803 #define TARGET_OPTIONS \
804 @{ @{ "short-data-", &m88k_short_data, \
805 N_("Specify the size of the short data section"), 0 @} \
806 @{ "short-data", &m88k_short_data, "", "64" @},
810 Here is an example which defines @option{-mno-alu}, @option{-malu1}, and
811 @option{-malu2} as a three-state switch, along with suitable macros for
812 checking the state of the option (documentation is elided for brevity).
816 char *chip_alu = ""; /* Specify default here. */
819 extern char *chip_alu;
820 #define TARGET_OPTIONS \
821 @{ @{ "no-alu", &chip_alu, "", "" @}, \
822 @{ "alu1", &chip_alu, "", "1" @}, \
823 @{ "alu2", &chip_alu, "", "2" @}, @}
824 #define TARGET_ALU (chip_alu[0] != '\0')
825 #define TARGET_ALU1 (chip_alu[0] == '1')
826 #define TARGET_ALU2 (chip_alu[0] == '2')
830 @defmac TARGET_VERSION
831 This macro is a C statement to print on @code{stderr} a string
832 describing the particular machine description choice. Every machine
833 description should define @code{TARGET_VERSION}. For example:
837 #define TARGET_VERSION \
838 fprintf (stderr, " (68k, Motorola syntax)");
840 #define TARGET_VERSION \
841 fprintf (stderr, " (68k, MIT syntax)");
846 @defmac OVERRIDE_OPTIONS
847 Sometimes certain combinations of command options do not make sense on
848 a particular target machine. You can define a macro
849 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
850 defined, is executed once just after all the command options have been
853 Don't use this macro to turn on various extra optimizations for
854 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
857 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
858 Some machines may desire to change what optimizations are performed for
859 various optimization levels. This macro, if defined, is executed once
860 just after the optimization level is determined and before the remainder
861 of the command options have been parsed. Values set in this macro are
862 used as the default values for the other command line options.
864 @var{level} is the optimization level specified; 2 if @option{-O2} is
865 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
867 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
869 You should not use this macro to change options that are not
870 machine-specific. These should uniformly selected by the same
871 optimization level on all supported machines. Use this macro to enable
872 machine-specific optimizations.
874 @strong{Do not examine @code{write_symbols} in
875 this macro!} The debugging options are not supposed to alter the
879 @defmac CAN_DEBUG_WITHOUT_FP
880 Define this macro if debugging can be performed even without a frame
881 pointer. If this macro is defined, GCC will turn on the
882 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
885 @node Per-Function Data
886 @section Defining data structures for per-function information.
887 @cindex per-function data
888 @cindex data structures
890 If the target needs to store information on a per-function basis, GCC
891 provides a macro and a couple of variables to allow this. Note, just
892 using statics to store the information is a bad idea, since GCC supports
893 nested functions, so you can be halfway through encoding one function
894 when another one comes along.
896 GCC defines a data structure called @code{struct function} which
897 contains all of the data specific to an individual function. This
898 structure contains a field called @code{machine} whose type is
899 @code{struct machine_function *}, which can be used by targets to point
900 to their own specific data.
902 If a target needs per-function specific data it should define the type
903 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
904 This macro should be used to initialize the function pointer
905 @code{init_machine_status}. This pointer is explained below.
907 One typical use of per-function, target specific data is to create an
908 RTX to hold the register containing the function's return address. This
909 RTX can then be used to implement the @code{__builtin_return_address}
910 function, for level 0.
912 Note---earlier implementations of GCC used a single data area to hold
913 all of the per-function information. Thus when processing of a nested
914 function began the old per-function data had to be pushed onto a
915 stack, and when the processing was finished, it had to be popped off the
916 stack. GCC used to provide function pointers called
917 @code{save_machine_status} and @code{restore_machine_status} to handle
918 the saving and restoring of the target specific information. Since the
919 single data area approach is no longer used, these pointers are no
922 @defmac INIT_EXPANDERS
923 Macro called to initialize any target specific information. This macro
924 is called once per function, before generation of any RTL has begun.
925 The intention of this macro is to allow the initialization of the
926 function pointer @code{init_machine_status}.
929 @deftypevar {void (*)(struct function *)} init_machine_status
930 If this function pointer is non-@code{NULL} it will be called once per
931 function, before function compilation starts, in order to allow the
932 target to perform any target specific initialization of the
933 @code{struct function} structure. It is intended that this would be
934 used to initialize the @code{machine} of that structure.
936 @code{struct machine_function} structures are expected to be freed by GC.
937 Generally, any memory that they reference must be allocated by using
938 @code{ggc_alloc}, including the structure itself.
942 @section Storage Layout
943 @cindex storage layout
945 Note that the definitions of the macros in this table which are sizes or
946 alignments measured in bits do not need to be constant. They can be C
947 expressions that refer to static variables, such as the @code{target_flags}.
948 @xref{Run-time Target}.
950 @defmac BITS_BIG_ENDIAN
951 Define this macro to have the value 1 if the most significant bit in a
952 byte has the lowest number; otherwise define it to have the value zero.
953 This means that bit-field instructions count from the most significant
954 bit. If the machine has no bit-field instructions, then this must still
955 be defined, but it doesn't matter which value it is defined to. This
956 macro need not be a constant.
958 This macro does not affect the way structure fields are packed into
959 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
962 @defmac BYTES_BIG_ENDIAN
963 Define this macro to have the value 1 if the most significant byte in a
964 word has the lowest number. This macro need not be a constant.
967 @defmac WORDS_BIG_ENDIAN
968 Define this macro to have the value 1 if, in a multiword object, the
969 most significant word has the lowest number. This applies to both
970 memory locations and registers; GCC fundamentally assumes that the
971 order of words in memory is the same as the order in registers. This
972 macro need not be a constant.
975 @defmac LIBGCC2_WORDS_BIG_ENDIAN
976 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
977 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
978 used only when compiling @file{libgcc2.c}. Typically the value will be set
979 based on preprocessor defines.
982 @defmac FLOAT_WORDS_BIG_ENDIAN
983 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
984 @code{TFmode} floating point numbers are stored in memory with the word
985 containing the sign bit at the lowest address; otherwise define it to
986 have the value 0. This macro need not be a constant.
988 You need not define this macro if the ordering is the same as for
992 @defmac BITS_PER_UNIT
993 Define this macro to be the number of bits in an addressable storage
994 unit (byte). If you do not define this macro the default is 8.
997 @defmac BITS_PER_WORD
998 Number of bits in a word. If you do not define this macro, the default
999 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
1002 @defmac MAX_BITS_PER_WORD
1003 Maximum number of bits in a word. If this is undefined, the default is
1004 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
1005 largest value that @code{BITS_PER_WORD} can have at run-time.
1008 @defmac UNITS_PER_WORD
1009 Number of storage units in a word; normally 4.
1012 @defmac MIN_UNITS_PER_WORD
1013 Minimum number of units in a word. If this is undefined, the default is
1014 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
1015 smallest value that @code{UNITS_PER_WORD} can have at run-time.
1018 @defmac POINTER_SIZE
1019 Width of a pointer, in bits. You must specify a value no wider than the
1020 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1021 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1022 a value the default is @code{BITS_PER_WORD}.
1025 @defmac POINTERS_EXTEND_UNSIGNED
1026 A C expression whose value is greater than zero if pointers that need to be
1027 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
1028 be zero-extended and zero if they are to be sign-extended. If the value
1029 is less then zero then there must be an "ptr_extend" instruction that
1030 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
1032 You need not define this macro if the @code{POINTER_SIZE} is equal
1033 to the width of @code{Pmode}.
1036 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1037 A macro to update @var{m} and @var{unsignedp} when an object whose type
1038 is @var{type} and which has the specified mode and signedness is to be
1039 stored in a register. This macro is only called when @var{type} is a
1042 On most RISC machines, which only have operations that operate on a full
1043 register, define this macro to set @var{m} to @code{word_mode} if
1044 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1045 cases, only integer modes should be widened because wider-precision
1046 floating-point operations are usually more expensive than their narrower
1049 For most machines, the macro definition does not change @var{unsignedp}.
1050 However, some machines, have instructions that preferentially handle
1051 either signed or unsigned quantities of certain modes. For example, on
1052 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1053 sign-extend the result to 64 bits. On such machines, set
1054 @var{unsignedp} according to which kind of extension is more efficient.
1056 Do not define this macro if it would never modify @var{m}.
1059 @defmac PROMOTE_FUNCTION_ARGS
1060 Define this macro if the promotion described by @code{PROMOTE_MODE}
1061 should also be done for outgoing function arguments.
1064 @defmac PROMOTE_FUNCTION_RETURN
1065 Define this macro if the promotion described by @code{PROMOTE_MODE}
1066 should also be done for the return value of functions.
1068 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
1069 promotions done by @code{PROMOTE_MODE}.
1072 @defmac PROMOTE_FOR_CALL_ONLY
1073 Define this macro if the promotion described by @code{PROMOTE_MODE}
1074 should @emph{only} be performed for outgoing function arguments or
1075 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
1076 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
1079 @defmac PARM_BOUNDARY
1080 Normal alignment required for function parameters on the stack, in
1081 bits. All stack parameters receive at least this much alignment
1082 regardless of data type. On most machines, this is the same as the
1086 @defmac STACK_BOUNDARY
1087 Define this macro to the minimum alignment enforced by hardware for the
1088 stack pointer on this machine. The definition is a C expression for the
1089 desired alignment (measured in bits). This value is used as a default
1090 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1091 this should be the same as @code{PARM_BOUNDARY}.
1094 @defmac PREFERRED_STACK_BOUNDARY
1095 Define this macro if you wish to preserve a certain alignment for the
1096 stack pointer, greater than what the hardware enforces. The definition
1097 is a C expression for the desired alignment (measured in bits). This
1098 macro must evaluate to a value equal to or larger than
1099 @code{STACK_BOUNDARY}.
1102 @defmac FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1103 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1104 not guaranteed by the runtime and we should emit code to align the stack
1105 at the beginning of @code{main}.
1107 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1108 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1109 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1110 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1111 be momentarily unaligned while pushing arguments.
1114 @defmac FUNCTION_BOUNDARY
1115 Alignment required for a function entry point, in bits.
1118 @defmac BIGGEST_ALIGNMENT
1119 Biggest alignment that any data type can require on this machine, in bits.
1122 @defmac MINIMUM_ATOMIC_ALIGNMENT
1123 If defined, the smallest alignment, in bits, that can be given to an
1124 object that can be referenced in one operation, without disturbing any
1125 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1126 on machines that don't have byte or half-word store operations.
1129 @defmac BIGGEST_FIELD_ALIGNMENT
1130 Biggest alignment that any structure or union field can require on this
1131 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1132 structure and union fields only, unless the field alignment has been set
1133 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1136 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1137 An expression for the alignment of a structure field @var{field} if the
1138 alignment computed in the usual way (including applying of
1139 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1140 alignment) is @var{computed}. It overrides alignment only if the
1141 field alignment has not been set by the
1142 @code{__attribute__ ((aligned (@var{n})))} construct.
1145 @defmac MAX_OFILE_ALIGNMENT
1146 Biggest alignment supported by the object file format of this machine.
1147 Use this macro to limit the alignment which can be specified using the
1148 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1149 the default value is @code{BIGGEST_ALIGNMENT}.
1152 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1153 If defined, a C expression to compute the alignment for a variable in
1154 the static store. @var{type} is the data type, and @var{basic-align} is
1155 the alignment that the object would ordinarily have. The value of this
1156 macro is used instead of that alignment to align the object.
1158 If this macro is not defined, then @var{basic-align} is used.
1161 One use of this macro is to increase alignment of medium-size data to
1162 make it all fit in fewer cache lines. Another is to cause character
1163 arrays to be word-aligned so that @code{strcpy} calls that copy
1164 constants to character arrays can be done inline.
1167 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1168 If defined, a C expression to compute the alignment given to a constant
1169 that is being placed in memory. @var{constant} is the constant and
1170 @var{basic-align} is the alignment that the object would ordinarily
1171 have. The value of this macro is used instead of that alignment to
1174 If this macro is not defined, then @var{basic-align} is used.
1176 The typical use of this macro is to increase alignment for string
1177 constants to be word aligned so that @code{strcpy} calls that copy
1178 constants can be done inline.
1181 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1182 If defined, a C expression to compute the alignment for a variable in
1183 the local store. @var{type} is the data type, and @var{basic-align} is
1184 the alignment that the object would ordinarily have. The value of this
1185 macro is used instead of that alignment to align the object.
1187 If this macro is not defined, then @var{basic-align} is used.
1189 One use of this macro is to increase alignment of medium-size data to
1190 make it all fit in fewer cache lines.
1193 @defmac EMPTY_FIELD_BOUNDARY
1194 Alignment in bits to be given to a structure bit-field that follows an
1195 empty field such as @code{int : 0;}.
1197 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1200 @defmac STRUCTURE_SIZE_BOUNDARY
1201 Number of bits which any structure or union's size must be a multiple of.
1202 Each structure or union's size is rounded up to a multiple of this.
1204 If you do not define this macro, the default is the same as
1205 @code{BITS_PER_UNIT}.
1208 @defmac STRICT_ALIGNMENT
1209 Define this macro to be the value 1 if instructions will fail to work
1210 if given data not on the nominal alignment. If instructions will merely
1211 go slower in that case, define this macro as 0.
1214 @defmac PCC_BITFIELD_TYPE_MATTERS
1215 Define this if you wish to imitate the way many other C compilers handle
1216 alignment of bit-fields and the structures that contain them.
1218 The behavior is that the type written for a named bit-field (@code{int},
1219 @code{short}, or other integer type) imposes an alignment for the entire
1220 structure, as if the structure really did contain an ordinary field of
1221 that type. In addition, the bit-field is placed within the structure so
1222 that it would fit within such a field, not crossing a boundary for it.
1224 Thus, on most machines, a named bit-field whose type is written as
1225 @code{int} would not cross a four-byte boundary, and would force
1226 four-byte alignment for the whole structure. (The alignment used may
1227 not be four bytes; it is controlled by the other alignment parameters.)
1229 An unnamed bit-field will not affect the alignment of the containing
1232 If the macro is defined, its definition should be a C expression;
1233 a nonzero value for the expression enables this behavior.
1235 Note that if this macro is not defined, or its value is zero, some
1236 bit-fields may cross more than one alignment boundary. The compiler can
1237 support such references if there are @samp{insv}, @samp{extv}, and
1238 @samp{extzv} insns that can directly reference memory.
1240 The other known way of making bit-fields work is to define
1241 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1242 Then every structure can be accessed with fullwords.
1244 Unless the machine has bit-field instructions or you define
1245 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1246 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1248 If your aim is to make GCC use the same conventions for laying out
1249 bit-fields as are used by another compiler, here is how to investigate
1250 what the other compiler does. Compile and run this program:
1269 printf ("Size of foo1 is %d\n",
1270 sizeof (struct foo1));
1271 printf ("Size of foo2 is %d\n",
1272 sizeof (struct foo2));
1277 If this prints 2 and 5, then the compiler's behavior is what you would
1278 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1281 @defmac BITFIELD_NBYTES_LIMITED
1282 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1283 to aligning a bit-field within the structure.
1286 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1287 Return 1 if a structure or array containing @var{field} should be accessed using
1290 If @var{field} is the only field in the structure, @var{mode} is its
1291 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1292 case where structures of one field would require the structure's mode to
1293 retain the field's mode.
1295 Normally, this is not needed. See the file @file{c4x.h} for an example
1296 of how to use this macro to prevent a structure having a floating point
1297 field from being accessed in an integer mode.
1300 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1301 Define this macro as an expression for the alignment of a type (given
1302 by @var{type} as a tree node) if the alignment computed in the usual
1303 way is @var{computed} and the alignment explicitly specified was
1306 The default is to use @var{specified} if it is larger; otherwise, use
1307 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1310 @defmac MAX_FIXED_MODE_SIZE
1311 An integer expression for the size in bits of the largest integer
1312 machine mode that should actually be used. All integer machine modes of
1313 this size or smaller can be used for structures and unions with the
1314 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1315 (DImode)} is assumed.
1318 @defmac VECTOR_MODE_SUPPORTED_P (@var{mode})
1319 Define this macro to be nonzero if the port is prepared to handle insns
1320 involving vector mode @var{mode}. At the very least, it must have move
1321 patterns for this mode.
1324 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1325 If defined, an expression of type @code{enum machine_mode} that
1326 specifies the mode of the save area operand of a
1327 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1328 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1329 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1330 having its mode specified.
1332 You need not define this macro if it always returns @code{Pmode}. You
1333 would most commonly define this macro if the
1334 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1338 @defmac STACK_SIZE_MODE
1339 If defined, an expression of type @code{enum machine_mode} that
1340 specifies the mode of the size increment operand of an
1341 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1343 You need not define this macro if it always returns @code{word_mode}.
1344 You would most commonly define this macro if the @code{allocate_stack}
1345 pattern needs to support both a 32- and a 64-bit mode.
1348 @defmac TARGET_FLOAT_FORMAT
1349 A code distinguishing the floating point format of the target machine.
1350 There are four defined values:
1353 @item IEEE_FLOAT_FORMAT
1354 This code indicates IEEE floating point. It is the default; there is no
1355 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1357 @item VAX_FLOAT_FORMAT
1358 This code indicates the ``F float'' (for @code{float}) and ``D float''
1359 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1361 @item IBM_FLOAT_FORMAT
1362 This code indicates the format used on the IBM System/370.
1364 @item C4X_FLOAT_FORMAT
1365 This code indicates the format used on the TMS320C3x/C4x.
1368 If your target uses a floating point format other than these, you must
1369 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1370 it to @file{real.c}.
1372 The ordering of the component words of floating point values stored in
1373 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1376 @defmac MODE_HAS_NANS (@var{mode})
1377 When defined, this macro should be true if @var{mode} has a NaN
1378 representation. The compiler assumes that NaNs are not equal to
1379 anything (including themselves) and that addition, subtraction,
1380 multiplication and division all return NaNs when one operand is
1383 By default, this macro is true if @var{mode} is a floating-point
1384 mode and the target floating-point format is IEEE@.
1387 @defmac MODE_HAS_INFINITIES (@var{mode})
1388 This macro should be true if @var{mode} can represent infinity. At
1389 present, the compiler uses this macro to decide whether @samp{x - x}
1390 is always defined. By default, the macro is true when @var{mode}
1391 is a floating-point mode and the target format is IEEE@.
1394 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1395 True if @var{mode} distinguishes between positive and negative zero.
1396 The rules are expected to follow the IEEE standard:
1400 @samp{x + x} has the same sign as @samp{x}.
1403 If the sum of two values with opposite sign is zero, the result is
1404 positive for all rounding modes expect towards @minus{}infinity, for
1405 which it is negative.
1408 The sign of a product or quotient is negative when exactly one
1409 of the operands is negative.
1412 The default definition is true if @var{mode} is a floating-point
1413 mode and the target format is IEEE@.
1416 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1417 If defined, this macro should be true for @var{mode} if it has at
1418 least one rounding mode in which @samp{x} and @samp{-x} can be
1419 rounded to numbers of different magnitude. Two such modes are
1420 towards @minus{}infinity and towards +infinity.
1422 The default definition of this macro is true if @var{mode} is
1423 a floating-point mode and the target format is IEEE@.
1426 @defmac ROUND_TOWARDS_ZERO
1427 If defined, this macro should be true if the prevailing rounding
1428 mode is towards zero. A true value has the following effects:
1432 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1435 @file{libgcc.a}'s floating-point emulator will round towards zero
1436 rather than towards nearest.
1439 The compiler's floating-point emulator will round towards zero after
1440 doing arithmetic, and when converting from the internal float format to
1444 The macro does not affect the parsing of string literals. When the
1445 primary rounding mode is towards zero, library functions like
1446 @code{strtod} might still round towards nearest, and the compiler's
1447 parser should behave like the target's @code{strtod} where possible.
1449 Not defining this macro is equivalent to returning zero.
1452 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1453 This macro should return true if floats with @var{size}
1454 bits do not have a NaN or infinity representation, but use the largest
1455 exponent for normal numbers instead.
1457 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1458 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1459 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1460 floating-point arithmetic.
1462 The default definition of this macro returns false for all sizes.
1465 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1466 This target hook should return @code{true} a vector is opaque. That
1467 is, if no cast is needed when copying a vector value of type
1468 @var{type} into another vector lvalue of the same size. Vector opaque
1469 types cannot be initialized. The default is that there are no such
1473 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1474 This target hook returns @code{true} if bit-fields in the given
1475 @var{record_type} are to be laid out following the rules of Microsoft
1476 Visual C/C++, namely: (i) a bit-field won't share the same storage
1477 unit with the previous bit-field if their underlying types have
1478 different sizes, and the bit-field will be aligned to the highest
1479 alignment of the underlying types of itself and of the previous
1480 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1481 the whole enclosing structure, even if it is unnamed; except that
1482 (iii) a zero-sized bit-field will be disregarded unless it follows
1483 another bit-field of nonzero size. If this hook returns @code{true},
1484 other macros that control bit-field layout are ignored.
1486 When a bit-field is inserted into a packed record, the whole size
1487 of the underlying type is used by one or more same-size adjacent
1488 bit-fields (that is, if its long:3, 32 bits is used in the record,
1489 and any additional adjacent long bit-fields are packed into the same
1490 chunk of 32 bits. However, if the size changes, a new field of that
1491 size is allocated). In an unpacked record, this is the same as using
1492 alignment, but not equivalent when packing.
1494 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1495 the latter will take precedence. If @samp{__attribute__((packed))} is
1496 used on a single field when MS bit-fields are in use, it will take
1497 precedence for that field, but the alignment of the rest of the structure
1498 may affect its placement.
1502 @section Layout of Source Language Data Types
1504 These macros define the sizes and other characteristics of the standard
1505 basic data types used in programs being compiled. Unlike the macros in
1506 the previous section, these apply to specific features of C and related
1507 languages, rather than to fundamental aspects of storage layout.
1509 @defmac INT_TYPE_SIZE
1510 A C expression for the size in bits of the type @code{int} on the
1511 target machine. If you don't define this, the default is one word.
1514 @defmac SHORT_TYPE_SIZE
1515 A C expression for the size in bits of the type @code{short} on the
1516 target machine. If you don't define this, the default is half a word.
1517 (If this would be less than one storage unit, it is rounded up to one
1521 @defmac LONG_TYPE_SIZE
1522 A C expression for the size in bits of the type @code{long} on the
1523 target machine. If you don't define this, the default is one word.
1526 @defmac ADA_LONG_TYPE_SIZE
1527 On some machines, the size used for the Ada equivalent of the type
1528 @code{long} by a native Ada compiler differs from that used by C. In
1529 that situation, define this macro to be a C expression to be used for
1530 the size of that type. If you don't define this, the default is the
1531 value of @code{LONG_TYPE_SIZE}.
1534 @defmac MAX_LONG_TYPE_SIZE
1535 Maximum number for the size in bits of the type @code{long} on the
1536 target machine. If this is undefined, the default is
1537 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1538 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1542 @defmac LONG_LONG_TYPE_SIZE
1543 A C expression for the size in bits of the type @code{long long} on the
1544 target machine. If you don't define this, the default is two
1545 words. If you want to support GNU Ada on your machine, the value of this
1546 macro must be at least 64.
1549 @defmac CHAR_TYPE_SIZE
1550 A C expression for the size in bits of the type @code{char} on the
1551 target machine. If you don't define this, the default is
1552 @code{BITS_PER_UNIT}.
1555 @defmac BOOL_TYPE_SIZE
1556 A C expression for the size in bits of the C++ type @code{bool} and
1557 C99 type @code{_Bool} on the target machine. If you don't define
1558 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1561 @defmac FLOAT_TYPE_SIZE
1562 A C expression for the size in bits of the type @code{float} on the
1563 target machine. If you don't define this, the default is one word.
1566 @defmac DOUBLE_TYPE_SIZE
1567 A C expression for the size in bits of the type @code{double} on the
1568 target machine. If you don't define this, the default is two
1572 @defmac LONG_DOUBLE_TYPE_SIZE
1573 A C expression for the size in bits of the type @code{long double} on
1574 the target machine. If you don't define this, the default is two
1578 @defmac MAX_LONG_DOUBLE_TYPE_SIZE
1579 Maximum number for the size in bits of the type @code{long double} on the
1580 target machine. If this is undefined, the default is
1581 @code{LONG_DOUBLE_TYPE_SIZE}. Otherwise, it is the constant value that is
1582 the largest value that @code{LONG_DOUBLE_TYPE_SIZE} can have at run-time.
1583 This is used in @code{cpp}.
1586 @defmac TARGET_FLT_EVAL_METHOD
1587 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1588 assuming, if applicable, that the floating-point control word is in its
1589 default state. If you do not define this macro the value of
1590 @code{FLT_EVAL_METHOD} will be zero.
1593 @defmac WIDEST_HARDWARE_FP_SIZE
1594 A C expression for the size in bits of the widest floating-point format
1595 supported by the hardware. If you define this macro, you must specify a
1596 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1597 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1601 @defmac DEFAULT_SIGNED_CHAR
1602 An expression whose value is 1 or 0, according to whether the type
1603 @code{char} should be signed or unsigned by default. The user can
1604 always override this default with the options @option{-fsigned-char}
1605 and @option{-funsigned-char}.
1608 @defmac DEFAULT_SHORT_ENUMS
1609 A C expression to determine whether to give an @code{enum} type
1610 only as many bytes as it takes to represent the range of possible values
1611 of that type. A nonzero value means to do that; a zero value means all
1612 @code{enum} types should be allocated like @code{int}.
1614 If you don't define the macro, the default is 0.
1618 A C expression for a string describing the name of the data type to use
1619 for size values. The typedef name @code{size_t} is defined using the
1620 contents of the string.
1622 The string can contain more than one keyword. If so, separate them with
1623 spaces, and write first any length keyword, then @code{unsigned} if
1624 appropriate, and finally @code{int}. The string must exactly match one
1625 of the data type names defined in the function
1626 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1627 omit @code{int} or change the order---that would cause the compiler to
1630 If you don't define this macro, the default is @code{"long unsigned
1634 @defmac PTRDIFF_TYPE
1635 A C expression for a string describing the name of the data type to use
1636 for the result of subtracting two pointers. The typedef name
1637 @code{ptrdiff_t} is defined using the contents of the string. See
1638 @code{SIZE_TYPE} above for more information.
1640 If you don't define this macro, the default is @code{"long int"}.
1644 A C expression for a string describing the name of the data type to use
1645 for wide characters. The typedef name @code{wchar_t} is defined using
1646 the contents of the string. See @code{SIZE_TYPE} above for more
1649 If you don't define this macro, the default is @code{"int"}.
1652 @defmac WCHAR_TYPE_SIZE
1653 A C expression for the size in bits of the data type for wide
1654 characters. This is used in @code{cpp}, which cannot make use of
1658 @defmac MAX_WCHAR_TYPE_SIZE
1659 Maximum number for the size in bits of the data type for wide
1660 characters. If this is undefined, the default is
1661 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1662 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1666 @defmac GCOV_TYPE_SIZE
1667 A C expression for the size in bits of the type used for gcov counters on the
1668 target machine. If you don't define this, the default is one
1669 @code{LONG_TYPE_SIZE} in case it is greater or equal to 64-bit and
1670 @code{LONG_LONG_TYPE_SIZE} otherwise. You may want to re-define the type to
1671 ensure atomicity for counters in multithreaded programs.
1675 A C expression for a string describing the name of the data type to
1676 use for wide characters passed to @code{printf} and returned from
1677 @code{getwc}. The typedef name @code{wint_t} is defined using the
1678 contents of the string. See @code{SIZE_TYPE} above for more
1681 If you don't define this macro, the default is @code{"unsigned int"}.
1685 A C expression for a string describing the name of the data type that
1686 can represent any value of any standard or extended signed integer type.
1687 The typedef name @code{intmax_t} is defined using the contents of the
1688 string. See @code{SIZE_TYPE} above for more information.
1690 If you don't define this macro, the default is the first of
1691 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1692 much precision as @code{long long int}.
1695 @defmac UINTMAX_TYPE
1696 A C expression for a string describing the name of the data type that
1697 can represent any value of any standard or extended unsigned integer
1698 type. The typedef name @code{uintmax_t} is defined using the contents
1699 of the string. See @code{SIZE_TYPE} above for more information.
1701 If you don't define this macro, the default is the first of
1702 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1703 unsigned int"} that has as much precision as @code{long long unsigned
1707 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1708 The C++ compiler represents a pointer-to-member-function with a struct
1715 ptrdiff_t vtable_index;
1722 The C++ compiler must use one bit to indicate whether the function that
1723 will be called through a pointer-to-member-function is virtual.
1724 Normally, we assume that the low-order bit of a function pointer must
1725 always be zero. Then, by ensuring that the vtable_index is odd, we can
1726 distinguish which variant of the union is in use. But, on some
1727 platforms function pointers can be odd, and so this doesn't work. In
1728 that case, we use the low-order bit of the @code{delta} field, and shift
1729 the remainder of the @code{delta} field to the left.
1731 GCC will automatically make the right selection about where to store
1732 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1733 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1734 set such that functions always start at even addresses, but the lowest
1735 bit of pointers to functions indicate whether the function at that
1736 address is in ARM or Thumb mode. If this is the case of your
1737 architecture, you should define this macro to
1738 @code{ptrmemfunc_vbit_in_delta}.
1740 In general, you should not have to define this macro. On architectures
1741 in which function addresses are always even, according to
1742 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1743 @code{ptrmemfunc_vbit_in_pfn}.
1746 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1747 Normally, the C++ compiler uses function pointers in vtables. This
1748 macro allows the target to change to use ``function descriptors''
1749 instead. Function descriptors are found on targets for whom a
1750 function pointer is actually a small data structure. Normally the
1751 data structure consists of the actual code address plus a data
1752 pointer to which the function's data is relative.
1754 If vtables are used, the value of this macro should be the number
1755 of words that the function descriptor occupies.
1758 @defmac TARGET_VTABLE_ENTRY_ALIGN
1759 By default, the vtable entries are void pointers, the so the alignment
1760 is the same as pointer alignment. The value of this macro specifies
1761 the alignment of the vtable entry in bits. It should be defined only
1762 when special alignment is necessary. */
1765 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1766 There are a few non-descriptor entries in the vtable at offsets below
1767 zero. If these entries must be padded (say, to preserve the alignment
1768 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1769 of words in each data entry.
1772 @node Escape Sequences
1773 @section Target Character Escape Sequences
1774 @cindex escape sequences
1776 By default, GCC assumes that the C character escape sequences take on
1777 their ASCII values for the target. If this is not correct, you must
1778 explicitly define all of the macros below. All of them must evaluate
1779 to constants; they are used in @code{case} statements.
1785 @findex TARGET_NEWLINE
1788 @multitable {@code{TARGET_NEWLINE}} {Escape} {ASCII character}
1789 @item Macro @tab Escape @tab ASCII character
1790 @item @code{TARGET_BELL} @tab @kbd{\a} @tab @code{07}, @code{BEL}
1791 @item @code{TARGET_CR} @tab @kbd{\r} @tab @code{0D}, @code{CR}
1792 @item @code{TARGET_ESC} @tab @kbd{\e}, @kbd{\E} @tab @code{1B}, @code{ESC}
1793 @item @code{TARGET_FF} @tab @kbd{\f} @tab @code{0C}, @code{FF}
1794 @item @code{TARGET_NEWLINE} @tab @kbd{\n} @tab @code{0A}, @code{LF}
1795 @item @code{TARGET_TAB} @tab @kbd{\t} @tab @code{09}, @code{HT}
1796 @item @code{TARGET_VT} @tab @kbd{\v} @tab @code{0B}, @code{VT}
1800 Note that the @kbd{\e} and @kbd{\E} escapes are GNU extensions, not
1801 part of the C standard.
1804 @section Register Usage
1805 @cindex register usage
1807 This section explains how to describe what registers the target machine
1808 has, and how (in general) they can be used.
1810 The description of which registers a specific instruction can use is
1811 done with register classes; see @ref{Register Classes}. For information
1812 on using registers to access a stack frame, see @ref{Frame Registers}.
1813 For passing values in registers, see @ref{Register Arguments}.
1814 For returning values in registers, see @ref{Scalar Return}.
1817 * Register Basics:: Number and kinds of registers.
1818 * Allocation Order:: Order in which registers are allocated.
1819 * Values in Registers:: What kinds of values each reg can hold.
1820 * Leaf Functions:: Renumbering registers for leaf functions.
1821 * Stack Registers:: Handling a register stack such as 80387.
1824 @node Register Basics
1825 @subsection Basic Characteristics of Registers
1827 @c prevent bad page break with this line
1828 Registers have various characteristics.
1830 @defmac FIRST_PSEUDO_REGISTER
1831 Number of hardware registers known to the compiler. They receive
1832 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1833 pseudo register's number really is assigned the number
1834 @code{FIRST_PSEUDO_REGISTER}.
1837 @defmac FIXED_REGISTERS
1838 @cindex fixed register
1839 An initializer that says which registers are used for fixed purposes
1840 all throughout the compiled code and are therefore not available for
1841 general allocation. These would include the stack pointer, the frame
1842 pointer (except on machines where that can be used as a general
1843 register when no frame pointer is needed), the program counter on
1844 machines where that is considered one of the addressable registers,
1845 and any other numbered register with a standard use.
1847 This information is expressed as a sequence of numbers, separated by
1848 commas and surrounded by braces. The @var{n}th number is 1 if
1849 register @var{n} is fixed, 0 otherwise.
