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 @code{libdir} as the default prefix to
453 try when searching for startup files such as @file{crt0.o}.
454 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
455 is built as a cross compiler.
458 @defmac MD_STARTFILE_PREFIX
459 If defined, this macro supplies an additional prefix to try after the
460 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
461 @option{-b} option is used, or when the compiler is built as a cross
465 @defmac MD_STARTFILE_PREFIX_1
466 If defined, this macro supplies yet another prefix to try after the
467 standard prefixes. It is not searched when the @option{-b} option is
468 used, or when the compiler is built as a cross compiler.
471 @defmac INIT_ENVIRONMENT
472 Define this macro as a C string constant if you wish to set environment
473 variables for programs called by the driver, such as the assembler and
474 loader. The driver passes the value of this macro to @code{putenv} to
475 initialize the necessary environment variables.
478 @defmac LOCAL_INCLUDE_DIR
479 Define this macro as a C string constant if you wish to override the
480 standard choice of @file{/usr/local/include} as the default prefix to
481 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
482 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
484 Cross compilers do not search either @file{/usr/local/include} or its
488 @defmac MODIFY_TARGET_NAME
489 Define this macro if you wish to define command-line switches that
490 modify the default target name.
492 For each switch, you can include a string to be appended to the first
493 part of the configuration name or a string to be deleted from the
494 configuration name, if present. The definition should be an initializer
495 for an array of structures. Each array element should have three
496 elements: the switch name (a string constant, including the initial
497 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
498 indicate whether the string should be inserted or deleted, and the string
499 to be inserted or deleted (a string constant).
501 For example, on a machine where @samp{64} at the end of the
502 configuration name denotes a 64-bit target and you want the @option{-32}
503 and @option{-64} switches to select between 32- and 64-bit targets, you would
507 #define MODIFY_TARGET_NAME \
508 @{ @{ "-32", DELETE, "64"@}, \
509 @{"-64", ADD, "64"@}@}
513 @defmac SYSTEM_INCLUDE_DIR
514 Define this macro as a C string constant if you wish to specify a
515 system-specific directory to search for header files before the standard
516 directory. @code{SYSTEM_INCLUDE_DIR} comes before
517 @code{STANDARD_INCLUDE_DIR} in the search order.
519 Cross compilers do not use this macro and do not search the directory
523 @defmac STANDARD_INCLUDE_DIR
524 Define this macro as a C string constant if you wish to override the
525 standard choice of @file{/usr/include} as the default prefix to
526 try when searching for header files.
528 Cross compilers ignore this macro and do not search either
529 @file{/usr/include} or its replacement.
532 @defmac STANDARD_INCLUDE_COMPONENT
533 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
534 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
535 If you do not define this macro, no component is used.
538 @defmac INCLUDE_DEFAULTS
539 Define this macro if you wish to override the entire default search path
540 for include files. For a native compiler, the default search path
541 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
542 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
543 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
544 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
545 and specify private search areas for GCC@. The directory
546 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
548 The definition should be an initializer for an array of structures.
549 Each array element should have four elements: the directory name (a
550 string constant), the component name (also a string constant), a flag
551 for C++-only directories,
552 and a flag showing that the includes in the directory don't need to be
553 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
554 the array with a null element.
556 The component name denotes what GNU package the include file is part of,
557 if any, in all uppercase letters. For example, it might be @samp{GCC}
558 or @samp{BINUTILS}. If the package is part of a vendor-supplied
559 operating system, code the component name as @samp{0}.
561 For example, here is the definition used for VAX/VMS:
564 #define INCLUDE_DEFAULTS \
566 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
567 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
568 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
575 Here is the order of prefixes tried for exec files:
579 Any prefixes specified by the user with @option{-B}.
582 The environment variable @code{GCC_EXEC_PREFIX}, if any.
585 The directories specified by the environment variable @code{COMPILER_PATH}.
588 The macro @code{STANDARD_EXEC_PREFIX}.
591 @file{/usr/lib/gcc/}.
594 The macro @code{MD_EXEC_PREFIX}, if any.
597 Here is the order of prefixes tried for startfiles:
601 Any prefixes specified by the user with @option{-B}.
604 The environment variable @code{GCC_EXEC_PREFIX}, if any.
607 The directories specified by the environment variable @code{LIBRARY_PATH}
608 (or port-specific name; native only, cross compilers do not use this).
611 The macro @code{STANDARD_EXEC_PREFIX}.
614 @file{/usr/lib/gcc/}.
617 The macro @code{MD_EXEC_PREFIX}, if any.
620 The macro @code{MD_STARTFILE_PREFIX}, if any.
623 The macro @code{STANDARD_STARTFILE_PREFIX}.
632 @node Run-time Target
633 @section Run-time Target Specification
634 @cindex run-time target specification
635 @cindex predefined macros
636 @cindex target specifications
638 @c prevent bad page break with this line
639 Here are run-time target specifications.
641 @defmac TARGET_CPU_CPP_BUILTINS ()
642 This function-like macro expands to a block of code that defines
643 built-in preprocessor macros and assertions for the target cpu, using
644 the functions @code{builtin_define}, @code{builtin_define_std} and
645 @code{builtin_assert}. When the front end
646 calls this macro it provides a trailing semicolon, and since it has
647 finished command line option processing your code can use those
650 @code{builtin_assert} takes a string in the form you pass to the
651 command-line option @option{-A}, such as @code{cpu=mips}, and creates
652 the assertion. @code{builtin_define} takes a string in the form
653 accepted by option @option{-D} and unconditionally defines the macro.
655 @code{builtin_define_std} takes a string representing the name of an
656 object-like macro. If it doesn't lie in the user's namespace,
657 @code{builtin_define_std} defines it unconditionally. Otherwise, it
658 defines a version with two leading underscores, and another version
659 with two leading and trailing underscores, and defines the original
660 only if an ISO standard was not requested on the command line. For
661 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
662 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
663 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
664 defines only @code{_ABI64}.
666 You can also test for the C dialect being compiled. The variable
667 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
668 or @code{clk_objective_c}. Note that if we are preprocessing
669 assembler, this variable will be @code{clk_c} but the function-like
670 macro @code{preprocessing_asm_p()} will return true, so you might want
671 to check for that first. If you need to check for strict ANSI, the
672 variable @code{flag_iso} can be used. The function-like macro
673 @code{preprocessing_trad_p()} can be used to check for traditional
677 @defmac TARGET_OS_CPP_BUILTINS ()
678 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
679 and is used for the target operating system instead.
682 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
683 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
684 and is used for the target object format. @file{elfos.h} uses this
685 macro to define @code{__ELF__}, so you probably do not need to define
689 @deftypevar {extern int} target_flags
690 This declaration should be present.
693 @cindex optional hardware or system features
694 @cindex features, optional, in system conventions
696 @defmac TARGET_@var{featurename}
697 This series of macros is to allow compiler command arguments to
698 enable or disable the use of optional features of the target machine.
699 For example, one machine description serves both the 68000 and
700 the 68020; a command argument tells the compiler whether it should
701 use 68020-only instructions or not. This command argument works
702 by means of a macro @code{TARGET_68020} that tests a bit in
705 Define a macro @code{TARGET_@var{featurename}} for each such option.
706 Its definition should test a bit in @code{target_flags}. It is
707 recommended that a helper macro @code{MASK_@var{featurename}}
708 is defined for each bit-value to test, and used in
709 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
713 #define TARGET_MASK_68020 1
714 #define TARGET_68020 (target_flags & MASK_68020)
717 One place where these macros are used is in the condition-expressions
718 of instruction patterns. Note how @code{TARGET_68020} appears
719 frequently in the 68000 machine description file, @file{m68k.md}.
720 Another place they are used is in the definitions of the other
721 macros in the @file{@var{machine}.h} file.
724 @defmac TARGET_SWITCHES
725 This macro defines names of command options to set and clear
726 bits in @code{target_flags}. Its definition is an initializer
727 with a subgrouping for each command option.
729 Each subgrouping contains a string constant, that defines the option
730 name, a number, which contains the bits to set in
731 @code{target_flags}, and a second string which is the description
732 displayed by @option{--help}. If the number is negative then the bits specified
733 by the number are cleared instead of being set. If the description
734 string is present but empty, then no help information will be displayed
735 for that option, but it will not count as an undocumented option. The
736 actual option name is made by appending @samp{-m} to the specified name.
737 Non-empty description strings should be marked with @code{N_(@dots{})} for
738 @command{xgettext}. Please do not mark empty strings because the empty
739 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
740 of the message catalog with meta information, not the empty string.
742 In addition to the description for @option{--help},
743 more detailed documentation for each option should be added to
746 One of the subgroupings should have a null string. The number in
747 this grouping is the default value for @code{target_flags}. Any
748 target options act starting with that value.
750 Here is an example which defines @option{-m68000} and @option{-m68020}
751 with opposite meanings, and picks the latter as the default:
754 #define TARGET_SWITCHES \
755 @{ @{ "68020", MASK_68020, "" @}, \
756 @{ "68000", -MASK_68020, \
757 N_("Compile for the 68000") @}, \
758 @{ "", MASK_68020, "" @}, \
763 @defmac TARGET_OPTIONS
764 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
765 options that have values. Its definition is an initializer with a
766 subgrouping for each command option.
768 Each subgrouping contains a string constant, that defines the option
769 name, the address of a variable, a description string, and a value.
770 Non-empty description strings should be marked with @code{N_(@dots{})}
771 for @command{xgettext}. Please do not mark empty strings because the
772 empty string is reserved by GNU gettext. @code{gettext("")} returns the
773 header entry of the message catalog with meta information, not the empty
776 If the value listed in the table is @code{NULL}, then the variable, type
777 @code{char *}, is set to the variable part of the given option if the
778 fixed part matches. In other words, if the first part of the option
779 matches what's in the table, the variable will be set to point to the
780 rest of the option. This allows the user to specify a value for that
781 option. The actual option name is made by appending @samp{-m} to the
782 specified name. Again, each option should also be documented in
785 If the value listed in the table is non-@code{NULL}, then the option
786 must match the option in the table exactly (with @samp{-m}), and the
787 variable is set to point to the value listed in the table.
789 Here is an example which defines @option{-mshort-data-@var{number}}. If the
790 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
791 will be set to the string @code{"512"}.
794 extern char *m88k_short_data;
795 #define TARGET_OPTIONS \
796 @{ @{ "short-data-", &m88k_short_data, \
797 N_("Specify the size of the short data section"), 0 @} @}
800 Here is a variant of the above that allows the user to also specify
801 just @option{-mshort-data} where a default of @code{"64"} is used.
804 extern char *m88k_short_data;
805 #define TARGET_OPTIONS \
806 @{ @{ "short-data-", &m88k_short_data, \
807 N_("Specify the size of the short data section"), 0 @} \
808 @{ "short-data", &m88k_short_data, "", "64" @},
812 Here is an example which defines @option{-mno-alu}, @option{-malu1}, and
813 @option{-malu2} as a three-state switch, along with suitable macros for
814 checking the state of the option (documentation is elided for brevity).