1851 The table initialized from this macro, and the table initialized by
1852 the following one, may be overridden at run time either automatically,
1853 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1854 the user with the command options @option{-ffixed-@var{reg}},
1855 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1858 @defmac CALL_USED_REGISTERS
1859 @cindex call-used register
1860 @cindex call-clobbered register
1861 @cindex call-saved register
1862 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1863 clobbered (in general) by function calls as well as for fixed
1864 registers. This macro therefore identifies the registers that are not
1865 available for general allocation of values that must live across
1868 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1869 automatically saves it on function entry and restores it on function
1870 exit, if the register is used within the function.
1873 @defmac CALL_REALLY_USED_REGISTERS
1874 @cindex call-used register
1875 @cindex call-clobbered register
1876 @cindex call-saved register
1877 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1878 that the entire set of @code{FIXED_REGISTERS} be included.
1879 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1880 This macro is optional. If not specified, it defaults to the value
1881 of @code{CALL_USED_REGISTERS}.
1884 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1885 @cindex call-used register
1886 @cindex call-clobbered register
1887 @cindex call-saved register
1888 A C expression that is nonzero if it is not permissible to store a
1889 value of mode @var{mode} in hard register number @var{regno} across a
1890 call without some part of it being clobbered. For most machines this
1891 macro need not be defined. It is only required for machines that do not
1892 preserve the entire contents of a register across a call.
1896 @findex call_used_regs
1899 @findex reg_class_contents
1900 @defmac CONDITIONAL_REGISTER_USAGE
1901 Zero or more C statements that may conditionally modify five variables
1902 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1903 @code{reg_names}, and @code{reg_class_contents}, to take into account
1904 any dependence of these register sets on target flags. The first three
1905 of these are of type @code{char []} (interpreted as Boolean vectors).
1906 @code{global_regs} is a @code{const char *[]}, and
1907 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1908 called, @code{fixed_regs}, @code{call_used_regs},
1909 @code{reg_class_contents}, and @code{reg_names} have been initialized
1910 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1911 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1912 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1913 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1914 command options have been applied.
1916 You need not define this macro if it has no work to do.
1918 @cindex disabling certain registers
1919 @cindex controlling register usage
1920 If the usage of an entire class of registers depends on the target
1921 flags, you may indicate this to GCC by using this macro to modify
1922 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1923 registers in the classes which should not be used by GCC@. Also define
1924 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1925 to return @code{NO_REGS} if it
1926 is called with a letter for a class that shouldn't be used.
1928 (However, if this class is not included in @code{GENERAL_REGS} and all
1929 of the insn patterns whose constraints permit this class are
1930 controlled by target switches, then GCC will automatically avoid using
1931 these registers when the target switches are opposed to them.)
1934 @defmac NON_SAVING_SETJMP
1935 If this macro is defined and has a nonzero value, it means that
1936 @code{setjmp} and related functions fail to save the registers, or that
1937 @code{longjmp} fails to restore them. To compensate, the compiler
1938 avoids putting variables in registers in functions that use
1942 @defmac INCOMING_REGNO (@var{out})
1943 Define this macro if the target machine has register windows. This C
1944 expression returns the register number as seen by the called function
1945 corresponding to the register number @var{out} as seen by the calling
1946 function. Return @var{out} if register number @var{out} is not an
1950 @defmac OUTGOING_REGNO (@var{in})
1951 Define this macro if the target machine has register windows. This C
1952 expression returns the register number as seen by the calling function
1953 corresponding to the register number @var{in} as seen by the called
1954 function. Return @var{in} if register number @var{in} is not an inbound
1958 @defmac LOCAL_REGNO (@var{regno})
1959 Define this macro if the target machine has register windows. This C
1960 expression returns true if the register is call-saved but is in the
1961 register window. Unlike most call-saved registers, such registers
1962 need not be explicitly restored on function exit or during non-local
1967 If the program counter has a register number, define this as that
1968 register number. Otherwise, do not define it.
1971 @node Allocation Order
1972 @subsection Order of Allocation of Registers
1973 @cindex order of register allocation
1974 @cindex register allocation order
1976 @c prevent bad page break with this line
1977 Registers are allocated in order.
1979 @defmac REG_ALLOC_ORDER
1980 If defined, an initializer for a vector of integers, containing the
1981 numbers of hard registers in the order in which GCC should prefer
1982 to use them (from most preferred to least).
1984 If this macro is not defined, registers are used lowest numbered first
1985 (all else being equal).
1987 One use of this macro is on machines where the highest numbered
1988 registers must always be saved and the save-multiple-registers
1989 instruction supports only sequences of consecutive registers. On such
1990 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1991 the highest numbered allocable register first.
1994 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
1995 A C statement (sans semicolon) to choose the order in which to allocate
1996 hard registers for pseudo-registers local to a basic block.
1998 Store the desired register order in the array @code{reg_alloc_order}.
1999 Element 0 should be the register to allocate first; element 1, the next
2000 register; and so on.
2002 The macro body should not assume anything about the contents of
2003 @code{reg_alloc_order} before execution of the macro.
2005 On most machines, it is not necessary to define this macro.
2008 @node Values in Registers
2009 @subsection How Values Fit in Registers
2011 This section discusses the macros that describe which kinds of values
2012 (specifically, which machine modes) each register can hold, and how many
2013 consecutive registers are needed for a given mode.
2015 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2016 A C expression for the number of consecutive hard registers, starting
2017 at register number @var{regno}, required to hold a value of mode
2020 On a machine where all registers are exactly one word, a suitable
2021 definition of this macro is
2024 #define HARD_REGNO_NREGS(REGNO, MODE) \
2025 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2030 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2031 A C expression that is nonzero if it is permissible to store a value
2032 of mode @var{mode} in hard register number @var{regno} (or in several
2033 registers starting with that one). For a machine where all registers
2034 are equivalent, a suitable definition is
2037 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2040 You need not include code to check for the numbers of fixed registers,
2041 because the allocation mechanism considers them to be always occupied.
2043 @cindex register pairs
2044 On some machines, double-precision values must be kept in even/odd
2045 register pairs. You can implement that by defining this macro to reject
2046 odd register numbers for such modes.
2048 The minimum requirement for a mode to be OK in a register is that the
2049 @samp{mov@var{mode}} instruction pattern support moves between the
2050 register and other hard register in the same class and that moving a
2051 value into the register and back out not alter it.
2053 Since the same instruction used to move @code{word_mode} will work for
2054 all narrower integer modes, it is not necessary on any machine for
2055 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2056 you define patterns @samp{movhi}, etc., to take advantage of this. This
2057 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2058 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2061 Many machines have special registers for floating point arithmetic.
2062 Often people assume that floating point machine modes are allowed only
2063 in floating point registers. This is not true. Any registers that
2064 can hold integers can safely @emph{hold} a floating point machine
2065 mode, whether or not floating arithmetic can be done on it in those
2066 registers. Integer move instructions can be used to move the values.
2068 On some machines, though, the converse is true: fixed-point machine
2069 modes may not go in floating registers. This is true if the floating
2070 registers normalize any value stored in them, because storing a
2071 non-floating value there would garble it. In this case,
2072 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2073 floating registers. But if the floating registers do not automatically
2074 normalize, if you can store any bit pattern in one and retrieve it
2075 unchanged without a trap, then any machine mode may go in a floating
2076 register, so you can define this macro to say so.
2078 The primary significance of special floating registers is rather that
2079 they are the registers acceptable in floating point arithmetic
2080 instructions. However, this is of no concern to
2081 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2082 constraints for those instructions.
2084 On some machines, the floating registers are especially slow to access,
2085 so that it is better to store a value in a stack frame than in such a
2086 register if floating point arithmetic is not being done. As long as the
2087 floating registers are not in class @code{GENERAL_REGS}, they will not
2088 be used unless some pattern's constraint asks for one.
2091 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2092 A C expression that is nonzero if a value of mode
2093 @var{mode1} is accessible in mode @var{mode2} without copying.
2095 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2096 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2097 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2098 should be nonzero. If they differ for any @var{r}, you should define
2099 this macro to return zero unless some other mechanism ensures the
2100 accessibility of the value in a narrower mode.
2102 You should define this macro to return nonzero in as many cases as
2103 possible since doing so will allow GCC to perform better register
2107 @defmac AVOID_CCMODE_COPIES
2108 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2109 registers. You should only define this macro if support for copying to/from
2110 @code{CCmode} is incomplete.
2113 @node Leaf Functions
2114 @subsection Handling Leaf Functions
2116 @cindex leaf functions
2117 @cindex functions, leaf
2118 On some machines, a leaf function (i.e., one which makes no calls) can run
2119 more efficiently if it does not make its own register window. Often this
2120 means it is required to receive its arguments in the registers where they
2121 are passed by the caller, instead of the registers where they would
2124 The special treatment for leaf functions generally applies only when
2125 other conditions are met; for example, often they may use only those
2126 registers for its own variables and temporaries. We use the term ``leaf
2127 function'' to mean a function that is suitable for this special
2128 handling, so that functions with no calls are not necessarily ``leaf
2131 GCC assigns register numbers before it knows whether the function is
2132 suitable for leaf function treatment. So it needs to renumber the
2133 registers in order to output a leaf function. The following macros
2136 @defmac LEAF_REGISTERS
2137 Name of a char vector, indexed by hard register number, which
2138 contains 1 for a register that is allowable in a candidate for leaf
2141 If leaf function treatment involves renumbering the registers, then the
2142 registers marked here should be the ones before renumbering---those that
2143 GCC would ordinarily allocate. The registers which will actually be
2144 used in the assembler code, after renumbering, should not be marked with 1
2147 Define this macro only if the target machine offers a way to optimize
2148 the treatment of leaf functions.
2151 @defmac LEAF_REG_REMAP (@var{regno})
2152 A C expression whose value is the register number to which @var{regno}
2153 should be renumbered, when a function is treated as a leaf function.
2155 If @var{regno} is a register number which should not appear in a leaf
2156 function before renumbering, then the expression should yield @minus{}1, which
2157 will cause the compiler to abort.
2159 Define this macro only if the target machine offers a way to optimize the
2160 treatment of leaf functions, and registers need to be renumbered to do
2164 @findex current_function_is_leaf
2165 @findex current_function_uses_only_leaf_regs
2166 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2167 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2168 specially. They can test the C variable @code{current_function_is_leaf}
2169 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2170 set prior to local register allocation and is valid for the remaining
2171 compiler passes. They can also test the C variable
2172 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2173 functions which only use leaf registers.
2174 @code{current_function_uses_only_leaf_regs} is valid after reload and is
2175 only useful if @code{LEAF_REGISTERS} is defined.
2176 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2177 @c of the next paragraph?! --mew 2feb93
2179 @node Stack Registers
2180 @subsection Registers That Form a Stack
2182 There are special features to handle computers where some of the
2183 ``registers'' form a stack. Stack registers are normally written by
2184 pushing onto the stack, and are numbered relative to the top of the
2187 Currently, GCC can only handle one group of stack-like registers, and
2188 they must be consecutively numbered. Furthermore, the existing
2189 support for stack-like registers is specific to the 80387 floating
2190 point coprocessor. If you have a new architecture that uses
2191 stack-like registers, you will need to do substantial work on
2192 @file{reg-stack.c} and write your machine description to cooperate
2193 with it, as well as defining these macros.
2196 Define this if the machine has any stack-like registers.
2199 @defmac FIRST_STACK_REG
2200 The number of the first stack-like register. This one is the top
2204 @defmac LAST_STACK_REG
2205 The number of the last stack-like register. This one is the bottom of
2209 @node Register Classes
2210 @section Register Classes
2211 @cindex register class definitions
2212 @cindex class definitions, register
2214 On many machines, the numbered registers are not all equivalent.
2215 For example, certain registers may not be allowed for indexed addressing;
2216 certain registers may not be allowed in some instructions. These machine
2217 restrictions are described to the compiler using @dfn{register classes}.
2219 You define a number of register classes, giving each one a name and saying
2220 which of the registers belong to it. Then you can specify register classes
2221 that are allowed as operands to particular instruction patterns.
2225 In general, each register will belong to several classes. In fact, one
2226 class must be named @code{ALL_REGS} and contain all the registers. Another
2227 class must be named @code{NO_REGS} and contain no registers. Often the
2228 union of two classes will be another class; however, this is not required.
2230 @findex GENERAL_REGS
2231 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2232 terribly special about the name, but the operand constraint letters
2233 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2234 the same as @code{ALL_REGS}, just define it as a macro which expands
2237 Order the classes so that if class @var{x} is contained in class @var{y}
2238 then @var{x} has a lower class number than @var{y}.
2240 The way classes other than @code{GENERAL_REGS} are specified in operand
2241 constraints is through machine-dependent operand constraint letters.
2242 You can define such letters to correspond to various classes, then use
2243 them in operand constraints.
2245 You should define a class for the union of two classes whenever some
2246 instruction allows both classes. For example, if an instruction allows
2247 either a floating point (coprocessor) register or a general register for a
2248 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2249 which includes both of them. Otherwise you will get suboptimal code.
2251 You must also specify certain redundant information about the register
2252 classes: for each class, which classes contain it and which ones are
2253 contained in it; for each pair of classes, the largest class contained
2256 When a value occupying several consecutive registers is expected in a
2257 certain class, all the registers used must belong to that class.
2258 Therefore, register classes cannot be used to enforce a requirement for
2259 a register pair to start with an even-numbered register. The way to
2260 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2262 Register classes used for input-operands of bitwise-and or shift
2263 instructions have a special requirement: each such class must have, for
2264 each fixed-point machine mode, a subclass whose registers can transfer that
2265 mode to or from memory. For example, on some machines, the operations for
2266 single-byte values (@code{QImode}) are limited to certain registers. When
2267 this is so, each register class that is used in a bitwise-and or shift
2268 instruction must have a subclass consisting of registers from which
2269 single-byte values can be loaded or stored. This is so that
2270 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2272 @deftp {Data type} {enum reg_class}
2273 An enumeral type that must be defined with all the register class names
2274 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
2275 must be the last register class, followed by one more enumeral value,
2276 @code{LIM_REG_CLASSES}, which is not a register class but rather
2277 tells how many classes there are.
2279 Each register class has a number, which is the value of casting
2280 the class name to type @code{int}. The number serves as an index
2281 in many of the tables described below.
2284 @defmac N_REG_CLASSES
2285 The number of distinct register classes, defined as follows:
2288 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2292 @defmac REG_CLASS_NAMES
2293 An initializer containing the names of the register classes as C string
2294 constants. These names are used in writing some of the debugging dumps.
2297 @defmac REG_CLASS_CONTENTS
2298 An initializer containing the contents of the register classes, as integers
2299 which are bit masks. The @var{n}th integer specifies the contents of class
2300 @var{n}. The way the integer @var{mask} is interpreted is that
2301 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2303 When the machine has more than 32 registers, an integer does not suffice.
2304 Then the integers are replaced by sub-initializers, braced groupings containing
2305 several integers. Each sub-initializer must be suitable as an initializer
2306 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2307 In this situation, the first integer in each sub-initializer corresponds to
2308 registers 0 through 31, the second integer to registers 32 through 63, and
2312 @defmac REGNO_REG_CLASS (@var{regno})
2313 A C expression whose value is a register class containing hard register
2314 @var{regno}. In general there is more than one such class; choose a class
2315 which is @dfn{minimal}, meaning that no smaller class also contains the
2319 @defmac BASE_REG_CLASS
2320 A macro whose definition is the name of the class to which a valid
2321 base register must belong. A base register is one used in an address
2322 which is the register value plus a displacement.
2325 @defmac MODE_BASE_REG_CLASS (@var{mode})
2326 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2327 the selection of a base register in a mode dependent manner. If
2328 @var{mode} is VOIDmode then it should return the same value as
2329 @code{BASE_REG_CLASS}.
2332 @defmac INDEX_REG_CLASS
2333 A macro whose definition is the name of the class to which a valid
2334 index register must belong. An index register is one used in an
2335 address where its value is either multiplied by a scale factor or
2336 added to another register (as well as added to a displacement).
2339 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2340 For the constraint at the start of @var{str}, which starts with the letter
2341 @var{c}, return the length. This allows you to have register class /
2342 constant / extra constraints that are longer than a single letter;
2343 you don't need to define this macro if you can do with single-letter
2344 constraints only. The definition of this macro should use
2345 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2346 to handle specially.
2347 There are some sanity checks in genoutput.c that check the constraint lengths
2348 for the md file, so you can also use this macro to help you while you are
2349 transitioning from a byzantine single-letter-constraint scheme: when you
2350 return a negative length for a constraint you want to re-use, genoutput
2351 will complain about every instance where it is used in the md file.
2354 @defmac REG_CLASS_FROM_LETTER (@var{char})
2355 A C expression which defines the machine-dependent operand constraint
2356 letters for register classes. If @var{char} is such a letter, the
2357 value should be the register class corresponding to it. Otherwise,
2358 the value should be @code{NO_REGS}. The register letter @samp{r},
2359 corresponding to class @code{GENERAL_REGS}, will not be passed
2360 to this macro; you do not need to handle it.
2363 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2364 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2365 passed in @var{str}, so that you can use suffixes to distinguish between
2369 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2370 A C expression which is nonzero if register number @var{num} is
2371 suitable for use as a base register in operand addresses. It may be
2372 either a suitable hard register or a pseudo register that has been
2373 allocated such a hard register.
2376 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2377 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2378 that expression may examine the mode of the memory reference in
2379 @var{mode}. You should define this macro if the mode of the memory
2380 reference affects whether a register may be used as a base register. If
2381 you define this macro, the compiler will use it instead of
2382 @code{REGNO_OK_FOR_BASE_P}.
2385 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2386 A C expression which is nonzero if register number @var{num} is
2387 suitable for use as an index register in operand addresses. It may be
2388 either a suitable hard register or a pseudo register that has been
2389 allocated such a hard register.
2391 The difference between an index register and a base register is that
2392 the index register may be scaled. If an address involves the sum of
2393 two registers, neither one of them scaled, then either one may be
2394 labeled the ``base'' and the other the ``index''; but whichever
2395 labeling is used must fit the machine's constraints of which registers
2396 may serve in each capacity. The compiler will try both labelings,
2397 looking for one that is valid, and will reload one or both registers
2398 only if neither labeling works.
2401 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2402 A C expression that places additional restrictions on the register class
2403 to use when it is necessary to copy value @var{x} into a register in class
2404 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2405 another, smaller class. On many machines, the following definition is
2409 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2412 Sometimes returning a more restrictive class makes better code. For
2413 example, on the 68000, when @var{x} is an integer constant that is in range
2414 for a @samp{moveq} instruction, the value of this macro is always
2415 @code{DATA_REGS} as long as @var{class} includes the data registers.
2416 Requiring a data register guarantees that a @samp{moveq} will be used.
2418 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2419 you can force @var{x} into a memory constant. This is useful on
2420 certain machines where immediate floating values cannot be loaded into
2421 certain kinds of registers.
2424 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2425 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2426 input reloads. If you don't define this macro, the default is to use
2427 @var{class}, unchanged.
2430 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2431 A C expression that places additional restrictions on the register class
2432 to use when it is necessary to be able to hold a value of mode
2433 @var{mode} in a reload register for which class @var{class} would
2436 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2437 there are certain modes that simply can't go in certain reload classes.
2439 The value is a register class; perhaps @var{class}, or perhaps another,
2442 Don't define this macro unless the target machine has limitations which
2443 require the macro to do something nontrivial.
2446 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2447 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2448 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2449 Many machines have some registers that cannot be copied directly to or
2450 from memory or even from other types of registers. An example is the
2451 @samp{MQ} register, which on most machines, can only be copied to or
2452 from general registers, but not memory. Some machines allow copying all
2453 registers to and from memory, but require a scratch register for stores
2454 to some memory locations (e.g., those with symbolic address on the RT,
2455 and those with certain symbolic address on the SPARC when compiling
2456 PIC)@. In some cases, both an intermediate and a scratch register are
2459 You should define these macros to indicate to the reload phase that it may
2460 need to allocate at least one register for a reload in addition to the
2461 register to contain the data. Specifically, if copying @var{x} to a
2462 register @var{class} in @var{mode} requires an intermediate register,
2463 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2464 largest register class all of whose registers can be used as
2465 intermediate registers or scratch registers.
2467 If copying a register @var{class} in @var{mode} to @var{x} requires an
2468 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2469 should be defined to return the largest register class required. If the
2470 requirements for input and output reloads are the same, the macro
2471 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2474 The values returned by these macros are often @code{GENERAL_REGS}.
2475 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2476 can be directly copied to or from a register of @var{class} in
2477 @var{mode} without requiring a scratch register. Do not define this
2478 macro if it would always return @code{NO_REGS}.
2480 If a scratch register is required (either with or without an
2481 intermediate register), you should define patterns for
2482 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2483 (@pxref{Standard Names}. These patterns, which will normally be
2484 implemented with a @code{define_expand}, should be similar to the
2485 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2488 Define constraints for the reload register and scratch register that
2489 contain a single register class. If the original reload register (whose
2490 class is @var{class}) can meet the constraint given in the pattern, the
2491 value returned by these macros is used for the class of the scratch
2492 register. Otherwise, two additional reload registers are required.
2493 Their classes are obtained from the constraints in the insn pattern.
2495 @var{x} might be a pseudo-register or a @code{subreg} of a
2496 pseudo-register, which could either be in a hard register or in memory.
2497 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2498 in memory and the hard register number if it is in a register.
2500 These macros should not be used in the case where a particular class of
2501 registers can only be copied to memory and not to another class of
2502 registers. In that case, secondary reload registers are not needed and
2503 would not be helpful. Instead, a stack location must be used to perform
2504 the copy and the @code{mov@var{m}} pattern should use memory as an
2505 intermediate storage. This case often occurs between floating-point and
2509 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2510 Certain machines have the property that some registers cannot be copied
2511 to some other registers without using memory. Define this macro on
2512 those machines to be a C expression that is nonzero if objects of mode
2513 @var{m} in registers of @var{class1} can only be copied to registers of
2514 class @var{class2} by storing a register of @var{class1} into memory
2515 and loading that memory location into a register of @var{class2}.
2517 Do not define this macro if its value would always be zero.
2520 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2521 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2522 allocates a stack slot for a memory location needed for register copies.
2523 If this macro is defined, the compiler instead uses the memory location
2524 defined by this macro.
2526 Do not define this macro if you do not define
2527 @code{SECONDARY_MEMORY_NEEDED}.
2530 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2531 When the compiler needs a secondary memory location to copy between two
2532 registers of mode @var{mode}, it normally allocates sufficient memory to
2533 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2534 load operations in a mode that many bits wide and whose class is the
2535 same as that of @var{mode}.
2537 This is right thing to do on most machines because it ensures that all
2538 bits of the register are copied and prevents accesses to the registers
2539 in a narrower mode, which some machines prohibit for floating-point
2542 However, this default behavior is not correct on some machines, such as
2543 the DEC Alpha, that store short integers in floating-point registers
2544 differently than in integer registers. On those machines, the default
2545 widening will not work correctly and you must define this macro to
2546 suppress that widening in some cases. See the file @file{alpha.h} for
2549 Do not define this macro if you do not define
2550 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2551 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2554 @defmac SMALL_REGISTER_CLASSES
2555 On some machines, it is risky to let hard registers live across arbitrary
2556 insns. Typically, these machines have instructions that require values
2557 to be in specific registers (like an accumulator), and reload will fail
2558 if the required hard register is used for another purpose across such an
2561 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2562 value on these machines. When this macro has a nonzero value, the
2563 compiler will try to minimize the lifetime of hard registers.
2565 It is always safe to define this macro with a nonzero value, but if you
2566 unnecessarily define it, you will reduce the amount of optimizations
2567 that can be performed in some cases. If you do not define this macro
2568 with a nonzero value when it is required, the compiler will run out of
2569 spill registers and print a fatal error message. For most machines, you
2570 should not define this macro at all.
2573 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2574 A C expression whose value is nonzero if pseudos that have been assigned
2575 to registers of class @var{class} would likely be spilled because
2576 registers of @var{class} are needed for spill registers.
2578 The default value of this macro returns 1 if @var{class} has exactly one
2579 register and zero otherwise. On most machines, this default should be
2580 used. Only define this macro to some other expression if pseudos
2581 allocated by @file{local-alloc.c} end up in memory because their hard
2582 registers were needed for spill registers. If this macro returns nonzero
2583 for those classes, those pseudos will only be allocated by
2584 @file{global.c}, which knows how to reallocate the pseudo to another
2585 register. If there would not be another register available for
2586 reallocation, you should not change the definition of this macro since
2587 the only effect of such a definition would be to slow down register
2591 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2592 A C expression for the maximum number of consecutive registers
2593 of class @var{class} needed to hold a value of mode @var{mode}.
2595 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2596 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2597 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2598 @var{mode})} for all @var{regno} values in the class @var{class}.
2600 This macro helps control the handling of multiple-word values
2604 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2605 If defined, a C expression that returns nonzero for a @var{class} for which
2606 a change from mode @var{from} to mode @var{to} is invalid.
2608 For the example, loading 32-bit integer or floating-point objects into
2609 floating-point registers on the Alpha extends them to 64 bits.
2610 Therefore loading a 64-bit object and then storing it as a 32-bit object
2611 does not store the low-order 32 bits, as would be the case for a normal
2612 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2616 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2617 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2618 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2622 Three other special macros describe which operands fit which constraint
2625 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2626 A C expression that defines the machine-dependent operand constraint
2627 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2628 particular ranges of integer values. If @var{c} is one of those
2629 letters, the expression should check that @var{value}, an integer, is in
2630 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2631 not one of those letters, the value should be 0 regardless of
2635 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2636 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2637 string passed in @var{str}, so that you can use suffixes to distinguish
2638 between different variants.
2641 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2642 A C expression that defines the machine-dependent operand constraint
2643 letters that specify particular ranges of @code{const_double} values
2644 (@samp{G} or @samp{H}).
2646 If @var{c} is one of those letters, the expression should check that
2647 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2648 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2649 letters, the value should be 0 regardless of @var{value}.
2651 @code{const_double} is used for all floating-point constants and for
2652 @code{DImode} fixed-point constants. A given letter can accept either
2653 or both kinds of values. It can use @code{GET_MODE} to distinguish
2654 between these kinds.
2657 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2658 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2659 string passed in @var{str}, so that you can use suffixes to distinguish
2660 between different variants.
2663 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2664 A C expression that defines the optional machine-dependent constraint
2665 letters that can be used to segregate specific types of operands, usually
2666 memory references, for the target machine. Any letter that is not
2667 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2668 @code{REG_CLASS_FROM_CONSTRAINT}
2669 may be used. Normally this macro will not be defined.
2671 If it is required for a particular target machine, it should return 1
2672 if @var{value} corresponds to the operand type represented by the
2673 constraint letter @var{c}. If @var{c} is not defined as an extra
2674 constraint, the value returned should be 0 regardless of @var{value}.
2676 For example, on the ROMP, load instructions cannot have their output
2677 in r0 if the memory reference contains a symbolic address. Constraint
2678 letter @samp{Q} is defined as representing a memory address that does
2679 @emph{not} contain a symbolic address. An alternative is specified with
2680 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2681 alternative specifies @samp{m} on the input and a register class that
2682 does not include r0 on the output.
2685 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2686 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2687 in @var{str}, so that you can use suffixes to distinguish between different
2691 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2692 A C expression that defines the optional machine-dependent constraint
2693 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2694 be treated like memory constraints by the reload pass.
2696 It should return 1 if the operand type represented by the constraint
2697 at the start of @var{str}, the first letter of which is the letter @var{c},
2698 comprises a subset of all memory references including
2699 all those whose address is simply a base register. This allows the reload
2700 pass to reload an operand, if it does not directly correspond to the operand
2701 type of @var{c}, by copying its address into a base register.
2703 For example, on the S/390, some instructions do not accept arbitrary
2704 memory references, but only those that do not make use of an index
2705 register. The constraint letter @samp{Q} is defined via
2706 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2707 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2708 a @samp{Q} constraint can handle any memory operand, because the
2709 reload pass knows it can be reloaded by copying the memory address
2710 into a base register if required. This is analogous to the way
2711 a @samp{o} constraint can handle any memory operand.
2714 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2715 A C expression that defines the optional machine-dependent constraint
2716 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2717 @code{EXTRA_CONSTRAINT_STR}, that should
2718 be treated like address constraints by the reload pass.
2720 It should return 1 if the operand type represented by the constraint
2721 at the start of @var{str}, which starts with the letter @var{c}, comprises
2722 a subset of all memory addresses including
2723 all those that consist of just a base register. This allows the reload
2724 pass to reload an operand, if it does not directly correspond to the operand
2725 type of @var{str}, by copying it into a base register.
2727 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2728 be used with the @code{address_operand} predicate. It is treated
2729 analogously to the @samp{p} constraint.
2732 @node Stack and Calling
2733 @section Stack Layout and Calling Conventions
2734 @cindex calling conventions
2736 @c prevent bad page break with this line
2737 This describes the stack layout and calling conventions.
2741 * Exception Handling::
2746 * Register Arguments::
2748 * Aggregate Return::
2756 @subsection Basic Stack Layout
2757 @cindex stack frame layout
2758 @cindex frame layout
2760 @c prevent bad page break with this line
2761 Here is the basic stack layout.
2763 @defmac STACK_GROWS_DOWNWARD
2764 Define this macro if pushing a word onto the stack moves the stack
2765 pointer to a smaller address.
2767 When we say, ``define this macro if @dots{},'' it means that the
2768 compiler checks this macro only with @code{#ifdef} so the precise
2769 definition used does not matter.
2772 @defmac STACK_PUSH_CODE
2773 This macro defines the operation used when something is pushed
2774 on the stack. In RTL, a push operation will be
2775 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2777 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2778 and @code{POST_INC}. Which of these is correct depends on
2779 the stack direction and on whether the stack pointer points
2780 to the last item on the stack or whether it points to the
2781 space for the next item on the stack.
2783 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2784 defined, which is almost always right, and @code{PRE_INC} otherwise,
2785 which is often wrong.
2788 @defmac FRAME_GROWS_DOWNWARD
2789 Define this macro if the addresses of local variable slots are at negative
2790 offsets from the frame pointer.
2793 @defmac ARGS_GROW_DOWNWARD
2794 Define this macro if successive arguments to a function occupy decreasing
2795 addresses on the stack.
2798 @defmac STARTING_FRAME_OFFSET
2799 Offset from the frame pointer to the first local variable slot to be allocated.
2801 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2802 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2803 Otherwise, it is found by adding the length of the first slot to the
2804 value @code{STARTING_FRAME_OFFSET}.
2805 @c i'm not sure if the above is still correct.. had to change it to get
2806 @c rid of an overfull. --mew 2feb93
2809 @defmac STACK_ALIGNMENT_NEEDED
2810 Define to zero to disable final alignment of the stack during reload.
2811 The nonzero default for this macro is suitable for most ports.
2813 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2814 is a register save block following the local block that doesn't require
2815 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2816 stack alignment and do it in the backend.
2819 @defmac STACK_POINTER_OFFSET
2820 Offset from the stack pointer register to the first location at which
2821 outgoing arguments are placed. If not specified, the default value of
2822 zero is used. This is the proper value for most machines.
2824 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2825 the first location at which outgoing arguments are placed.
2828 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2829 Offset from the argument pointer register to the first argument's
2830 address. On some machines it may depend on the data type of the
2833 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2834 the first argument's address.
2837 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2838 Offset from the stack pointer register to an item dynamically allocated
2839 on the stack, e.g., by @code{alloca}.
2841 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2842 length of the outgoing arguments. The default is correct for most
2843 machines. See @file{function.c} for details.
2846 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2847 A C expression whose value is RTL representing the address in a stack
2848 frame where the pointer to the caller's frame is stored. Assume that
2849 @var{frameaddr} is an RTL expression for the address of the stack frame
2852 If you don't define this macro, the default is to return the value
2853 of @var{frameaddr}---that is, the stack frame address is also the
2854 address of the stack word that points to the previous frame.
2857 @defmac SETUP_FRAME_ADDRESSES
2858 If defined, a C expression that produces the machine-specific code to
2859 setup the stack so that arbitrary frames can be accessed. For example,
2860 on the SPARC, we must flush all of the register windows to the stack
2861 before we can access arbitrary stack frames. You will seldom need to
2865 @defmac BUILTIN_SETJMP_FRAME_VALUE
2866 If defined, a C expression that contains an rtx that is used to store
2867 the address of the current frame into the built in @code{setjmp} buffer.
2868 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2869 machines. One reason you may need to define this macro is if
2870 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2873 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2874 A C expression whose value is RTL representing the value of the return
2875 address for the frame @var{count} steps up from the current frame, after
2876 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2877 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2878 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2880 The value of the expression must always be the correct address when
2881 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2882 determine the return address of other frames.
2885 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2886 Define this if the return address of a particular stack frame is accessed
2887 from the frame pointer of the previous stack frame.
2890 @defmac INCOMING_RETURN_ADDR_RTX
2891 A C expression whose value is RTL representing the location of the
2892 incoming return address at the beginning of any function, before the
2893 prologue. This RTL is either a @code{REG}, indicating that the return
2894 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2897 You only need to define this macro if you want to support call frame
2898 debugging information like that provided by DWARF 2.