818 char *chip_alu = ""; /* Specify default here. */
821 extern char *chip_alu;
822 #define TARGET_OPTIONS \
823 @{ @{ "no-alu", &chip_alu, "", "" @}, \
824 @{ "alu1", &chip_alu, "", "1" @}, \
825 @{ "alu2", &chip_alu, "", "2" @}, @}
826 #define TARGET_ALU (chip_alu[0] != '\0')
827 #define TARGET_ALU1 (chip_alu[0] == '1')
828 #define TARGET_ALU2 (chip_alu[0] == '2')
832 @defmac TARGET_VERSION
833 This macro is a C statement to print on @code{stderr} a string
834 describing the particular machine description choice. Every machine
835 description should define @code{TARGET_VERSION}. For example:
839 #define TARGET_VERSION \
840 fprintf (stderr, " (68k, Motorola syntax)");
842 #define TARGET_VERSION \
843 fprintf (stderr, " (68k, MIT syntax)");
848 @defmac OVERRIDE_OPTIONS
849 Sometimes certain combinations of command options do not make sense on
850 a particular target machine. You can define a macro
851 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
852 defined, is executed once just after all the command options have been
855 Don't use this macro to turn on various extra optimizations for
856 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
859 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
860 Some machines may desire to change what optimizations are performed for
861 various optimization levels. This macro, if defined, is executed once
862 just after the optimization level is determined and before the remainder
863 of the command options have been parsed. Values set in this macro are
864 used as the default values for the other command line options.
866 @var{level} is the optimization level specified; 2 if @option{-O2} is
867 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
869 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
871 You should not use this macro to change options that are not
872 machine-specific. These should uniformly selected by the same
873 optimization level on all supported machines. Use this macro to enable
874 machine-specific optimizations.
876 @strong{Do not examine @code{write_symbols} in
877 this macro!} The debugging options are not supposed to alter the
881 @defmac CAN_DEBUG_WITHOUT_FP
882 Define this macro if debugging can be performed even without a frame
883 pointer. If this macro is defined, GCC will turn on the
884 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
887 @node Per-Function Data
888 @section Defining data structures for per-function information.
889 @cindex per-function data
890 @cindex data structures
892 If the target needs to store information on a per-function basis, GCC
893 provides a macro and a couple of variables to allow this. Note, just
894 using statics to store the information is a bad idea, since GCC supports
895 nested functions, so you can be halfway through encoding one function
896 when another one comes along.
898 GCC defines a data structure called @code{struct function} which
899 contains all of the data specific to an individual function. This
900 structure contains a field called @code{machine} whose type is
901 @code{struct machine_function *}, which can be used by targets to point
902 to their own specific data.
904 If a target needs per-function specific data it should define the type
905 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
906 This macro should be used to initialize the function pointer
907 @code{init_machine_status}. This pointer is explained below.
909 One typical use of per-function, target specific data is to create an
910 RTX to hold the register containing the function's return address. This
911 RTX can then be used to implement the @code{__builtin_return_address}
912 function, for level 0.
914 Note---earlier implementations of GCC used a single data area to hold
915 all of the per-function information. Thus when processing of a nested
916 function began the old per-function data had to be pushed onto a
917 stack, and when the processing was finished, it had to be popped off the
918 stack. GCC used to provide function pointers called
919 @code{save_machine_status} and @code{restore_machine_status} to handle
920 the saving and restoring of the target specific information. Since the
921 single data area approach is no longer used, these pointers are no
924 @defmac INIT_EXPANDERS
925 Macro called to initialize any target specific information. This macro
926 is called once per function, before generation of any RTL has begun.
927 The intention of this macro is to allow the initialization of the
928 function pointer @code{init_machine_status}.
931 @deftypevar {void (*)(struct function *)} init_machine_status
932 If this function pointer is non-@code{NULL} it will be called once per
933 function, before function compilation starts, in order to allow the
934 target to perform any target specific initialization of the
935 @code{struct function} structure. It is intended that this would be
936 used to initialize the @code{machine} of that structure.
938 @code{struct machine_function} structures are expected to be freed by GC.
939 Generally, any memory that they reference must be allocated by using
940 @code{ggc_alloc}, including the structure itself.
944 @section Storage Layout
945 @cindex storage layout
947 Note that the definitions of the macros in this table which are sizes or
948 alignments measured in bits do not need to be constant. They can be C
949 expressions that refer to static variables, such as the @code{target_flags}.
950 @xref{Run-time Target}.
952 @defmac BITS_BIG_ENDIAN
953 Define this macro to have the value 1 if the most significant bit in a
954 byte has the lowest number; otherwise define it to have the value zero.
955 This means that bit-field instructions count from the most significant
956 bit. If the machine has no bit-field instructions, then this must still
957 be defined, but it doesn't matter which value it is defined to. This
958 macro need not be a constant.
960 This macro does not affect the way structure fields are packed into
961 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
964 @defmac BYTES_BIG_ENDIAN
965 Define this macro to have the value 1 if the most significant byte in a
966 word has the lowest number. This macro need not be a constant.
969 @defmac WORDS_BIG_ENDIAN
970 Define this macro to have the value 1 if, in a multiword object, the
971 most significant word has the lowest number. This applies to both
972 memory locations and registers; GCC fundamentally assumes that the
973 order of words in memory is the same as the order in registers. This
974 macro need not be a constant.
977 @defmac LIBGCC2_WORDS_BIG_ENDIAN
978 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
979 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
980 used only when compiling @file{libgcc2.c}. Typically the value will be set
981 based on preprocessor defines.
984 @defmac FLOAT_WORDS_BIG_ENDIAN
985 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
986 @code{TFmode} floating point numbers are stored in memory with the word
987 containing the sign bit at the lowest address; otherwise define it to
988 have the value 0. This macro need not be a constant.
990 You need not define this macro if the ordering is the same as for
994 @defmac BITS_PER_UNIT
995 Define this macro to be the number of bits in an addressable storage
996 unit (byte). If you do not define this macro the default is 8.
999 @defmac BITS_PER_WORD
1000 Number of bits in a word. If you do not define this macro, the default
1001 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
1004 @defmac MAX_BITS_PER_WORD
1005 Maximum number of bits in a word. If this is undefined, the default is
1006 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
1007 largest value that @code{BITS_PER_WORD} can have at run-time.
1010 @defmac UNITS_PER_WORD
1011 Number of storage units in a word; normally 4.
1014 @defmac MIN_UNITS_PER_WORD
1015 Minimum number of units in a word. If this is undefined, the default is
1016 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
1017 smallest value that @code{UNITS_PER_WORD} can have at run-time.
1020 @defmac POINTER_SIZE
1021 Width of a pointer, in bits. You must specify a value no wider than the
1022 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1023 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1024 a value the default is @code{BITS_PER_WORD}.
1027 @defmac POINTERS_EXTEND_UNSIGNED
1028 A C expression whose value is greater than zero if pointers that need to be
1029 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
1030 be zero-extended and zero if they are to be sign-extended. If the value
1031 is less then zero then there must be an "ptr_extend" instruction that
1032 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
1034 You need not define this macro if the @code{POINTER_SIZE} is equal
1035 to the width of @code{Pmode}.
1038 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1039 A macro to update @var{m} and @var{unsignedp} when an object whose type
1040 is @var{type} and which has the specified mode and signedness is to be
1041 stored in a register. This macro is only called when @var{type} is a
1044 On most RISC machines, which only have operations that operate on a full
1045 register, define this macro to set @var{m} to @code{word_mode} if
1046 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1047 cases, only integer modes should be widened because wider-precision
1048 floating-point operations are usually more expensive than their narrower
1051 For most machines, the macro definition does not change @var{unsignedp}.
1052 However, some machines, have instructions that preferentially handle
1053 either signed or unsigned quantities of certain modes. For example, on
1054 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1055 sign-extend the result to 64 bits. On such machines, set
1056 @var{unsignedp} according to which kind of extension is more efficient.
1058 Do not define this macro if it would never modify @var{m}.
1061 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1062 This target hook should return @code{true} if the promotion described by
1063 @code{PROMOTE_MODE} should also be done for outgoing function arguments.
1066 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1067 This target hook should return @code{true} if the promotion described by
1068 @code{PROMOTE_MODE} should also be done for the return value of
1071 If this target hook returns @code{true}, @code{FUNCTION_VALUE} must
1072 perform the same promotions done by @code{PROMOTE_MODE}.
1075 @defmac PROMOTE_FOR_CALL_ONLY
1076 Define this macro if the promotion described by @code{PROMOTE_MODE}
1077 should @emph{only} be performed for outgoing function arguments or
1078 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1079 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1082 @defmac PARM_BOUNDARY
1083 Normal alignment required for function parameters on the stack, in
1084 bits. All stack parameters receive at least this much alignment
1085 regardless of data type. On most machines, this is the same as the
1089 @defmac STACK_BOUNDARY
1090 Define this macro to the minimum alignment enforced by hardware for the
1091 stack pointer on this machine. The definition is a C expression for the
1092 desired alignment (measured in bits). This value is used as a default
1093 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1094 this should be the same as @code{PARM_BOUNDARY}.
1097 @defmac PREFERRED_STACK_BOUNDARY
1098 Define this macro if you wish to preserve a certain alignment for the
1099 stack pointer, greater than what the hardware enforces. The definition
1100 is a C expression for the desired alignment (measured in bits). This
1101 macro must evaluate to a value equal to or larger than
1102 @code{STACK_BOUNDARY}.
1105 @defmac FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1106 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1107 not guaranteed by the runtime and we should emit code to align the stack
1108 at the beginning of @code{main}.
1110 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1111 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1112 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1113 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1114 be momentarily unaligned while pushing arguments.
1117 @defmac FUNCTION_BOUNDARY
1118 Alignment required for a function entry point, in bits.
1121 @defmac BIGGEST_ALIGNMENT
1122 Biggest alignment that any data type can require on this machine, in bits.
1125 @defmac MINIMUM_ATOMIC_ALIGNMENT
1126 If defined, the smallest alignment, in bits, that can be given to an
1127 object that can be referenced in one operation, without disturbing any
1128 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1129 on machines that don't have byte or half-word store operations.
1132 @defmac BIGGEST_FIELD_ALIGNMENT
1133 Biggest alignment that any structure or union field can require on this
1134 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1135 structure and union fields only, unless the field alignment has been set
1136 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1139 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1140 An expression for the alignment of a structure field @var{field} if the
1141 alignment computed in the usual way (including applying of
1142 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1143 alignment) is @var{computed}. It overrides alignment only if the
1144 field alignment has not been set by the
1145 @code{__attribute__ ((aligned (@var{n})))} construct.
1148 @defmac MAX_OFILE_ALIGNMENT
1149 Biggest alignment supported by the object file format of this machine.
1150 Use this macro to limit the alignment which can be specified using the
1151 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1152 the default value is @code{BIGGEST_ALIGNMENT}.
1155 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1156 If defined, a C expression to compute the alignment for a variable in
1157 the static store. @var{type} is the data type, and @var{basic-align} is
1158 the alignment that the object would ordinarily have. The value of this
1159 macro is used instead of that alignment to align the object.
1161 If this macro is not defined, then @var{basic-align} is used.
1164 One use of this macro is to increase alignment of medium-size data to
1165 make it all fit in fewer cache lines. Another is to cause character
1166 arrays to be word-aligned so that @code{strcpy} calls that copy
1167 constants to character arrays can be done inline.
1170 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1171 If defined, a C expression to compute the alignment given to a constant
1172 that is being placed in memory. @var{constant} is the constant and
1173 @var{basic-align} is the alignment that the object would ordinarily
1174 have. The value of this macro is used instead of that alignment to
1177 If this macro is not defined, then @var{basic-align} is used.
1179 The typical use of this macro is to increase alignment for string
1180 constants to be word aligned so that @code{strcpy} calls that copy
1181 constants can be done inline.
1184 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1185 If defined, a C expression to compute the alignment for a variable in
1186 the local store. @var{type} is the data type, and @var{basic-align} is
1187 the alignment that the object would ordinarily have. The value of this
1188 macro is used instead of that alignment to align the object.
1190 If this macro is not defined, then @var{basic-align} is used.
1192 One use of this macro is to increase alignment of medium-size data to
1193 make it all fit in fewer cache lines.
1196 @defmac EMPTY_FIELD_BOUNDARY
1197 Alignment in bits to be given to a structure bit-field that follows an
1198 empty field such as @code{int : 0;}.
1200 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1203 @defmac STRUCTURE_SIZE_BOUNDARY
1204 Number of bits which any structure or union's size must be a multiple of.
1205 Each structure or union's size is rounded up to a multiple of this.
1207 If you do not define this macro, the default is the same as
1208 @code{BITS_PER_UNIT}.
1211 @defmac STRICT_ALIGNMENT
1212 Define this macro to be the value 1 if instructions will fail to work
1213 if given data not on the nominal alignment. If instructions will merely
1214 go slower in that case, define this macro as 0.
1217 @defmac PCC_BITFIELD_TYPE_MATTERS
1218 Define this if you wish to imitate the way many other C compilers handle
1219 alignment of bit-fields and the structures that contain them.
1221 The behavior is that the type written for a named bit-field (@code{int},
1222 @code{short}, or other integer type) imposes an alignment for the entire
1223 structure, as if the structure really did contain an ordinary field of
1224 that type. In addition, the bit-field is placed within the structure so
1225 that it would fit within such a field, not crossing a boundary for it.
1227 Thus, on most machines, a named bit-field whose type is written as
1228 @code{int} would not cross a four-byte boundary, and would force
1229 four-byte alignment for the whole structure. (The alignment used may
1230 not be four bytes; it is controlled by the other alignment parameters.)
1232 An unnamed bit-field will not affect the alignment of the containing
1235 If the macro is defined, its definition should be a C expression;
1236 a nonzero value for the expression enables this behavior.
1238 Note that if this macro is not defined, or its value is zero, some
1239 bit-fields may cross more than one alignment boundary. The compiler can
1240 support such references if there are @samp{insv}, @samp{extv}, and
1241 @samp{extzv} insns that can directly reference memory.
1243 The other known way of making bit-fields work is to define
1244 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1245 Then every structure can be accessed with fullwords.
1247 Unless the machine has bit-field instructions or you define
1248 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1249 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1251 If your aim is to make GCC use the same conventions for laying out
1252 bit-fields as are used by another compiler, here is how to investigate
1253 what the other compiler does. Compile and run this program:
1272 printf ("Size of foo1 is %d\n",
1273 sizeof (struct foo1));
1274 printf ("Size of foo2 is %d\n",
1275 sizeof (struct foo2));
1280 If this prints 2 and 5, then the compiler's behavior is what you would
1281 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1284 @defmac BITFIELD_NBYTES_LIMITED
1285 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1286 to aligning a bit-field within the structure.
1289 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1290 Return 1 if a structure or array containing @var{field} should be accessed using
1293 If @var{field} is the only field in the structure, @var{mode} is its
1294 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1295 case where structures of one field would require the structure's mode to
1296 retain the field's mode.
1298 Normally, this is not needed. See the file @file{c4x.h} for an example
1299 of how to use this macro to prevent a structure having a floating point
1300 field from being accessed in an integer mode.
1303 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1304 Define this macro as an expression for the alignment of a type (given
1305 by @var{type} as a tree node) if the alignment computed in the usual
1306 way is @var{computed} and the alignment explicitly specified was
1309 The default is to use @var{specified} if it is larger; otherwise, use
1310 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1313 @defmac MAX_FIXED_MODE_SIZE
1314 An integer expression for the size in bits of the largest integer
1315 machine mode that should actually be used. All integer machine modes of
1316 this size or smaller can be used for structures and unions with the
1317 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1318 (DImode)} is assumed.
1321 @defmac VECTOR_MODE_SUPPORTED_P (@var{mode})
1322 Define this macro to be nonzero if the port is prepared to handle insns
1323 involving vector mode @var{mode}. At the very least, it must have move
1324 patterns for this mode.
1327 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1328 If defined, an expression of type @code{enum machine_mode} that
1329 specifies the mode of the save area operand of a
1330 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1331 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1332 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1333 having its mode specified.
1335 You need not define this macro if it always returns @code{Pmode}. You
1336 would most commonly define this macro if the
1337 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1341 @defmac STACK_SIZE_MODE
1342 If defined, an expression of type @code{enum machine_mode} that
1343 specifies the mode of the size increment operand of an
1344 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1346 You need not define this macro if it always returns @code{word_mode}.
1347 You would most commonly define this macro if the @code{allocate_stack}
1348 pattern needs to support both a 32- and a 64-bit mode.
1351 @defmac TARGET_FLOAT_FORMAT
1352 A code distinguishing the floating point format of the target machine.
1353 There are four defined values:
1356 @item IEEE_FLOAT_FORMAT
1357 This code indicates IEEE floating point. It is the default; there is no
1358 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1360 @item VAX_FLOAT_FORMAT
1361 This code indicates the ``F float'' (for @code{float}) and ``D float''
1362 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1364 @item IBM_FLOAT_FORMAT
1365 This code indicates the format used on the IBM System/370.
1367 @item C4X_FLOAT_FORMAT
1368 This code indicates the format used on the TMS320C3x/C4x.
1371 If your target uses a floating point format other than these, you must
1372 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1373 it to @file{real.c}.
1375 The ordering of the component words of floating point values stored in
1376 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1379 @defmac MODE_HAS_NANS (@var{mode})
1380 When defined, this macro should be true if @var{mode} has a NaN
1381 representation. The compiler assumes that NaNs are not equal to
1382 anything (including themselves) and that addition, subtraction,
1383 multiplication and division all return NaNs when one operand is
1386 By default, this macro is true if @var{mode} is a floating-point
1387 mode and the target floating-point format is IEEE@.
1390 @defmac MODE_HAS_INFINITIES (@var{mode})
1391 This macro should be true if @var{mode} can represent infinity. At
1392 present, the compiler uses this macro to decide whether @samp{x - x}
1393 is always defined. By default, the macro is true when @var{mode}
1394 is a floating-point mode and the target format is IEEE@.
1397 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1398 True if @var{mode} distinguishes between positive and negative zero.
1399 The rules are expected to follow the IEEE standard:
1403 @samp{x + x} has the same sign as @samp{x}.
1406 If the sum of two values with opposite sign is zero, the result is
1407 positive for all rounding modes expect towards @minus{}infinity, for
1408 which it is negative.
1411 The sign of a product or quotient is negative when exactly one
1412 of the operands is negative.
1415 The default definition is true if @var{mode} is a floating-point
1416 mode and the target format is IEEE@.
1419 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1420 If defined, this macro should be true for @var{mode} if it has at
1421 least one rounding mode in which @samp{x} and @samp{-x} can be
1422 rounded to numbers of different magnitude. Two such modes are
1423 towards @minus{}infinity and towards +infinity.
1425 The default definition of this macro is true if @var{mode} is
1426 a floating-point mode and the target format is IEEE@.
1429 @defmac ROUND_TOWARDS_ZERO
1430 If defined, this macro should be true if the prevailing rounding
1431 mode is towards zero. A true value has the following effects:
1435 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1438 @file{libgcc.a}'s floating-point emulator will round towards zero
1439 rather than towards nearest.
1442 The compiler's floating-point emulator will round towards zero after
1443 doing arithmetic, and when converting from the internal float format to
1447 The macro does not affect the parsing of string literals. When the
1448 primary rounding mode is towards zero, library functions like
1449 @code{strtod} might still round towards nearest, and the compiler's
1450 parser should behave like the target's @code{strtod} where possible.
1452 Not defining this macro is equivalent to returning zero.
1455 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1456 This macro should return true if floats with @var{size}
1457 bits do not have a NaN or infinity representation, but use the largest
1458 exponent for normal numbers instead.
1460 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1461 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1462 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1463 floating-point arithmetic.
1465 The default definition of this macro returns false for all sizes.
1468 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1469 This target hook should return @code{true} a vector is opaque. That
1470 is, if no cast is needed when copying a vector value of type
1471 @var{type} into another vector lvalue of the same size. Vector opaque
1472 types cannot be initialized. The default is that there are no such
1476 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1477 This target hook returns @code{true} if bit-fields in the given
1478 @var{record_type} are to be laid out following the rules of Microsoft
1479 Visual C/C++, namely: (i) a bit-field won't share the same storage
1480 unit with the previous bit-field if their underlying types have
1481 different sizes, and the bit-field will be aligned to the highest
1482 alignment of the underlying types of itself and of the previous
1483 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1484 the whole enclosing structure, even if it is unnamed; except that
1485 (iii) a zero-sized bit-field will be disregarded unless it follows
1486 another bit-field of nonzero size. If this hook returns @code{true},
1487 other macros that control bit-field layout are ignored.
1489 When a bit-field is inserted into a packed record, the whole size
1490 of the underlying type is used by one or more same-size adjacent
1491 bit-fields (that is, if its long:3, 32 bits is used in the record,
1492 and any additional adjacent long bit-fields are packed into the same
1493 chunk of 32 bits. However, if the size changes, a new field of that
1494 size is allocated). In an unpacked record, this is the same as using
1495 alignment, but not equivalent when packing.
1497 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1498 the latter will take precedence. If @samp{__attribute__((packed))} is
1499 used on a single field when MS bit-fields are in use, it will take
1500 precedence for that field, but the alignment of the rest of the structure
1501 may affect its placement.
1505 @section Layout of Source Language Data Types
1507 These macros define the sizes and other characteristics of the standard
1508 basic data types used in programs being compiled. Unlike the macros in
1509 the previous section, these apply to specific features of C and related
1510 languages, rather than to fundamental aspects of storage layout.
1512 @defmac INT_TYPE_SIZE
1513 A C expression for the size in bits of the type @code{int} on the
1514 target machine. If you don't define this, the default is one word.
1517 @defmac SHORT_TYPE_SIZE
1518 A C expression for the size in bits of the type @code{short} on the
1519 target machine. If you don't define this, the default is half a word.
1520 (If this would be less than one storage unit, it is rounded up to one
1524 @defmac LONG_TYPE_SIZE
1525 A C expression for the size in bits of the type @code{long} on the
1526 target machine. If you don't define this, the default is one word.