2900 If this RTL is a @code{REG}, you should also define
2901 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2904 @defmac INCOMING_FRAME_SP_OFFSET
2905 A C expression whose value is an integer giving the offset, in bytes,
2906 from the value of the stack pointer register to the top of the stack
2907 frame at the beginning of any function, before the prologue. The top of
2908 the frame is defined to be the value of the stack pointer in the
2909 previous frame, just before the call instruction.
2911 You only need to define this macro if you want to support call frame
2912 debugging information like that provided by DWARF 2.
2915 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2916 A C expression whose value is an integer giving the offset, in bytes,
2917 from the argument pointer to the canonical frame address (cfa). The
2918 final value should coincide with that calculated by
2919 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2920 during virtual register instantiation.
2922 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2923 which is correct for most machines; in general, the arguments are found
2924 immediately before the stack frame. Note that this is not the case on
2925 some targets that save registers into the caller's frame, such as SPARC
2926 and rs6000, and so such targets need to define this macro.
2928 You only need to define this macro if the default is incorrect, and you
2929 want to support call frame debugging information like that provided by
2933 @node Exception Handling
2934 @subsection Exception Handling Support
2935 @cindex exception handling
2937 @defmac EH_RETURN_DATA_REGNO (@var{N})
2938 A C expression whose value is the @var{N}th register number used for
2939 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2940 @var{N} registers are usable.
2942 The exception handling library routines communicate with the exception
2943 handlers via a set of agreed upon registers. Ideally these registers
2944 should be call-clobbered; it is possible to use call-saved registers,
2945 but may negatively impact code size. The target must support at least
2946 2 data registers, but should define 4 if there are enough free registers.
2948 You must define this macro if you want to support call frame exception
2949 handling like that provided by DWARF 2.
2952 @defmac EH_RETURN_STACKADJ_RTX
2953 A C expression whose value is RTL representing a location in which
2954 to store a stack adjustment to be applied before function return.
2955 This is used to unwind the stack to an exception handler's call frame.
2956 It will be assigned zero on code paths that return normally.
2958 Typically this is a call-clobbered hard register that is otherwise
2959 untouched by the epilogue, but could also be a stack slot.
2961 Do not define this macro if the stack pointer is saved and restored
2962 by the regular prolog and epilog code in the call frame itself; in
2963 this case, the exception handling library routines will update the
2964 stack location to be restored in place. Otherwise, you must define
2965 this macro if you want to support call frame exception handling like
2966 that provided by DWARF 2.
2969 @defmac EH_RETURN_HANDLER_RTX
2970 A C expression whose value is RTL representing a location in which
2971 to store the address of an exception handler to which we should
2972 return. It will not be assigned on code paths that return normally.
2974 Typically this is the location in the call frame at which the normal
2975 return address is stored. For targets that return by popping an
2976 address off the stack, this might be a memory address just below
2977 the @emph{target} call frame rather than inside the current call
2978 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
2979 been assigned, so it may be used to calculate the location of the
2982 Some targets have more complex requirements than storing to an
2983 address calculable during initial code generation. In that case
2984 the @code{eh_return} instruction pattern should be used instead.
2986 If you want to support call frame exception handling, you must
2987 define either this macro or the @code{eh_return} instruction pattern.
2990 @defmac RETURN_ADDR_OFFSET
2991 If defined, an integer-valued C expression for which rtl will be generated
2992 to add it to the exception handler address before it is searched in the
2993 exception handling tables, and to subtract it again from the address before
2994 using it to return to the exception handler.
2997 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
2998 This macro chooses the encoding of pointers embedded in the exception
2999 handling sections. If at all possible, this should be defined such
3000 that the exception handling section will not require dynamic relocations,
3001 and so may be read-only.
3003 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3004 @var{global} is true if the symbol may be affected by dynamic relocations.
3005 The macro should return a combination of the @code{DW_EH_PE_*} defines
3006 as found in @file{dwarf2.h}.
3008 If this macro is not defined, pointers will not be encoded but
3009 represented directly.
3012 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3013 This macro allows the target to emit whatever special magic is required
3014 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3015 Generic code takes care of pc-relative and indirect encodings; this must
3016 be defined if the target uses text-relative or data-relative encodings.
3018 This is a C statement that branches to @var{done} if the format was
3019 handled. @var{encoding} is the format chosen, @var{size} is the number
3020 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3024 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs}, @var{success})
3025 This macro allows the target to add cpu and operating system specific
3026 code to the call-frame unwinder for use when there is no unwind data
3027 available. The most common reason to implement this macro is to unwind
3028 through signal frames.
3030 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3031 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3032 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3033 for the address of the code being executed and @code{context->cfa} for
3034 the stack pointer value. If the frame can be decoded, the register save
3035 addresses should be updated in @var{fs} and the macro should branch to
3036 @var{success}. If the frame cannot be decoded, the macro should do
3039 For proper signal handling in Java this macro is accompanied by
3040 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3043 @node Stack Checking
3044 @subsection Specifying How Stack Checking is Done
3046 GCC will check that stack references are within the boundaries of
3047 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3051 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3052 will assume that you have arranged for stack checking to be done at
3053 appropriate places in the configuration files, e.g., in
3054 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3058 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3059 called @code{check_stack} in your @file{md} file, GCC will call that
3060 pattern with one argument which is the address to compare the stack
3061 value against. You must arrange for this pattern to report an error if
3062 the stack pointer is out of range.
3065 If neither of the above are true, GCC will generate code to periodically
3066 ``probe'' the stack pointer using the values of the macros defined below.
3069 Normally, you will use the default values of these macros, so GCC
3070 will use the third approach.
3072 @defmac STACK_CHECK_BUILTIN
3073 A nonzero value if stack checking is done by the configuration files in a
3074 machine-dependent manner. You should define this macro if stack checking
3075 is require by the ABI of your machine or if you would like to have to stack
3076 checking in some more efficient way than GCC's portable approach.
3077 The default value of this macro is zero.
3080 @defmac STACK_CHECK_PROBE_INTERVAL
3081 An integer representing the interval at which GCC must generate stack
3082 probe instructions. You will normally define this macro to be no larger
3083 than the size of the ``guard pages'' at the end of a stack area. The
3084 default value of 4096 is suitable for most systems.
3087 @defmac STACK_CHECK_PROBE_LOAD
3088 A integer which is nonzero if GCC should perform the stack probe
3089 as a load instruction and zero if GCC should use a store instruction.
3090 The default is zero, which is the most efficient choice on most systems.
3093 @defmac STACK_CHECK_PROTECT
3094 The number of bytes of stack needed to recover from a stack overflow,
3095 for languages where such a recovery is supported. The default value of
3096 75 words should be adequate for most machines.
3099 @defmac STACK_CHECK_MAX_FRAME_SIZE
3100 The maximum size of a stack frame, in bytes. GCC will generate probe
3101 instructions in non-leaf functions to ensure at least this many bytes of
3102 stack are available. If a stack frame is larger than this size, stack
3103 checking will not be reliable and GCC will issue a warning. The
3104 default is chosen so that GCC only generates one instruction on most
3105 systems. You should normally not change the default value of this macro.
3108 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3109 GCC uses this value to generate the above warning message. It
3110 represents the amount of fixed frame used by a function, not including
3111 space for any callee-saved registers, temporaries and user variables.
3112 You need only specify an upper bound for this amount and will normally
3113 use the default of four words.
3116 @defmac STACK_CHECK_MAX_VAR_SIZE
3117 The maximum size, in bytes, of an object that GCC will place in the
3118 fixed area of the stack frame when the user specifies
3119 @option{-fstack-check}.
3120 GCC computed the default from the values of the above macros and you will
3121 normally not need to override that default.
3125 @node Frame Registers
3126 @subsection Registers That Address the Stack Frame
3128 @c prevent bad page break with this line
3129 This discusses registers that address the stack frame.
3131 @defmac STACK_POINTER_REGNUM
3132 The register number of the stack pointer register, which must also be a
3133 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3134 the hardware determines which register this is.
3137 @defmac FRAME_POINTER_REGNUM
3138 The register number of the frame pointer register, which is used to
3139 access automatic variables in the stack frame. On some machines, the
3140 hardware determines which register this is. On other machines, you can
3141 choose any register you wish for this purpose.
3144 @defmac HARD_FRAME_POINTER_REGNUM
3145 On some machines the offset between the frame pointer and starting
3146 offset of the automatic variables is not known until after register
3147 allocation has been done (for example, because the saved registers are
3148 between these two locations). On those machines, define
3149 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3150 be used internally until the offset is known, and define
3151 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3152 used for the frame pointer.
3154 You should define this macro only in the very rare circumstances when it
3155 is not possible to calculate the offset between the frame pointer and
3156 the automatic variables until after register allocation has been
3157 completed. When this macro is defined, you must also indicate in your
3158 definition of @code{ELIMINABLE_REGS} how to eliminate
3159 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3160 or @code{STACK_POINTER_REGNUM}.
3162 Do not define this macro if it would be the same as
3163 @code{FRAME_POINTER_REGNUM}.
3166 @defmac ARG_POINTER_REGNUM
3167 The register number of the arg pointer register, which is used to access
3168 the function's argument list. On some machines, this is the same as the
3169 frame pointer register. On some machines, the hardware determines which
3170 register this is. On other machines, you can choose any register you
3171 wish for this purpose. If this is not the same register as the frame
3172 pointer register, then you must mark it as a fixed register according to
3173 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3174 (@pxref{Elimination}).
3177 @defmac RETURN_ADDRESS_POINTER_REGNUM
3178 The register number of the return address pointer register, which is used to
3179 access the current function's return address from the stack. On some
3180 machines, the return address is not at a fixed offset from the frame
3181 pointer or stack pointer or argument pointer. This register can be defined
3182 to point to the return address on the stack, and then be converted by
3183 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3185 Do not define this macro unless there is no other way to get the return
3186 address from the stack.
3189 @defmac STATIC_CHAIN_REGNUM
3190 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3191 Register numbers used for passing a function's static chain pointer. If
3192 register windows are used, the register number as seen by the called
3193 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3194 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3195 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3198 The static chain register need not be a fixed register.
3200 If the static chain is passed in memory, these macros should not be
3201 defined; instead, the next two macros should be defined.
3204 @defmac STATIC_CHAIN
3205 @defmacx STATIC_CHAIN_INCOMING
3206 If the static chain is passed in memory, these macros provide rtx giving
3207 @code{mem} expressions that denote where they are stored.
3208 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3209 as seen by the calling and called functions, respectively. Often the former
3210 will be at an offset from the stack pointer and the latter at an offset from
3213 @findex stack_pointer_rtx
3214 @findex frame_pointer_rtx
3215 @findex arg_pointer_rtx
3216 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3217 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3218 macros and should be used to refer to those items.
3220 If the static chain is passed in a register, the two previous macros should
3224 @defmac DWARF_FRAME_REGISTERS
3225 This macro specifies the maximum number of hard registers that can be
3226 saved in a call frame. This is used to size data structures used in
3227 DWARF2 exception handling.
3229 Prior to GCC 3.0, this macro was needed in order to establish a stable
3230 exception handling ABI in the face of adding new hard registers for ISA
3231 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3232 in the number of hard registers. Nevertheless, this macro can still be
3233 used to reduce the runtime memory requirements of the exception handling
3234 routines, which can be substantial if the ISA contains a lot of
3235 registers that are not call-saved.
3237 If this macro is not defined, it defaults to
3238 @code{FIRST_PSEUDO_REGISTER}.
3241 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3243 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3244 for backward compatibility in pre GCC 3.0 compiled code.
3246 If this macro is not defined, it defaults to
3247 @code{DWARF_FRAME_REGISTERS}.
3250 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3252 Define this macro if the target's representation for dwarf registers
3253 is different than the internal representation for unwind column.
3254 Given a dwarf register, this macro should return the internal unwind
3255 column number to use instead.
3257 See the PowerPC's SPE target for an example.
3261 @subsection Eliminating Frame Pointer and Arg Pointer
3263 @c prevent bad page break with this line
3264 This is about eliminating the frame pointer and arg pointer.
3266 @defmac FRAME_POINTER_REQUIRED
3267 A C expression which is nonzero if a function must have and use a frame
3268 pointer. This expression is evaluated in the reload pass. If its value is
3269 nonzero the function will have a frame pointer.
3271 The expression can in principle examine the current function and decide
3272 according to the facts, but on most machines the constant 0 or the
3273 constant 1 suffices. Use 0 when the machine allows code to be generated
3274 with no frame pointer, and doing so saves some time or space. Use 1
3275 when there is no possible advantage to avoiding a frame pointer.
3277 In certain cases, the compiler does not know how to produce valid code
3278 without a frame pointer. The compiler recognizes those cases and
3279 automatically gives the function a frame pointer regardless of what
3280 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3283 In a function that does not require a frame pointer, the frame pointer
3284 register can be allocated for ordinary usage, unless you mark it as a
3285 fixed register. See @code{FIXED_REGISTERS} for more information.
3288 @findex get_frame_size
3289 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3290 A C statement to store in the variable @var{depth-var} the difference
3291 between the frame pointer and the stack pointer values immediately after
3292 the function prologue. The value would be computed from information
3293 such as the result of @code{get_frame_size ()} and the tables of
3294 registers @code{regs_ever_live} and @code{call_used_regs}.
3296 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3297 need not be defined. Otherwise, it must be defined even if
3298 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3299 case, you may set @var{depth-var} to anything.
3302 @defmac ELIMINABLE_REGS
3303 If defined, this macro specifies a table of register pairs used to
3304 eliminate unneeded registers that point into the stack frame. If it is not
3305 defined, the only elimination attempted by the compiler is to replace
3306 references to the frame pointer with references to the stack pointer.
3308 The definition of this macro is a list of structure initializations, each
3309 of which specifies an original and replacement register.
3311 On some machines, the position of the argument pointer is not known until
3312 the compilation is completed. In such a case, a separate hard register
3313 must be used for the argument pointer. This register can be eliminated by
3314 replacing it with either the frame pointer or the argument pointer,
3315 depending on whether or not the frame pointer has been eliminated.
3317 In this case, you might specify:
3319 #define ELIMINABLE_REGS \
3320 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3321 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3322 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3325 Note that the elimination of the argument pointer with the stack pointer is
3326 specified first since that is the preferred elimination.
3329 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3330 A C expression that returns nonzero if the compiler is allowed to try
3331 to replace register number @var{from-reg} with register number
3332 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3333 is defined, and will usually be the constant 1, since most of the cases
3334 preventing register elimination are things that the compiler already
3338 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3339 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3340 specifies the initial difference between the specified pair of
3341 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3345 @node Stack Arguments
3346 @subsection Passing Function Arguments on the Stack
3347 @cindex arguments on stack
3348 @cindex stack arguments
3350 The macros in this section control how arguments are passed
3351 on the stack. See the following section for other macros that
3352 control passing certain arguments in registers.
3354 @defmac PROMOTE_PROTOTYPES
3355 A C expression whose value is nonzero if an argument declared in
3356 a prototype as an integral type smaller than @code{int} should
3357 actually be passed as an @code{int}. In addition to avoiding
3358 errors in certain cases of mismatch, it also makes for better
3359 code on certain machines. If the macro is not defined in target
3360 header files, it defaults to 0.
3364 A C expression. If nonzero, push insns will be used to pass
3366 If the target machine does not have a push instruction, set it to zero.
3367 That directs GCC to use an alternate strategy: to
3368 allocate the entire argument block and then store the arguments into
3369 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3372 @defmac PUSH_ARGS_REVERSED
3373 A C expression. If nonzero, function arguments will be evaluated from
3374 last to first, rather than from first to last. If this macro is not
3375 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3376 and args grow in opposite directions, and 0 otherwise.
3379 @defmac PUSH_ROUNDING (@var{npushed})
3380 A C expression that is the number of bytes actually pushed onto the
3381 stack when an instruction attempts to push @var{npushed} bytes.
3383 On some machines, the definition
3386 #define PUSH_ROUNDING(BYTES) (BYTES)
3390 will suffice. But on other machines, instructions that appear
3391 to push one byte actually push two bytes in an attempt to maintain
3392 alignment. Then the definition should be
3395 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3399 @findex current_function_outgoing_args_size
3400 @defmac ACCUMULATE_OUTGOING_ARGS
3401 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3402 will be computed and placed into the variable
3403 @code{current_function_outgoing_args_size}. No space will be pushed
3404 onto the stack for each call; instead, the function prologue should
3405 increase the stack frame size by this amount.
3407 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3411 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3412 Define this macro if functions should assume that stack space has been
3413 allocated for arguments even when their values are passed in
3416 The value of this macro is the size, in bytes, of the area reserved for
3417 arguments passed in registers for the function represented by @var{fndecl},
3418 which can be zero if GCC is calling a library function.
3420 This space can be allocated by the caller, or be a part of the
3421 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3424 @c above is overfull. not sure what to do. --mew 5feb93 did
3425 @c something, not sure if it looks good. --mew 10feb93
3427 @defmac MAYBE_REG_PARM_STACK_SPACE
3428 @defmacx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
3429 Define these macros in addition to the one above if functions might
3430 allocate stack space for arguments even when their values are passed
3431 in registers. These should be used when the stack space allocated
3432 for arguments in registers is not a simple constant independent of the
3433 function declaration.
3435 The value of the first macro is the size, in bytes, of the area that
3436 we should initially assume would be reserved for arguments passed in registers.
3438 The value of the second macro is the actual size, in bytes, of the area
3439 that will be reserved for arguments passed in registers. This takes two
3440 arguments: an integer representing the number of bytes of fixed sized
3441 arguments on the stack, and a tree representing the number of bytes of
3442 variable sized arguments on the stack.
3444 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
3445 called for libcall functions, the current function, or for a function
3446 being called when it is known that such stack space must be allocated.
3447 In each case this value can be easily computed.
3449 When deciding whether a called function needs such stack space, and how
3450 much space to reserve, GCC uses these two macros instead of
3451 @code{REG_PARM_STACK_SPACE}.
3454 @defmac OUTGOING_REG_PARM_STACK_SPACE
3455 Define this if it is the responsibility of the caller to allocate the area
3456 reserved for arguments passed in registers.
3458 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3459 whether the space for these arguments counts in the value of
3460 @code{current_function_outgoing_args_size}.
3463 @defmac STACK_PARMS_IN_REG_PARM_AREA
3464 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3465 stack parameters don't skip the area specified by it.
3466 @c i changed this, makes more sens and it should have taken care of the
3467 @c overfull.. not as specific, tho. --mew 5feb93
3469 Normally, when a parameter is not passed in registers, it is placed on the
3470 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3471 suppresses this behavior and causes the parameter to be passed on the
3472 stack in its natural location.
3475 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3476 A C expression that should indicate the number of bytes of its own
3477 arguments that a function pops on returning, or 0 if the
3478 function pops no arguments and the caller must therefore pop them all
3479 after the function returns.
3481 @var{fundecl} is a C variable whose value is a tree node that describes
3482 the function in question. Normally it is a node of type
3483 @code{FUNCTION_DECL} that describes the declaration of the function.
3484 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3486 @var{funtype} is a C variable whose value is a tree node that
3487 describes the function in question. Normally it is a node of type
3488 @code{FUNCTION_TYPE} that describes the data type of the function.
3489 From this it is possible to obtain the data types of the value and
3490 arguments (if known).
3492 When a call to a library function is being considered, @var{fundecl}
3493 will contain an identifier node for the library function. Thus, if
3494 you need to distinguish among various library functions, you can do so
3495 by their names. Note that ``library function'' in this context means
3496 a function used to perform arithmetic, whose name is known specially
3497 in the compiler and was not mentioned in the C code being compiled.
3499 @var{stack-size} is the number of bytes of arguments passed on the
3500 stack. If a variable number of bytes is passed, it is zero, and
3501 argument popping will always be the responsibility of the calling function.
3503 On the VAX, all functions always pop their arguments, so the definition
3504 of this macro is @var{stack-size}. On the 68000, using the standard
3505 calling convention, no functions pop their arguments, so the value of
3506 the macro is always 0 in this case. But an alternative calling
3507 convention is available in which functions that take a fixed number of
3508 arguments pop them but other functions (such as @code{printf}) pop
3509 nothing (the caller pops all). When this convention is in use,
3510 @var{funtype} is examined to determine whether a function takes a fixed
3511 number of arguments.
3514 @defmac CALL_POPS_ARGS (@var{cum})
3515 A C expression that should indicate the number of bytes a call sequence
3516 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3517 when compiling a function call.
3519 @var{cum} is the variable in which all arguments to the called function
3520 have been accumulated.
3522 On certain architectures, such as the SH5, a call trampoline is used
3523 that pops certain registers off the stack, depending on the arguments
3524 that have been passed to the function. Since this is a property of the
3525 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3529 @node Register Arguments
3530 @subsection Passing Arguments in Registers
3531 @cindex arguments in registers
3532 @cindex registers arguments
3534 This section describes the macros which let you control how various
3535 types of arguments are passed in registers or how they are arranged in
3538 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3539 A C expression that controls whether a function argument is passed
3540 in a register, and which register.
3542 The arguments are @var{cum}, which summarizes all the previous
3543 arguments; @var{mode}, the machine mode of the argument; @var{type},
3544 the data type of the argument as a tree node or 0 if that is not known
3545 (which happens for C support library functions); and @var{named},
3546 which is 1 for an ordinary argument and 0 for nameless arguments that
3547 correspond to @samp{@dots{}} in the called function's prototype.
3548 @var{type} can be an incomplete type if a syntax error has previously
3551 The value of the expression is usually either a @code{reg} RTX for the
3552 hard register in which to pass the argument, or zero to pass the
3553 argument on the stack.
3555 For machines like the VAX and 68000, where normally all arguments are
3556 pushed, zero suffices as a definition.
3558 The value of the expression can also be a @code{parallel} RTX@. This is
3559 used when an argument is passed in multiple locations. The mode of the
3560 @code{parallel} should be the mode of the entire argument. The
3561 @code{parallel} holds any number of @code{expr_list} pairs; each one
3562 describes where part of the argument is passed. In each
3563 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3564 register in which to pass this part of the argument, and the mode of the
3565 register RTX indicates how large this part of the argument is. The
3566 second operand of the @code{expr_list} is a @code{const_int} which gives
3567 the offset in bytes into the entire argument of where this part starts.
3568 As a special exception the first @code{expr_list} in the @code{parallel}
3569 RTX may have a first operand of zero. This indicates that the entire
3570 argument is also stored on the stack.
3572 The last time this macro is called, it is called with @code{MODE ==
3573 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3574 pattern as operands 2 and 3 respectively.
3576 @cindex @file{stdarg.h} and register arguments
3577 The usual way to make the ISO library @file{stdarg.h} work on a machine
3578 where some arguments are usually passed in registers, is to cause
3579 nameless arguments to be passed on the stack instead. This is done
3580 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3582 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3583 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3584 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3585 in the definition of this macro to determine if this argument is of a
3586 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3587 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3588 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3589 defined, the argument will be computed in the stack and then loaded into
3593 @defmac MUST_PASS_IN_STACK (@var{mode}, @var{type})
3594 Define as a C expression that evaluates to nonzero if we do not know how
3595 to pass TYPE solely in registers. The file @file{expr.h} defines a
3596 definition that is usually appropriate, refer to @file{expr.h} for additional
3600 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3601 Define this macro if the target machine has ``register windows'', so
3602 that the register in which a function sees an arguments is not
3603 necessarily the same as the one in which the caller passed the
3606 For such machines, @code{FUNCTION_ARG} computes the register in which
3607 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3608 be defined in a similar fashion to tell the function being called
3609 where the arguments will arrive.
3611 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3612 serves both purposes.
3615 @defmac FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3616 A C expression for the number of words, at the beginning of an
3617 argument, that must be put in registers. The value must be zero for
3618 arguments that are passed entirely in registers or that are entirely
3619 pushed on the stack.
3621 On some machines, certain arguments must be passed partially in
3622 registers and partially in memory. On these machines, typically the
3623 first @var{n} words of arguments are passed in registers, and the rest
3624 on the stack. If a multi-word argument (a @code{double} or a
3625 structure) crosses that boundary, its first few words must be passed
3626 in registers and the rest must be pushed. This macro tells the
3627 compiler when this occurs, and how many of the words should go in
3630 @code{FUNCTION_ARG} for these arguments should return the first
3631 register to be used by the caller for this argument; likewise
3632 @code{FUNCTION_INCOMING_ARG}, for the called function.
3635 @defmac FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3636 A C expression that indicates when an argument must be passed by reference.
3637 If nonzero for an argument, a copy of that argument is made in memory and a
3638 pointer to the argument is passed instead of the argument itself.
3639 The pointer is passed in whatever way is appropriate for passing a pointer
3642 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3643 definition of this macro might be
3645 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3646 (CUM, MODE, TYPE, NAMED) \
3647 MUST_PASS_IN_STACK (MODE, TYPE)
3649 @c this is *still* too long. --mew 5feb93
3652 @defmac FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3653 If defined, a C expression that indicates when it is the called function's
3654 responsibility to make a copy of arguments passed by invisible reference.
3655 Normally, the caller makes a copy and passes the address of the copy to the
3656 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3657 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3658 ``live'' value. The called function must not modify this value. If it can be
3659 determined that the value won't be modified, it need not make a copy;
3660 otherwise a copy must be made.
3663 @defmac CUMULATIVE_ARGS
3664 A C type for declaring a variable that is used as the first argument of
3665 @code{FUNCTION_ARG} and other related values. For some target machines,
3666 the type @code{int} suffices and can hold the number of bytes of
3669 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3670 arguments that have been passed on the stack. The compiler has other
3671 variables to keep track of that. For target machines on which all
3672 arguments are passed on the stack, there is no need to store anything in
3673 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3674 should not be empty, so use @code{int}.
3677 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl})
3678 A C statement (sans semicolon) for initializing the variable
3679 @var{cum} for the state at the beginning of the argument list. The
3680 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3681 is the tree node for the data type of the function which will receive
3682 the args, or 0 if the args are to a compiler support library function.
3683 For direct calls that are not libcalls, @var{fndecl} contain the
3684 declaration node of the function. @var{fndecl} is also set when
3685 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3688 When processing a call to a compiler support library function,
3689 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3690 contains the name of the function, as a string. @var{libname} is 0 when
3691 an ordinary C function call is being processed. Thus, each time this
3692 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3693 never both of them at once.
3696 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3697 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3698 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3699 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3700 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3701 0)} is used instead.
3704 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3705 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3706 finding the arguments for the function being compiled. If this macro is
3707 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3709 The value passed for @var{libname} is always 0, since library routines
3710 with special calling conventions are never compiled with GCC@. The
3711 argument @var{libname} exists for symmetry with
3712 @code{INIT_CUMULATIVE_ARGS}.
3713 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3714 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3717 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3718 A C statement (sans semicolon) to update the summarizer variable
3719 @var{cum} to advance past an argument in the argument list. The
3720 values @var{mode}, @var{type} and @var{named} describe that argument.
3721 Once this is done, the variable @var{cum} is suitable for analyzing
3722 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3724 This macro need not do anything if the argument in question was passed
3725 on the stack. The compiler knows how to track the amount of stack space
3726 used for arguments without any special help.
3729 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3730 If defined, a C expression which determines whether, and in which direction,
3731 to pad out an argument with extra space. The value should be of type
3732 @code{enum direction}: either @code{upward} to pad above the argument,
3733 @code{downward} to pad below, or @code{none} to inhibit padding.
3735 The @emph{amount} of padding is always just enough to reach the next
3736 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3739 This macro has a default definition which is right for most systems.
3740 For little-endian machines, the default is to pad upward. For
3741 big-endian machines, the default is to pad downward for an argument of
3742 constant size shorter than an @code{int}, and upward otherwise.
3745 @defmac PAD_VARARGS_DOWN
3746 If defined, a C expression which determines whether the default
3747 implementation of va_arg will attempt to pad down before reading the
3748 next argument, if that argument is smaller than its aligned space as
3749 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3750 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3753 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3754 Specify padding for the last element of a block move between registers and
3755 memory. @var{first} is nonzero if this is the only element. Defining this
3756 macro allows better control of register function parameters on big-endian
3757 machines, without using @code{PARALLEL} rtl. In particular,
3758 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3759 registers, as there is no longer a "wrong" part of a register; For example,
3760 a three byte aggregate may be passed in the high part of a register if so
3764 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3765 If defined, a C expression that gives the alignment boundary, in bits,
3766 of an argument with the specified mode and type. If it is not defined,
3767 @code{PARM_BOUNDARY} is used for all arguments.
3770 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3771 A C expression that is nonzero if @var{regno} is the number of a hard
3772 register in which function arguments are sometimes passed. This does
3773 @emph{not} include implicit arguments such as the static chain and
3774 the structure-value address. On many machines, no registers can be
3775 used for this purpose since all function arguments are pushed on the
3779 @defmac SPLIT_COMPLEX_ARGS
3781 Define this macro to a nonzero value if complex function arguments
3782 should be split into their corresponding components. By default, GCC
3783 will attempt to pack complex arguments into the target's word size.
3784 Some ABIs require complex arguments to be split and treated as their
3785 individual components. For example, on AIX64, complex floats should
3786 be passed in a pair of floating point registers, even though a complex
3787 float would fit in one 64-bit floating point register.
3790 @defmac LOAD_ARGS_REVERSED
3791 If defined, the order in which arguments are loaded into their
3792 respective argument registers is reversed so that the last
3793 argument is loaded first. This macro only affects arguments
3794 passed in registers.
3798 @subsection How Scalar Function Values Are Returned
3799 @cindex return values in registers
3800 @cindex values, returned by functions
3801 @cindex scalars, returned as values
3803 This section discusses the macros that control returning scalars as
3804 values---values that can fit in registers.
3806 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3807 A C expression to create an RTX representing the place where a
3808 function returns a value of data type @var{valtype}. @var{valtype} is
3809 a tree node representing a data type. Write @code{TYPE_MODE
3810 (@var{valtype})} to get the machine mode used to represent that type.
3811 On many machines, only the mode is relevant. (Actually, on most
3812 machines, scalar values are returned in the same place regardless of
3815 The value of the expression is usually a @code{reg} RTX for the hard
3816 register where the return value is stored. The value can also be a
3817 @code{parallel} RTX, if the return value is in multiple places. See
3818 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3820 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3821 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3824 If the precise function being called is known, @var{func} is a tree
3825 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3826 pointer. This makes it possible to use a different value-returning
3827 convention for specific functions when all their calls are
3830 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3831 types, because these are returned in another way. See
3832 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3835 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3836 Define this macro if the target machine has ``register windows''
3837 so that the register in which a function returns its value is not
3838 the same as the one in which the caller sees the value.
3840 For such machines, @code{FUNCTION_VALUE} computes the register in which
3841 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3842 defined in a similar fashion to tell the function where to put the
3845 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3846 @code{FUNCTION_VALUE} serves both purposes.
3848 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3849 aggregate data types, because these are returned in another way. See
3850 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3853 @defmac LIBCALL_VALUE (@var{mode})
3854 A C expression to create an RTX representing the place where a library
3855 function returns a value of mode @var{mode}. If the precise function
3856 being called is known, @var{func} is a tree node
3857 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3858 pointer. This makes it possible to use a different value-returning
3859 convention for specific functions when all their calls are
3862 Note that ``library function'' in this context means a compiler
3863 support routine, used to perform arithmetic, whose name is known
3864 specially by the compiler and was not mentioned in the C code being
3867 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3868 data types, because none of the library functions returns such types.
3871 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3872 A C expression that is nonzero if @var{regno} is the number of a hard
3873 register in which the values of called function may come back.
3875 A register whose use for returning values is limited to serving as the
3876 second of a pair (for a value of type @code{double}, say) need not be
3877 recognized by this macro. So for most machines, this definition
3881 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3884 If the machine has register windows, so that the caller and the called
3885 function use different registers for the return value, this macro
3886 should recognize only the caller's register numbers.
3889 @defmac APPLY_RESULT_SIZE
3890 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3891 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3892 saving and restoring an arbitrary return value.
3895 @node Aggregate Return
3896 @subsection How Large Values Are Returned
3897 @cindex aggregates as return values
3898 @cindex large return values
3899 @cindex returning aggregate values
3900 @cindex structure value address
3902 When a function value's mode is @code{BLKmode} (and in some other
3903 cases), the value is not returned according to @code{FUNCTION_VALUE}
3904 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3905 block of memory in which the value should be stored. This address
3906 is called the @dfn{structure value address}.
3908 This section describes how to control returning structure values in
3911 @defmac RETURN_IN_MEMORY (@var{type})
3912 A C expression which can inhibit the returning of certain function
3913 values in registers, based on the type of value. A nonzero value says
3914 to return the function value in memory, just as large structures are
3915 always returned. Here @var{type} will be a C expression of type
3916 @code{tree}, representing the data type of the value.
3918 Note that values of mode @code{BLKmode} must be explicitly handled
3919 by this macro. Also, the option @option{-fpcc-struct-return}
3920 takes effect regardless of this macro. On most systems, it is
3921 possible to leave the macro undefined; this causes a default
3922 definition to be used, whose value is the constant 1 for @code{BLKmode}
3923 values, and 0 otherwise.
3925 Do not use this macro to indicate that structures and unions should always
3926 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3930 @defmac DEFAULT_PCC_STRUCT_RETURN
3931 Define this macro to be 1 if all structure and union return values must be
3932 in memory. Since this results in slower code, this should be defined
3933 only if needed for compatibility with other compilers or with an ABI@.
3934 If you define this macro to be 0, then the conventions used for structure
3935 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3937 If not defined, this defaults to the value 1.