1529 @defmac ADA_LONG_TYPE_SIZE
1530 On some machines, the size used for the Ada equivalent of the type
1531 @code{long} by a native Ada compiler differs from that used by C. In
1532 that situation, define this macro to be a C expression to be used for
1533 the size of that type. If you don't define this, the default is the
1534 value of @code{LONG_TYPE_SIZE}.
1537 @defmac MAX_LONG_TYPE_SIZE
1538 Maximum number for the size in bits of the type @code{long} on the
1539 target machine. If this is undefined, the default is
1540 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1541 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1545 @defmac LONG_LONG_TYPE_SIZE
1546 A C expression for the size in bits of the type @code{long long} on the
1547 target machine. If you don't define this, the default is two
1548 words. If you want to support GNU Ada on your machine, the value of this
1549 macro must be at least 64.
1552 @defmac CHAR_TYPE_SIZE
1553 A C expression for the size in bits of the type @code{char} on the
1554 target machine. If you don't define this, the default is
1555 @code{BITS_PER_UNIT}.
1558 @defmac BOOL_TYPE_SIZE
1559 A C expression for the size in bits of the C++ type @code{bool} and
1560 C99 type @code{_Bool} on the target machine. If you don't define
1561 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1564 @defmac FLOAT_TYPE_SIZE
1565 A C expression for the size in bits of the type @code{float} on the
1566 target machine. If you don't define this, the default is one word.
1569 @defmac DOUBLE_TYPE_SIZE
1570 A C expression for the size in bits of the type @code{double} on the
1571 target machine. If you don't define this, the default is two
1575 @defmac LONG_DOUBLE_TYPE_SIZE
1576 A C expression for the size in bits of the type @code{long double} on
1577 the target machine. If you don't define this, the default is two
1581 @defmac MAX_LONG_DOUBLE_TYPE_SIZE
1582 Maximum number for the size in bits of the type @code{long double} on the
1583 target machine. If this is undefined, the default is
1584 @code{LONG_DOUBLE_TYPE_SIZE}. Otherwise, it is the constant value that is
1585 the largest value that @code{LONG_DOUBLE_TYPE_SIZE} can have at run-time.
1586 This is used in @code{cpp}.
1589 @defmac TARGET_FLT_EVAL_METHOD
1590 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1591 assuming, if applicable, that the floating-point control word is in its
1592 default state. If you do not define this macro the value of
1593 @code{FLT_EVAL_METHOD} will be zero.
1596 @defmac WIDEST_HARDWARE_FP_SIZE
1597 A C expression for the size in bits of the widest floating-point format
1598 supported by the hardware. If you define this macro, you must specify a
1599 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1600 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1604 @defmac DEFAULT_SIGNED_CHAR
1605 An expression whose value is 1 or 0, according to whether the type
1606 @code{char} should be signed or unsigned by default. The user can
1607 always override this default with the options @option{-fsigned-char}
1608 and @option{-funsigned-char}.
1611 @defmac DEFAULT_SHORT_ENUMS
1612 A C expression to determine whether to give an @code{enum} type
1613 only as many bytes as it takes to represent the range of possible values
1614 of that type. A nonzero value means to do that; a zero value means all
1615 @code{enum} types should be allocated like @code{int}.
1617 If you don't define the macro, the default is 0.
1621 A C expression for a string describing the name of the data type to use
1622 for size values. The typedef name @code{size_t} is defined using the
1623 contents of the string.
1625 The string can contain more than one keyword. If so, separate them with
1626 spaces, and write first any length keyword, then @code{unsigned} if
1627 appropriate, and finally @code{int}. The string must exactly match one
1628 of the data type names defined in the function
1629 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1630 omit @code{int} or change the order---that would cause the compiler to
1633 If you don't define this macro, the default is @code{"long unsigned
1637 @defmac PTRDIFF_TYPE
1638 A C expression for a string describing the name of the data type to use
1639 for the result of subtracting two pointers. The typedef name
1640 @code{ptrdiff_t} is defined using the contents of the string. See
1641 @code{SIZE_TYPE} above for more information.
1643 If you don't define this macro, the default is @code{"long int"}.
1647 A C expression for a string describing the name of the data type to use
1648 for wide characters. The typedef name @code{wchar_t} is defined using
1649 the contents of the string. See @code{SIZE_TYPE} above for more
1652 If you don't define this macro, the default is @code{"int"}.
1655 @defmac WCHAR_TYPE_SIZE
1656 A C expression for the size in bits of the data type for wide
1657 characters. This is used in @code{cpp}, which cannot make use of
1661 @defmac MAX_WCHAR_TYPE_SIZE
1662 Maximum number for the size in bits of the data type for wide
1663 characters. If this is undefined, the default is
1664 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1665 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1669 @defmac GCOV_TYPE_SIZE
1670 A C expression for the size in bits of the type used for gcov counters on the
1671 target machine. If you don't define this, the default is one
1672 @code{LONG_TYPE_SIZE} in case it is greater or equal to 64-bit and
1673 @code{LONG_LONG_TYPE_SIZE} otherwise. You may want to re-define the type to
1674 ensure atomicity for counters in multithreaded programs.
1678 A C expression for a string describing the name of the data type to
1679 use for wide characters passed to @code{printf} and returned from
1680 @code{getwc}. The typedef name @code{wint_t} is defined using the
1681 contents of the string. See @code{SIZE_TYPE} above for more
1684 If you don't define this macro, the default is @code{"unsigned int"}.
1688 A C expression for a string describing the name of the data type that
1689 can represent any value of any standard or extended signed integer type.
1690 The typedef name @code{intmax_t} is defined using the contents of the
1691 string. See @code{SIZE_TYPE} above for more information.
1693 If you don't define this macro, the default is the first of
1694 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1695 much precision as @code{long long int}.
1698 @defmac UINTMAX_TYPE
1699 A C expression for a string describing the name of the data type that
1700 can represent any value of any standard or extended unsigned integer
1701 type. The typedef name @code{uintmax_t} is defined using the contents
1702 of the string. See @code{SIZE_TYPE} above for more information.
1704 If you don't define this macro, the default is the first of
1705 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1706 unsigned int"} that has as much precision as @code{long long unsigned
1710 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1711 The C++ compiler represents a pointer-to-member-function with a struct
1718 ptrdiff_t vtable_index;
1725 The C++ compiler must use one bit to indicate whether the function that
1726 will be called through a pointer-to-member-function is virtual.
1727 Normally, we assume that the low-order bit of a function pointer must
1728 always be zero. Then, by ensuring that the vtable_index is odd, we can
1729 distinguish which variant of the union is in use. But, on some
1730 platforms function pointers can be odd, and so this doesn't work. In
1731 that case, we use the low-order bit of the @code{delta} field, and shift
1732 the remainder of the @code{delta} field to the left.
1734 GCC will automatically make the right selection about where to store
1735 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1736 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1737 set such that functions always start at even addresses, but the lowest
1738 bit of pointers to functions indicate whether the function at that
1739 address is in ARM or Thumb mode. If this is the case of your
1740 architecture, you should define this macro to
1741 @code{ptrmemfunc_vbit_in_delta}.
1743 In general, you should not have to define this macro. On architectures
1744 in which function addresses are always even, according to
1745 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1746 @code{ptrmemfunc_vbit_in_pfn}.
1749 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1750 Normally, the C++ compiler uses function pointers in vtables. This
1751 macro allows the target to change to use ``function descriptors''
1752 instead. Function descriptors are found on targets for whom a
1753 function pointer is actually a small data structure. Normally the
1754 data structure consists of the actual code address plus a data
1755 pointer to which the function's data is relative.
1757 If vtables are used, the value of this macro should be the number
1758 of words that the function descriptor occupies.
1761 @defmac TARGET_VTABLE_ENTRY_ALIGN
1762 By default, the vtable entries are void pointers, the so the alignment
1763 is the same as pointer alignment. The value of this macro specifies
1764 the alignment of the vtable entry in bits. It should be defined only
1765 when special alignment is necessary. */
1768 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1769 There are a few non-descriptor entries in the vtable at offsets below
1770 zero. If these entries must be padded (say, to preserve the alignment
1771 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1772 of words in each data entry.
1775 @node Escape Sequences
1776 @section Target Character Escape Sequences
1777 @cindex escape sequences
1779 By default, GCC assumes that the C character escape sequences take on
1780 their ASCII values for the target. If this is not correct, you must
1781 explicitly define all of the macros below. All of them must evaluate
1782 to constants; they are used in @code{case} statements.
1788 @findex TARGET_NEWLINE
1791 @multitable {@code{TARGET_NEWLINE}} {Escape} {ASCII character}
1792 @item Macro @tab Escape @tab ASCII character
1793 @item @code{TARGET_BELL} @tab @kbd{\a} @tab @code{07}, @code{BEL}
1794 @item @code{TARGET_CR} @tab @kbd{\r} @tab @code{0D}, @code{CR}
1795 @item @code{TARGET_ESC} @tab @kbd{\e}, @kbd{\E} @tab @code{1B}, @code{ESC}
1796 @item @code{TARGET_FF} @tab @kbd{\f} @tab @code{0C}, @code{FF}
1797 @item @code{TARGET_NEWLINE} @tab @kbd{\n} @tab @code{0A}, @code{LF}
1798 @item @code{TARGET_TAB} @tab @kbd{\t} @tab @code{09}, @code{HT}
1799 @item @code{TARGET_VT} @tab @kbd{\v} @tab @code{0B}, @code{VT}
1803 Note that the @kbd{\e} and @kbd{\E} escapes are GNU extensions, not
1804 part of the C standard.
1807 @section Register Usage
1808 @cindex register usage
1810 This section explains how to describe what registers the target machine
1811 has, and how (in general) they can be used.
1813 The description of which registers a specific instruction can use is
1814 done with register classes; see @ref{Register Classes}. For information
1815 on using registers to access a stack frame, see @ref{Frame Registers}.
1816 For passing values in registers, see @ref{Register Arguments}.
1817 For returning values in registers, see @ref{Scalar Return}.
1820 * Register Basics:: Number and kinds of registers.
1821 * Allocation Order:: Order in which registers are allocated.
1822 * Values in Registers:: What kinds of values each reg can hold.
1823 * Leaf Functions:: Renumbering registers for leaf functions.
1824 * Stack Registers:: Handling a register stack such as 80387.
1827 @node Register Basics
1828 @subsection Basic Characteristics of Registers
1830 @c prevent bad page break with this line
1831 Registers have various characteristics.
1833 @defmac FIRST_PSEUDO_REGISTER
1834 Number of hardware registers known to the compiler. They receive
1835 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1836 pseudo register's number really is assigned the number
1837 @code{FIRST_PSEUDO_REGISTER}.
1840 @defmac FIXED_REGISTERS
1841 @cindex fixed register
1842 An initializer that says which registers are used for fixed purposes
1843 all throughout the compiled code and are therefore not available for
1844 general allocation. These would include the stack pointer, the frame
1845 pointer (except on machines where that can be used as a general
1846 register when no frame pointer is needed), the program counter on
1847 machines where that is considered one of the addressable registers,
1848 and any other numbered register with a standard use.
1850 This information is expressed as a sequence of numbers, separated by
1851 commas and surrounded by braces. The @var{n}th number is 1 if
1852 register @var{n} is fixed, 0 otherwise.
1854 The table initialized from this macro, and the table initialized by
1855 the following one, may be overridden at run time either automatically,
1856 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1857 the user with the command options @option{-ffixed-@var{reg}},
1858 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1861 @defmac CALL_USED_REGISTERS
1862 @cindex call-used register
1863 @cindex call-clobbered register
1864 @cindex call-saved register
1865 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1866 clobbered (in general) by function calls as well as for fixed
1867 registers. This macro therefore identifies the registers that are not
1868 available for general allocation of values that must live across
1871 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1872 automatically saves it on function entry and restores it on function
1873 exit, if the register is used within the function.
1876 @defmac CALL_REALLY_USED_REGISTERS
1877 @cindex call-used register
1878 @cindex call-clobbered register
1879 @cindex call-saved register
1880 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1881 that the entire set of @code{FIXED_REGISTERS} be included.
1882 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1883 This macro is optional. If not specified, it defaults to the value
1884 of @code{CALL_USED_REGISTERS}.
1887 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1888 @cindex call-used register
1889 @cindex call-clobbered register
1890 @cindex call-saved register
1891 A C expression that is nonzero if it is not permissible to store a
1892 value of mode @var{mode} in hard register number @var{regno} across a
1893 call without some part of it being clobbered. For most machines this
1894 macro need not be defined. It is only required for machines that do not
1895 preserve the entire contents of a register across a call.
1899 @findex call_used_regs
1902 @findex reg_class_contents
1903 @defmac CONDITIONAL_REGISTER_USAGE
1904 Zero or more C statements that may conditionally modify five variables
1905 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1906 @code{reg_names}, and @code{reg_class_contents}, to take into account
1907 any dependence of these register sets on target flags. The first three
1908 of these are of type @code{char []} (interpreted as Boolean vectors).
1909 @code{global_regs} is a @code{const char *[]}, and
1910 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1911 called, @code{fixed_regs}, @code{call_used_regs},
1912 @code{reg_class_contents}, and @code{reg_names} have been initialized
1913 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1914 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1915 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1916 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1917 command options have been applied.
1919 You need not define this macro if it has no work to do.
1921 @cindex disabling certain registers
1922 @cindex controlling register usage
1923 If the usage of an entire class of registers depends on the target
1924 flags, you may indicate this to GCC by using this macro to modify
1925 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1926 registers in the classes which should not be used by GCC@. Also define
1927 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1928 to return @code{NO_REGS} if it
1929 is called with a letter for a class that shouldn't be used.
1931 (However, if this class is not included in @code{GENERAL_REGS} and all
1932 of the insn patterns whose constraints permit this class are
1933 controlled by target switches, then GCC will automatically avoid using
1934 these registers when the target switches are opposed to them.)
1937 @defmac NON_SAVING_SETJMP
1938 If this macro is defined and has a nonzero value, it means that
1939 @code{setjmp} and related functions fail to save the registers, or that
1940 @code{longjmp} fails to restore them. To compensate, the compiler
1941 avoids putting variables in registers in functions that use
1945 @defmac INCOMING_REGNO (@var{out})
1946 Define this macro if the target machine has register windows. This C
1947 expression returns the register number as seen by the called function
1948 corresponding to the register number @var{out} as seen by the calling
1949 function. Return @var{out} if register number @var{out} is not an
1953 @defmac OUTGOING_REGNO (@var{in})
1954 Define this macro if the target machine has register windows. This C
1955 expression returns the register number as seen by the calling function
1956 corresponding to the register number @var{in} as seen by the called
1957 function. Return @var{in} if register number @var{in} is not an inbound
1961 @defmac LOCAL_REGNO (@var{regno})
1962 Define this macro if the target machine has register windows. This C
1963 expression returns true if the register is call-saved but is in the
1964 register window. Unlike most call-saved registers, such registers
1965 need not be explicitly restored on function exit or during non-local
1970 If the program counter has a register number, define this as that
1971 register number. Otherwise, do not define it.
1974 @node Allocation Order
1975 @subsection Order of Allocation of Registers
1976 @cindex order of register allocation
1977 @cindex register allocation order
1979 @c prevent bad page break with this line
1980 Registers are allocated in order.
1982 @defmac REG_ALLOC_ORDER
1983 If defined, an initializer for a vector of integers, containing the
1984 numbers of hard registers in the order in which GCC should prefer
1985 to use them (from most preferred to least).
1987 If this macro is not defined, registers are used lowest numbered first
1988 (all else being equal).
1990 One use of this macro is on machines where the highest numbered
1991 registers must always be saved and the save-multiple-registers
1992 instruction supports only sequences of consecutive registers. On such
1993 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1994 the highest numbered allocable register first.
1997 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
1998 A C statement (sans semicolon) to choose the order in which to allocate
1999 hard registers for pseudo-registers local to a basic block.
2001 Store the desired register order in the array @code{reg_alloc_order}.
2002 Element 0 should be the register to allocate first; element 1, the next
2003 register; and so on.
2005 The macro body should not assume anything about the contents of
2006 @code{reg_alloc_order} before execution of the macro.
2008 On most machines, it is not necessary to define this macro.
2011 @node Values in Registers
2012 @subsection How Values Fit in Registers
2014 This section discusses the macros that describe which kinds of values
2015 (specifically, which machine modes) each register can hold, and how many
2016 consecutive registers are needed for a given mode.
2018 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2019 A C expression for the number of consecutive hard registers, starting
2020 at register number @var{regno}, required to hold a value of mode
2023 On a machine where all registers are exactly one word, a suitable
2024 definition of this macro is
2027 #define HARD_REGNO_NREGS(REGNO, MODE) \
2028 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2033 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2034 A C expression that is nonzero if it is permissible to store a value
2035 of mode @var{mode} in hard register number @var{regno} (or in several
2036 registers starting with that one). For a machine where all registers
2037 are equivalent, a suitable definition is
2040 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2043 You need not include code to check for the numbers of fixed registers,
2044 because the allocation mechanism considers them to be always occupied.
2046 @cindex register pairs
2047 On some machines, double-precision values must be kept in even/odd
2048 register pairs. You can implement that by defining this macro to reject
2049 odd register numbers for such modes.
2051 The minimum requirement for a mode to be OK in a register is that the
2052 @samp{mov@var{mode}} instruction pattern support moves between the
2053 register and other hard register in the same class and that moving a
2054 value into the register and back out not alter it.
2056 Since the same instruction used to move @code{word_mode} will work for
2057 all narrower integer modes, it is not necessary on any machine for
2058 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2059 you define patterns @samp{movhi}, etc., to take advantage of this. This
2060 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2061 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2064 Many machines have special registers for floating point arithmetic.
2065 Often people assume that floating point machine modes are allowed only
2066 in floating point registers. This is not true. Any registers that
2067 can hold integers can safely @emph{hold} a floating point machine
2068 mode, whether or not floating arithmetic can be done on it in those
2069 registers. Integer move instructions can be used to move the values.
2071 On some machines, though, the converse is true: fixed-point machine
2072 modes may not go in floating registers. This is true if the floating
2073 registers normalize any value stored in them, because storing a
2074 non-floating value there would garble it. In this case,
2075 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2076 floating registers. But if the floating registers do not automatically
2077 normalize, if you can store any bit pattern in one and retrieve it
2078 unchanged without a trap, then any machine mode may go in a floating
2079 register, so you can define this macro to say so.
2081 The primary significance of special floating registers is rather that
2082 they are the registers acceptable in floating point arithmetic
2083 instructions. However, this is of no concern to
2084 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2085 constraints for those instructions.
2087 On some machines, the floating registers are especially slow to access,
2088 so that it is better to store a value in a stack frame than in such a
2089 register if floating point arithmetic is not being done. As long as the
2090 floating registers are not in class @code{GENERAL_REGS}, they will not
2091 be used unless some pattern's constraint asks for one.
2094 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2095 A C expression that is nonzero if a value of mode
2096 @var{mode1} is accessible in mode @var{mode2} without copying.
2098 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2099 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2100 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2101 should be nonzero. If they differ for any @var{r}, you should define
2102 this macro to return zero unless some other mechanism ensures the
2103 accessibility of the value in a narrower mode.
2105 You should define this macro to return nonzero in as many cases as
2106 possible since doing so will allow GCC to perform better register
2110 @defmac AVOID_CCMODE_COPIES
2111 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2112 registers. You should only define this macro if support for copying to/from
2113 @code{CCmode} is incomplete.
2116 @node Leaf Functions
2117 @subsection Handling Leaf Functions
2119 @cindex leaf functions
2120 @cindex functions, leaf
2121 On some machines, a leaf function (i.e., one which makes no calls) can run
2122 more efficiently if it does not make its own register window. Often this
2123 means it is required to receive its arguments in the registers where they
2124 are passed by the caller, instead of the registers where they would
2127 The special treatment for leaf functions generally applies only when
2128 other conditions are met; for example, often they may use only those
2129 registers for its own variables and temporaries. We use the term ``leaf
2130 function'' to mean a function that is suitable for this special
2131 handling, so that functions with no calls are not necessarily ``leaf
2134 GCC assigns register numbers before it knows whether the function is
2135 suitable for leaf function treatment. So it needs to renumber the
2136 registers in order to output a leaf function. The following macros
2139 @defmac LEAF_REGISTERS
2140 Name of a char vector, indexed by hard register number, which
2141 contains 1 for a register that is allowable in a candidate for leaf
2144 If leaf function treatment involves renumbering the registers, then the
2145 registers marked here should be the ones before renumbering---those that
2146 GCC would ordinarily allocate. The registers which will actually be
2147 used in the assembler code, after renumbering, should not be marked with 1
2150 Define this macro only if the target machine offers a way to optimize
2151 the treatment of leaf functions.
2154 @defmac LEAF_REG_REMAP (@var{regno})
2155 A C expression whose value is the register number to which @var{regno}
2156 should be renumbered, when a function is treated as a leaf function.
2158 If @var{regno} is a register number which should not appear in a leaf
2159 function before renumbering, then the expression should yield @minus{}1, which
2160 will cause the compiler to abort.
2162 Define this macro only if the target machine offers a way to optimize the
2163 treatment of leaf functions, and registers need to be renumbered to do
2167 @findex current_function_is_leaf
2168 @findex current_function_uses_only_leaf_regs
2169 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2170 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2171 specially. They can test the C variable @code{current_function_is_leaf}
2172 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2173 set prior to local register allocation and is valid for the remaining
2174 compiler passes. They can also test the C variable
2175 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2176 functions which only use leaf registers.
2177 @code{current_function_uses_only_leaf_regs} is valid after reload and is
2178 only useful if @code{LEAF_REGISTERS} is defined.