3940 @defmac STRUCT_VALUE_REGNUM
3941 If the structure value address is passed in a register, then
3942 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3945 @defmac STRUCT_VALUE
3946 If the structure value address is not passed in a register, define
3947 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3948 where the address is passed. If it returns 0, the address is passed as
3949 an ``invisible'' first argument.
3952 @defmac STRUCT_VALUE_INCOMING_REGNUM
3953 On some architectures the place where the structure value address
3954 is found by the called function is not the same place that the
3955 caller put it. This can be due to register windows, or it could
3956 be because the function prologue moves it to a different place.
3958 If the incoming location of the structure value address is in a
3959 register, define this macro as the register number.
3962 @defmac STRUCT_VALUE_INCOMING
3963 If the incoming location is not a register, then you should define
3964 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3965 called function should find the value. If it should find the value on
3966 the stack, define this to create a @code{mem} which refers to the frame
3967 pointer. A definition of 0 means that the address is passed as an
3968 ``invisible'' first argument.
3971 @defmac PCC_STATIC_STRUCT_RETURN
3972 Define this macro if the usual system convention on the target machine
3973 for returning structures and unions is for the called function to return
3974 the address of a static variable containing the value.
3976 Do not define this if the usual system convention is for the caller to
3977 pass an address to the subroutine.
3979 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3980 nothing when you use @option{-freg-struct-return} mode.
3984 @subsection Caller-Saves Register Allocation
3986 If you enable it, GCC can save registers around function calls. This
3987 makes it possible to use call-clobbered registers to hold variables that
3988 must live across calls.
3990 @defmac DEFAULT_CALLER_SAVES
3991 Define this macro if function calls on the target machine do not preserve
3992 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3993 for all registers. When defined, this macro enables @option{-fcaller-saves}
3994 by default for all optimization levels. It has no effect for optimization
3995 levels 2 and higher, where @option{-fcaller-saves} is the default.
3998 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3999 A C expression to determine whether it is worthwhile to consider placing
4000 a pseudo-register in a call-clobbered hard register and saving and
4001 restoring it around each function call. The expression should be 1 when
4002 this is worth doing, and 0 otherwise.
4004 If you don't define this macro, a default is used which is good on most
4005 machines: @code{4 * @var{calls} < @var{refs}}.
4008 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4009 A C expression specifying which mode is required for saving @var{nregs}
4010 of a pseudo-register in call-clobbered hard register @var{regno}. If
4011 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4012 returned. For most machines this macro need not be defined since GCC
4013 will select the smallest suitable mode.
4016 @node Function Entry
4017 @subsection Function Entry and Exit
4018 @cindex function entry and exit
4022 This section describes the macros that output function entry
4023 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4025 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4026 If defined, a function that outputs the assembler code for entry to a
4027 function. The prologue is responsible for setting up the stack frame,
4028 initializing the frame pointer register, saving registers that must be
4029 saved, and allocating @var{size} additional bytes of storage for the
4030 local variables. @var{size} is an integer. @var{file} is a stdio
4031 stream to which the assembler code should be output.
4033 The label for the beginning of the function need not be output by this
4034 macro. That has already been done when the macro is run.
4036 @findex regs_ever_live
4037 To determine which registers to save, the macro can refer to the array
4038 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4039 @var{r} is used anywhere within the function. This implies the function
4040 prologue should save register @var{r}, provided it is not one of the
4041 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4042 @code{regs_ever_live}.)
4044 On machines that have ``register windows'', the function entry code does
4045 not save on the stack the registers that are in the windows, even if
4046 they are supposed to be preserved by function calls; instead it takes
4047 appropriate steps to ``push'' the register stack, if any non-call-used
4048 registers are used in the function.
4050 @findex frame_pointer_needed
4051 On machines where functions may or may not have frame-pointers, the
4052 function entry code must vary accordingly; it must set up the frame
4053 pointer if one is wanted, and not otherwise. To determine whether a
4054 frame pointer is in wanted, the macro can refer to the variable
4055 @code{frame_pointer_needed}. The variable's value will be 1 at run
4056 time in a function that needs a frame pointer. @xref{Elimination}.
4058 The function entry code is responsible for allocating any stack space
4059 required for the function. This stack space consists of the regions
4060 listed below. In most cases, these regions are allocated in the
4061 order listed, with the last listed region closest to the top of the
4062 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4063 the highest address if it is not defined). You can use a different order
4064 for a machine if doing so is more convenient or required for
4065 compatibility reasons. Except in cases where required by standard
4066 or by a debugger, there is no reason why the stack layout used by GCC
4067 need agree with that used by other compilers for a machine.
4070 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4071 If defined, a function that outputs assembler code at the end of a
4072 prologue. This should be used when the function prologue is being
4073 emitted as RTL, and you have some extra assembler that needs to be
4074 emitted. @xref{prologue instruction pattern}.
4077 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4078 If defined, a function that outputs assembler code at the start of an
4079 epilogue. This should be used when the function epilogue is being
4080 emitted as RTL, and you have some extra assembler that needs to be
4081 emitted. @xref{epilogue instruction pattern}.
4084 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4085 If defined, a function that outputs the assembler code for exit from a
4086 function. The epilogue is responsible for restoring the saved
4087 registers and stack pointer to their values when the function was
4088 called, and returning control to the caller. This macro takes the
4089 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4090 registers to restore are determined from @code{regs_ever_live} and
4091 @code{CALL_USED_REGISTERS} in the same way.
4093 On some machines, there is a single instruction that does all the work
4094 of returning from the function. On these machines, give that
4095 instruction the name @samp{return} and do not define the macro
4096 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4098 Do not define a pattern named @samp{return} if you want the
4099 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4100 switches to control whether return instructions or epilogues are used,
4101 define a @samp{return} pattern with a validity condition that tests the
4102 target switches appropriately. If the @samp{return} pattern's validity
4103 condition is false, epilogues will be used.
4105 On machines where functions may or may not have frame-pointers, the
4106 function exit code must vary accordingly. Sometimes the code for these
4107 two cases is completely different. To determine whether a frame pointer
4108 is wanted, the macro can refer to the variable
4109 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4110 a function that needs a frame pointer.
4112 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4113 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4114 The C variable @code{current_function_is_leaf} is nonzero for such a
4115 function. @xref{Leaf Functions}.
4117 On some machines, some functions pop their arguments on exit while
4118 others leave that for the caller to do. For example, the 68020 when
4119 given @option{-mrtd} pops arguments in functions that take a fixed
4120 number of arguments.
4122 @findex current_function_pops_args
4123 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4124 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4125 needs to know what was decided. The variable that is called
4126 @code{current_function_pops_args} is the number of bytes of its
4127 arguments that a function should pop. @xref{Scalar Return}.
4128 @c what is the "its arguments" in the above sentence referring to, pray
4129 @c tell? --mew 5feb93
4134 @findex current_function_pretend_args_size
4135 A region of @code{current_function_pretend_args_size} bytes of
4136 uninitialized space just underneath the first argument arriving on the
4137 stack. (This may not be at the very start of the allocated stack region
4138 if the calling sequence has pushed anything else since pushing the stack
4139 arguments. But usually, on such machines, nothing else has been pushed
4140 yet, because the function prologue itself does all the pushing.) This
4141 region is used on machines where an argument may be passed partly in
4142 registers and partly in memory, and, in some cases to support the
4143 features in @code{<stdarg.h>}.
4146 An area of memory used to save certain registers used by the function.
4147 The size of this area, which may also include space for such things as
4148 the return address and pointers to previous stack frames, is
4149 machine-specific and usually depends on which registers have been used
4150 in the function. Machines with register windows often do not require
4154 A region of at least @var{size} bytes, possibly rounded up to an allocation
4155 boundary, to contain the local variables of the function. On some machines,
4156 this region and the save area may occur in the opposite order, with the
4157 save area closer to the top of the stack.
4160 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4161 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4162 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4163 argument lists of the function. @xref{Stack Arguments}.
4166 Normally, it is necessary for the macros
4167 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4168 @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially.
4169 The C variable @code{current_function_is_leaf} is nonzero for such a
4172 @defmac EXIT_IGNORE_STACK
4173 Define this macro as a C expression that is nonzero if the return
4174 instruction or the function epilogue ignores the value of the stack
4175 pointer; in other words, if it is safe to delete an instruction to
4176 adjust the stack pointer before a return from the function.
4178 Note that this macro's value is relevant only for functions for which
4179 frame pointers are maintained. It is never safe to delete a final
4180 stack adjustment in a function that has no frame pointer, and the
4181 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4184 @defmac EPILOGUE_USES (@var{regno})
4185 Define this macro as a C expression that is nonzero for registers that are
4186 used by the epilogue or the @samp{return} pattern. The stack and frame
4187 pointer registers are already be assumed to be used as needed.
4190 @defmac EH_USES (@var{regno})
4191 Define this macro as a C expression that is nonzero for registers that are
4192 used by the exception handling mechanism, and so should be considered live
4193 on entry to an exception edge.
4196 @defmac DELAY_SLOTS_FOR_EPILOGUE
4197 Define this macro if the function epilogue contains delay slots to which
4198 instructions from the rest of the function can be ``moved''. The
4199 definition should be a C expression whose value is an integer
4200 representing the number of delay slots there.
4203 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4204 A C expression that returns 1 if @var{insn} can be placed in delay
4205 slot number @var{n} of the epilogue.
4207 The argument @var{n} is an integer which identifies the delay slot now
4208 being considered (since different slots may have different rules of
4209 eligibility). It is never negative and is always less than the number
4210 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4211 If you reject a particular insn for a given delay slot, in principle, it
4212 may be reconsidered for a subsequent delay slot. Also, other insns may
4213 (at least in principle) be considered for the so far unfilled delay
4216 @findex current_function_epilogue_delay_list
4217 @findex final_scan_insn
4218 The insns accepted to fill the epilogue delay slots are put in an RTL
4219 list made with @code{insn_list} objects, stored in the variable
4220 @code{current_function_epilogue_delay_list}. The insn for the first
4221 delay slot comes first in the list. Your definition of the macro
4222 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4223 outputting the insns in this list, usually by calling
4224 @code{final_scan_insn}.
4226 You need not define this macro if you did not define
4227 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4230 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, tree @var{function})
4231 A function that outputs the assembler code for a thunk
4232 function, used to implement C++ virtual function calls with multiple
4233 inheritance. The thunk acts as a wrapper around a virtual function,
4234 adjusting the implicit object parameter before handing control off to
4237 First, emit code to add the integer @var{delta} to the location that
4238 contains the incoming first argument. Assume that this argument
4239 contains a pointer, and is the one used to pass the @code{this} pointer
4240 in C++. This is the incoming argument @emph{before} the function prologue,
4241 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4242 all other incoming arguments.
4244 After the addition, emit code to jump to @var{function}, which is a
4245 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4246 not touch the return address. Hence returning from @var{FUNCTION} will
4247 return to whoever called the current @samp{thunk}.
4249 The effect must be as if @var{function} had been called directly with
4250 the adjusted first argument. This macro is responsible for emitting all
4251 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4252 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4254 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4255 have already been extracted from it.) It might possibly be useful on
4256 some targets, but probably not.
4258 If you do not define this macro, the target-independent code in the C++
4259 front end will generate a less efficient heavyweight thunk that calls
4260 @var{function} instead of jumping to it. The generic approach does
4261 not support varargs.
4264 @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})
4265 A function like @code{TARGET_ASM_OUTPUT_MI_THUNK}, except that if
4266 @var{vcall_offset} is nonzero, an additional adjustment should be made
4267 after adding @code{delta}. In particular, if @var{p} is the
4268 adjusted pointer, the following adjustment should be made:
4271 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4275 If this function is defined, it will always be used in place of
4276 @code{TARGET_ASM_OUTPUT_MI_THUNK}.
4280 @subsection Generating Code for Profiling
4281 @cindex profiling, code generation
4283 These macros will help you generate code for profiling.
4285 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4286 A C statement or compound statement to output to @var{file} some
4287 assembler code to call the profiling subroutine @code{mcount}.
4290 The details of how @code{mcount} expects to be called are determined by
4291 your operating system environment, not by GCC@. To figure them out,
4292 compile a small program for profiling using the system's installed C
4293 compiler and look at the assembler code that results.
4295 Older implementations of @code{mcount} expect the address of a counter
4296 variable to be loaded into some register. The name of this variable is
4297 @samp{LP} followed by the number @var{labelno}, so you would generate
4298 the name using @samp{LP%d} in a @code{fprintf}.
4301 @defmac PROFILE_HOOK
4302 A C statement or compound statement to output to @var{file} some assembly
4303 code to call the profiling subroutine @code{mcount} even the target does
4304 not support profiling.
4307 @defmac NO_PROFILE_COUNTERS
4308 Define this macro if the @code{mcount} subroutine on your system does
4309 not need a counter variable allocated for each function. This is true
4310 for almost all modern implementations. If you define this macro, you
4311 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4314 @defmac PROFILE_BEFORE_PROLOGUE
4315 Define this macro if the code for function profiling should come before
4316 the function prologue. Normally, the profiling code comes after.
4320 @subsection Permitting tail calls
4323 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4324 True if it is ok to do sibling call optimization for the specified
4325 call expression @var{exp}. @var{decl} will be the called function,
4326 or @code{NULL} if this is an indirect call.
4328 It is not uncommon for limitations of calling conventions to prevent
4329 tail calls to functions outside the current unit of translation, or
4330 during PIC compilation. The hook is used to enforce these restrictions,
4331 as the @code{sibcall} md pattern can not fail, or fall over to a
4332 ``normal'' call. The criteria for successful sibling call optimization
4333 may vary greatly between different architectures.
4337 @section Implementing the Varargs Macros
4338 @cindex varargs implementation
4340 GCC comes with an implementation of @code{<varargs.h>} and
4341 @code{<stdarg.h>} that work without change on machines that pass arguments
4342 on the stack. Other machines require their own implementations of
4343 varargs, and the two machine independent header files must have
4344 conditionals to include it.
4346 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4347 the calling convention for @code{va_start}. The traditional
4348 implementation takes just one argument, which is the variable in which
4349 to store the argument pointer. The ISO implementation of
4350 @code{va_start} takes an additional second argument. The user is
4351 supposed to write the last named argument of the function here.
4353 However, @code{va_start} should not use this argument. The way to find
4354 the end of the named arguments is with the built-in functions described
4357 @defmac __builtin_saveregs ()
4358 Use this built-in function to save the argument registers in memory so
4359 that the varargs mechanism can access them. Both ISO and traditional
4360 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4361 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
4363 On some machines, @code{__builtin_saveregs} is open-coded under the
4364 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
4365 it calls a routine written in assembler language, found in
4368 Code generated for the call to @code{__builtin_saveregs} appears at the
4369 beginning of the function, as opposed to where the call to
4370 @code{__builtin_saveregs} is written, regardless of what the code is.
4371 This is because the registers must be saved before the function starts
4372 to use them for its own purposes.
4373 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4377 @defmac __builtin_args_info (@var{category})
4378 Use this built-in function to find the first anonymous arguments in
4381 In general, a machine may have several categories of registers used for
4382 arguments, each for a particular category of data types. (For example,
4383 on some machines, floating-point registers are used for floating-point
4384 arguments while other arguments are passed in the general registers.)
4385 To make non-varargs functions use the proper calling convention, you
4386 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4387 registers in each category have been used so far
4389 @code{__builtin_args_info} accesses the same data structure of type
4390 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4391 with it, with @var{category} specifying which word to access. Thus, the
4392 value indicates the first unused register in a given category.
4394 Normally, you would use @code{__builtin_args_info} in the implementation
4395 of @code{va_start}, accessing each category just once and storing the
4396 value in the @code{va_list} object. This is because @code{va_list} will
4397 have to update the values, and there is no way to alter the
4398 values accessed by @code{__builtin_args_info}.
4401 @defmac __builtin_next_arg (@var{lastarg})
4402 This is the equivalent of @code{__builtin_args_info}, for stack
4403 arguments. It returns the address of the first anonymous stack
4404 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4405 returns the address of the location above the first anonymous stack
4406 argument. Use it in @code{va_start} to initialize the pointer for
4407 fetching arguments from the stack. Also use it in @code{va_start} to
4408 verify that the second parameter @var{lastarg} is the last named argument
4409 of the current function.
4412 @defmac __builtin_classify_type (@var{object})
4413 Since each machine has its own conventions for which data types are
4414 passed in which kind of register, your implementation of @code{va_arg}
4415 has to embody these conventions. The easiest way to categorize the
4416 specified data type is to use @code{__builtin_classify_type} together
4417 with @code{sizeof} and @code{__alignof__}.
4419 @code{__builtin_classify_type} ignores the value of @var{object},
4420 considering only its data type. It returns an integer describing what
4421 kind of type that is---integer, floating, pointer, structure, and so on.
4423 The file @file{typeclass.h} defines an enumeration that you can use to
4424 interpret the values of @code{__builtin_classify_type}.
4427 These machine description macros help implement varargs:
4429 @defmac EXPAND_BUILTIN_SAVEREGS ()
4430 If defined, is a C expression that produces the machine-specific code
4431 for a call to @code{__builtin_saveregs}. This code will be moved to the
4432 very beginning of the function, before any parameter access are made.
4433 The return value of this function should be an RTX that contains the
4434 value to use as the return of @code{__builtin_saveregs}.
4437 @defmac SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
4438 This macro offers an alternative to using @code{__builtin_saveregs} and
4439 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
4440 anonymous register arguments into the stack so that all the arguments
4441 appear to have been passed consecutively on the stack. Once this is
4442 done, you can use the standard implementation of varargs that works for
4443 machines that pass all their arguments on the stack.
4445 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
4446 structure, containing the values that are obtained after processing the
4447 named arguments. The arguments @var{mode} and @var{type} describe the
4448 last named argument---its machine mode and its data type as a tree node.
4450 The macro implementation should do two things: first, push onto the
4451 stack all the argument registers @emph{not} used for the named
4452 arguments, and second, store the size of the data thus pushed into the
4453 @code{int}-valued variable whose name is supplied as the argument
4454 @var{pretend_args_size}. The value that you store here will serve as
4455 additional offset for setting up the stack frame.
4457 Because you must generate code to push the anonymous arguments at
4458 compile time without knowing their data types,
4459 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
4460 a single category of argument register and use it uniformly for all data
4463 If the argument @var{second_time} is nonzero, it means that the
4464 arguments of the function are being analyzed for the second time. This
4465 happens for an inline function, which is not actually compiled until the
4466 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
4467 not generate any instructions in this case.
4470 @defmac STRICT_ARGUMENT_NAMING
4471 Define this macro to be a nonzero value if the location where a function
4472 argument is passed depends on whether or not it is a named argument.
4474 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
4475 is set for varargs and stdarg functions. If this macro returns a
4476 nonzero value, the @var{named} argument is always true for named
4477 arguments, and false for unnamed arguments. If it returns a value of
4478 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
4479 are treated as named. Otherwise, all named arguments except the last
4480 are treated as named.
4482 You need not define this macro if it always returns zero.
4485 @defmac PRETEND_OUTGOING_VARARGS_NAMED
4486 If you need to conditionally change ABIs so that one works with
4487 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
4488 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
4489 defined, then define this macro to return nonzero if
4490 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
4491 Otherwise, you should not define this macro.
4495 @section Trampolines for Nested Functions
4496 @cindex trampolines for nested functions
4497 @cindex nested functions, trampolines for
4499 A @dfn{trampoline} is a small piece of code that is created at run time
4500 when the address of a nested function is taken. It normally resides on
4501 the stack, in the stack frame of the containing function. These macros
4502 tell GCC how to generate code to allocate and initialize a
4505 The instructions in the trampoline must do two things: load a constant
4506 address into the static chain register, and jump to the real address of
4507 the nested function. On CISC machines such as the m68k, this requires
4508 two instructions, a move immediate and a jump. Then the two addresses
4509 exist in the trampoline as word-long immediate operands. On RISC
4510 machines, it is often necessary to load each address into a register in
4511 two parts. Then pieces of each address form separate immediate
4514 The code generated to initialize the trampoline must store the variable
4515 parts---the static chain value and the function address---into the
4516 immediate operands of the instructions. On a CISC machine, this is
4517 simply a matter of copying each address to a memory reference at the
4518 proper offset from the start of the trampoline. On a RISC machine, it
4519 may be necessary to take out pieces of the address and store them
4522 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4523 A C statement to output, on the stream @var{file}, assembler code for a
4524 block of data that contains the constant parts of a trampoline. This
4525 code should not include a label---the label is taken care of
4528 If you do not define this macro, it means no template is needed
4529 for the target. Do not define this macro on systems where the block move
4530 code to copy the trampoline into place would be larger than the code
4531 to generate it on the spot.
4534 @defmac TRAMPOLINE_SECTION
4535 The name of a subroutine to switch to the section in which the
4536 trampoline template is to be placed (@pxref{Sections}). The default is
4537 a value of @samp{readonly_data_section}, which places the trampoline in
4538 the section containing read-only data.
4541 @defmac TRAMPOLINE_SIZE
4542 A C expression for the size in bytes of the trampoline, as an integer.
4545 @defmac TRAMPOLINE_ALIGNMENT
4546 Alignment required for trampolines, in bits.
4548 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4549 is used for aligning trampolines.
4552 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4553 A C statement to initialize the variable parts of a trampoline.
4554 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4555 an RTX for the address of the nested function; @var{static_chain} is an
4556 RTX for the static chain value that should be passed to the function
4560 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4561 A C statement that should perform any machine-specific adjustment in
4562 the address of the trampoline. Its argument contains the address that
4563 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4564 used for a function call should be different from the address in which
4565 the template was stored, the different address should be assigned to
4566 @var{addr}. If this macro is not defined, @var{addr} will be used for
4569 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4570 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4571 If this macro is not defined, by default the trampoline is allocated as
4572 a stack slot. This default is right for most machines. The exceptions
4573 are machines where it is impossible to execute instructions in the stack
4574 area. On such machines, you may have to implement a separate stack,
4575 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4576 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4578 @var{fp} points to a data structure, a @code{struct function}, which
4579 describes the compilation status of the immediate containing function of
4580 the function which the trampoline is for. The stack slot for the
4581 trampoline is in the stack frame of this containing function. Other
4582 allocation strategies probably must do something analogous with this
4586 Implementing trampolines is difficult on many machines because they have
4587 separate instruction and data caches. Writing into a stack location
4588 fails to clear the memory in the instruction cache, so when the program
4589 jumps to that location, it executes the old contents.
4591 Here are two possible solutions. One is to clear the relevant parts of
4592 the instruction cache whenever a trampoline is set up. The other is to
4593 make all trampolines identical, by having them jump to a standard
4594 subroutine. The former technique makes trampoline execution faster; the
4595 latter makes initialization faster.
4597 To clear the instruction cache when a trampoline is initialized, define
4598 the following macro.
4600 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4601 If defined, expands to a C expression clearing the @emph{instruction
4602 cache} in the specified interval. The definition of this macro would
4603 typically be a series of @code{asm} statements. Both @var{beg} and
4604 @var{end} are both pointer expressions.
4607 To use a standard subroutine, define the following macro. In addition,
4608 you must make sure that the instructions in a trampoline fill an entire
4609 cache line with identical instructions, or else ensure that the
4610 beginning of the trampoline code is always aligned at the same point in
4611 its cache line. Look in @file{m68k.h} as a guide.
4613 @defmac TRANSFER_FROM_TRAMPOLINE
4614 Define this macro if trampolines need a special subroutine to do their
4615 work. The macro should expand to a series of @code{asm} statements
4616 which will be compiled with GCC@. They go in a library function named
4617 @code{__transfer_from_trampoline}.
4619 If you need to avoid executing the ordinary prologue code of a compiled
4620 C function when you jump to the subroutine, you can do so by placing a
4621 special label of your own in the assembler code. Use one @code{asm}
4622 statement to generate an assembler label, and another to make the label
4623 global. Then trampolines can use that label to jump directly to your
4624 special assembler code.
4628 @section Implicit Calls to Library Routines
4629 @cindex library subroutine names
4630 @cindex @file{libgcc.a}
4632 @c prevent bad page break with this line
4633 Here is an explanation of implicit calls to library routines.
4635 @defmac MULSI3_LIBCALL
4636 A C string constant giving the name of the function to call for
4637 multiplication of one signed full-word by another. If you do not
4638 define this macro, the default name is used, which is @code{__mulsi3},
4639 a function defined in @file{libgcc.a}.
4642 @defmac DIVSI3_LIBCALL
4643 A C string constant giving the name of the function to call for
4644 division of one signed full-word by another. If you do not define
4645 this macro, the default name is used, which is @code{__divsi3}, a
4646 function defined in @file{libgcc.a}.
4649 @defmac UDIVSI3_LIBCALL
4650 A C string constant giving the name of the function to call for
4651 division of one unsigned full-word by another. If you do not define
4652 this macro, the default name is used, which is @code{__udivsi3}, a
4653 function defined in @file{libgcc.a}.
4656 @defmac MODSI3_LIBCALL
4657 A C string constant giving the name of the function to call for the
4658 remainder in division of one signed full-word by another. If you do
4659 not define this macro, the default name is used, which is
4660 @code{__modsi3}, a function defined in @file{libgcc.a}.
4663 @defmac UMODSI3_LIBCALL
4664 A C string constant giving the name of the function to call for the
4665 remainder in division of one unsigned full-word by another. If you do
4666 not define this macro, the default name is used, which is
4667 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4670 @defmac MULDI3_LIBCALL
4671 A C string constant giving the name of the function to call for
4672 multiplication of one signed double-word by another. If you do not
4673 define this macro, the default name is used, which is @code{__muldi3},
4674 a function defined in @file{libgcc.a}.
4677 @defmac DIVDI3_LIBCALL
4678 A C string constant giving the name of the function to call for
4679 division of one signed double-word by another. If you do not define
4680 this macro, the default name is used, which is @code{__divdi3}, a
4681 function defined in @file{libgcc.a}.
4684 @defmac UDIVDI3_LIBCALL
4685 A C string constant giving the name of the function to call for
4686 division of one unsigned full-word by another. If you do not define
4687 this macro, the default name is used, which is @code{__udivdi3}, a
4688 function defined in @file{libgcc.a}.
4691 @defmac MODDI3_LIBCALL
4692 A C string constant giving the name of the function to call for the
4693 remainder in division of one signed double-word by another. If you do
4694 not define this macro, the default name is used, which is
4695 @code{__moddi3}, a function defined in @file{libgcc.a}.
4698 @defmac UMODDI3_LIBCALL
4699 A C string constant giving the name of the function to call for the
4700 remainder in division of one unsigned full-word by another. If you do
4701 not define this macro, the default name is used, which is
4702 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4705 @defmac DECLARE_LIBRARY_RENAMES
4706 This macro, if defined, should expand to a piece of C code that will get
4707 expanded when compiling functions for libgcc.a. It can be used to
4708 provide alternate names for gcc's internal library functions if there
4709 are ABI-mandated names that the compiler should provide.
4712 @defmac INIT_TARGET_OPTABS
4713 Define this macro as a C statement that declares additional library
4714 routines renames existing ones. @code{init_optabs} calls this macro after
4715 initializing all the normal library routines.
4718 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4719 Define this macro as a C statement that returns nonzero if a call to
4720 the floating point comparison library function will return a boolean
4721 value that indicates the result of the comparison. It should return
4722 zero if one of gcc's own libgcc functions is called.
4724 Most ports don't need to define this macro.
4727 @cindex @code{EDOM}, implicit usage
4730 The value of @code{EDOM} on the target machine, as a C integer constant
4731 expression. If you don't define this macro, GCC does not attempt to
4732 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4733 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4736 If you do not define @code{TARGET_EDOM}, then compiled code reports
4737 domain errors by calling the library function and letting it report the
4738 error. If mathematical functions on your system use @code{matherr} when
4739 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4740 that @code{matherr} is used normally.
4743 @cindex @code{errno}, implicit usage
4744 @defmac GEN_ERRNO_RTX
4745 Define this macro as a C expression to create an rtl expression that
4746 refers to the global ``variable'' @code{errno}. (On certain systems,
4747 @code{errno} may not actually be a variable.) If you don't define this
4748 macro, a reasonable default is used.
4751 @cindex @code{bcopy}, implicit usage
4752 @cindex @code{memcpy}, implicit usage
4753 @cindex @code{memmove}, implicit usage
4754 @cindex @code{bzero}, implicit usage
4755 @cindex @code{memset}, implicit usage
4756 @defmac TARGET_MEM_FUNCTIONS
4757 Define this macro if GCC should generate calls to the ISO C
4758 (and System V) library functions @code{memcpy}, @code{memmove} and
4759 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4762 @cindex C99 math functions, implicit usage
4763 @defmac TARGET_C99_FUNCTIONS
4764 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4765 @code{sinf} and similarly for other functions defined by C99 standard. The
4766 default is nonzero that should be proper value for most modern systems, however
4767 number of existing systems lacks support for these functions in the runtime so
4768 they needs this macro to be redefined to 0.
4771 @defmac LIBGCC_NEEDS_DOUBLE
4772 Define this macro if @code{float} arguments cannot be passed to library
4773 routines (so they must be converted to @code{double}). This macro
4774 affects both how library calls are generated and how the library
4775 routines in @file{libgcc.a} accept their arguments. It is useful on
4776 machines where floating and fixed point arguments are passed
4777 differently, such as the i860.
4780 @defmac NEXT_OBJC_RUNTIME
4781 Define this macro to generate code for Objective-C message sending using
4782 the calling convention of the NeXT system. This calling convention
4783 involves passing the object, the selector and the method arguments all
4784 at once to the method-lookup library function.
4786 The default calling convention passes just the object and the selector
4787 to the lookup function, which returns a pointer to the method.
4790 @node Addressing Modes
4791 @section Addressing Modes
4792 @cindex addressing modes
4794 @c prevent bad page break with this line
4795 This is about addressing modes.
4797 @defmac HAVE_PRE_INCREMENT
4798 @defmacx HAVE_PRE_DECREMENT
4799 @defmacx HAVE_POST_INCREMENT
4800 @defmacx HAVE_POST_DECREMENT
4801 A C expression that is nonzero if the machine supports pre-increment,
4802 pre-decrement, post-increment, or post-decrement addressing respectively.
4805 @defmac HAVE_PRE_MODIFY_DISP
4806 @defmacx HAVE_POST_MODIFY_DISP
4807 A C expression that is nonzero if the machine supports pre- or
4808 post-address side-effect generation involving constants other than
4809 the size of the memory operand.
4812 @defmac HAVE_PRE_MODIFY_REG
4813 @defmacx HAVE_POST_MODIFY_REG
4814 A C expression that is nonzero if the machine supports pre- or
4815 post-address side-effect generation involving a register displacement.
4818 @defmac CONSTANT_ADDRESS_P (@var{x})
4819 A C expression that is 1 if the RTX @var{x} is a constant which
4820 is a valid address. On most machines, this can be defined as
4821 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4822 in which constant addresses are supported.
4825 @defmac CONSTANT_P (@var{x})
4826 @code{CONSTANT_P}, which is defined by target-independent code,
4827 accepts integer-values expressions whose values are not explicitly
4828 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4829 expressions and @code{const} arithmetic expressions, in addition to
4830 @code{const_int} and @code{const_double} expressions.
4833 @defmac MAX_REGS_PER_ADDRESS
4834 A number, the maximum number of registers that can appear in a valid
4835 memory address. Note that it is up to you to specify a value equal to
4836 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4840 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4841 A C compound statement with a conditional @code{goto @var{label};}
4842 executed if @var{x} (an RTX) is a legitimate memory address on the
4843 target machine for a memory operand of mode @var{mode}.
4845 It usually pays to define several simpler macros to serve as
4846 subroutines for this one. Otherwise it may be too complicated to
4849 This macro must exist in two variants: a strict variant and a
4850 non-strict one. The strict variant is used in the reload pass. It
4851 must be defined so that any pseudo-register that has not been
4852 allocated a hard register is considered a memory reference. In
4853 contexts where some kind of register is required, a pseudo-register
4854 with no hard register must be rejected.
4856 The non-strict variant is used in other passes. It must be defined to
4857 accept all pseudo-registers in every context where some kind of
4858 register is required.
4860 @findex REG_OK_STRICT
4861 Compiler source files that want to use the strict variant of this
4862 macro define the macro @code{REG_OK_STRICT}. You should use an
4863 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4864 in that case and the non-strict variant otherwise.
4866 Subroutines to check for acceptable registers for various purposes (one
4867 for base registers, one for index registers, and so on) are typically
4868 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4869 Then only these subroutine macros need have two variants; the higher
4870 levels of macros may be the same whether strict or not.
4872 Normally, constant addresses which are the sum of a @code{symbol_ref}
4873 and an integer are stored inside a @code{const} RTX to mark them as
4874 constant. Therefore, there is no need to recognize such sums
4875 specifically as legitimate addresses. Normally you would simply
4876 recognize any @code{const} as legitimate.
4878 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4879 sums that are not marked with @code{const}. It assumes that a naked
4880 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4881 naked constant sums as illegitimate addresses, so that none of them will
4882 be given to @code{PRINT_OPERAND_ADDRESS}.
4884 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4885 On some machines, whether a symbolic address is legitimate depends on
4886 the section that the address refers to. On these machines, define the
4887 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4888 into the @code{symbol_ref}, and then check for it here. When you see a
4889 @code{const}, you will have to look inside it to find the
4890 @code{symbol_ref} in order to determine the section. @xref{Assembler
4894 @defmac REG_OK_FOR_BASE_P (@var{x})
4895 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4896 RTX) is valid for use as a base register. For hard registers, it
4897 should always accept those which the hardware permits and reject the
4898 others. Whether the macro accepts or rejects pseudo registers must be
4899 controlled by @code{REG_OK_STRICT} as described above. This usually
4900 requires two variant definitions, of which @code{REG_OK_STRICT}
4901 controls the one actually used.