2179 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2180 @c of the next paragraph?! --mew 2feb93
2182 @node Stack Registers
2183 @subsection Registers That Form a Stack
2185 There are special features to handle computers where some of the
2186 ``registers'' form a stack. Stack registers are normally written by
2187 pushing onto the stack, and are numbered relative to the top of the
2190 Currently, GCC can only handle one group of stack-like registers, and
2191 they must be consecutively numbered. Furthermore, the existing
2192 support for stack-like registers is specific to the 80387 floating
2193 point coprocessor. If you have a new architecture that uses
2194 stack-like registers, you will need to do substantial work on
2195 @file{reg-stack.c} and write your machine description to cooperate
2196 with it, as well as defining these macros.
2199 Define this if the machine has any stack-like registers.
2202 @defmac FIRST_STACK_REG
2203 The number of the first stack-like register. This one is the top
2207 @defmac LAST_STACK_REG
2208 The number of the last stack-like register. This one is the bottom of
2212 @node Register Classes
2213 @section Register Classes
2214 @cindex register class definitions
2215 @cindex class definitions, register
2217 On many machines, the numbered registers are not all equivalent.
2218 For example, certain registers may not be allowed for indexed addressing;
2219 certain registers may not be allowed in some instructions. These machine
2220 restrictions are described to the compiler using @dfn{register classes}.
2222 You define a number of register classes, giving each one a name and saying
2223 which of the registers belong to it. Then you can specify register classes
2224 that are allowed as operands to particular instruction patterns.
2228 In general, each register will belong to several classes. In fact, one
2229 class must be named @code{ALL_REGS} and contain all the registers. Another
2230 class must be named @code{NO_REGS} and contain no registers. Often the
2231 union of two classes will be another class; however, this is not required.
2233 @findex GENERAL_REGS
2234 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2235 terribly special about the name, but the operand constraint letters
2236 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2237 the same as @code{ALL_REGS}, just define it as a macro which expands
2240 Order the classes so that if class @var{x} is contained in class @var{y}
2241 then @var{x} has a lower class number than @var{y}.
2243 The way classes other than @code{GENERAL_REGS} are specified in operand
2244 constraints is through machine-dependent operand constraint letters.
2245 You can define such letters to correspond to various classes, then use
2246 them in operand constraints.
2248 You should define a class for the union of two classes whenever some
2249 instruction allows both classes. For example, if an instruction allows
2250 either a floating point (coprocessor) register or a general register for a
2251 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2252 which includes both of them. Otherwise you will get suboptimal code.
2254 You must also specify certain redundant information about the register
2255 classes: for each class, which classes contain it and which ones are
2256 contained in it; for each pair of classes, the largest class contained
2259 When a value occupying several consecutive registers is expected in a
2260 certain class, all the registers used must belong to that class.
2261 Therefore, register classes cannot be used to enforce a requirement for
2262 a register pair to start with an even-numbered register. The way to
2263 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2265 Register classes used for input-operands of bitwise-and or shift
2266 instructions have a special requirement: each such class must have, for
2267 each fixed-point machine mode, a subclass whose registers can transfer that
2268 mode to or from memory. For example, on some machines, the operations for
2269 single-byte values (@code{QImode}) are limited to certain registers. When
2270 this is so, each register class that is used in a bitwise-and or shift
2271 instruction must have a subclass consisting of registers from which
2272 single-byte values can be loaded or stored. This is so that
2273 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2275 @deftp {Data type} {enum reg_class}
2276 An enumeral type that must be defined with all the register class names
2277 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
2278 must be the last register class, followed by one more enumeral value,
2279 @code{LIM_REG_CLASSES}, which is not a register class but rather
2280 tells how many classes there are.
2282 Each register class has a number, which is the value of casting
2283 the class name to type @code{int}. The number serves as an index
2284 in many of the tables described below.
2287 @defmac N_REG_CLASSES
2288 The number of distinct register classes, defined as follows:
2291 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2295 @defmac REG_CLASS_NAMES
2296 An initializer containing the names of the register classes as C string
2297 constants. These names are used in writing some of the debugging dumps.
2300 @defmac REG_CLASS_CONTENTS
2301 An initializer containing the contents of the register classes, as integers
2302 which are bit masks. The @var{n}th integer specifies the contents of class
2303 @var{n}. The way the integer @var{mask} is interpreted is that
2304 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2306 When the machine has more than 32 registers, an integer does not suffice.
2307 Then the integers are replaced by sub-initializers, braced groupings containing
2308 several integers. Each sub-initializer must be suitable as an initializer
2309 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2310 In this situation, the first integer in each sub-initializer corresponds to
2311 registers 0 through 31, the second integer to registers 32 through 63, and
2315 @defmac REGNO_REG_CLASS (@var{regno})
2316 A C expression whose value is a register class containing hard register
2317 @var{regno}. In general there is more than one such class; choose a class
2318 which is @dfn{minimal}, meaning that no smaller class also contains the
2322 @defmac BASE_REG_CLASS
2323 A macro whose definition is the name of the class to which a valid
2324 base register must belong. A base register is one used in an address
2325 which is the register value plus a displacement.
2328 @defmac MODE_BASE_REG_CLASS (@var{mode})
2329 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2330 the selection of a base register in a mode dependent manner. If
2331 @var{mode} is VOIDmode then it should return the same value as
2332 @code{BASE_REG_CLASS}.
2335 @defmac INDEX_REG_CLASS
2336 A macro whose definition is the name of the class to which a valid
2337 index register must belong. An index register is one used in an
2338 address where its value is either multiplied by a scale factor or
2339 added to another register (as well as added to a displacement).
2342 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2343 For the constraint at the start of @var{str}, which starts with the letter
2344 @var{c}, return the length. This allows you to have register class /
2345 constant / extra constraints that are longer than a single letter;
2346 you don't need to define this macro if you can do with single-letter
2347 constraints only. The definition of this macro should use
2348 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2349 to handle specially.
2350 There are some sanity checks in genoutput.c that check the constraint lengths
2351 for the md file, so you can also use this macro to help you while you are
2352 transitioning from a byzantine single-letter-constraint scheme: when you
2353 return a negative length for a constraint you want to re-use, genoutput
2354 will complain about every instance where it is used in the md file.
2357 @defmac REG_CLASS_FROM_LETTER (@var{char})
2358 A C expression which defines the machine-dependent operand constraint
2359 letters for register classes. If @var{char} is such a letter, the
2360 value should be the register class corresponding to it. Otherwise,
2361 the value should be @code{NO_REGS}. The register letter @samp{r},
2362 corresponding to class @code{GENERAL_REGS}, will not be passed
2363 to this macro; you do not need to handle it.
2366 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2367 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2368 passed in @var{str}, so that you can use suffixes to distinguish between
2372 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2373 A C expression which is nonzero if register number @var{num} is
2374 suitable for use as a base register in operand addresses. It may be
2375 either a suitable hard register or a pseudo register that has been
2376 allocated such a hard register.
2379 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2380 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2381 that expression may examine the mode of the memory reference in
2382 @var{mode}. You should define this macro if the mode of the memory
2383 reference affects whether a register may be used as a base register. If
2384 you define this macro, the compiler will use it instead of
2385 @code{REGNO_OK_FOR_BASE_P}.
2388 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2389 A C expression which is nonzero if register number @var{num} is
2390 suitable for use as an index register in operand addresses. It may be
2391 either a suitable hard register or a pseudo register that has been
2392 allocated such a hard register.
2394 The difference between an index register and a base register is that
2395 the index register may be scaled. If an address involves the sum of
2396 two registers, neither one of them scaled, then either one may be
2397 labeled the ``base'' and the other the ``index''; but whichever
2398 labeling is used must fit the machine's constraints of which registers
2399 may serve in each capacity. The compiler will try both labelings,
2400 looking for one that is valid, and will reload one or both registers
2401 only if neither labeling works.
2404 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2405 A C expression that places additional restrictions on the register class
2406 to use when it is necessary to copy value @var{x} into a register in class
2407 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2408 another, smaller class. On many machines, the following definition is
2412 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2415 Sometimes returning a more restrictive class makes better code. For
2416 example, on the 68000, when @var{x} is an integer constant that is in range
2417 for a @samp{moveq} instruction, the value of this macro is always
2418 @code{DATA_REGS} as long as @var{class} includes the data registers.
2419 Requiring a data register guarantees that a @samp{moveq} will be used.
2421 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2422 you can force @var{x} into a memory constant. This is useful on
2423 certain machines where immediate floating values cannot be loaded into
2424 certain kinds of registers.
2427 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2428 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2429 input reloads. If you don't define this macro, the default is to use
2430 @var{class}, unchanged.
2433 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2434 A C expression that places additional restrictions on the register class
2435 to use when it is necessary to be able to hold a value of mode
2436 @var{mode} in a reload register for which class @var{class} would
2439 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2440 there are certain modes that simply can't go in certain reload classes.
2442 The value is a register class; perhaps @var{class}, or perhaps another,
2445 Don't define this macro unless the target machine has limitations which
2446 require the macro to do something nontrivial.
2449 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2450 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2451 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2452 Many machines have some registers that cannot be copied directly to or
2453 from memory or even from other types of registers. An example is the
2454 @samp{MQ} register, which on most machines, can only be copied to or
2455 from general registers, but not memory. Some machines allow copying all
2456 registers to and from memory, but require a scratch register for stores
2457 to some memory locations (e.g., those with symbolic address on the RT,
2458 and those with certain symbolic address on the SPARC when compiling
2459 PIC)@. In some cases, both an intermediate and a scratch register are
2462 You should define these macros to indicate to the reload phase that it may
2463 need to allocate at least one register for a reload in addition to the
2464 register to contain the data. Specifically, if copying @var{x} to a
2465 register @var{class} in @var{mode} requires an intermediate register,
2466 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2467 largest register class all of whose registers can be used as
2468 intermediate registers or scratch registers.
2470 If copying a register @var{class} in @var{mode} to @var{x} requires an
2471 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2472 should be defined to return the largest register class required. If the
2473 requirements for input and output reloads are the same, the macro
2474 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2477 The values returned by these macros are often @code{GENERAL_REGS}.
2478 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2479 can be directly copied to or from a register of @var{class} in
2480 @var{mode} without requiring a scratch register. Do not define this
2481 macro if it would always return @code{NO_REGS}.
2483 If a scratch register is required (either with or without an
2484 intermediate register), you should define patterns for
2485 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2486 (@pxref{Standard Names}. These patterns, which will normally be
2487 implemented with a @code{define_expand}, should be similar to the
2488 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2491 Define constraints for the reload register and scratch register that
2492 contain a single register class. If the original reload register (whose
2493 class is @var{class}) can meet the constraint given in the pattern, the
2494 value returned by these macros is used for the class of the scratch
2495 register. Otherwise, two additional reload registers are required.
2496 Their classes are obtained from the constraints in the insn pattern.
2498 @var{x} might be a pseudo-register or a @code{subreg} of a
2499 pseudo-register, which could either be in a hard register or in memory.
2500 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2501 in memory and the hard register number if it is in a register.