4904 @defmac REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4905 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4906 that expression may examine the mode of the memory reference in
4907 @var{mode}. You should define this macro if the mode of the memory
4908 reference affects whether a register may be used as a base register. If
4909 you define this macro, the compiler will use it instead of
4910 @code{REG_OK_FOR_BASE_P}.
4913 @defmac REG_OK_FOR_INDEX_P (@var{x})
4914 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4915 RTX) is valid for use as an index register.
4917 The difference between an index register and a base register is that
4918 the index register may be scaled. If an address involves the sum of
4919 two registers, neither one of them scaled, then either one may be
4920 labeled the ``base'' and the other the ``index''; but whichever
4921 labeling is used must fit the machine's constraints of which registers
4922 may serve in each capacity. The compiler will try both labelings,
4923 looking for one that is valid, and will reload one or both registers
4924 only if neither labeling works.
4927 @defmac FIND_BASE_TERM (@var{x})
4928 A C expression to determine the base term of address @var{x}.
4929 This macro is used in only one place: `find_base_term' in alias.c.
4931 It is always safe for this macro to not be defined. It exists so
4932 that alias analysis can understand machine-dependent addresses.
4934 The typical use of this macro is to handle addresses containing
4935 a label_ref or symbol_ref within an UNSPEC@.
4938 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4939 A C compound statement that attempts to replace @var{x} with a valid
4940 memory address for an operand of mode @var{mode}. @var{win} will be a
4941 C statement label elsewhere in the code; the macro definition may use
4944 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4948 to avoid further processing if the address has become legitimate.
4950 @findex break_out_memory_refs
4951 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4952 and @var{oldx} will be the operand that was given to that function to produce
4955 The code generated by this macro should not alter the substructure of
4956 @var{x}. If it transforms @var{x} into a more legitimate form, it
4957 should assign @var{x} (which will always be a C variable) a new value.
4959 It is not necessary for this macro to come up with a legitimate
4960 address. The compiler has standard ways of doing so in all cases. In
4961 fact, it is safe for this macro to do nothing. But often a
4962 machine-dependent strategy can generate better code.
4965 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4966 A C compound statement that attempts to replace @var{x}, which is an address
4967 that needs reloading, with a valid memory address for an operand of mode
4968 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4969 It is not necessary to define this macro, but it might be useful for
4970 performance reasons.
4972 For example, on the i386, it is sometimes possible to use a single
4973 reload register instead of two by reloading a sum of two pseudo
4974 registers into a register. On the other hand, for number of RISC
4975 processors offsets are limited so that often an intermediate address
4976 needs to be generated in order to address a stack slot. By defining
4977 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4978 generated for adjacent some stack slots can be made identical, and thus
4981 @emph{Note}: This macro should be used with caution. It is necessary
4982 to know something of how reload works in order to effectively use this,
4983 and it is quite easy to produce macros that build in too much knowledge
4984 of reload internals.
4986 @emph{Note}: This macro must be able to reload an address created by a
4987 previous invocation of this macro. If it fails to handle such addresses
4988 then the compiler may generate incorrect code or abort.
4991 The macro definition should use @code{push_reload} to indicate parts that
4992 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4993 suitable to be passed unaltered to @code{push_reload}.
4995 The code generated by this macro must not alter the substructure of
4996 @var{x}. If it transforms @var{x} into a more legitimate form, it
4997 should assign @var{x} (which will always be a C variable) a new value.
4998 This also applies to parts that you change indirectly by calling
5001 @findex strict_memory_address_p
5002 The macro definition may use @code{strict_memory_address_p} to test if
5003 the address has become legitimate.
5006 If you want to change only a part of @var{x}, one standard way of doing
5007 this is to use @code{copy_rtx}. Note, however, that is unshares only a
5008 single level of rtl. Thus, if the part to be changed is not at the
5009 top level, you'll need to replace first the top level.
5010 It is not necessary for this macro to come up with a legitimate
5011 address; but often a machine-dependent strategy can generate better code.
5014 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5015 A C statement or compound statement with a conditional @code{goto
5016 @var{label};} executed if memory address @var{x} (an RTX) can have
5017 different meanings depending on the machine mode of the memory
5018 reference it is used for or if the address is valid for some modes
5021 Autoincrement and autodecrement addresses typically have mode-dependent
5022 effects because the amount of the increment or decrement is the size
5023 of the operand being addressed. Some machines have other mode-dependent
5024 addresses. Many RISC machines have no mode-dependent addresses.
5026 You may assume that @var{addr} is a valid address for the machine.
5029 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5030 A C expression that is nonzero if @var{x} is a legitimate constant for
5031 an immediate operand on the target machine. You can assume that
5032 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5033 @samp{1} is a suitable definition for this macro on machines where
5034 anything @code{CONSTANT_P} is valid.
5037 @node Condition Code
5038 @section Condition Code Status
5039 @cindex condition code status
5041 @c prevent bad page break with this line
5042 This describes the condition code status.
5045 The file @file{conditions.h} defines a variable @code{cc_status} to
5046 describe how the condition code was computed (in case the interpretation of
5047 the condition code depends on the instruction that it was set by). This
5048 variable contains the RTL expressions on which the condition code is
5049 currently based, and several standard flags.
5051 Sometimes additional machine-specific flags must be defined in the machine
5052 description header file. It can also add additional machine-specific
5053 information by defining @code{CC_STATUS_MDEP}.
5055 @defmac CC_STATUS_MDEP
5056 C code for a data type which is used for declaring the @code{mdep}
5057 component of @code{cc_status}. It defaults to @code{int}.
5059 This macro is not used on machines that do not use @code{cc0}.
5062 @defmac CC_STATUS_MDEP_INIT
5063 A C expression to initialize the @code{mdep} field to ``empty''.
5064 The default definition does nothing, since most machines don't use
5065 the field anyway. If you want to use the field, you should probably
5066 define this macro to initialize it.
5068 This macro is not used on machines that do not use @code{cc0}.
5071 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5072 A C compound statement to set the components of @code{cc_status}
5073 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5074 this macro's responsibility to recognize insns that set the condition
5075 code as a byproduct of other activity as well as those that explicitly
5078 This macro is not used on machines that do not use @code{cc0}.
5080 If there are insns that do not set the condition code but do alter
5081 other machine registers, this macro must check to see whether they
5082 invalidate the expressions that the condition code is recorded as
5083 reflecting. For example, on the 68000, insns that store in address
5084 registers do not set the condition code, which means that usually
5085 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5086 insns. But suppose that the previous insn set the condition code
5087 based on location @samp{a4@@(102)} and the current insn stores a new
5088 value in @samp{a4}. Although the condition code is not changed by
5089 this, it will no longer be true that it reflects the contents of
5090 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5091 @code{cc_status} in this case to say that nothing is known about the
5092 condition code value.
5094 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5095 with the results of peephole optimization: insns whose patterns are
5096 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5097 constants which are just the operands. The RTL structure of these
5098 insns is not sufficient to indicate what the insns actually do. What
5099 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5100 @code{CC_STATUS_INIT}.
5102 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5103 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5104 @samp{cc}. This avoids having detailed information about patterns in
5105 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5108 @defmac EXTRA_CC_MODES
5109 Condition codes are represented in registers by machine modes of class
5110 @code{MODE_CC}. By default, there is just one mode, @code{CCmode}, with
5111 this class. If you need more such modes, create a file named
5112 @file{@var{machine}-modes.def} in your @file{config/@var{machine}}
5113 directory (@pxref{Back End, , Anatomy of a Target Back End}), containing
5114 a list of these modes. Each entry in the list should be a call to the
5115 macro @code{CC}. This macro takes one argument, which is the name of
5116 the mode: it should begin with @samp{CC}. Do not put quotation marks
5117 around the name, or include the trailing @samp{mode}; these are
5118 automatically added. There should not be anything else in the file
5121 A sample @file{@var{machine}-modes.def} file might look like this:
5124 CC (CC_NOOV) /* @r{Comparison only valid if there was no overflow.} */
5125 CC (CCFP) /* @r{Floating point comparison that cannot trap.} */
5126 CC (CCFPE) /* @r{Floating point comparison that may trap.} */
5129 When you create this file, the macro @code{EXTRA_CC_MODES} is
5130 automatically defined by @command{configure}, with value @samp{1}.
5133 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5134 Returns a mode from class @code{MODE_CC} to be used when comparison
5135 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5136 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5137 @pxref{Jump Patterns} for a description of the reason for this
5141 #define SELECT_CC_MODE(OP,X,Y) \
5142 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5143 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5144 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5145 || GET_CODE (X) == NEG) \
5146 ? CC_NOOVmode : CCmode))
5149 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
5152 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5153 On some machines not all possible comparisons are defined, but you can
5154 convert an invalid comparison into a valid one. For example, the Alpha
5155 does not have a @code{GT} comparison, but you can use an @code{LT}
5156 comparison instead and swap the order of the operands.
5158 On such machines, define this macro to be a C statement to do any
5159 required conversions. @var{code} is the initial comparison code
5160 and @var{op0} and @var{op1} are the left and right operands of the
5161 comparison, respectively. You should modify @var{code}, @var{op0}, and
5162 @var{op1} as required.
5164 GCC will not assume that the comparison resulting from this macro is
5165 valid but will see if the resulting insn matches a pattern in the
5168 You need not define this macro if it would never change the comparison
5172 @defmac REVERSIBLE_CC_MODE (@var{mode})
5173 A C expression whose value is one if it is always safe to reverse a
5174 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5175 can ever return @var{mode} for a floating-point inequality comparison,
5176 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5178 You need not define this macro if it would always returns zero or if the
5179 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5180 For example, here is the definition used on the SPARC, where floating-point
5181 inequality comparisons are always given @code{CCFPEmode}:
5184 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5188 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5189 A C expression whose value is reversed condition code of the @var{code} for
5190 comparison done in CC_MODE @var{mode}. The macro is used only in case
5191 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5192 machine has some non-standard way how to reverse certain conditionals. For
5193 instance in case all floating point conditions are non-trapping, compiler may
5194 freely convert unordered compares to ordered one. Then definition may look
5198 #define REVERSE_CONDITION(CODE, MODE) \
5199 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5200 : reverse_condition_maybe_unordered (CODE))
5204 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
5205 A C expression that returns true if the conditional execution predicate
5206 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
5207 return 0 if the target has conditional execution predicates that cannot be
5208 reversed safely. If no expansion is specified, this macro is defined as
5212 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5213 ((x) == reverse_condition (y))
5218 @section Describing Relative Costs of Operations
5219 @cindex costs of instructions
5220 @cindex relative costs
5221 @cindex speed of instructions
5223 These macros let you describe the relative speed of various operations
5224 on the target machine.
5226 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5227 A C expression for the cost of moving data of mode @var{mode} from a
5228 register in class @var{from} to one in class @var{to}. The classes are
5229 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5230 value of 2 is the default; other values are interpreted relative to
5233 It is not required that the cost always equal 2 when @var{from} is the
5234 same as @var{to}; on some machines it is expensive to move between
5235 registers if they are not general registers.
5237 If reload sees an insn consisting of a single @code{set} between two
5238 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5239 classes returns a value of 2, reload does not check to ensure that the
5240 constraints of the insn are met. Setting a cost of other than 2 will
5241 allow reload to verify that the constraints are met. You should do this
5242 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5245 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5246 A C expression for the cost of moving data of mode @var{mode} between a
5247 register of class @var{class} and memory; @var{in} is zero if the value
5248 is to be written to memory, nonzero if it is to be read in. This cost
5249 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5250 registers and memory is more expensive than between two registers, you
5251 should define this macro to express the relative cost.
5253 If you do not define this macro, GCC uses a default cost of 4 plus
5254 the cost of copying via a secondary reload register, if one is
5255 needed. If your machine requires a secondary reload register to copy
5256 between memory and a register of @var{class} but the reload mechanism is
5257 more complex than copying via an intermediate, define this macro to
5258 reflect the actual cost of the move.
5260 GCC defines the function @code{memory_move_secondary_cost} if
5261 secondary reloads are needed. It computes the costs due to copying via
5262 a secondary register. If your machine copies from memory using a
5263 secondary register in the conventional way but the default base value of
5264 4 is not correct for your machine, define this macro to add some other
5265 value to the result of that function. The arguments to that function
5266 are the same as to this macro.
5270 A C expression for the cost of a branch instruction. A value of 1 is
5271 the default; other values are interpreted relative to that.
5274 Here are additional macros which do not specify precise relative costs,
5275 but only that certain actions are more expensive than GCC would
5278 @defmac SLOW_BYTE_ACCESS
5279 Define this macro as a C expression which is nonzero if accessing less
5280 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5281 faster than accessing a word of memory, i.e., if such access
5282 require more than one instruction or if there is no difference in cost
5283 between byte and (aligned) word loads.
5285 When this macro is not defined, the compiler will access a field by
5286 finding the smallest containing object; when it is defined, a fullword
5287 load will be used if alignment permits. Unless bytes accesses are
5288 faster than word accesses, using word accesses is preferable since it
5289 may eliminate subsequent memory access if subsequent accesses occur to
5290 other fields in the same word of the structure, but to different bytes.
5293 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5294 Define this macro to be the value 1 if memory accesses described by the
5295 @var{mode} and @var{alignment} parameters have a cost many times greater
5296 than aligned accesses, for example if they are emulated in a trap
5299 When this macro is nonzero, the compiler will act as if
5300 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5301 moves. This can cause significantly more instructions to be produced.
5302 Therefore, do not set this macro nonzero if unaligned accesses only add a
5303 cycle or two to the time for a memory access.
5305 If the value of this macro is always zero, it need not be defined. If
5306 this macro is defined, it should produce a nonzero value when
5307 @code{STRICT_ALIGNMENT} is nonzero.
5311 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5312 which a sequence of insns should be generated instead of a
5313 string move insn or a library call. Increasing the value will always
5314 make code faster, but eventually incurs high cost in increased code size.
5316 Note that on machines where the corresponding move insn is a
5317 @code{define_expand} that emits a sequence of insns, this macro counts
5318 the number of such sequences.
5320 If you don't define this, a reasonable default is used.
5323 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5324 A C expression used to determine whether @code{move_by_pieces} will be used to
5325 copy a chunk of memory, or whether some other block move mechanism
5326 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5327 than @code{MOVE_RATIO}.
5330 @defmac MOVE_MAX_PIECES
5331 A C expression used by @code{move_by_pieces} to determine the largest unit
5332 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5336 The threshold of number of scalar move insns, @emph{below} which a sequence
5337 of insns should be generated to clear memory instead of a string clear insn
5338 or a library call. Increasing the value will always make code faster, but
5339 eventually incurs high cost in increased code size.
5341 If you don't define this, a reasonable default is used.
5344 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5345 A C expression used to determine whether @code{clear_by_pieces} will be used
5346 to clear a chunk of memory, or whether some other block clear mechanism
5347 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5348 than @code{CLEAR_RATIO}.
5351 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5352 A C expression used to determine whether @code{store_by_pieces} will be
5353 used to set a chunk of memory to a constant value, or whether some other
5354 mechanism will be used. Used by @code{__builtin_memset} when storing
5355 values other than constant zero and by @code{__builtin_strcpy} when
5356 when called with a constant source string.
5357 Defaults to @code{MOVE_BY_PIECES_P}.
5360 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5361 A C expression used to determine whether a load postincrement is a good
5362 thing to use for a given mode. Defaults to the value of
5363 @code{HAVE_POST_INCREMENT}.
5366 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5367 A C expression used to determine whether a load postdecrement is a good
5368 thing to use for a given mode. Defaults to the value of
5369 @code{HAVE_POST_DECREMENT}.
5372 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5373 A C expression used to determine whether a load preincrement is a good
5374 thing to use for a given mode. Defaults to the value of
5375 @code{HAVE_PRE_INCREMENT}.
5378 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5379 A C expression used to determine whether a load predecrement is a good
5380 thing to use for a given mode. Defaults to the value of
5381 @code{HAVE_PRE_DECREMENT}.
5384 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5385 A C expression used to determine whether a store postincrement is a good
5386 thing to use for a given mode. Defaults to the value of
5387 @code{HAVE_POST_INCREMENT}.
5390 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5391 A C expression used to determine whether a store postdecrement is a good
5392 thing to use for a given mode. Defaults to the value of
5393 @code{HAVE_POST_DECREMENT}.
5396 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5397 This macro is used to determine whether a store preincrement is a good
5398 thing to use for a given mode. Defaults to the value of
5399 @code{HAVE_PRE_INCREMENT}.
5402 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5403 This macro is used to determine whether a store predecrement is a good
5404 thing to use for a given mode. Defaults to the value of
5405 @code{HAVE_PRE_DECREMENT}.
5408 @defmac NO_FUNCTION_CSE
5409 Define this macro if it is as good or better to call a constant
5410 function address than to call an address kept in a register.
5413 @defmac NO_RECURSIVE_FUNCTION_CSE
5414 Define this macro if it is as good or better for a function to call
5415 itself with an explicit address than to call an address kept in a
5419 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5420 Define this macro if a non-short-circuit operation produced by
5421 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5422 @code{BRANCH_COST} is greater than or equal to the value 2.
5425 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5426 This target hook describes the relative costs of RTL expressions.
5428 The cost may depend on the precise form of the expression, which is
5429 available for examination in @var{x}, and the rtx code of the expression
5430 in which it is contained, found in @var{outer_code}. @var{code} is the
5431 expression code---redundant, since it can be obtained with
5432 @code{GET_CODE (@var{x})}.
5434 In implementing this hook, you can use the construct
5435 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5438 On entry to the hook, @code{*@var{total}} contains a default estimate
5439 for the cost of the expression. The hook should modify this value as
5442 The hook returns true when all subexpressions of @var{x} have been
5443 processed, and false when @code{rtx_cost} should recurse.
5446 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5447 This hook computes the cost of an addressing mode that contains
5448 @var{address}. If not defined, the cost is computed from
5449 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5451 For most CISC machines, the default cost is a good approximation of the
5452 true cost of the addressing mode. However, on RISC machines, all
5453 instructions normally have the same length and execution time. Hence
5454 all addresses will have equal costs.
5456 In cases where more than one form of an address is known, the form with
5457 the lowest cost will be used. If multiple forms have the same, lowest,
5458 cost, the one that is the most complex will be used.
5460 For example, suppose an address that is equal to the sum of a register
5461 and a constant is used twice in the same basic block. When this macro
5462 is not defined, the address will be computed in a register and memory
5463 references will be indirect through that register. On machines where
5464 the cost of the addressing mode containing the sum is no higher than
5465 that of a simple indirect reference, this will produce an additional
5466 instruction and possibly require an additional register. Proper
5467 specification of this macro eliminates this overhead for such machines.
5469 This hook is never called with an invalid address.
5471 On machines where an address involving more than one register is as
5472 cheap as an address computation involving only one register, defining
5473 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5474 be live over a region of code where only one would have been if
5475 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5476 should be considered in the definition of this macro. Equivalent costs
5477 should probably only be given to addresses with different numbers of
5478 registers on machines with lots of registers.
5482 @section Adjusting the Instruction Scheduler
5484 The instruction scheduler may need a fair amount of machine-specific
5485 adjustment in order to produce good code. GCC provides several target
5486 hooks for this purpose. It is usually enough to define just a few of
5487 them: try the first ones in this list first.
5489 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5490 This hook returns the maximum number of instructions that can ever
5491 issue at the same time on the target machine. The default is one.
5492 Although the insn scheduler can define itself the possibility of issue
5493 an insn on the same cycle, the value can serve as an additional
5494 constraint to issue insns on the same simulated processor cycle (see
5495 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5496 This value must be constant over the entire compilation. If you need
5497 it to vary depending on what the instructions are, you must use
5498 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5500 For the automaton based pipeline interface, you could define this hook
5501 to return the value of the macro @code{MAX_DFA_ISSUE_RATE}.
5504 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5505 This hook is executed by the scheduler after it has scheduled an insn
5506 from the ready list. It should return the number of insns which can
5507 still be issued in the current cycle. The default is
5508 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5509 @code{USE}, which normally are not counted against the issue rate.
5510 You should define this hook if some insns take more machine resources
5511 than others, so that fewer insns can follow them in the same cycle.
5512 @var{file} is either a null pointer, or a stdio stream to write any
5513 debug output to. @var{verbose} is the verbose level provided by
5514 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5518 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5519 This function corrects the value of @var{cost} based on the
5520 relationship between @var{insn} and @var{dep_insn} through the
5521 dependence @var{link}. It should return the new value. The default
5522 is to make no adjustment to @var{cost}. This can be used for example
5523 to specify to the scheduler using the traditional pipeline description
5524 that an output- or anti-dependence does not incur the same cost as a
5525 data-dependence. If the scheduler using the automaton based pipeline
5526 description, the cost of anti-dependence is zero and the cost of
5527 output-dependence is maximum of one and the difference of latency
5528 times of the first and the second insns. If these values are not
5529 acceptable, you could use the hook to modify them too. See also
5530 @pxref{Automaton pipeline description}.
5533 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5534 This hook adjusts the integer scheduling priority @var{priority} of
5535 @var{insn}. It should return the new priority. Reduce the priority to
5536 execute @var{insn} earlier, increase the priority to execute @var{insn}
5537 later. Do not define this hook if you do not need to adjust the
5538 scheduling priorities of insns.
5541 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5542 This hook is executed by the scheduler after it has scheduled the ready
5543 list, to allow the machine description to reorder it (for example to
5544 combine two small instructions together on @samp{VLIW} machines).
5545 @var{file} is either a null pointer, or a stdio stream to write any
5546 debug output to. @var{verbose} is the verbose level provided by
5547 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5548 list of instructions that are ready to be scheduled. @var{n_readyp} is
5549 a pointer to the number of elements in the ready list. The scheduler
5550 reads the ready list in reverse order, starting with
5551 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5552 is the timer tick of the scheduler. You may modify the ready list and
5553 the number of ready insns. The return value is the number of insns that
5554 can issue this cycle; normally this is just @code{issue_rate}. See also
5555 @samp{TARGET_SCHED_REORDER2}.
5558 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5559 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5560 function is called whenever the scheduler starts a new cycle. This one
5561 is called once per iteration over a cycle, immediately after
5562 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5563 return the number of insns to be scheduled in the same cycle. Defining
5564 this hook can be useful if there are frequent situations where
5565 scheduling one insn causes other insns to become ready in the same
5566 cycle. These other insns can then be taken into account properly.
5569 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5570 This hook is called after evaluation forward dependencies of insns in
5571 chain given by two parameter values (@var{head} and @var{tail}
5572 correspondingly) but before insns scheduling of the insn chain. For
5573 example, it can be used for better insn classification if it requires
5574 analysis of dependencies. This hook can use backward and forward
5575 dependencies of the insn scheduler because they are already
5579 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5580 This hook is executed by the scheduler at the beginning of each block of
5581 instructions that are to be scheduled. @var{file} is either a null
5582 pointer, or a stdio stream to write any debug output to. @var{verbose}
5583 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5584 @var{max_ready} is the maximum number of insns in the current scheduling
5585 region that can be live at the same time. This can be used to allocate
5586 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5589 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5590 This hook is executed by the scheduler at the end of each block of
5591 instructions that are to be scheduled. It can be used to perform
5592 cleanup of any actions done by the other scheduling hooks. @var{file}
5593 is either a null pointer, or a stdio stream to write any debug output
5594 to. @var{verbose} is the verbose level provided by
5595 @option{-fsched-verbose-@var{n}}.
5598 @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void)
5599 This hook is called many times during insn scheduling. If the hook
5600 returns nonzero, the automaton based pipeline description is used for
5601 insn scheduling. Otherwise the traditional pipeline description is
5602 used. The default is usage of the traditional pipeline description.
5604 You should also remember that to simplify the insn scheduler sources
5605 an empty traditional pipeline description interface is generated even
5606 if there is no a traditional pipeline description in the @file{.md}
5607 file. The same is true for the automaton based pipeline description.
5608 That means that you should be accurate in defining the hook.
5611 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5612 The hook returns an RTL insn. The automaton state used in the
5613 pipeline hazard recognizer is changed as if the insn were scheduled
5614 when the new simulated processor cycle starts. Usage of the hook may
5615 simplify the automaton pipeline description for some @acronym{VLIW}
5616 processors. If the hook is defined, it is used only for the automaton
5617 based pipeline description. The default is not to change the state
5618 when the new simulated processor cycle starts.
5621 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5622 The hook can be used to initialize data used by the previous hook.
5625 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5626 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5627 to changed the state as if the insn were scheduled when the new
5628 simulated processor cycle finishes.
5631 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5632 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5633 used to initialize data used by the previous hook.
5636 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5637 This hook controls better choosing an insn from the ready insn queue
5638 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5639 chooses the first insn from the queue. If the hook returns a positive
5640 value, an additional scheduler code tries all permutations of
5641 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5642 subsequent ready insns to choose an insn whose issue will result in
5643 maximal number of issued insns on the same cycle. For the
5644 @acronym{VLIW} processor, the code could actually solve the problem of
5645 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5646 rules of @acronym{VLIW} packing are described in the automaton.
5648 This code also could be used for superscalar @acronym{RISC}
5649 processors. Let us consider a superscalar @acronym{RISC} processor
5650 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5651 @var{B}, some insns can be executed only in pipelines @var{B} or
5652 @var{C}, and one insn can be executed in pipeline @var{B}. The
5653 processor may issue the 1st insn into @var{A} and the 2nd one into
5654 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5655 until the next cycle. If the scheduler issues the 3rd insn the first,
5656 the processor could issue all 3 insns per cycle.
5658 Actually this code demonstrates advantages of the automaton based
5659 pipeline hazard recognizer. We try quickly and easy many insn
5660 schedules to choose the best one.
5662 The default is no multipass scheduling.
5665 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5667 This hook controls what insns from the ready insn queue will be
5668 considered for the multipass insn scheduling. If the hook returns
5669 zero for insn passed as the parameter, the insn will be not chosen to
5672 The default is that any ready insns can be chosen to be issued.
5675 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5677 This hook is called by the insn scheduler before issuing insn passed
5678 as the third parameter on given cycle. If the hook returns nonzero,
5679 the insn is not issued on given processors cycle. Instead of that,
5680 the processor cycle is advanced. If the value passed through the last
5681 parameter is zero, the insn ready queue is not sorted on the new cycle
5682 start as usually. The first parameter passes file for debugging
5683 output. The second one passes the scheduler verbose level of the
5684 debugging output. The forth and the fifth parameter values are
5685 correspondingly processor cycle on which the previous insn has been
5686 issued and the current processor cycle.
5689 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void)
5690 The @acronym{DFA}-based scheduler could take the insertion of nop
5691 operations for better insn scheduling into account. It can be done
5692 only if the multi-pass insn scheduling works (see hook
5693 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}).
5695 Let us consider a @acronym{VLIW} processor insn with 3 slots. Each
5696 insn can be placed only in one of the three slots. We have 3 ready
5697 insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be
5698 placed only in the 1st slot, @var{B} can be placed only in the 3rd
5699 slot. We described the automaton which does not permit empty slot
5700 gaps between insns (usually such description is simpler). Without
5701 this code the scheduler would place each insn in 3 separate
5702 @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd
5703 slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is
5704 the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook
5705 @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or
5706 create the nop insns.
5708 You should remember that the scheduler does not insert the nop insns.
5709 It is not wise because of the following optimizations. The scheduler
5710 only considers such possibility to improve the result schedule. The
5711 nop insns should be inserted lately, e.g. on the final phase.
5714 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index})
5715 This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert
5716 nop operations for better insn scheduling when @acronym{DFA}-based
5717 scheduler makes multipass insn scheduling (see also description of
5718 hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook
5719 returns a nop insn with given @var{index}. The indexes start with
5720 zero. The hook should return @code{NULL} if there are no more nop
5721 insns with indexes greater than given index.
5724 Macros in the following table are generated by the program
5725 @file{genattr} and can be useful for writing the hooks.
5727 @defmac TRADITIONAL_PIPELINE_INTERFACE
5728 The macro definition is generated if there is a traditional pipeline
5729 description in @file{.md} file. You should also remember that to
5730 simplify the insn scheduler sources an empty traditional pipeline
5731 description interface is generated even if there is no a traditional
5732 pipeline description in the @file{.md} file. The macro can be used to
5733 distinguish the two types of the traditional interface.
5736 @defmac DFA_PIPELINE_INTERFACE
5737 The macro definition is generated if there is an automaton pipeline
5738 description in @file{.md} file. You should also remember that to
5739 simplify the insn scheduler sources an empty automaton pipeline
5740 description interface is generated even if there is no an automaton
5741 pipeline description in the @file{.md} file. The macro can be used to
5742 distinguish the two types of the automaton interface.
5745 @defmac MAX_DFA_ISSUE_RATE
5746 The macro definition is generated in the automaton based pipeline
5747 description interface. Its value is calculated from the automaton
5748 based pipeline description and is equal to maximal number of all insns
5749 described in constructions @samp{define_insn_reservation} which can be
5750 issued on the same processor cycle.
5754 @section Dividing the Output into Sections (Texts, Data, @dots{})
5755 @c the above section title is WAY too long. maybe cut the part between
5756 @c the (...)? --mew 10feb93
5758 An object file is divided into sections containing different types of
5759 data. In the most common case, there are three sections: the @dfn{text
5760 section}, which holds instructions and read-only data; the @dfn{data
5761 section}, which holds initialized writable data; and the @dfn{bss
5762 section}, which holds uninitialized data. Some systems have other kinds
5765 The compiler must tell the assembler when to switch sections. These
5766 macros control what commands to output to tell the assembler this. You
5767 can also define additional sections.
5769 @defmac TEXT_SECTION_ASM_OP
5770 A C expression whose value is a string, including spacing, containing the
5771 assembler operation that should precede instructions and read-only data.
5772 Normally @code{"\t.text"} is right.
5775 @defmac TEXT_SECTION
5776 A C statement that switches to the default section containing instructions.
5777 Normally this is not needed, as simply defining @code{TEXT_SECTION_ASM_OP}
5778 is enough. The MIPS port uses this to sort all functions after all data
5782 @defmac HOT_TEXT_SECTION_NAME
5783 If defined, a C string constant for the name of the section containing most
5784 frequently executed functions of the program. If not defined, GCC will provide
5785 a default definition if the target supports named sections.
5788 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5789 If defined, a C string constant for the name of the section containing unlikely
5790 executed functions in the program.
5793 @defmac DATA_SECTION_ASM_OP
5794 A C expression whose value is a string, including spacing, containing the
5795 assembler operation to identify the following data as writable initialized
5796 data. Normally @code{"\t.data"} is right.
5799 @defmac READONLY_DATA_SECTION_ASM_OP
5800 A C expression whose value is a string, including spacing, containing the
5801 assembler operation to identify the following data as read-only initialized
5805 @defmac READONLY_DATA_SECTION
5806 A macro naming a function to call to switch to the proper section for
5807 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5808 if defined, else fall back to @code{text_section}.
5810 The most common definition will be @code{data_section}, if the target
5811 does not have a special read-only data section, and does not put data
5812 in the text section.
5815 @defmac SHARED_SECTION_ASM_OP
5816 If defined, a C expression whose value is a string, including spacing,
5817 containing the assembler operation to identify the following data as
5818 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5821 @defmac BSS_SECTION_ASM_OP
5822 If defined, a C expression whose value is a string, including spacing,
5823 containing the assembler operation to identify the following data as
5824 uninitialized global data. If not defined, and neither
5825 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5826 uninitialized global data will be output in the data section if
5827 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5831 @defmac SHARED_BSS_SECTION_ASM_OP
5832 If defined, a C expression whose value is a string, including spacing,
5833 containing the assembler operation to identify the following data as
5834 uninitialized global shared data. If not defined, and
5835 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5838 @defmac INIT_SECTION_ASM_OP
5839 If defined, a C expression whose value is a string, including spacing,
5840 containing the assembler operation to identify the following data as
5841 initialization code. If not defined, GCC will assume such a section does
5845 @defmac FINI_SECTION_ASM_OP
5846 If defined, a C expression whose value is a string, including spacing,
5847 containing the assembler operation to identify the following data as
5848 finalization code. If not defined, GCC will assume such a section does
5852 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5853 If defined, an ASM statement that switches to a different section
5854 via @var{section_op}, calls @var{function}, and switches back to
5855 the text section. This is used in @file{crtstuff.c} if
5856 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5857 to initialization and finalization functions from the init and fini
5858 sections. By default, this macro uses a simple function call. Some
5859 ports need hand-crafted assembly code to avoid dependencies on
5860 registers initialized in the function prologue or to ensure that
5861 constant pools don't end up too far way in the text section.
5864 @defmac FORCE_CODE_SECTION_ALIGN
5865 If defined, an ASM statement that aligns a code section to some
5866 arbitrary boundary. This is used to force all fragments of the
5867 @code{.init} and @code{.fini} sections to have to same alignment
5868 and thus prevent the linker from having to add any padding.
5873 @defmac EXTRA_SECTIONS
5874 A list of names for sections other than the standard two, which are
5875 @code{in_text} and @code{in_data}. You need not define this macro
5876 on a system with no other sections (that GCC needs to use).