2503 These macros should not be used in the case where a particular class of
2504 registers can only be copied to memory and not to another class of
2505 registers. In that case, secondary reload registers are not needed and
2506 would not be helpful. Instead, a stack location must be used to perform
2507 the copy and the @code{mov@var{m}} pattern should use memory as an
2508 intermediate storage. This case often occurs between floating-point and
2512 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2513 Certain machines have the property that some registers cannot be copied
2514 to some other registers without using memory. Define this macro on
2515 those machines to be a C expression that is nonzero if objects of mode
2516 @var{m} in registers of @var{class1} can only be copied to registers of
2517 class @var{class2} by storing a register of @var{class1} into memory
2518 and loading that memory location into a register of @var{class2}.
2520 Do not define this macro if its value would always be zero.
2523 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2524 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2525 allocates a stack slot for a memory location needed for register copies.
2526 If this macro is defined, the compiler instead uses the memory location
2527 defined by this macro.
2529 Do not define this macro if you do not define
2530 @code{SECONDARY_MEMORY_NEEDED}.
2533 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2534 When the compiler needs a secondary memory location to copy between two
2535 registers of mode @var{mode}, it normally allocates sufficient memory to
2536 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2537 load operations in a mode that many bits wide and whose class is the
2538 same as that of @var{mode}.
2540 This is right thing to do on most machines because it ensures that all
2541 bits of the register are copied and prevents accesses to the registers
2542 in a narrower mode, which some machines prohibit for floating-point
2545 However, this default behavior is not correct on some machines, such as
2546 the DEC Alpha, that store short integers in floating-point registers
2547 differently than in integer registers. On those machines, the default
2548 widening will not work correctly and you must define this macro to
2549 suppress that widening in some cases. See the file @file{alpha.h} for
2552 Do not define this macro if you do not define
2553 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2554 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2557 @defmac SMALL_REGISTER_CLASSES
2558 On some machines, it is risky to let hard registers live across arbitrary
2559 insns. Typically, these machines have instructions that require values
2560 to be in specific registers (like an accumulator), and reload will fail
2561 if the required hard register is used for another purpose across such an
2564 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2565 value on these machines. When this macro has a nonzero value, the
2566 compiler will try to minimize the lifetime of hard registers.
2568 It is always safe to define this macro with a nonzero value, but if you
2569 unnecessarily define it, you will reduce the amount of optimizations
2570 that can be performed in some cases. If you do not define this macro
2571 with a nonzero value when it is required, the compiler will run out of
2572 spill registers and print a fatal error message. For most machines, you
2573 should not define this macro at all.
2576 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2577 A C expression whose value is nonzero if pseudos that have been assigned
2578 to registers of class @var{class} would likely be spilled because
2579 registers of @var{class} are needed for spill registers.
2581 The default value of this macro returns 1 if @var{class} has exactly one
2582 register and zero otherwise. On most machines, this default should be
2583 used. Only define this macro to some other expression if pseudos
2584 allocated by @file{local-alloc.c} end up in memory because their hard
2585 registers were needed for spill registers. If this macro returns nonzero
2586 for those classes, those pseudos will only be allocated by
2587 @file{global.c}, which knows how to reallocate the pseudo to another
2588 register. If there would not be another register available for
2589 reallocation, you should not change the definition of this macro since
2590 the only effect of such a definition would be to slow down register
2594 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2595 A C expression for the maximum number of consecutive registers
2596 of class @var{class} needed to hold a value of mode @var{mode}.
2598 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2599 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2600 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2601 @var{mode})} for all @var{regno} values in the class @var{class}.
2603 This macro helps control the handling of multiple-word values
2607 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2608 If defined, a C expression that returns nonzero for a @var{class} for which
2609 a change from mode @var{from} to mode @var{to} is invalid.
2611 For the example, loading 32-bit integer or floating-point objects into
2612 floating-point registers on the Alpha extends them to 64 bits.
2613 Therefore loading a 64-bit object and then storing it as a 32-bit object
2614 does not store the low-order 32 bits, as would be the case for a normal
2615 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2619 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2620 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2621 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2625 Three other special macros describe which operands fit which constraint
2628 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2629 A C expression that defines the machine-dependent operand constraint
2630 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2631 particular ranges of integer values. If @var{c} is one of those
2632 letters, the expression should check that @var{value}, an integer, is in
2633 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2634 not one of those letters, the value should be 0 regardless of
2638 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2639 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2640 string passed in @var{str}, so that you can use suffixes to distinguish
2641 between different variants.
2644 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2645 A C expression that defines the machine-dependent operand constraint
2646 letters that specify particular ranges of @code{const_double} values
2647 (@samp{G} or @samp{H}).
2649 If @var{c} is one of those letters, the expression should check that
2650 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2651 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2652 letters, the value should be 0 regardless of @var{value}.
2654 @code{const_double} is used for all floating-point constants and for
2655 @code{DImode} fixed-point constants. A given letter can accept either
2656 or both kinds of values. It can use @code{GET_MODE} to distinguish
2657 between these kinds.
2660 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2661 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2662 string passed in @var{str}, so that you can use suffixes to distinguish
2663 between different variants.
2666 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2667 A C expression that defines the optional machine-dependent constraint
2668 letters that can be used to segregate specific types of operands, usually
2669 memory references, for the target machine. Any letter that is not
2670 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2671 @code{REG_CLASS_FROM_CONSTRAINT}
2672 may be used. Normally this macro will not be defined.
2674 If it is required for a particular target machine, it should return 1
2675 if @var{value} corresponds to the operand type represented by the
2676 constraint letter @var{c}. If @var{c} is not defined as an extra
2677 constraint, the value returned should be 0 regardless of @var{value}.
2679 For example, on the ROMP, load instructions cannot have their output
2680 in r0 if the memory reference contains a symbolic address. Constraint
2681 letter @samp{Q} is defined as representing a memory address that does
2682 @emph{not} contain a symbolic address. An alternative is specified with
2683 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2684 alternative specifies @samp{m} on the input and a register class that
2685 does not include r0 on the output.
2688 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2689 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2690 in @var{str}, so that you can use suffixes to distinguish between different
2694 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2695 A C expression that defines the optional machine-dependent constraint
2696 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2697 be treated like memory constraints by the reload pass.
2699 It should return 1 if the operand type represented by the constraint
2700 at the start of @var{str}, the first letter of which is the letter @var{c},
2701 comprises a subset of all memory references including
2702 all those whose address is simply a base register. This allows the reload
2703 pass to reload an operand, if it does not directly correspond to the operand
2704 type of @var{c}, by copying its address into a base register.
2706 For example, on the S/390, some instructions do not accept arbitrary
2707 memory references, but only those that do not make use of an index
2708 register. The constraint letter @samp{Q} is defined via
2709 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2710 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2711 a @samp{Q} constraint can handle any memory operand, because the
2712 reload pass knows it can be reloaded by copying the memory address
2713 into a base register if required. This is analogous to the way
2714 a @samp{o} constraint can handle any memory operand.
2717 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2718 A C expression that defines the optional machine-dependent constraint
2719 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2720 @code{EXTRA_CONSTRAINT_STR}, that should
2721 be treated like address constraints by the reload pass.
2723 It should return 1 if the operand type represented by the constraint
2724 at the start of @var{str}, which starts with the letter @var{c}, comprises
2725 a subset of all memory addresses including
2726 all those that consist of just a base register. This allows the reload
2727 pass to reload an operand, if it does not directly correspond to the operand
2728 type of @var{str}, by copying it into a base register.
2730 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2731 be used with the @code{address_operand} predicate. It is treated
2732 analogously to the @samp{p} constraint.
2735 @node Stack and Calling
2736 @section Stack Layout and Calling Conventions
2737 @cindex calling conventions
2739 @c prevent bad page break with this line
2740 This describes the stack layout and calling conventions.
2744 * Exception Handling::
2749 * Register Arguments::
2751 * Aggregate Return::
2759 @subsection Basic Stack Layout
2760 @cindex stack frame layout
2761 @cindex frame layout
2763 @c prevent bad page break with this line
2764 Here is the basic stack layout.
2766 @defmac STACK_GROWS_DOWNWARD
2767 Define this macro if pushing a word onto the stack moves the stack
2768 pointer to a smaller address.
2770 When we say, ``define this macro if @dots{},'' it means that the
2771 compiler checks this macro only with @code{#ifdef} so the precise
2772 definition used does not matter.
2775 @defmac STACK_PUSH_CODE
2776 This macro defines the operation used when something is pushed
2777 on the stack. In RTL, a push operation will be
2778 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2780 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2781 and @code{POST_INC}. Which of these is correct depends on
2782 the stack direction and on whether the stack pointer points
2783 to the last item on the stack or whether it points to the
2784 space for the next item on the stack.
2786 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2787 defined, which is almost always right, and @code{PRE_INC} otherwise,
2788 which is often wrong.
2791 @defmac FRAME_GROWS_DOWNWARD
2792 Define this macro if the addresses of local variable slots are at negative
2793 offsets from the frame pointer.
2796 @defmac ARGS_GROW_DOWNWARD
2797 Define this macro if successive arguments to a function occupy decreasing
2798 addresses on the stack.
2801 @defmac STARTING_FRAME_OFFSET
2802 Offset from the frame pointer to the first local variable slot to be allocated.
2804 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2805 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2806 Otherwise, it is found by adding the length of the first slot to the
2807 value @code{STARTING_FRAME_OFFSET}.
2808 @c i'm not sure if the above is still correct.. had to change it to get
2809 @c rid of an overfull. --mew 2feb93
2812 @defmac STACK_ALIGNMENT_NEEDED
2813 Define to zero to disable final alignment of the stack during reload.
2814 The nonzero default for this macro is suitable for most ports.
2816 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2817 is a register save block following the local block that doesn't require
2818 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2819 stack alignment and do it in the backend.
2822 @defmac STACK_POINTER_OFFSET
2823 Offset from the stack pointer register to the first location at which
2824 outgoing arguments are placed. If not specified, the default value of
2825 zero is used. This is the proper value for most machines.
2827 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2828 the first location at which outgoing arguments are placed.
2831 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2832 Offset from the argument pointer register to the first argument's
2833 address. On some machines it may depend on the data type of the
2836 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2837 the first argument's address.
2840 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2841 Offset from the stack pointer register to an item dynamically allocated
2842 on the stack, e.g., by @code{alloca}.
2844 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2845 length of the outgoing arguments. The default is correct for most
2846 machines. See @file{function.c} for details.
2849 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2850 A C expression whose value is RTL representing the address in a stack
2851 frame where the pointer to the caller's frame is stored. Assume that
2852 @var{frameaddr} is an RTL expression for the address of the stack frame
2855 If you don't define this macro, the default is to return the value
2856 of @var{frameaddr}---that is, the stack frame address is also the
2857 address of the stack word that points to the previous frame.
2860 @defmac SETUP_FRAME_ADDRESSES
2861 If defined, a C expression that produces the machine-specific code to
2862 setup the stack so that arbitrary frames can be accessed. For example,
2863 on the SPARC, we must flush all of the register windows to the stack
2864 before we can access arbitrary stack frames. You will seldom need to
2868 @defmac BUILTIN_SETJMP_FRAME_VALUE
2869 If defined, a C expression that contains an rtx that is used to store
2870 the address of the current frame into the built in @code{setjmp} buffer.