5879 @findex text_section
5880 @findex data_section
5881 @defmac EXTRA_SECTION_FUNCTIONS
5882 One or more functions to be defined in @file{varasm.c}. These
5883 functions should do jobs analogous to those of @code{text_section} and
5884 @code{data_section}, for your additional sections. Do not define this
5885 macro if you do not define @code{EXTRA_SECTIONS}.
5888 @defmac JUMP_TABLES_IN_TEXT_SECTION
5889 Define this macro to be an expression with a nonzero value if jump
5890 tables (for @code{tablejump} insns) should be output in the text
5891 section, along with the assembler instructions. Otherwise, the
5892 readonly data section is used.
5894 This macro is irrelevant if there is no separate readonly data section.
5897 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5898 Switches to the appropriate section for output of @var{exp}. You can
5899 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5900 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5901 requires link-time relocations. Bit 0 is set when variable contains
5902 local relocations only, while bit 1 is set for global relocations.
5903 Select the section by calling @code{data_section} or one of the
5904 alternatives for other sections. @var{align} is the constant alignment
5907 The default version of this function takes care of putting read-only
5908 variables in @code{readonly_data_section}.
5911 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5912 Build up a unique section name, expressed as a @code{STRING_CST} node,
5913 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5914 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5915 the initial value of @var{exp} requires link-time relocations.
5917 The default version of this function appends the symbol name to the
5918 ELF section name that would normally be used for the symbol. For
5919 example, the function @code{foo} would be placed in @code{.text.foo}.
5920 Whatever the actual target object format, this is often good enough.
5923 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
5924 Switches to the appropriate section for output of constant pool entry
5925 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
5926 constant in RTL@. The argument @var{mode} is redundant except in the
5927 case of a @code{const_int} rtx. Select the section by calling
5928 @code{readonly_data_section} or one of the alternatives for other
5929 sections. @var{align} is the constant alignment in bits.
5931 The default version of this function takes care of putting symbolic
5932 constants in @code{flag_pic} mode in @code{data_section} and everything
5933 else in @code{readonly_data_section}.
5936 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
5937 Define this hook if references to a symbol or a constant must be
5938 treated differently depending on something about the variable or
5939 function named by the symbol (such as what section it is in).
5941 The hook is executed immediately after rtl has been created for
5942 @var{decl}, which may be a variable or function declaration or
5943 an entry in the constant pool. In either case, @var{rtl} is the
5944 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
5945 in this hook; that field may not have been initialized yet.
5947 In the case of a constant, it is safe to assume that the rtl is
5948 a @code{mem} whose address is a @code{symbol_ref}. Most decls
5949 will also have this form, but that is not guaranteed. Global
5950 register variables, for instance, will have a @code{reg} for their
5951 rtl. (Normally the right thing to do with such unusual rtl is
5954 The @var{new_decl_p} argument will be true if this is the first time
5955 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
5956 be false for subsequent invocations, which will happen for duplicate
5957 declarations. Whether or not anything must be done for the duplicate
5958 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
5959 @var{new_decl_p} is always true when the hook is called for a constant.
5961 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
5962 The usual thing for this hook to do is to record flags in the
5963 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
5964 Historically, the name string was modified if it was necessary to
5965 encode more than one bit of information, but this practice is now
5966 discouraged; use @code{SYMBOL_REF_FLAGS}.
5968 The default definition of this hook, @code{default_encode_section_info}
5969 in @file{varasm.c}, sets a number of commonly-useful bits in
5970 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
5971 before overriding it.
5974 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
5975 Decode @var{name} and return the real name part, sans
5976 the characters that @code{TARGET_ENCODE_SECTION_INFO}
5980 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
5981 Returns true if @var{exp} should be placed into a ``small data'' section.
5982 The default version of this hook always returns false.
5985 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
5986 Contains the value true if the target places read-only
5987 ``small data'' into a separate section. The default value is false.
5990 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
5991 Returns true if @var{exp} names an object for which name resolution
5992 rules must resolve to the current ``module'' (dynamic shared library
5993 or executable image).
5995 The default version of this hook implements the name resolution rules
5996 for ELF, which has a looser model of global name binding than other
5997 currently supported object file formats.
6000 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6001 Contains the value true if the target supports thread-local storage.
6002 The default value is false.
6007 @section Position Independent Code
6008 @cindex position independent code
6011 This section describes macros that help implement generation of position
6012 independent code. Simply defining these macros is not enough to
6013 generate valid PIC; you must also add support to the macros
6014 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6015 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6016 @samp{movsi} to do something appropriate when the source operand
6017 contains a symbolic address. You may also need to alter the handling of
6018 switch statements so that they use relative addresses.
6019 @c i rearranged the order of the macros above to try to force one of
6020 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6022 @defmac PIC_OFFSET_TABLE_REGNUM
6023 The register number of the register used to address a table of static
6024 data addresses in memory. In some cases this register is defined by a
6025 processor's ``application binary interface'' (ABI)@. When this macro
6026 is defined, RTL is generated for this register once, as with the stack
6027 pointer and frame pointer registers. If this macro is not defined, it
6028 is up to the machine-dependent files to allocate such a register (if
6029 necessary). Note that this register must be fixed when in use (e.g.@:
6030 when @code{flag_pic} is true).
6033 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6034 Define this macro if the register defined by
6035 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6036 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6039 @defmac FINALIZE_PIC
6040 By generating position-independent code, when two different programs (A
6041 and B) share a common library (libC.a), the text of the library can be
6042 shared whether or not the library is linked at the same address for both
6043 programs. In some of these environments, position-independent code
6044 requires not only the use of different addressing modes, but also
6045 special code to enable the use of these addressing modes.
6047 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
6048 codes once the function is being compiled into assembly code, but not
6049 before. (It is not done before, because in the case of compiling an
6050 inline function, it would lead to multiple PIC prologues being
6051 included in functions which used inline functions and were compiled to
6055 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6056 A C expression that is nonzero if @var{x} is a legitimate immediate
6057 operand on the target machine when generating position independent code.
6058 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6059 check this. You can also assume @var{flag_pic} is true, so you need not
6060 check it either. You need not define this macro if all constants
6061 (including @code{SYMBOL_REF}) can be immediate operands when generating
6062 position independent code.
6065 @node Assembler Format
6066 @section Defining the Output Assembler Language
6068 This section describes macros whose principal purpose is to describe how
6069 to write instructions in assembler language---rather than what the
6073 * File Framework:: Structural information for the assembler file.
6074 * Data Output:: Output of constants (numbers, strings, addresses).
6075 * Uninitialized Data:: Output of uninitialized variables.
6076 * Label Output:: Output and generation of labels.
6077 * Initialization:: General principles of initialization
6078 and termination routines.
6079 * Macros for Initialization::
6080 Specific macros that control the handling of
6081 initialization and termination routines.
6082 * Instruction Output:: Output of actual instructions.
6083 * Dispatch Tables:: Output of jump tables.
6084 * Exception Region Output:: Output of exception region code.
6085 * Alignment Output:: Pseudo ops for alignment and skipping data.
6088 @node File Framework
6089 @subsection The Overall Framework of an Assembler File
6090 @cindex assembler format
6091 @cindex output of assembler code
6093 @c prevent bad page break with this line
6094 This describes the overall framework of an assembly file.
6096 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6097 @findex default_file_start
6098 Output to @code{asm_out_file} any text which the assembler expects to
6099 find at the beginning of a file. The default behavior is controlled
6100 by two flags, documented below. Unless your target's assembler is
6101 quite unusual, if you override the default, you should call
6102 @code{default_file_start} at some point in your target hook. This
6103 lets other target files rely on these variables.
6106 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6107 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6108 printed as the very first line in the assembly file, unless
6109 @option{-fverbose-asm} is in effect. (If that macro has been defined
6110 to the empty string, this variable has no effect.) With the normal
6111 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6112 assembler that it need not bother stripping comments or extra
6113 whitespace from its input. This allows it to work a bit faster.
6115 The default is false. You should not set it to true unless you have
6116 verified that your port does not generate any extra whitespace or
6117 comments that will cause GAS to issue errors in NO_APP mode.
6120 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6121 If this flag is true, @code{output_file_directive} will be called
6122 for the primary source file, immediately after printing
6123 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6124 this to be done. The default is false.
6127 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6128 Output to @code{asm_out_file} any text which the assembler expects
6129 to find at the end of a file. The default is to output nothing.
6132 @deftypefun void file_end_indicate_exec_stack ()
6133 Some systems use a common convention, the @samp{.note.GNU-stack}
6134 special section, to indicate whether or not an object file relies on
6135 the stack being executable. If your system uses this convention, you
6136 should define @code{TARGET_ASM_FILE_END} to this function. If you
6137 need to do other things in that hook, have your hook function call
6141 @defmac ASM_COMMENT_START
6142 A C string constant describing how to begin a comment in the target
6143 assembler language. The compiler assumes that the comment will end at
6144 the end of the line.
6148 A C string constant for text to be output before each @code{asm}
6149 statement or group of consecutive ones. Normally this is
6150 @code{"#APP"}, which is a comment that has no effect on most
6151 assemblers but tells the GNU assembler that it must check the lines
6152 that follow for all valid assembler constructs.
6156 A C string constant for text to be output after each @code{asm}
6157 statement or group of consecutive ones. Normally this is
6158 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6159 time-saving assumptions that are valid for ordinary compiler output.
6162 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6163 A C statement to output COFF information or DWARF debugging information
6164 which indicates that filename @var{name} is the current source file to
6165 the stdio stream @var{stream}.
6167 This macro need not be defined if the standard form of output
6168 for the file format in use is appropriate.
6171 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6172 A C statement to output the string @var{string} to the stdio stream
6173 @var{stream}. If you do not call the function @code{output_quoted_string}
6174 in your config files, GCC will only call it to output filenames to
6175 the assembler source. So you can use it to canonicalize the format
6176 of the filename using this macro.
6179 @defmac ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
6180 A C statement to output DBX or SDB debugging information before code
6181 for line number @var{line} of the current source file to the
6182 stdio stream @var{stream}. @var{counter} is the number of time the
6183 macro was invoked, including the current invocation; it is intended
6184 to generate unique labels in the assembly output.
6186 This macro need not be defined if the standard form of debugging
6187 information for the debugger in use is appropriate.
6190 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6191 A C statement to output something to the assembler file to handle a
6192 @samp{#ident} directive containing the text @var{string}. If this
6193 macro is not defined, nothing is output for a @samp{#ident} directive.
6196 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6197 Output assembly directives to switch to section @var{name}. The section
6198 should have attributes as specified by @var{flags}, which is a bit mask
6199 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6200 is nonzero, it contains an alignment in bytes to be used for the section,
6201 otherwise some target default should be used. Only targets that must
6202 specify an alignment within the section directive need pay attention to
6203 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6206 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6207 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6210 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6211 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6212 based on a variable or function decl, a section name, and whether or not the
6213 declaration's initializer may contain runtime relocations. @var{decl} may be
6214 null, in which case read-write data should be assumed.
6216 The default version if this function handles choosing code vs data,
6217 read-only vs read-write data, and @code{flag_pic}. You should only
6218 need to override this if your target has special flags that might be
6219 set via @code{__attribute__}.
6224 @subsection Output of Data
6227 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6228 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6229 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6230 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6231 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6232 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6233 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6234 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6235 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6236 These hooks specify assembly directives for creating certain kinds
6237 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6238 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6239 aligned two-byte object, and so on. Any of the hooks may be
6240 @code{NULL}, indicating that no suitable directive is available.
6242 The compiler will print these strings at the start of a new line,
6243 followed immediately by the object's initial value. In most cases,
6244 the string should contain a tab, a pseudo-op, and then another tab.
6247 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6248 The @code{assemble_integer} function uses this hook to output an
6249 integer object. @var{x} is the object's value, @var{size} is its size
6250 in bytes and @var{aligned_p} indicates whether it is aligned. The
6251 function should return @code{true} if it was able to output the
6252 object. If it returns false, @code{assemble_integer} will try to
6253 split the object into smaller parts.
6255 The default implementation of this hook will use the
6256 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6257 when the relevant string is @code{NULL}.
6260 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6261 A C statement to recognize @var{rtx} patterns that
6262 @code{output_addr_const} can't deal with, and output assembly code to
6263 @var{stream} corresponding to the pattern @var{x}. This may be used to
6264 allow machine-dependent @code{UNSPEC}s to appear within constants.
6266 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6267 @code{goto fail}, so that a standard error message is printed. If it
6268 prints an error message itself, by calling, for example,
6269 @code{output_operand_lossage}, it may just complete normally.
6272 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6273 A C statement to output to the stdio stream @var{stream} an assembler
6274 instruction to assemble a string constant containing the @var{len}
6275 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6276 @code{char *} and @var{len} a C expression of type @code{int}.
6278 If the assembler has a @code{.ascii} pseudo-op as found in the
6279 Berkeley Unix assembler, do not define the macro
6280 @code{ASM_OUTPUT_ASCII}.
6283 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6284 A C statement to output word @var{n} of a function descriptor for
6285 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6286 is defined, and is otherwise unused.
6289 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6290 You may define this macro as a C expression. You should define the
6291 expression to have a nonzero value if GCC should output the constant
6292 pool for a function before the code for the function, or a zero value if
6293 GCC should output the constant pool after the function. If you do
6294 not define this macro, the usual case, GCC will output the constant
6295 pool before the function.
6298 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6299 A C statement to output assembler commands to define the start of the
6300 constant pool for a function. @var{funname} is a string giving
6301 the name of the function. Should the return type of the function
6302 be required, it can be obtained via @var{fundecl}. @var{size}
6303 is the size, in bytes, of the constant pool that will be written
6304 immediately after this call.
6306 If no constant-pool prefix is required, the usual case, this macro need
6310 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6311 A C statement (with or without semicolon) to output a constant in the
6312 constant pool, if it needs special treatment. (This macro need not do
6313 anything for RTL expressions that can be output normally.)
6315 The argument @var{file} is the standard I/O stream to output the
6316 assembler code on. @var{x} is the RTL expression for the constant to
6317 output, and @var{mode} is the machine mode (in case @var{x} is a
6318 @samp{const_int}). @var{align} is the required alignment for the value
6319 @var{x}; you should output an assembler directive to force this much
6322 The argument @var{labelno} is a number to use in an internal label for
6323 the address of this pool entry. The definition of this macro is
6324 responsible for outputting the label definition at the proper place.
6325 Here is how to do this:
6328 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6331 When you output a pool entry specially, you should end with a
6332 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6333 entry from being output a second time in the usual manner.
6335 You need not define this macro if it would do nothing.
6338 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6339 A C statement to output assembler commands to at the end of the constant
6340 pool for a function. @var{funname} is a string giving the name of the
6341 function. Should the return type of the function be required, you can
6342 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6343 constant pool that GCC wrote immediately before this call.
6345 If no constant-pool epilogue is required, the usual case, you need not
6349 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6350 Define this macro as a C expression which is nonzero if @var{C} is
6351 used as a logical line separator by the assembler.
6353 If you do not define this macro, the default is that only
6354 the character @samp{;} is treated as a logical line separator.
6357 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6358 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6359 These target hooks are C string constants, describing the syntax in the
6360 assembler for grouping arithmetic expressions. If not overridden, they
6361 default to normal parentheses, which is correct for most assemblers.
6364 These macros are provided by @file{real.h} for writing the definitions
6365 of @code{ASM_OUTPUT_DOUBLE} and the like:
6367 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6368 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6369 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6370 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6371 floating point representation, and store its bit pattern in the variable
6372 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6373 be a simple @code{long int}. For the others, it should be an array of
6374 @code{long int}. The number of elements in this array is determined by
6375 the size of the desired target floating point data type: 32 bits of it
6376 go in each @code{long int} array element. Each array element holds 32
6377 bits of the result, even if @code{long int} is wider than 32 bits on the
6380 The array element values are designed so that you can print them out
6381 using @code{fprintf} in the order they should appear in the target
6385 @node Uninitialized Data
6386 @subsection Output of Uninitialized Variables
6388 Each of the macros in this section is used to do the whole job of
6389 outputting a single uninitialized variable.
6391 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6392 A C statement (sans semicolon) to output to the stdio stream
6393 @var{stream} the assembler definition of a common-label named
6394 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6395 is the size rounded up to whatever alignment the caller wants.
6397 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6398 output the name itself; before and after that, output the additional
6399 assembler syntax for defining the name, and a newline.
6401 This macro controls how the assembler definitions of uninitialized
6402 common global variables are output.
6405 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6406 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6407 separate, explicit argument. If you define this macro, it is used in
6408 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6409 handling the required alignment of the variable. The alignment is specified
6410 as the number of bits.
6413 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6414 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6415 variable to be output, if there is one, or @code{NULL_TREE} if there
6416 is no corresponding variable. If you define this macro, GCC will use it
6417 in place of both @code{ASM_OUTPUT_COMMON} and
6418 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6419 the variable's decl in order to chose what to output.
6422 @defmac ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6423 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6424 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6428 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6429 A C statement (sans semicolon) to output to the stdio stream
6430 @var{stream} the assembler definition of uninitialized global @var{decl} named
6431 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6432 is the size rounded up to whatever alignment the caller wants.
6434 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6435 defining this macro. If unable, use the expression
6436 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6437 before and after that, output the additional assembler syntax for defining
6438 the name, and a newline.
6440 This macro controls how the assembler definitions of uninitialized global
6441 variables are output. This macro exists to properly support languages like
6442 C++ which do not have @code{common} data. However, this macro currently
6443 is not defined for all targets. If this macro and
6444 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6445 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6446 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6449 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6450 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6451 separate, explicit argument. If you define this macro, it is used in
6452 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6453 handling the required alignment of the variable. The alignment is specified
6454 as the number of bits.
6456 Try to use function @code{asm_output_aligned_bss} defined in file
6457 @file{varasm.c} when defining this macro.
6460 @defmac ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6461 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6462 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6466 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6467 A C statement (sans semicolon) to output to the stdio stream
6468 @var{stream} the assembler definition of a local-common-label named
6469 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6470 is the size rounded up to whatever alignment the caller wants.
6472 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6473 output the name itself; before and after that, output the additional
6474 assembler syntax for defining the name, and a newline.
6476 This macro controls how the assembler definitions of uninitialized
6477 static variables are output.
6480 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6481 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6482 separate, explicit argument. If you define this macro, it is used in
6483 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6484 handling the required alignment of the variable. The alignment is specified
6485 as the number of bits.
6488 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6489 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6490 variable to be output, if there is one, or @code{NULL_TREE} if there
6491 is no corresponding variable. If you define this macro, GCC will use it
6492 in place of both @code{ASM_OUTPUT_DECL} and
6493 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6494 the variable's decl in order to chose what to output.
6497 @defmac ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6498 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6499 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6504 @subsection Output and Generation of Labels
6506 @c prevent bad page break with this line
6507 This is about outputting labels.
6509 @findex assemble_name
6510 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6511 A C statement (sans semicolon) to output to the stdio stream
6512 @var{stream} the assembler definition of a label named @var{name}.
6513 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6514 output the name itself; before and after that, output the additional
6515 assembler syntax for defining the name, and a newline. A default
6516 definition of this macro is provided which is correct for most systems.
6520 A C string containing the appropriate assembler directive to specify the
6521 size of a symbol, without any arguments. On systems that use ELF, the
6522 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6523 systems, the default is not to define this macro.
6525 Define this macro only if it is correct to use the default definitions
6526 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6527 for your system. If you need your own custom definitions of those
6528 macros, or if you do not need explicit symbol sizes at all, do not
6532 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6533 A C statement (sans semicolon) to output to the stdio stream
6534 @var{stream} a directive telling the assembler that the size of the
6535 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6536 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6540 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6541 A C statement (sans semicolon) to output to the stdio stream
6542 @var{stream} a directive telling the assembler to calculate the size of
6543 the symbol @var{name} by subtracting its address from the current
6546 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6547 provided. The default assumes that the assembler recognizes a special
6548 @samp{.} symbol as referring to the current address, and can calculate
6549 the difference between this and another symbol. If your assembler does
6550 not recognize @samp{.} or cannot do calculations with it, you will need
6551 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6555 A C string containing the appropriate assembler directive to specify the
6556 type of a symbol, without any arguments. On systems that use ELF, the
6557 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6558 systems, the default is not to define this macro.
6560 Define this macro only if it is correct to use the default definition of
6561 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6562 custom definition of this macro, or if you do not need explicit symbol
6563 types at all, do not define this macro.
6566 @defmac TYPE_OPERAND_FMT
6567 A C string which specifies (using @code{printf} syntax) the format of
6568 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6569 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6570 the default is not to define this macro.
6572 Define this macro only if it is correct to use the default definition of
6573 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6574 custom definition of this macro, or if you do not need explicit symbol
6575 types at all, do not define this macro.
6578 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6579 A C statement (sans semicolon) to output to the stdio stream
6580 @var{stream} a directive telling the assembler that the type of the
6581 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6582 that string is always either @samp{"function"} or @samp{"object"}, but
6583 you should not count on this.
6585 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6586 definition of this macro is provided.
6589 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6590 A C statement (sans semicolon) to output to the stdio stream
6591 @var{stream} any text necessary for declaring the name @var{name} of a
6592 function which is being defined. This macro is responsible for
6593 outputting the label definition (perhaps using
6594 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6595 @code{FUNCTION_DECL} tree node representing the function.
6597 If this macro is not defined, then the function name is defined in the
6598 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6600 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6604 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6605 A C statement (sans semicolon) to output to the stdio stream
6606 @var{stream} any text necessary for declaring the size of a function
6607 which is being defined. The argument @var{name} is the name of the
6608 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6609 representing the function.
6611 If this macro is not defined, then the function size is not defined.
6613 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6617 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6618 A C statement (sans semicolon) to output to the stdio stream
6619 @var{stream} any text necessary for declaring the name @var{name} of an
6620 initialized variable which is being defined. This macro must output the
6621 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6622 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6624 If this macro is not defined, then the variable name is defined in the
6625 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6627 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6628 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6631 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
6632 A C statement (sans semicolon) to output to the stdio stream
6633 @var{stream} any text necessary for declaring the name @var{name} of a
6634 constant which is being defined. This macro is responsible for
6635 outputting the label definition (perhaps using
6636 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
6637 value of the constant, and @var{size} is the size of the constant
6638 in bytes. @var{name} will be an internal label.
6640 If this macro is not defined, then the @var{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} in the definition
6647 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6648 A C statement (sans semicolon) to output to the stdio stream
6649 @var{stream} any text necessary for claiming a register @var{regno}
6650 for a global variable @var{decl} with name @var{name}.
6652 If you don't define this macro, that is equivalent to defining it to do
6656 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6657 A C statement (sans semicolon) to finish up declaring a variable name
6658 once the compiler has processed its initializer fully and thus has had a
6659 chance to determine the size of an array when controlled by an
6660 initializer. This is used on systems where it's necessary to declare
6661 something about the size of the object.
6663 If you don't define this macro, that is equivalent to defining it to do
6666 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6667 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6670 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6671 This target hook is a function to output to the stdio stream
6672 @var{stream} some commands that will make the label @var{name} global;
6673 that is, available for reference from other files.
6675 The default implementation relies on a proper definition of
6676 @code{GLOBAL_ASM_OP}.
6679 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6680 A C statement (sans semicolon) to output to the stdio stream
6681 @var{stream} some commands that will make the label @var{name} weak;
6682 that is, available for reference from other files but only used if
6683 no other definition is available. Use the expression
6684 @code{assemble_name (@var{stream}, @var{name})} to output the name
6685 itself; before and after that, output the additional assembler syntax
6686 for making that name weak, and a newline.
6688 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6689 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6693 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6694 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6695 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6696 or variable decl. If @var{value} is not @code{NULL}, this C statement
6697 should output to the stdio stream @var{stream} assembler code which
6698 defines (equates) the weak symbol @var{name} to have the value
6699 @var{value}. If @var{value} is @code{NULL}, it should output commands
6700 to make @var{name} weak.
6703 @defmac SUPPORTS_WEAK
6704 A C expression which evaluates to true if the target supports weak symbols.
6706 If you don't define this macro, @file{defaults.h} provides a default
6707 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6708 is defined, the default definition is @samp{1}; otherwise, it is
6709 @samp{0}. Define this macro if you want to control weak symbol support
6710 with a compiler flag such as @option{-melf}.
6713 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
6714 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6715 public symbol such that extra copies in multiple translation units will
6716 be discarded by the linker. Define this macro if your object file
6717 format provides support for this concept, such as the @samp{COMDAT}
6718 section flags in the Microsoft Windows PE/COFF format, and this support
6719 requires changes to @var{decl}, such as putting it in a separate section.
6722 @defmac SUPPORTS_ONE_ONLY
6723 A C expression which evaluates to true if the target supports one-only
6726 If you don't define this macro, @file{varasm.c} provides a default
6727 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6728 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6729 you want to control one-only symbol support with a compiler flag, or if
6730 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6731 be emitted as one-only.
6734 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6735 This target hook is a function to output to @var{asm_out_file} some
6736 commands that will make the symbol(s) associated with @var{decl} have
6737 hidden, protected or internal visibility as specified by @var{visibility}.
6740 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6741 A C statement (sans semicolon) to output to the stdio stream
6742 @var{stream} any text necessary for declaring the name of an external
6743 symbol named @var{name} which is referenced in this compilation but
6744 not defined. The value of @var{decl} is the tree node for the
6747 This macro need not be defined if it does not need to output anything.
6748 The GNU assembler and most Unix assemblers don't require anything.
6751 @defmac ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
6752 A C statement (sans semicolon) to output on @var{stream} an assembler
6753 pseudo-op to declare a library function name external. The name of the
6754 library function is given by @var{symref}, which has type @code{rtx} and
6755 is a @code{symbol_ref}.
6757 This macro need not be defined if it does not need to output anything.
6758 The GNU assembler and most Unix assemblers don't require anything.
6761 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6762 A C statement (sans semicolon) to output to the stdio stream
6763 @var{stream} a reference in assembler syntax to a label named
6764 @var{name}. This should add @samp{_} to the front of the name, if that
6765 is customary on your operating system, as it is in most Berkeley Unix
6766 systems. This macro is used in @code{assemble_name}.
6769 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6770 A C statement (sans semicolon) to output a reference to
6771 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6772 will be used to output the name of the symbol. This macro may be used
6773 to modify the way a symbol is referenced depending on information
6774 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6777 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6778 A C statement (sans semicolon) to output a reference to @var{buf}, the
6779 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6780 @code{assemble_name} will be used to output the name of the symbol.
6781 This macro is not used by @code{output_asm_label}, or the @code{%l}
6782 specifier that calls it; the intention is that this macro should be set
6783 when it is necessary to output a label differently when its address is
6787 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6788 A function to output to the stdio stream @var{stream} a label whose
6789 name is made from the string @var{prefix} and the number @var{labelno}.
6791 It is absolutely essential that these labels be distinct from the labels
6792 used for user-level functions and variables. Otherwise, certain programs
6793 will have name conflicts with internal labels.
6795 It is desirable to exclude internal labels from the symbol table of the
6796 object file. Most assemblers have a naming convention for labels that
6797 should be excluded; on many systems, the letter @samp{L} at the
6798 beginning of a label has this effect. You should find out what
6799 convention your system uses, and follow it.
6801 The default version of this function utilizes ASM_GENERATE_INTERNAL_LABEL.
6804 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6805 A C statement to output to the stdio stream @var{stream} a debug info
6806 label whose name is made from the string @var{prefix} and the number
6807 @var{num}. This is useful for VLIW targets, where debug info labels
6808 may need to be treated differently than branch target labels. On some
6809 systems, branch target labels must be at the beginning of instruction
6810 bundles, but debug info labels can occur in the middle of instruction
6813 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
6817 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6818 A C statement to store into the string @var{string} a label whose name
6819 is made from the string @var{prefix} and the number @var{num}.
6821 This string, when output subsequently by @code{assemble_name}, should
6822 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
6823 with the same @var{prefix} and @var{num}.
6825 If the string begins with @samp{*}, then @code{assemble_name} will
6826 output the rest of the string unchanged. It is often convenient for
6827 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6828 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6829 to output the string, and may change it. (Of course,
6830 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6831 you should know what it does on your machine.)
6834 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6835 A C expression to assign to @var{outvar} (which is a variable of type
6836 @code{char *}) a newly allocated string made from the string
6837 @var{name} and the number @var{number}, with some suitable punctuation
6838 added. Use @code{alloca} to get space for the string.
6840 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6841 produce an assembler label for an internal static variable whose name is
6842 @var{name}. Therefore, the string must be such as to result in valid
6843 assembler code. The argument @var{number} is different each time this
6844 macro is executed; it prevents conflicts between similarly-named
6845 internal static variables in different scopes.
6847 Ideally this string should not be a valid C identifier, to prevent any
6848 conflict with the user's own symbols. Most assemblers allow periods
6849 or percent signs in assembler symbols; putting at least one of these
6850 between the name and the number will suffice.
6852 If this macro is not defined, a default definition will be provided
6853 which is correct for most systems.
6856 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6857 A C statement to output to the stdio stream @var{stream} assembler code
6858 which defines (equates) the symbol @var{name} to have the value @var{value}.
6861 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6862 correct for most systems.
6865 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6866 A C statement to output to the stdio stream @var{stream} assembler code
6867 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6868 to have the value of the tree node @var{decl_of_value}. This macro will
6869 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6870 the tree nodes are available.
6873 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6874 correct for most systems.
6877 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6878 A C statement to output to the stdio stream @var{stream} assembler code
6879 which defines (equates) the weak symbol @var{name} to have the value
6880 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6881 an undefined weak symbol.
6883 Define this macro if the target only supports weak aliases; define
6884 @code{ASM_OUTPUT_DEF} instead if possible.
6887 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6888 Define this macro to override the default assembler names used for
6889 Objective-C methods.
6891 The default name is a unique method number followed by the name of the
6892 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6893 the category is also included in the assembler name (e.g.@:
6896 These names are safe on most systems, but make debugging difficult since
6897 the method's selector is not present in the name. Therefore, particular
6898 systems define other ways of computing names.
6900 @var{buf} is an expression of type @code{char *} which gives you a
6901 buffer in which to store the name; its length is as long as
6902 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6903 50 characters extra.
6905 The argument @var{is_inst} specifies whether the method is an instance
6906 method or a class method; @var{class_name} is the name of the class;
6907 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6908 in a category); and @var{sel_name} is the name of the selector.
6910 On systems where the assembler can handle quoted names, you can use this
6911 macro to provide more human-readable names.
6914 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
6915 A C statement (sans semicolon) to output to the stdio stream
6916 @var{stream} commands to declare that the label @var{name} is an
6917 Objective-C class reference. This is only needed for targets whose
6918 linkers have special support for NeXT-style runtimes.
6921 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6922 A C statement (sans semicolon) to output to the stdio stream
6923 @var{stream} commands to declare that the label @var{name} is an
6924 unresolved Objective-C class reference. This is only needed for targets
6925 whose linkers have special support for NeXT-style runtimes.
6928 @node Initialization
6929 @subsection How Initialization Functions Are Handled
6930 @cindex initialization routines
6931 @cindex termination routines
6932 @cindex constructors, output of
6933 @cindex destructors, output of
6935 The compiled code for certain languages includes @dfn{constructors}
6936 (also called @dfn{initialization routines})---functions to initialize
6937 data in the program when the program is started. These functions need
6938 to be called before the program is ``started''---that is to say, before
6939 @code{main} is called.
6941 Compiling some languages generates @dfn{destructors} (also called
6942 @dfn{termination routines}) that should be called when the program
6945 To make the initialization and termination functions work, the compiler
6946 must output something in the assembler code to cause those functions to
6947 be called at the appropriate time. When you port the compiler to a new
6948 system, you need to specify how to do this.
6950 There are two major ways that GCC currently supports the execution of
6951 initialization and termination functions. Each way has two variants.
6952 Much of the structure is common to all four variations.
6954 @findex __CTOR_LIST__
6955 @findex __DTOR_LIST__
6956 The linker must build two lists of these functions---a list of
6957 initialization functions, called @code{__CTOR_LIST__}, and a list of
6958 termination functions, called @code{__DTOR_LIST__}.
6960 Each list always begins with an ignored function pointer (which may hold
6961 0, @minus{}1, or a count of the function pointers after it, depending on
6962 the environment). This is followed by a series of zero or more function
6963 pointers to constructors (or destructors), followed by a function
6964 pointer containing zero.
6966 Depending on the operating system and its executable file format, either
6967 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6968 time and exit time. Constructors are called in reverse order of the
6969 list; destructors in forward order.
6971 The best way to handle static constructors works only for object file
6972 formats which provide arbitrarily-named sections. A section is set
6973 aside for a list of constructors, and another for a list of destructors.
6974 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6975 object file that defines an initialization function also puts a word in
6976 the constructor section to point to that function. The linker
6977 accumulates all these words into one contiguous @samp{.ctors} section.
6978 Termination functions are handled similarly.
6980 This method will be chosen as the default by @file{target-def.h} if
6981 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
6982 support arbitrary sections, but does support special designated
6983 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
6984 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
6986 When arbitrary sections are available, there are two variants, depending
6987 upon how the code in @file{crtstuff.c} is called. On systems that
6988 support a @dfn{.init} section which is executed at program startup,
6989 parts of @file{crtstuff.c} are compiled into that section. The
6990 program is linked by the @command{gcc} driver like this:
6993 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
6996 The prologue of a function (@code{__init}) appears in the @code{.init}
6997 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
6998 for the function @code{__fini} in the @dfn{.fini} section. Normally these
6999 files are provided by the operating system or by the GNU C library, but
7000 are provided by GCC for a few targets.
7002 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7003 compiled from @file{crtstuff.c}. They contain, among other things, code
7004 fragments within the @code{.init} and @code{.fini} sections that branch
7005 to routines in the @code{.text} section. The linker will pull all parts
7006 of a section together, which results in a complete @code{__init} function
7007 that invokes the routines we need at startup.
7009 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7012 If no init section is available, when GCC compiles any function called
7013 @code{main} (or more accurately, any function designated as a program
7014 entry point by the language front end calling @code{expand_main_function}),
7015 it inserts a procedure call to @code{__main} as the first executable code
7016 after the function prologue. The @code{__main} function is defined
7017 in @file{libgcc2.c} and runs the global constructors.
7019 In file formats that don't support arbitrary sections, there are again
7020 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7021 and an `a.out' format must be used. In this case,
7022 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7023 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7024 and with the address of the void function containing the initialization
7025 code as its value. The GNU linker recognizes this as a request to add
7026 the value to a @dfn{set}; the values are accumulated, and are eventually
7027 placed in the executable as a vector in the format described above, with
7028 a leading (ignored) count and a trailing zero element.
7029 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7030 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7031 the compilation of @code{main} to call @code{__main} as above, starting
7032 the initialization process.
7034 The last variant uses neither arbitrary sections nor the GNU linker.
7035 This is preferable when you want to do dynamic linking and when using
7036 file formats which the GNU linker does not support, such as `ECOFF'@. In
7037 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7038 termination functions are recognized simply by their names. This requires
7039 an extra program in the linkage step, called @command{collect2}. This program
7040 pretends to be the linker, for use with GCC; it does its job by running
7041 the ordinary linker, but also arranges to include the vectors of
7042 initialization and termination functions. These functions are called
7043 via @code{__main} as described above. In order to use this method,
7044 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7047 The following section describes the specific macros that control and
7048 customize the handling of initialization and termination functions.
7051 @node Macros for Initialization
7052 @subsection Macros Controlling Initialization Routines
7054 Here are the macros that control how the compiler handles initialization
7055 and termination functions:
7057 @defmac INIT_SECTION_ASM_OP
7058 If defined, a C string constant, including spacing, for the assembler
7059 operation to identify the following data as initialization code. If not
7060 defined, GCC will assume such a section does not exist. When you are
7061 using special sections for initialization and termination functions, this
7062 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7063 run the initialization functions.
7066 @defmac HAS_INIT_SECTION
7067 If defined, @code{main} will not call @code{__main} as described above.
7068 This macro should be defined for systems that control start-up code
7069 on a symbol-by-symbol basis, such as OSF/1, and should not
7070 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7073 @defmac LD_INIT_SWITCH
7074 If defined, a C string constant for a switch that tells the linker that
7075 the following symbol is an initialization routine.
7078 @defmac LD_FINI_SWITCH
7079 If defined, a C string constant for a switch that tells the linker that
7080 the following symbol is a finalization routine.
7083 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7084 If defined, a C statement that will write a function that can be
7085 automatically called when a shared library is loaded. The function
7086 should call @var{func}, which takes no arguments. If not defined, and
7087 the object format requires an explicit initialization function, then a
7088 function called @code{_GLOBAL__DI} will be generated.
7090 This function and the following one are used by collect2 when linking a
7091 shared library that needs constructors or destructors, or has DWARF2
7092 exception tables embedded in the code.
7095 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7096 If defined, a C statement that will write a function that can be
7097 automatically called when a shared library is unloaded. The function
7098 should call @var{func}, which takes no arguments. If not defined, and
7099 the object format requires an explicit finalization function, then a
7100 function called @code{_GLOBAL__DD} will be generated.
7103 @defmac INVOKE__main
7104 If defined, @code{main} will call @code{__main} despite the presence of
7105 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7106 where the init section is not actually run automatically, but is still
7107 useful for collecting the lists of constructors and destructors.
7110 @defmac SUPPORTS_INIT_PRIORITY
7111 If nonzero, the C++ @code{init_priority} attribute is supported and the
7112 compiler should emit instructions to control the order of initialization
7113 of objects. If zero, the compiler will issue an error message upon
7114 encountering an @code{init_priority} attribute.
7117 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7118 This value is true if the target supports some ``native'' method of
7119 collecting constructors and destructors to be run at startup and exit.
7120 It is false if we must use @command{collect2}.
7123 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7124 If defined, a function that outputs assembler code to arrange to call
7125 the function referenced by @var{symbol} at initialization time.
7127 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7128 no arguments and with no return value. If the target supports initialization
7129 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7130 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7132 If this macro is not defined by the target, a suitable default will
7133 be chosen if (1) the target supports arbitrary section names, (2) the
7134 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7138 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7139 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7140 functions rather than initialization functions.
7143 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7144 generated for the generated object file will have static linkage.
7146 If your system uses @command{collect2} as the means of processing
7147 constructors, then that program normally uses @command{nm} to scan
7148 an object file for constructor functions to be called.
7150 On certain kinds of systems, you can define this macro to make
7151 @command{collect2} work faster (and, in some cases, make it work at all):
7153 @defmac OBJECT_FORMAT_COFF
7154 Define this macro if the system uses COFF (Common Object File Format)
7155 object files, so that @command{collect2} can assume this format and scan
7156 object files directly for dynamic constructor/destructor functions.
7158 This macro is effective only in a native compiler; @command{collect2} as
7159 part of a cross compiler always uses @command{nm} for the target machine.
7162 @defmac REAL_NM_FILE_NAME
7163 Define this macro as a C string constant containing the file name to use
7164 to execute @command{nm}. The default is to search the path normally for
7167 If your system supports shared libraries and has a program to list the
7168 dynamic dependencies of a given library or executable, you can define
7169 these macros to enable support for running initialization and
7170 termination functions in shared libraries:
7174 Define this macro to a C string constant containing the name of the program
7175 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7178 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7179 Define this macro to be C code that extracts filenames from the output
7180 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7181 of type @code{char *} that points to the beginning of a line of output
7182 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7183 code must advance @var{ptr} to the beginning of the filename on that
7184 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7187 @node Instruction Output
7188 @subsection Output of Assembler Instructions
7190 @c prevent bad page break with this line
7191 This describes assembler instruction output.
7193 @defmac REGISTER_NAMES
7194 A C initializer containing the assembler's names for the machine
7195 registers, each one as a C string constant. This is what translates
7196 register numbers in the compiler into assembler language.
7199 @defmac ADDITIONAL_REGISTER_NAMES
7200 If defined, a C initializer for an array of structures containing a name
7201 and a register number. This macro defines additional names for hard
7202 registers, thus allowing the @code{asm} option in declarations to refer
7203 to registers using alternate names.
7206 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7207 Define this macro if you are using an unusual assembler that
7208 requires different names for the machine instructions.
7210 The definition is a C statement or statements which output an
7211 assembler instruction opcode to the stdio stream @var{stream}. The
7212 macro-operand @var{ptr} is a variable of type @code{char *} which
7213 points to the opcode name in its ``internal'' form---the form that is
7214 written in the machine description. The definition should output the
7215 opcode name to @var{stream}, performing any translation you desire, and
7216 increment the variable @var{ptr} to point at the end of the opcode
7217 so that it will not be output twice.
7219 In fact, your macro definition may process less than the entire opcode
7220 name, or more than the opcode name; but if you want to process text
7221 that includes @samp{%}-sequences to substitute operands, you must take
7222 care of the substitution yourself. Just be sure to increment
7223 @var{ptr} over whatever text should not be output normally.
7225 @findex recog_data.operand
7226 If you need to look at the operand values, they can be found as the
7227 elements of @code{recog_data.operand}.
7229 If the macro definition does nothing, the instruction is output
7233 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7234 If defined, a C statement to be executed just prior to the output of
7235 assembler code for @var{insn}, to modify the extracted operands so
7236 they will be output differently.
7238 Here the argument @var{opvec} is the vector containing the operands
7239 extracted from @var{insn}, and @var{noperands} is the number of
7240 elements of the vector which contain meaningful data for this insn.
7241 The contents of this vector are what will be used to convert the insn
7242 template into assembler code, so you can change the assembler output
7243 by changing the contents of the vector.
7245 This macro is useful when various assembler syntaxes share a single
7246 file of instruction patterns; by defining this macro differently, you
7247 can cause a large class of instructions to be output differently (such
7248 as with rearranged operands). Naturally, variations in assembler
7249 syntax affecting individual insn patterns ought to be handled by
7250 writing conditional output routines in those patterns.
7252 If this macro is not defined, it is equivalent to a null statement.
7255 @defmac FINAL_PRESCAN_LABEL
7256 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
7257 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
7258 @var{noperands} will be zero.
7261 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7262 A C compound statement to output to stdio stream @var{stream} the
7263 assembler syntax for an instruction operand @var{x}. @var{x} is an
7266 @var{code} is a value that can be used to specify one of several ways
7267 of printing the operand. It is used when identical operands must be
7268 printed differently depending on the context. @var{code} comes from
7269 the @samp{%} specification that was used to request printing of the
7270 operand. If the specification was just @samp{%@var{digit}} then
7271 @var{code} is 0; if the specification was @samp{%@var{ltr}
7272 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7275 If @var{x} is a register, this macro should print the register's name.
7276 The names can be found in an array @code{reg_names} whose type is
7277 @code{char *[]}. @code{reg_names} is initialized from
7278 @code{REGISTER_NAMES}.
7280 When the machine description has a specification @samp{%@var{punct}}
7281 (a @samp{%} followed by a punctuation character), this macro is called
7282 with a null pointer for @var{x} and the punctuation character for
7286 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7287 A C expression which evaluates to true if @var{code} is a valid
7288 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7289 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7290 punctuation characters (except for the standard one, @samp{%}) are used
7294 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7295 A C compound statement to output to stdio stream @var{stream} the
7296 assembler syntax for an instruction operand that is a memory reference
7297 whose address is @var{x}. @var{x} is an RTL expression.
7299 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7300 On some machines, the syntax for a symbolic address depends on the
7301 section that the address refers to. On these machines, define the hook
7302 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7303 @code{symbol_ref}, and then check for it here. @xref{Assembler
7307 @findex dbr_sequence_length
7308 @defmac DBR_OUTPUT_SEQEND (@var{file})
7309 A C statement, to be executed after all slot-filler instructions have
7310 been output. If necessary, call @code{dbr_sequence_length} to
7311 determine the number of slots filled in a sequence (zero if not
7312 currently outputting a sequence), to decide how many no-ops to output,
7315 Don't define this macro if it has nothing to do, but it is helpful in
7316 reading assembly output if the extent of the delay sequence is made
7317 explicit (e.g.@: with white space).
7320 @findex final_sequence
7321 Note that output routines for instructions with delay slots must be
7322 prepared to deal with not being output as part of a sequence
7323 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7324 found.) The variable @code{final_sequence} is null when not
7325 processing a sequence, otherwise it contains the @code{sequence} rtx
7329 @defmac REGISTER_PREFIX
7330 @defmacx LOCAL_LABEL_PREFIX
7331 @defmacx USER_LABEL_PREFIX
7332 @defmacx IMMEDIATE_PREFIX
7333 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7334 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7335 @file{final.c}). These are useful when a single @file{md} file must
7336 support multiple assembler formats. In that case, the various @file{tm.h}
7337 files can define these macros differently.
7340 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
7341 If defined this macro should expand to a series of @code{case}
7342 statements which will be parsed inside the @code{switch} statement of
7343 the @code{asm_fprintf} function. This allows targets to define extra
7344 printf formats which may useful when generating their assembler
7345 statements. Note that uppercase letters are reserved for future
7346 generic extensions to asm_fprintf, and so are not available to target
7347 specific code. The output file is given by the parameter @var{file}.
7348 The varargs input pointer is @var{argptr} and the rest of the format
7349 string, starting the character after the one that is being switched
7350 upon, is pointed to by @var{format}.
7353 @defmac ASSEMBLER_DIALECT
7354 If your target supports multiple dialects of assembler language (such as
7355 different opcodes), define this macro as a C expression that gives the
7356 numeric index of the assembler language dialect to use, with zero as the
7359 If this macro is defined, you may use constructs of the form
7361 @samp{@{option0|option1|option2@dots{}@}}
7364 in the output templates of patterns (@pxref{Output Template}) or in the
7365 first argument of @code{asm_fprintf}. This construct outputs
7366 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7367 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7368 within these strings retain their usual meaning. If there are fewer
7369 alternatives within the braces than the value of
7370 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7372 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7373 @samp{@}} do not have any special meaning when used in templates or
7374 operands to @code{asm_fprintf}.
7376 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7377 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7378 the variations in assembler language syntax with that mechanism. Define
7379 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7380 if the syntax variant are larger and involve such things as different
7381 opcodes or operand order.
7384 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7385 A C expression to output to @var{stream} some assembler code
7386 which will push hard register number @var{regno} onto the stack.
7387 The code need not be optimal, since this macro is used only when
7391 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7392 A C expression to output to @var{stream} some assembler code
7393 which will pop hard register number @var{regno} off of the stack.
7394 The code need not be optimal, since this macro is used only when
7398 @node Dispatch Tables
7399 @subsection Output of Dispatch Tables
7401 @c prevent bad page break with this line
7402 This concerns dispatch tables.
7404 @cindex dispatch table
7405 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7406 A C statement to output to the stdio stream @var{stream} an assembler
7407 pseudo-instruction to generate a difference between two labels.
7408 @var{value} and @var{rel} are the numbers of two internal labels. The
7409 definitions of these labels are output using
7410 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7411 way here. For example,
7414 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7415 @var{value}, @var{rel})
7418 You must provide this macro on machines where the addresses in a
7419 dispatch table are relative to the table's own address. If defined, GCC
7420 will also use this macro on all machines when producing PIC@.
7421 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7422 mode and flags can be read.
7425 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7426 This macro should be provided on machines where the addresses
7427 in a dispatch table are absolute.
7429 The definition should be a C statement to output to the stdio stream
7430 @var{stream} an assembler pseudo-instruction to generate a reference to
7431 a label. @var{value} is the number of an internal label whose
7432 definition is output using @code{(*targetm.asm_out.internal_label)}.
7436 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7440 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7441 Define this if the label before a jump-table needs to be output
7442 specially. The first three arguments are the same as for
7443 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7444 jump-table which follows (a @code{jump_insn} containing an
7445 @code{addr_vec} or @code{addr_diff_vec}).
7447 This feature is used on system V to output a @code{swbeg} statement
7450 If this macro is not defined, these labels are output with
7451 @code{(*targetm.asm_out.internal_label)}.
7454 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7455 Define this if something special must be output at the end of a
7456 jump-table. The definition should be a C statement to be executed
7457 after the assembler code for the table is written. It should write
7458 the appropriate code to stdio stream @var{stream}. The argument
7459 @var{table} is the jump-table insn, and @var{num} is the label-number
7460 of the preceding label.
7462 If this macro is not defined, nothing special is output at the end of
7466 @node Exception Region Output
7467 @subsection Assembler Commands for Exception Regions
7469 @c prevent bad page break with this line
7471 This describes commands marking the start and the end of an exception
7474 @defmac EH_FRAME_SECTION_NAME
7475 If defined, a C string constant for the name of the section containing
7476 exception handling frame unwind information. If not defined, GCC will
7477 provide a default definition if the target supports named sections.
7478 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7480 You should define this symbol if your target supports DWARF 2 frame
7481 unwind information and the default definition does not work.
7484 @defmac EH_FRAME_IN_DATA_SECTION
7485 If defined, DWARF 2 frame unwind information will be placed in the
7486 data section even though the target supports named sections. This
7487 might be necessary, for instance, if the system linker does garbage
7488 collection and sections cannot be marked as not to be collected.
7490 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7494 @defmac MASK_RETURN_ADDR
7495 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7496 that it does not contain any extraneous set bits in it.
7499 @defmac DWARF2_UNWIND_INFO
7500 Define this macro to 0 if your target supports DWARF 2 frame unwind
7501 information, but it does not yet work with exception handling.
7502 Otherwise, if your target supports this information (if it defines
7503 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7504 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7507 If this macro is defined to 1, the DWARF 2 unwinder will be the default
7508 exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by
7511 If this macro is defined to anything, the DWARF 2 unwinder will be used
7512 instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case.
7515 @defmac DWARF_CIE_DATA_ALIGNMENT
7516 This macro need only be defined if the target might save registers in the
7517 function prologue at an offset to the stack pointer that is not aligned to
7518 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7519 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7520 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7521 the target supports DWARF 2 frame unwind information.
7524 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7525 If defined, a function that switches to the section in which the main
7526 exception table is to be placed (@pxref{Sections}). The default is a
7527 function that switches to a section named @code{.gcc_except_table} on
7528 machines that support named sections via
7529 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7530 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7531 @code{readonly_data_section}.
7534 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7535 If defined, a function that switches to the section in which the DWARF 2
7536 frame unwind information to be placed (@pxref{Sections}). The default
7537 is a function that outputs a standard GAS section directive, if
7538 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7539 directive followed by a synthetic label.
7542 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7543 Contains the value true if the target should add a zero word onto the
7544 end of a Dwarf-2 frame info section when used for exception handling.
7545 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7549 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
7550 Given a register, this hook should return a parallel of registers to
7551 represent where to find the register pieces. Define this hook if the
7552 register and its mode are represented in Dwarf in non-contiguous
7553 locations, or if the register should be represented in more than one
7554 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
7555 If not defined, the default is to return @code{NULL_RTX}.
7558 @node Alignment Output
7559 @subsection Assembler Commands for Alignment
7561 @c prevent bad page break with this line
7562 This describes commands for alignment.
7564 @defmac JUMP_ALIGN (@var{label})
7565 The alignment (log base 2) to put in front of @var{label}, which is
7566 a common destination of jumps and has no fallthru incoming edge.
7568 This macro need not be defined if you don't want any special alignment
7569 to be done at such a time. Most machine descriptions do not currently
7572 Unless it's necessary to inspect the @var{label} parameter, it is better
7573 to set the variable @var{align_jumps} in the target's
7574 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7575 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7578 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
7579 The alignment (log base 2) to put in front of @var{label}, which follows
7582 This macro need not be defined if you don't want any special alignment
7583 to be done at such a time. Most machine descriptions do not currently
7587 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7588 The maximum number of bytes to skip when applying
7589 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7590 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7593 @defmac LOOP_ALIGN (@var{label})
7594 The alignment (log base 2) to put in front of @var{label}, which follows
7595 a @code{NOTE_INSN_LOOP_BEG} note.
7597 This macro need not be defined if you don't want any special alignment
7598 to be done at such a time. Most machine descriptions do not currently
7601 Unless it's necessary to inspect the @var{label} parameter, it is better
7602 to set the variable @code{align_loops} in the target's
7603 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7604 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7607 @defmac LOOP_ALIGN_MAX_SKIP
7608 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7609 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7612 @defmac LABEL_ALIGN (@var{label})
7613 The alignment (log base 2) to put in front of @var{label}.
7614 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7615 the maximum of the specified values is used.
7617 Unless it's necessary to inspect the @var{label} parameter, it is better
7618 to set the variable @code{align_labels} in the target's
7619 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7620 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7623 @defmac LABEL_ALIGN_MAX_SKIP
7624 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7625 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7628 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7629 A C statement to output to the stdio stream @var{stream} an assembler
7630 instruction to advance the location counter by @var{nbytes} bytes.
7631 Those bytes should be zero when loaded. @var{nbytes} will be a C
7632 expression of type @code{int}.
7635 @defmac ASM_NO_SKIP_IN_TEXT
7636 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7637 text section because it fails to put zeros in the bytes that are skipped.
7638 This is true on many Unix systems, where the pseudo--op to skip bytes
7639 produces no-op instructions rather than zeros when used in the text
7643 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7644 A C statement to output to the stdio stream @var{stream} an assembler
7645 command to advance the location counter to a multiple of 2 to the
7646 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7649 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7650 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7651 for padding, if necessary.
7654 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7655 A C statement to output to the stdio stream @var{stream} an assembler
7656 command to advance the location counter to a multiple of 2 to the
7657 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7658 satisfy the alignment request. @var{power} and @var{max_skip} will be
7659 a C expression of type @code{int}.
7663 @node Debugging Info
7664 @section Controlling Debugging Information Format
7666 @c prevent bad page break with this line
7667 This describes how to specify debugging information.
7670 * All Debuggers:: Macros that affect all debugging formats uniformly.
7671 * DBX Options:: Macros enabling specific options in DBX format.
7672 * DBX Hooks:: Hook macros for varying DBX format.
7673 * File Names and DBX:: Macros controlling output of file names in DBX format.
7674 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7675 * VMS Debug:: Macros for VMS debug format.
7679 @subsection Macros Affecting All Debugging Formats
7681 @c prevent bad page break with this line
7682 These macros affect all debugging formats.
7684 @defmac DBX_REGISTER_NUMBER (@var{regno})
7685 A C expression that returns the DBX register number for the compiler
7686 register number @var{regno}. In the default macro provided, the value
7687 of this expression will be @var{regno} itself. But sometimes there are
7688 some registers that the compiler knows about and DBX does not, or vice
7689 versa. In such cases, some register may need to have one number in the
7690 compiler and another for DBX@.
7692 If two registers have consecutive numbers inside GCC, and they can be
7693 used as a pair to hold a multiword value, then they @emph{must} have
7694 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7695 Otherwise, debuggers will be unable to access such a pair, because they
7696 expect register pairs to be consecutive in their own numbering scheme.
7698 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7699 does not preserve register pairs, then what you must do instead is
7700 redefine the actual register numbering scheme.
7703 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
7704 A C expression that returns the integer offset value for an automatic
7705 variable having address @var{x} (an RTL expression). The default
7706 computation assumes that @var{x} is based on the frame-pointer and
7707 gives the offset from the frame-pointer. This is required for targets
7708 that produce debugging output for DBX or COFF-style debugging output
7709 for SDB and allow the frame-pointer to be eliminated when the
7710 @option{-g} options is used.
7713 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7714 A C expression that returns the integer offset value for an argument
7715 having address @var{x} (an RTL expression). The nominal offset is
7719 @defmac PREFERRED_DEBUGGING_TYPE
7720 A C expression that returns the type of debugging output GCC should
7721 produce when the user specifies just @option{-g}. Define
7722 this if you have arranged for GCC to support more than one format of
7723 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7724 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7725 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7727 When the user specifies @option{-ggdb}, GCC normally also uses the
7728 value of this macro to select the debugging output format, but with two
7729 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7730 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7731 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7732 defined, GCC uses @code{DBX_DEBUG}.
7734 The value of this macro only affects the default debugging output; the
7735 user can always get a specific type of output by using @option{-gstabs},
7736 @option{-gcoff}, @option{-gdwarf-1}, @option{-gdwarf-2}, @option{-gxcoff},
7741 @subsection Specific Options for DBX Output
7743 @c prevent bad page break with this line
7744 These are specific options for DBX output.
7746 @defmac DBX_DEBUGGING_INFO
7747 Define this macro if GCC should produce debugging output for DBX
7748 in response to the @option{-g} option.
7751 @defmac XCOFF_DEBUGGING_INFO
7752 Define this macro if GCC should produce XCOFF format debugging output
7753 in response to the @option{-g} option. This is a variant of DBX format.
7756 @defmac DEFAULT_GDB_EXTENSIONS
7757 Define this macro to control whether GCC should by default generate
7758 GDB's extended version of DBX debugging information (assuming DBX-format
7759 debugging information is enabled at all). If you don't define the
7760 macro, the default is 1: always generate the extended information
7761 if there is any occasion to.
7764 @defmac DEBUG_SYMS_TEXT
7765 Define this macro if all @code{.stabs} commands should be output while
7766 in the text section.
7769 @defmac ASM_STABS_OP
7770 A C string constant, including spacing, naming the assembler pseudo op to
7771 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7772 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7773 applies only to DBX debugging information format.
7776 @defmac ASM_STABD_OP
7777 A C string constant, including spacing, naming the assembler pseudo op to
7778 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7779 value is the current location. If you don't define this macro,
7780 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7784 @defmac ASM_STABN_OP
7785 A C string constant, including spacing, naming the assembler pseudo op to
7786 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7787 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7788 macro applies only to DBX debugging information format.
7791 @defmac DBX_NO_XREFS
7792 Define this macro if DBX on your system does not support the construct
7793 @samp{xs@var{tagname}}. On some systems, this construct is used to
7794 describe a forward reference to a structure named @var{tagname}.
7795 On other systems, this construct is not supported at all.
7798 @defmac DBX_CONTIN_LENGTH
7799 A symbol name in DBX-format debugging information is normally
7800 continued (split into two separate @code{.stabs} directives) when it
7801 exceeds a certain length (by default, 80 characters). On some
7802 operating systems, DBX requires this splitting; on others, splitting
7803 must not be done. You can inhibit splitting by defining this macro
7804 with the value zero. You can override the default splitting-length by
7805 defining this macro as an expression for the length you desire.
7808 @defmac DBX_CONTIN_CHAR
7809 Normally continuation is indicated by adding a @samp{\} character to
7810 the end of a @code{.stabs} string when a continuation follows. To use
7811 a different character instead, define this macro as a character
7812 constant for the character you want to use. Do not define this macro
7813 if backslash is correct for your system.
7816 @defmac DBX_STATIC_STAB_DATA_SECTION
7817 Define this macro if it is necessary to go to the data section before
7818 outputting the @samp{.stabs} pseudo-op for a non-global static
7822 @defmac DBX_TYPE_DECL_STABS_CODE
7823 The value to use in the ``code'' field of the @code{.stabs} directive
7824 for a typedef. The default is @code{N_LSYM}.
7827 @defmac DBX_STATIC_CONST_VAR_CODE
7828 The value to use in the ``code'' field of the @code{.stabs} directive
7829 for a static variable located in the text section. DBX format does not
7830 provide any ``right'' way to do this. The default is @code{N_FUN}.
7833 @defmac DBX_REGPARM_STABS_CODE
7834 The value to use in the ``code'' field of the @code{.stabs} directive
7835 for a parameter passed in registers. DBX format does not provide any
7836 ``right'' way to do this. The default is @code{N_RSYM}.
7839 @defmac DBX_REGPARM_STABS_LETTER
7840 The letter to use in DBX symbol data to identify a symbol as a parameter
7841 passed in registers. DBX format does not customarily provide any way to
7842 do this. The default is @code{'P'}.
7845 @defmac DBX_MEMPARM_STABS_LETTER
7846 The letter to use in DBX symbol data to identify a symbol as a stack
7847 parameter. The default is @code{'p'}.
7850 @defmac DBX_FUNCTION_FIRST
7851 Define this macro if the DBX information for a function and its
7852 arguments should precede the assembler code for the function. Normally,
7853 in DBX format, the debugging information entirely follows the assembler
7857 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
7858 Define this macro if the value of a symbol describing the scope of a
7859 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7860 of the enclosing function. Normally, GCC uses an absolute address.
7863 @defmac DBX_USE_BINCL
7864 Define this macro if GCC should generate @code{N_BINCL} and
7865 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7866 macro also directs GCC to output a type number as a pair of a file
7867 number and a type number within the file. Normally, GCC does not
7868 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7869 number for a type number.
7873 @subsection Open-Ended Hooks for DBX Format
7875 @c prevent bad page break with this line
7876 These are hooks for DBX format.
7878 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7879 Define this macro to say how to output to @var{stream} the debugging
7880 information for the start of a scope level for variable names. The
7881 argument @var{name} is the name of an assembler symbol (for use with
7882 @code{assemble_name}) whose value is the address where the scope begins.
7885 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7886 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7889 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
7890 Define this macro if the target machine requires special handling to
7891 output an @code{N_FUN} entry for the function @var{decl}.
7894 @defmac DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7895 Define this macro if the target machine requires special output at the
7896 end of the debugging information for a function. The definition should
7897 be a C statement (sans semicolon) to output the appropriate information
7898 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7902 @defmac DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7903 Define this macro if you need to control the order of output of the
7904 standard data types at the beginning of compilation. The argument
7905 @var{syms} is a @code{tree} which is a chain of all the predefined
7906 global symbols, including names of data types.
7908 Normally, DBX output starts with definitions of the types for integers
7909 and characters, followed by all the other predefined types of the
7910 particular language in no particular order.
7912 On some machines, it is necessary to output different particular types
7913 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
7914 those symbols in the necessary order. Any predefined types that you
7915 don't explicitly output will be output afterward in no particular order.
7917 Be careful not to define this macro so that it works only for C@. There
7918 are no global variables to access most of the built-in types, because
7919 another language may have another set of types. The way to output a
7920 particular type is to look through @var{syms} to see if you can find it.
7926 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7927 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
7929 dbxout_symbol (decl);
7935 This does nothing if the expected type does not exist.
7937 See the function @code{init_decl_processing} in @file{c-decl.c} to find
7938 the names to use for all the built-in C types.
7940 Here is another way of finding a particular type:
7942 @c this is still overfull. --mew 10feb93
7946 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7947 if (TREE_CODE (decl) == TYPE_DECL
7948 && (TREE_CODE (TREE_TYPE (decl))
7950 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
7951 && TYPE_UNSIGNED (TREE_TYPE (decl)))
7953 /* @r{This must be @code{unsigned short}.} */
7954 dbxout_symbol (decl);
7961 @defmac NO_DBX_FUNCTION_END
7962 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7963 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
7964 On those machines, define this macro to turn this feature off without
7965 disturbing the rest of the gdb extensions.
7968 @node File Names and DBX
7969 @subsection File Names in DBX Format
7971 @c prevent bad page break with this line
7972 This describes file names in DBX format.
7974 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7975 A C statement to output DBX debugging information to the stdio stream
7976 @var{stream} which indicates that file @var{name} is the main source
7977 file---the file specified as the input file for compilation.
7978 This macro is called only once, at the beginning of compilation.
7980 This macro need not be defined if the standard form of output
7981 for DBX debugging information is appropriate.
7984 @defmac DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
7985 A C statement to output DBX debugging information to the stdio stream
7986 @var{stream} which indicates that the current directory during
7987 compilation is named @var{name}.
7989 This macro need not be defined if the standard form of output
7990 for DBX debugging information is appropriate.
7993 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7994 A C statement to output DBX debugging information at the end of
7995 compilation of the main source file @var{name}.
7997 If you don't define this macro, nothing special is output at the end
7998 of compilation, which is correct for most machines.
8003 @subsection Macros for SDB and DWARF Output
8005 @c prevent bad page break with this line
8006 Here are macros for SDB and DWARF output.
8008 @defmac SDB_DEBUGGING_INFO
8009 Define this macro if GCC should produce COFF-style debugging output
8010 for SDB in response to the @option{-g} option.
8013 @defmac DWARF_DEBUGGING_INFO
8014 Define this macro if GCC should produce dwarf format debugging output
8015 in response to the @option{-g} option.
8018 @defmac DWARF2_DEBUGGING_INFO
8019 Define this macro if GCC should produce dwarf version 2 format
8020 debugging output in response to the @option{-g} option.
8022 To support optional call frame debugging information, you must also
8023 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8024 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8025 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8026 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8029 @defmac DWARF2_FRAME_INFO
8030 Define this macro to a nonzero value if GCC should always output
8031 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8032 (@pxref{Exception Region Output} is nonzero, GCC will output this
8033 information not matter how you define @code{DWARF2_FRAME_INFO}.
8036 @defmac LINKER_DOES_NOT_WORK_WITH_DWARF2
8037 Define this macro if the linker does not work with Dwarf version 2.
8038 Normally, if the user specifies only @option{-ggdb} GCC will use Dwarf
8039 version 2 if available; this macro disables this. See the description
8040 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
8043 @defmac DWARF2_GENERATE_TEXT_SECTION_LABEL
8044 By default, the Dwarf 2 debugging information generator will generate a
8045 label to mark the beginning of the text section. If it is better simply
8046 to use the name of the text section itself, rather than an explicit label,
8047 to indicate the beginning of the text section, define this macro to zero.
8050 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8051 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8052 line debug info sections. This will result in much more compact line number
8053 tables, and hence is desirable if it works.
8056 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8057 A C statement to issue assembly directives that create a difference
8058 between the two given labels, using an integer of the given size.
8061 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8062 A C statement to issue assembly directives that create a
8063 section-relative reference to the given label, using an integer of the
8067 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8068 A C statement to issue assembly directives that create a self-relative
8069 reference to the given label, using an integer of the given size.
8072 @defmac PUT_SDB_@dots{}
8073 Define these macros to override the assembler syntax for the special
8074 SDB assembler directives. See @file{sdbout.c} for a list of these
8075 macros and their arguments. If the standard syntax is used, you need
8076 not define them yourself.
8080 Some assemblers do not support a semicolon as a delimiter, even between
8081 SDB assembler directives. In that case, define this macro to be the
8082 delimiter to use (usually @samp{\n}). It is not necessary to define
8083 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8087 @defmac SDB_GENERATE_FAKE
8088 Define this macro to override the usual method of constructing a dummy
8089 name for anonymous structure and union types. See @file{sdbout.c} for
8093 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8094 Define this macro to allow references to unknown structure,
8095 union, or enumeration tags to be emitted. Standard COFF does not
8096 allow handling of unknown references, MIPS ECOFF has support for
8100 @defmac SDB_ALLOW_FORWARD_REFERENCES
8101 Define this macro to allow references to structure, union, or
8102 enumeration tags that have not yet been seen to be handled. Some
8103 assemblers choke if forward tags are used, while some require it.
8108 @subsection Macros for VMS Debug Format
8110 @c prevent bad page break with this line
8111 Here are macros for VMS debug format.
8113 @defmac VMS_DEBUGGING_INFO
8114 Define this macro if GCC should produce debugging output for VMS
8115 in response to the @option{-g} option. The default behavior for VMS
8116 is to generate minimal debug info for a traceback in the absence of
8117 @option{-g} unless explicitly overridden with @option{-g0}. This
8118 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8119 @code{OVERRIDE_OPTIONS}.
8122 @node Floating Point
8123 @section Cross Compilation and Floating Point
8124 @cindex cross compilation and floating point
8125 @cindex floating point and cross compilation
8127 While all modern machines use twos-complement representation for integers,
8128 there are a variety of representations for floating point numbers. This
8129 means that in a cross-compiler the representation of floating point numbers
8130 in the compiled program may be different from that used in the machine
8131 doing the compilation.
8133 Because different representation systems may offer different amounts of
8134 range and precision, all floating point constants must be represented in
8135 the target machine's format. Therefore, the cross compiler cannot
8136 safely use the host machine's floating point arithmetic; it must emulate
8137 the target's arithmetic. To ensure consistency, GCC always uses
8138 emulation to work with floating point values, even when the host and
8139 target floating point formats are identical.
8141 The following macros are provided by @file{real.h} for the compiler to
8142 use. All parts of the compiler which generate or optimize
8143 floating-point calculations must use these macros. They may evaluate
8144 their operands more than once, so operands must not have side effects.
8146 @defmac REAL_VALUE_TYPE
8147 The C data type to be used to hold a floating point value in the target
8148 machine's format. Typically this is a @code{struct} containing an
8149 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8153 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8154 Compares for equality the two values, @var{x} and @var{y}. If the target
8155 floating point format supports negative zeroes and/or NaNs,
8156 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8157 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8160 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8161 Tests whether @var{x} is less than @var{y}.
8164 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8165 Truncates @var{x} to a signed integer, rounding toward zero.
8168 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8169 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8170 @var{x} is negative, returns zero.
8173 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8174 Converts @var{string} into a floating point number in the target machine's
8175 representation for mode @var{mode}. This routine can handle both
8176 decimal and hexadecimal floating point constants, using the syntax
8177 defined by the C language for both.
8180 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8181 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8184 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8185 Determines whether @var{x} represents infinity (positive or negative).
8188 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8189 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8192 @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})
8193 Calculates an arithmetic operation on the two floating point values
8194 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8197 The operation to be performed is specified by @var{code}. Only the
8198 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8199 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8201 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8202 target's floating point format cannot represent infinity, it will call
8203 @code{abort}. Callers should check for this situation first, using
8204 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8207 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8208 Returns the negative of the floating point value @var{x}.
8211 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8212 Returns the absolute value of @var{x}.
8215 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8216 Truncates the floating point value @var{x} to fit in @var{mode}. The
8217 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8218 appropriate bit pattern to be output asa floating constant whose
8219 precision accords with mode @var{mode}.
8222 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8223 Converts a floating point value @var{x} into a double-precision integer
8224 which is then stored into @var{low} and @var{high}. If the value is not
8225 integral, it is truncated.
8228 @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})
8229 Converts a double-precision integer found in @var{low} and @var{high},
8230 into a floating point value which is then stored into @var{x}. The
8231 value is truncated to fit in mode @var{mode}.
8234 @node Mode Switching
8235 @section Mode Switching Instructions
8236 @cindex mode switching
8237 The following macros control mode switching optimizations:
8239 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8240 Define this macro if the port needs extra instructions inserted for mode
8241 switching in an optimizing compilation.
8243 For an example, the SH4 can perform both single and double precision
8244 floating point operations, but to perform a single precision operation,
8245 the FPSCR PR bit has to be cleared, while for a double precision
8246 operation, this bit has to be set. Changing the PR bit requires a general
8247 purpose register as a scratch register, hence these FPSCR sets have to
8248 be inserted before reload, i.e.@: you can't put this into instruction emitting
8249 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
8251 You can have multiple entities that are mode-switched, and select at run time
8252 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8253 return nonzero for any @var{entity} that needs mode-switching.
8254 If you define this macro, you also have to define
8255 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8256 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8257 @code{NORMAL_MODE} is optional.
8260 @defmac NUM_MODES_FOR_MODE_SWITCHING
8261 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8262 initializer for an array of integers. Each initializer element
8263 N refers to an entity that needs mode switching, and specifies the number
8264 of different modes that might need to be set for this entity.
8265 The position of the initializer in the initializer - starting counting at
8266 zero - determines the integer that is used to refer to the mode-switched
8268 In macros that take mode arguments / yield a mode result, modes are
8269 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8270 switch is needed / supplied.
8273 @defmac MODE_NEEDED (@var{entity}, @var{insn})
8274 @var{entity} is an integer specifying a mode-switched entity. If
8275 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8276 return an integer value not larger than the corresponding element in
8277 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8278 be switched into prior to the execution of @var{insn}.
8281 @defmac NORMAL_MODE (@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 entry and exit.
8287 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8288 This macro specifies the order in which modes for @var{entity} are processed.
8289 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8290 lowest. The value of the macro should be an integer designating a mode
8291 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8292 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8293 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8296 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8297 Generate one or more insns to set @var{entity} to @var{mode}.
8298 @var{hard_reg_live} is the set of hard registers live at the point where
8299 the insn(s) are to be inserted.
8302 @node Target Attributes
8303 @section Defining target-specific uses of @code{__attribute__}
8304 @cindex target attributes
8305 @cindex machine attributes
8306 @cindex attributes, target-specific
8308 Target-specific attributes may be defined for functions, data and types.
8309 These are described using the following target hooks; they also need to
8310 be documented in @file{extend.texi}.
8312 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8313 If defined, this target hook points to an array of @samp{struct
8314 attribute_spec} (defined in @file{tree.h}) specifying the machine
8315 specific attributes for this target and some of the restrictions on the
8316 entities to which these attributes are applied and the arguments they
8320 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8321 If defined, this target hook is a function which returns zero if the attributes on
8322 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8323 and two if they are nearly compatible (which causes a warning to be
8324 generated). If this is not defined, machine-specific attributes are
8325 supposed always to be compatible.
8328 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8329 If defined, this target hook is a function which assigns default attributes to
8330 newly defined @var{type}.
8333 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8334 Define this target hook if the merging of type attributes needs special
8335 handling. If defined, the result is a list of the combined
8336 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8337 that @code{comptypes} has already been called and returned 1. This
8338 function may call @code{merge_attributes} to handle machine-independent
8342 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8343 Define this target hook if the merging of decl attributes needs special
8344 handling. If defined, the result is a list of the combined
8345 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8346 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8347 when this is needed are when one attribute overrides another, or when an
8348 attribute is nullified by a subsequent definition. This function may
8349 call @code{merge_attributes} to handle machine-independent merging.
8351 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8352 If the only target-specific handling you require is @samp{dllimport} for
8353 Windows targets, you should define the macro
8354 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function
8355 called @code{merge_dllimport_decl_attributes} which can then be defined
8356 as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done
8357 in @file{i386/cygwin.h} and @file{i386/i386.c}, for example.
8360 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8361 Define this target hook if you want to be able to add attributes to a decl
8362 when it is being created. This is normally useful for back ends which
8363 wish to implement a pragma by using the attributes which correspond to
8364 the pragma's effect. The @var{node} argument is the decl which is being
8365 created. The @var{attr_ptr} argument is a pointer to the attribute list
8366 for this decl. The list itself should not be modified, since it may be
8367 shared with other decls, but attributes may be chained on the head of
8368 the list and @code{*@var{attr_ptr}} modified to point to the new
8369 attributes, or a copy of the list may be made if further changes are
8373 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8375 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8376 into the current function, despite its having target-specific
8377 attributes, @code{false} otherwise. By default, if a function has a
8378 target specific attribute attached to it, it will not be inlined.
8381 @node MIPS Coprocessors
8382 @section Defining coprocessor specifics for MIPS targets.
8383 @cindex MIPS coprocessor-definition macros
8385 The MIPS specification allows MIPS implementations to have as many as 4
8386 coprocessors, each with as many as 32 private registers. gcc supports
8387 accessing these registers and transferring values between the registers
8388 and memory using asm-ized variables. For example:
8391 register unsigned int cp0count asm ("c0r1");
8397 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8398 names may be added as described below, or the default names may be
8399 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8401 Coprocessor registers are assumed to be epilogue-used; sets to them will
8402 be preserved even if it does not appear that the register is used again
8403 later in the function.
8405 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8406 the FPU. One accesses COP1 registers through standard mips
8407 floating-point support; they are not included in this mechanism.
8409 There is one macro used in defining the MIPS coprocessor interface which
8410 you may want to override in subtargets; it is described below.
8412 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
8413 A comma-separated list (with leading comma) of pairs describing the
8414 alternate names of coprocessor registers. The format of each entry should be
8416 @{ @var{alternatename}, @var{register_number}@}
8422 @section Miscellaneous Parameters
8423 @cindex parameters, miscellaneous
8425 @c prevent bad page break with this line
8426 Here are several miscellaneous parameters.
8428 @defmac PREDICATE_CODES
8429 Define this if you have defined special-purpose predicates in the file
8430 @file{@var{machine}.c}. This macro is called within an initializer of an
8431 array of structures. The first field in the structure is the name of a
8432 predicate and the second field is an array of rtl codes. For each
8433 predicate, list all rtl codes that can be in expressions matched by the
8434 predicate. The list should have a trailing comma. Here is an example
8435 of two entries in the list for a typical RISC machine:
8438 #define PREDICATE_CODES \
8439 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8440 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8443 Defining this macro does not affect the generated code (however,
8444 incorrect definitions that omit an rtl code that may be matched by the
8445 predicate can cause the compiler to malfunction). Instead, it allows
8446 the table built by @file{genrecog} to be more compact and efficient,
8447 thus speeding up the compiler. The most important predicates to include
8448 in the list specified by this macro are those used in the most insn
8451 For each predicate function named in @code{PREDICATE_CODES}, a
8452 declaration will be generated in @file{insn-codes.h}.
8455 @defmac SPECIAL_MODE_PREDICATES
8456 Define this if you have special predicates that know special things
8457 about modes. Genrecog will warn about certain forms of
8458 @code{match_operand} without a mode; if the operand predicate is
8459 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8462 Here is an example from the IA-32 port (@code{ext_register_operand}
8463 specially checks for @code{HImode} or @code{SImode} in preparation
8464 for a byte extraction from @code{%ah} etc.).
8467 #define SPECIAL_MODE_PREDICATES \
8468 "ext_register_operand",
8472 @defmac CASE_VECTOR_MODE
8473 An alias for a machine mode name. This is the machine mode that
8474 elements of a jump-table should have.
8477 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8478 Optional: return the preferred mode for an @code{addr_diff_vec}
8479 when the minimum and maximum offset are known. If you define this,
8480 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8481 To make this work, you also have to define @code{INSN_ALIGN} and
8482 make the alignment for @code{addr_diff_vec} explicit.
8483 The @var{body} argument is provided so that the offset_unsigned and scale
8484 flags can be updated.
8487 @defmac CASE_VECTOR_PC_RELATIVE
8488 Define this macro to be a C expression to indicate when jump-tables
8489 should contain relative addresses. If jump-tables never contain
8490 relative addresses, then you need not define this macro.
8493 @defmac CASE_DROPS_THROUGH
8494 Define this if control falls through a @code{case} insn when the index
8495 value is out of range. This means the specified default-label is
8496 actually ignored by the @code{case} insn proper.
8499 @defmac CASE_VALUES_THRESHOLD
8500 Define this to be the smallest number of different values for which it
8501 is best to use a jump-table instead of a tree of conditional branches.
8502 The default is four for machines with a @code{casesi} instruction and
8503 five otherwise. This is best for most machines.
8506 @defmac CASE_USE_BIT_TESTS
8507 Define this macro to be a C expression to indicate whether C switch
8508 statements may be implemented by a sequence of bit tests. This is
8509 advantageous on processors that can efficiently implement left shift
8510 of 1 by the number of bits held in a register, but inappropriate on
8511 targets that would require a loop. By default, this macro returns
8512 @code{true} if the target defines an @code{ashlsi3} pattern, and
8513 @code{false} otherwise.
8516 @defmac WORD_REGISTER_OPERATIONS
8517 Define this macro if operations between registers with integral mode
8518 smaller than a word are always performed on the entire register.
8519 Most RISC machines have this property and most CISC machines do not.
8522 @defmac LOAD_EXTEND_OP (@var{mode})
8523 Define this macro to be a C expression indicating when insns that read
8524 memory in @var{mode}, an integral mode narrower than a word, set the
8525 bits outside of @var{mode} to be either the sign-extension or the
8526 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8527 of @var{mode} for which the
8528 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8529 @code{NIL} for other modes.
8531 This macro is not called with @var{mode} non-integral or with a width
8532 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8533 value in this case. Do not define this macro if it would always return
8534 @code{NIL}. On machines where this macro is defined, you will normally
8535 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8538 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
8539 Define this macro if loading short immediate values into registers sign
8543 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
8544 Define this macro if the same instructions that convert a floating
8545 point number to a signed fixed point number also convert validly to an
8550 The maximum number of bytes that a single instruction can move quickly
8551 between memory and registers or between two memory locations.
8554 @defmac MAX_MOVE_MAX
8555 The maximum number of bytes that a single instruction can move quickly
8556 between memory and registers or between two memory locations. If this
8557 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8558 constant value that is the largest value that @code{MOVE_MAX} can have
8562 @defmac SHIFT_COUNT_TRUNCATED
8563 A C expression that is nonzero if on this machine the number of bits
8564 actually used for the count of a shift operation is equal to the number
8565 of bits needed to represent the size of the object being shifted. When
8566 this macro is nonzero, the compiler will assume that it is safe to omit
8567 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8568 truncates the count of a shift operation. On machines that have
8569 instructions that act on bit-fields at variable positions, which may
8570 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8571 also enables deletion of truncations of the values that serve as
8572 arguments to bit-field instructions.
8574 If both types of instructions truncate the count (for shifts) and
8575 position (for bit-field operations), or if no variable-position bit-field
8576 instructions exist, you should define this macro.
8578 However, on some machines, such as the 80386 and the 680x0, truncation
8579 only applies to shift operations and not the (real or pretended)
8580 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8581 such machines. Instead, add patterns to the @file{md} file that include
8582 the implied truncation of the shift instructions.
8584 You need not define this macro if it would always have the value of zero.
8587 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8588 A C expression which is nonzero if on this machine it is safe to
8589 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8590 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8591 operating on it as if it had only @var{outprec} bits.
8593 On many machines, this expression can be 1.
8595 @c rearranged this, removed the phrase "it is reported that". this was
8596 @c to fix an overfull hbox. --mew 10feb93
8597 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8598 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8599 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8600 such cases may improve things.
8603 @defmac STORE_FLAG_VALUE
8604 A C expression describing the value returned by a comparison operator
8605 with an integral mode and stored by a store-flag instruction
8606 (@samp{s@var{cond}}) when the condition is true. This description must
8607 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8608 comparison operators whose results have a @code{MODE_INT} mode.
8610 A value of 1 or @minus{}1 means that the instruction implementing the
8611 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8612 and 0 when the comparison is false. Otherwise, the value indicates
8613 which bits of the result are guaranteed to be 1 when the comparison is
8614 true. This value is interpreted in the mode of the comparison
8615 operation, which is given by the mode of the first operand in the
8616 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8617 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8620 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8621 generate code that depends only on the specified bits. It can also
8622 replace comparison operators with equivalent operations if they cause
8623 the required bits to be set, even if the remaining bits are undefined.
8624 For example, on a machine whose comparison operators return an
8625 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8626 @samp{0x80000000}, saying that just the sign bit is relevant, the
8630 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8637 (ashift:SI @var{x} (const_int @var{n}))
8641 where @var{n} is the appropriate shift count to move the bit being
8642 tested into the sign bit.
8644 There is no way to describe a machine that always sets the low-order bit
8645 for a true value, but does not guarantee the value of any other bits,
8646 but we do not know of any machine that has such an instruction. If you
8647 are trying to port GCC to such a machine, include an instruction to
8648 perform a logical-and of the result with 1 in the pattern for the
8649 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8651 Often, a machine will have multiple instructions that obtain a value
8652 from a comparison (or the condition codes). Here are rules to guide the
8653 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8658 Use the shortest sequence that yields a valid definition for
8659 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8660 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8661 comparison operators to do so because there may be opportunities to
8662 combine the normalization with other operations.
8665 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8666 slightly preferred on machines with expensive jumps and 1 preferred on
8670 As a second choice, choose a value of @samp{0x80000001} if instructions
8671 exist that set both the sign and low-order bits but do not define the
8675 Otherwise, use a value of @samp{0x80000000}.
8678 Many machines can produce both the value chosen for
8679 @code{STORE_FLAG_VALUE} and its negation in the same number of
8680 instructions. On those machines, you should also define a pattern for
8681 those cases, e.g., one matching
8684 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8687 Some machines can also perform @code{and} or @code{plus} operations on
8688 condition code values with less instructions than the corresponding
8689 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8690 machines, define the appropriate patterns. Use the names @code{incscc}
8691 and @code{decscc}, respectively, for the patterns which perform
8692 @code{plus} or @code{minus} operations on condition code values. See
8693 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8694 find such instruction sequences on other machines.
8696 If this macro is not defined, the default value, 1, is used. You need
8697 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8698 instructions, or if the value generated by these instructions is 1.
8701 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
8702 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8703 returned when comparison operators with floating-point results are true.
8704 Define this macro on machine that have comparison operations that return
8705 floating-point values. If there are no such operations, do not define
8709 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8710 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8711 A C expression that evaluates to true if the architecture defines a value
8712 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
8713 should be set to this value. If this macro is not defined, the value of
8714 @code{clz} or @code{ctz} is assumed to be undefined.
8716 This macro must be defined if the target's expansion for @code{ffs}
8717 relies on a particular value to get correct results. Otherwise it
8718 is not necessary, though it may be used to optimize some corner cases.
8720 Note that regardless of this macro the ``definedness'' of @code{clz}
8721 and @code{ctz} at zero do @emph{not} extend to the builtin functions
8722 visible to the user. Thus one may be free to adjust the value at will
8723 to match the target expansion of these operations without fear of
8728 An alias for the machine mode for pointers. On most machines, define
8729 this to be the integer mode corresponding to the width of a hardware
8730 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8731 On some machines you must define this to be one of the partial integer
8732 modes, such as @code{PSImode}.
8734 The width of @code{Pmode} must be at least as large as the value of
8735 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8736 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8740 @defmac FUNCTION_MODE
8741 An alias for the machine mode used for memory references to functions
8742 being called, in @code{call} RTL expressions. On most machines this
8743 should be @code{QImode}.
8746 @defmac INTEGRATE_THRESHOLD (@var{decl})
8747 A C expression for the maximum number of instructions above which the
8748 function @var{decl} should not be inlined. @var{decl} is a
8749 @code{FUNCTION_DECL} node.
8751 The default definition of this macro is 64 plus 8 times the number of
8752 arguments that the function accepts. Some people think a larger
8753 threshold should be used on RISC machines.
8756 @defmac STDC_0_IN_SYSTEM_HEADERS
8757 In normal operation, the preprocessor expands @code{__STDC__} to the
8758 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8759 hosts, like Solaris, the system compiler uses a different convention,
8760 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8761 strict conformance to the C Standard.
8763 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8764 convention when processing system header files, but when processing user
8765 files @code{__STDC__} will always expand to 1.
8768 @defmac NO_IMPLICIT_EXTERN_C
8769 Define this macro if the system header files support C++ as well as C@.
8770 This macro inhibits the usual method of using system header files in
8771 C++, which is to pretend that the file's contents are enclosed in
8772 @samp{extern "C" @{@dots{}@}}.
8777 @defmac REGISTER_TARGET_PRAGMAS ()
8778 Define this macro if you want to implement any target-specific pragmas.
8779 If defined, it is a C expression which makes a series of calls to
8780 @code{c_register_pragma} for each pragma. The macro may also do any
8781 setup required for the pragmas.
8783 The primary reason to define this macro is to provide compatibility with
8784 other compilers for the same target. In general, we discourage
8785 definition of target-specific pragmas for GCC@.
8787 If the pragma can be implemented by attributes then you should consider
8788 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8790 Preprocessor macros that appear on pragma lines are not expanded. All
8791 @samp{#pragma} directives that do not match any registered pragma are
8792 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8795 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
8797 Each call to @code{c_register_pragma} establishes one pragma. The
8798 @var{callback} routine will be called when the preprocessor encounters a
8802 #pragma [@var{space}] @var{name} @dots{}
8805 @var{space} is the case-sensitive namespace of the pragma, or
8806 @code{NULL} to put the pragma in the global namespace. The callback
8807 routine receives @var{pfile} as its first argument, which can be passed
8808 on to cpplib's functions if necessary. You can lex tokens after the
8809 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8810 callback will be silently ignored. The end of the line is indicated by
8811 a token of type @code{CPP_EOF}
8813 For an example use of this routine, see @file{c4x.h} and the callback
8814 routines defined in @file{c4x-c.c}.
8816 Note that the use of @code{c_lex} is specific to the C and C++
8817 compilers. It will not work in the Java or Fortran compilers, or any
8818 other language compilers for that matter. Thus if @code{c_lex} is going
8819 to be called from target-specific code, it must only be done so when
8820 building the C and C++ compilers. This can be done by defining the
8821 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8822 target entry in the @file{config.gcc} file. These variables should name
8823 the target-specific, language-specific object file which contains the
8824 code that uses @code{c_lex}. Note it will also be necessary to add a
8825 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8826 how to build this object file.
8831 @defmac HANDLE_SYSV_PRAGMA
8832 Define this macro (to a value of 1) if you want the System V style
8833 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8834 [=<value>]} to be supported by gcc.
8836 The pack pragma specifies the maximum alignment (in bytes) of fields
8837 within a structure, in much the same way as the @samp{__aligned__} and
8838 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8839 the behavior to the default.
8841 A subtlety for Microsoft Visual C/C++ style bit-field packing
8842 (e.g. -mms-bitfields) for targets that support it:
8843 When a bit-field is inserted into a packed record, the whole size
8844 of the underlying type is used by one or more same-size adjacent
8845 bit-fields (that is, if its long:3, 32 bits is used in the record,
8846 and any additional adjacent long bit-fields are packed into the same
8847 chunk of 32 bits. However, if the size changes, a new field of that
8850 If both MS bit-fields and @samp{__attribute__((packed))} are used,
8851 the latter will take precedence. If @samp{__attribute__((packed))} is
8852 used on a single field when MS bit-fields are in use, it will take
8853 precedence for that field, but the alignment of the rest of the structure
8854 may affect its placement.
8856 The weak pragma only works if @code{SUPPORTS_WEAK} and
8857 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8858 of specifically named weak labels, optionally with a value.
8863 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
8864 Define this macro (to a value of 1) if you want to support the Win32
8865 style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma
8866 pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment
8867 (in bytes) of fields within a structure, in much the same way as the
8868 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8869 pack value of zero resets the behavior to the default. Successive
8870 invocations of this pragma cause the previous values to be stacked, so
8871 that invocations of @samp{#pragma pack(pop)} will return to the previous
8875 @defmac DOLLARS_IN_IDENTIFIERS
8876 Define this macro to control use of the character @samp{$} in
8877 identifier names for the C family of languages. 0 means @samp{$} is
8878 not allowed by default; 1 means it is allowed. 1 is the default;
8879 there is no need to define this macro in that case.
8882 @defmac NO_DOLLAR_IN_LABEL
8883 Define this macro if the assembler does not accept the character
8884 @samp{$} in label names. By default constructors and destructors in
8885 G++ have @samp{$} in the identifiers. If this macro is defined,
8886 @samp{.} is used instead.
8889 @defmac NO_DOT_IN_LABEL
8890 Define this macro if the assembler does not accept the character
8891 @samp{.} in label names. By default constructors and destructors in G++
8892 have names that use @samp{.}. If this macro is defined, these names
8893 are rewritten to avoid @samp{.}.
8896 @defmac DEFAULT_MAIN_RETURN
8897 Define this macro if the target system expects every program's @code{main}
8898 function to return a standard ``success'' value by default (if no other
8899 value is explicitly returned).
8901 The definition should be a C statement (sans semicolon) to generate the
8902 appropriate rtl instructions. It is used only when compiling the end of
8906 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
8907 Define this macro as a C expression that is nonzero if it is safe for the
8908 delay slot scheduler to place instructions in the delay slot of @var{insn},
8909 even if they appear to use a resource set or clobbered in @var{insn}.
8910 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8911 every @code{call_insn} has this behavior. On machines where some @code{insn}
8912 or @code{jump_insn} is really a function call and hence has this behavior,
8913 you should define this macro.
8915 You need not define this macro if it would always return zero.
8918 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
8919 Define this macro as a C expression that is nonzero if it is safe for the
8920 delay slot scheduler to place instructions in the delay slot of @var{insn},
8921 even if they appear to set or clobber a resource referenced in @var{insn}.
8922 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8923 some @code{insn} or @code{jump_insn} is really a function call and its operands
8924 are registers whose use is actually in the subroutine it calls, you should
8925 define this macro. Doing so allows the delay slot scheduler to move
8926 instructions which copy arguments into the argument registers into the delay
8929 You need not define this macro if it would always return zero.
8932 @defmac MULTIPLE_SYMBOL_SPACES
8933 Define this macro if in some cases global symbols from one translation
8934 unit may not be bound to undefined symbols in another translation unit
8935 without user intervention. For instance, under Microsoft Windows
8936 symbols must be explicitly imported from shared libraries (DLLs).
8939 @defmac MD_ASM_CLOBBERS (@var{clobbers})
8940 A C statement that adds to @var{clobbers} @code{STRING_CST} trees for
8941 any hard regs the port wishes to automatically clobber for all asms.
8944 @defmac MAX_INTEGER_COMPUTATION_MODE
8945 Define this to the largest integer machine mode which can be used for
8946 operations other than load, store and copy operations.
8948 You need only define this macro if the target holds values larger than
8949 @code{word_mode} in general purpose registers. Most targets should not define
8953 @defmac MATH_LIBRARY
8954 Define this macro as a C string constant for the linker argument to link
8955 in the system math library, or @samp{""} if the target does not have a
8956 separate math library.
8958 You need only define this macro if the default of @samp{"-lm"} is wrong.
8961 @defmac LIBRARY_PATH_ENV
8962 Define this macro as a C string constant for the environment variable that
8963 specifies where the linker should look for libraries.
8965 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
8969 @defmac TARGET_HAS_F_SETLKW
8970 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
8971 Note that this functionality is part of POSIX@.
8972 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
8973 to use file locking when exiting a program, which avoids race conditions
8974 if the program has forked.
8977 @defmac MAX_CONDITIONAL_EXECUTE
8979 A C expression for the maximum number of instructions to execute via
8980 conditional execution instructions instead of a branch. A value of
8981 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8982 1 if it does use cc0.
8985 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
8986 Used if the target needs to perform machine-dependent modifications on the
8987 conditionals used for turning basic blocks into conditionally executed code.
8988 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
8989 contains information about the currently processed blocks. @var{true_expr}
8990 and @var{false_expr} are the tests that are used for converting the
8991 then-block and the else-block, respectively. Set either @var{true_expr} or
8992 @var{false_expr} to a null pointer if the tests cannot be converted.
8995 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
8996 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
8997 if-statements into conditions combined by @code{and} and @code{or} operations.
8998 @var{bb} contains the basic block that contains the test that is currently
8999 being processed and about to be turned into a condition.
9002 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
9003 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9004 be converted to conditional execution format. @var{ce_info} points to
9005 a data structure, @code{struct ce_if_block}, which contains information
9006 about the currently processed blocks.
9009 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
9010 A C expression to perform any final machine dependent modifications in
9011 converting code to conditional execution. The involved basic blocks
9012 can be found in the @code{struct ce_if_block} structure that is pointed
9013 to by @var{ce_info}.
9016 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
9017 A C expression to cancel any machine dependent modifications in
9018 converting code to conditional execution. The involved basic blocks
9019 can be found in the @code{struct ce_if_block} structure that is pointed
9020 to by @var{ce_info}.
9023 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
9024 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9025 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9028 @defmac IFCVT_EXTRA_FIELDS
9029 If defined, it should expand to a set of field declarations that will be
9030 added to the @code{struct ce_if_block} structure. These should be initialized
9031 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9034 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
9035 If non-null, this hook performs a target-specific pass over the
9036 instruction stream. The compiler will run it at all optimization levels,
9037 just before the point at which it normally does delayed-branch scheduling.
9039 The exact purpose of the hook varies from target to target. Some use
9040 it to do transformations that are necessary for correctness, such as
9041 laying out in-function constant pools or avoiding hardware hazards.
9042 Others use it as an opportunity to do some machine-dependent optimizations.
9044 You need not implement the hook if it has nothing to do. The default
9048 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9049 Define this hook if you have any machine-specific built-in functions
9050 that need to be defined. It should be a function that performs the
9053 Machine specific built-in functions can be useful to expand special machine
9054 instructions that would otherwise not normally be generated because
9055 they have no equivalent in the source language (for example, SIMD vector
9056 instructions or prefetch instructions).
9058 To create a built-in function, call the function @code{builtin_function}
9059 which is defined by the language front end. You can use any type nodes set
9060 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9061 only language front ends that use those two functions will call
9062 @samp{TARGET_INIT_BUILTINS}.
9065 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9067 Expand a call to a machine specific built-in function that was set up by
9068 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9069 function call; the result should go to @var{target} if that is
9070 convenient, and have mode @var{mode} if that is convenient.
9071 @var{subtarget} may be used as the target for computing one of
9072 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9073 ignored. This function should return the result of the call to the
9077 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
9079 Take a branch insn in @var{branch1} and another in @var{branch2}.
9080 Return true if redirecting @var{branch1} to the destination of
9081 @var{branch2} is possible.
9083 On some targets, branches may have a limited range. Optimizing the
9084 filling of delay slots can result in branches being redirected, and this
9085 may in turn cause a branch offset to overflow.
9088 @defmac ALLOCATE_INITIAL_VALUE (@var{hard_reg})
9090 When the initial value of a hard register has been copied in a pseudo
9091 register, it is often not necessary to actually allocate another register
9092 to this pseudo register, because the original hard register or a stack slot
9093 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9094 defined, is called at the start of register allocation once for each
9095 hard register that had its initial value copied by using
9096 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9097 Possible values are @code{NULL_RTX}, if you don't want
9098 to do any special allocation, a @code{REG} rtx---that would typically be
9099 the hard register itself, if it is known not to be clobbered---or a
9101 If you are returning a @code{MEM}, this is only a hint for the allocator;
9102 it might decide to use another register anyways.
9103 You may use @code{current_function_leaf_function} in the definition of the
9104 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9105 register in question will not be clobbered.
9108 @defmac TARGET_OBJECT_SUFFIX
9109 Define this macro to be a C string representing the suffix for object
9110 files on your target machine. If you do not define this macro, GCC will
9111 use @samp{.o} as the suffix for object files.
9114 @defmac TARGET_EXECUTABLE_SUFFIX
9115 Define this macro to be a C string representing the suffix to be
9116 automatically added to executable files on your target machine. If you
9117 do not define this macro, GCC will use the null string as the suffix for
9121 @defmac COLLECT_EXPORT_LIST
9122 If defined, @code{collect2} will scan the individual object files
9123 specified on its command line and create an export list for the linker.
9124 Define this macro for systems like AIX, where the linker discards
9125 object files that are not referenced from @code{main} and uses export
9129 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
9130 Define this macro to a C expression representing a variant of the
9131 method call @var{mdecl}, if Java Native Interface (JNI) methods
9132 must be invoked differently from other methods on your target.
9133 For example, on 32-bit Windows, JNI methods must be invoked using
9134 the @code{stdcall} calling convention and this macro is then
9135 defined as this expression:
9138 build_type_attribute_variant (@var{mdecl},
9140 (get_identifier ("stdcall"),
9145 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9146 This target hook returns @code{true} past the point in which new jump
9147 instructions could be created. On machines that require a register for
9148 every jump such as the SHmedia ISA of SH5, this point would typically be
9149 reload, so this target hook should be defined to a function such as:
9153 cannot_modify_jumps_past_reload_p ()
9155 return (reload_completed || reload_in_progress);
9160 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
9161 This target hook returns a register class for which branch target register
9162 optimizations should be applied. All registers in this class should be
9163 usable interchangeably. After reload, registers in this class will be
9164 re-allocated and loads will be hoisted out of loops and be subjected
9165 to inter-block scheduling.
9168 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
9169 Branch target register optimization will by default exclude callee-saved
9171 that are not already live during the current function; if this target hook
9172 returns true, they will be included. The target code must than make sure
9173 that all target registers in the class returned by
9174 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
9175 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
9176 epilogues have already been generated. Note, even if you only return
9177 true when @var{after_prologue_epilogue_gen} is false, you still are likely
9178 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
9179 to reserve space for caller-saved target registers.
9182 @defmac POWI_MAX_MULTS
9183 If defined, this macro is interpreted as a signed integer C expression
9184 that specifies the maximum number of floating point multiplications
9185 that should be emitted when expanding exponentiation by an integer
9186 constant inline. When this value is defined, exponentiation requiring
9187 more than this number of multiplications is implemented by calling the
9188 system library's @code{pow}, @code{powf} or @code{powl} routines.
9189 The default value places no upper bound on the multiplication count.