2871 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2872 machines. One reason you may need to define this macro is if
2873 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2876 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2877 A C expression whose value is RTL representing the value of the return
2878 address for the frame @var{count} steps up from the current frame, after
2879 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2880 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2881 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2883 The value of the expression must always be the correct address when
2884 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2885 determine the return address of other frames.
2888 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2889 Define this if the return address of a particular stack frame is accessed
2890 from the frame pointer of the previous stack frame.
2893 @defmac INCOMING_RETURN_ADDR_RTX
2894 A C expression whose value is RTL representing the location of the
2895 incoming return address at the beginning of any function, before the
2896 prologue. This RTL is either a @code{REG}, indicating that the return
2897 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2900 You only need to define this macro if you want to support call frame
2901 debugging information like that provided by DWARF 2.
2903 If this RTL is a @code{REG}, you should also define
2904 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2907 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2908 A C expression whose value is an integer giving a DWARF 2 column
2909 number that may be used as an alternate return column. This should
2910 be defined only if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2911 general register, but an alternate column needs to be used for
2915 @defmac INCOMING_FRAME_SP_OFFSET
2916 A C expression whose value is an integer giving the offset, in bytes,
2917 from the value of the stack pointer register to the top of the stack
2918 frame at the beginning of any function, before the prologue. The top of
2919 the frame is defined to be the value of the stack pointer in the
2920 previous frame, just before the call instruction.
2922 You only need to define this macro if you want to support call frame
2923 debugging information like that provided by DWARF 2.
2926 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2927 A C expression whose value is an integer giving the offset, in bytes,
2928 from the argument pointer to the canonical frame address (cfa). The
2929 final value should coincide with that calculated by
2930 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2931 during virtual register instantiation.
2933 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2934 which is correct for most machines; in general, the arguments are found
2935 immediately before the stack frame. Note that this is not the case on
2936 some targets that save registers into the caller's frame, such as SPARC
2937 and rs6000, and so such targets need to define this macro.
2939 You only need to define this macro if the default is incorrect, and you
2940 want to support call frame debugging information like that provided by
2944 @node Exception Handling
2945 @subsection Exception Handling Support
2946 @cindex exception handling
2948 @defmac EH_RETURN_DATA_REGNO (@var{N})
2949 A C expression whose value is the @var{N}th register number used for
2950 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2951 @var{N} registers are usable.
2953 The exception handling library routines communicate with the exception
2954 handlers via a set of agreed upon registers. Ideally these registers
2955 should be call-clobbered; it is possible to use call-saved registers,
2956 but may negatively impact code size. The target must support at least
2957 2 data registers, but should define 4 if there are enough free registers.
2959 You must define this macro if you want to support call frame exception
2960 handling like that provided by DWARF 2.
2963 @defmac EH_RETURN_STACKADJ_RTX
2964 A C expression whose value is RTL representing a location in which
2965 to store a stack adjustment to be applied before function return.
2966 This is used to unwind the stack to an exception handler's call frame.
2967 It will be assigned zero on code paths that return normally.
2969 Typically this is a call-clobbered hard register that is otherwise
2970 untouched by the epilogue, but could also be a stack slot.
2972 Do not define this macro if the stack pointer is saved and restored
2973 by the regular prolog and epilog code in the call frame itself; in
2974 this case, the exception handling library routines will update the
2975 stack location to be restored in place. Otherwise, you must define
2976 this macro if you want to support call frame exception handling like
2977 that provided by DWARF 2.
2980 @defmac EH_RETURN_HANDLER_RTX
2981 A C expression whose value is RTL representing a location in which
2982 to store the address of an exception handler to which we should
2983 return. It will not be assigned on code paths that return normally.
2985 Typically this is the location in the call frame at which the normal
2986 return address is stored. For targets that return by popping an
2987 address off the stack, this might be a memory address just below
2988 the @emph{target} call frame rather than inside the current call
2989 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
2990 been assigned, so it may be used to calculate the location of the
2993 Some targets have more complex requirements than storing to an
2994 address calculable during initial code generation. In that case
2995 the @code{eh_return} instruction pattern should be used instead.
2997 If you want to support call frame exception handling, you must
2998 define either this macro or the @code{eh_return} instruction pattern.
3001 @defmac RETURN_ADDR_OFFSET
3002 If defined, an integer-valued C expression for which rtl will be generated
3003 to add it to the exception handler address before it is searched in the
3004 exception handling tables, and to subtract it again from the address before
3005 using it to return to the exception handler.
3008 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3009 This macro chooses the encoding of pointers embedded in the exception
3010 handling sections. If at all possible, this should be defined such
3011 that the exception handling section will not require dynamic relocations,
3012 and so may be read-only.
3014 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3015 @var{global} is true if the symbol may be affected by dynamic relocations.
3016 The macro should return a combination of the @code{DW_EH_PE_*} defines
3017 as found in @file{dwarf2.h}.
3019 If this macro is not defined, pointers will not be encoded but
3020 represented directly.
3023 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3024 This macro allows the target to emit whatever special magic is required
3025 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3026 Generic code takes care of pc-relative and indirect encodings; this must
3027 be defined if the target uses text-relative or data-relative encodings.
3029 This is a C statement that branches to @var{done} if the format was
3030 handled. @var{encoding} is the format chosen, @var{size} is the number
3031 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3035 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs}, @var{success})
3036 This macro allows the target to add cpu and operating system specific
3037 code to the call-frame unwinder for use when there is no unwind data
3038 available. The most common reason to implement this macro is to unwind
3039 through signal frames.
3041 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3042 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3043 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3044 for the address of the code being executed and @code{context->cfa} for
3045 the stack pointer value. If the frame can be decoded, the register save
3046 addresses should be updated in @var{fs} and the macro should branch to
3047 @var{success}. If the frame cannot be decoded, the macro should do
3050 For proper signal handling in Java this macro is accompanied by
3051 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3054 @node Stack Checking
3055 @subsection Specifying How Stack Checking is Done
3057 GCC will check that stack references are within the boundaries of
3058 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3062 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3063 will assume that you have arranged for stack checking to be done at
3064 appropriate places in the configuration files, e.g., in
3065 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3069 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3070 called @code{check_stack} in your @file{md} file, GCC will call that
3071 pattern with one argument which is the address to compare the stack
3072 value against. You must arrange for this pattern to report an error if
3073 the stack pointer is out of range.
3076 If neither of the above are true, GCC will generate code to periodically
3077 ``probe'' the stack pointer using the values of the macros defined below.
3080 Normally, you will use the default values of these macros, so GCC
3081 will use the third approach.
3083 @defmac STACK_CHECK_BUILTIN
3084 A nonzero value if stack checking is done by the configuration files in a
3085 machine-dependent manner. You should define this macro if stack checking
3086 is require by the ABI of your machine or if you would like to have to stack
3087 checking in some more efficient way than GCC's portable approach.
3088 The default value of this macro is zero.
3091 @defmac STACK_CHECK_PROBE_INTERVAL
3092 An integer representing the interval at which GCC must generate stack
3093 probe instructions. You will normally define this macro to be no larger
3094 than the size of the ``guard pages'' at the end of a stack area. The
3095 default value of 4096 is suitable for most systems.
3098 @defmac STACK_CHECK_PROBE_LOAD
3099 A integer which is nonzero if GCC should perform the stack probe
3100 as a load instruction and zero if GCC should use a store instruction.
3101 The default is zero, which is the most efficient choice on most systems.
3104 @defmac STACK_CHECK_PROTECT
3105 The number of bytes of stack needed to recover from a stack overflow,
3106 for languages where such a recovery is supported. The default value of
3107 75 words should be adequate for most machines.
3110 @defmac STACK_CHECK_MAX_FRAME_SIZE
3111 The maximum size of a stack frame, in bytes. GCC will generate probe
3112 instructions in non-leaf functions to ensure at least this many bytes of
3113 stack are available. If a stack frame is larger than this size, stack
3114 checking will not be reliable and GCC will issue a warning. The
3115 default is chosen so that GCC only generates one instruction on most
3116 systems. You should normally not change the default value of this macro.
3119 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3120 GCC uses this value to generate the above warning message. It
3121 represents the amount of fixed frame used by a function, not including
3122 space for any callee-saved registers, temporaries and user variables.
3123 You need only specify an upper bound for this amount and will normally
3124 use the default of four words.
3127 @defmac STACK_CHECK_MAX_VAR_SIZE
3128 The maximum size, in bytes, of an object that GCC will place in the
3129 fixed area of the stack frame when the user specifies
3130 @option{-fstack-check}.
3131 GCC computed the default from the values of the above macros and you will
3132 normally not need to override that default.
3136 @node Frame Registers
3137 @subsection Registers That Address the Stack Frame
3139 @c prevent bad page break with this line
3140 This discusses registers that address the stack frame.
3142 @defmac STACK_POINTER_REGNUM
3143 The register number of the stack pointer register, which must also be a
3144 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3145 the hardware determines which register this is.
3148 @defmac FRAME_POINTER_REGNUM
3149 The register number of the frame pointer register, which is used to
3150 access automatic variables in the stack frame. On some machines, the
3151 hardware determines which register this is. On other machines, you can
3152 choose any register you wish for this purpose.
3155 @defmac HARD_FRAME_POINTER_REGNUM
3156 On some machines the offset between the frame pointer and starting
3157 offset of the automatic variables is not known until after register
3158 allocation has been done (for example, because the saved registers are
3159 between these two locations). On those machines, define
3160 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3161 be used internally until the offset is known, and define
3162 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3163 used for the frame pointer.
3165 You should define this macro only in the very rare circumstances when it
3166 is not possible to calculate the offset between the frame pointer and
3167 the automatic variables until after register allocation has been
3168 completed. When this macro is defined, you must also indicate in your
3169 definition of @code{ELIMINABLE_REGS} how to eliminate
3170 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3171 or @code{STACK_POINTER_REGNUM}.
3173 Do not define this macro if it would be the same as
3174 @code{FRAME_POINTER_REGNUM}.
3177 @defmac ARG_POINTER_REGNUM
3178 The register number of the arg pointer register, which is used to access
3179 the function's argument list. On some machines, this is the same as the
3180 frame pointer register. On some machines, the hardware determines which
3181 register this is. On other machines, you can choose any register you
3182 wish for this purpose. If this is not the same register as the frame
3183 pointer register, then you must mark it as a fixed register according to
3184 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3185 (@pxref{Elimination}).
3188 @defmac RETURN_ADDRESS_POINTER_REGNUM
3189 The register number of the return address pointer register, which is used to
3190 access the current function's return address from the stack. On some
3191 machines, the return address is not at a fixed offset from the frame
3192 pointer or stack pointer or argument pointer. This register can be defined
3193 to point to the return address on the stack, and then be converted by
3194 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3196 Do not define this macro unless there is no other way to get the return
3197 address from the stack.
3200 @defmac STATIC_CHAIN_REGNUM
3201 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3202 Register numbers used for passing a function's static chain pointer. If
3203 register windows are used, the register number as seen by the called
3204 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3205 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3206 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3209 The static chain register need not be a fixed register.
3211 If the static chain is passed in memory, these macros should not be
3212 defined; instead, the next two macros should be defined.
3215 @defmac STATIC_CHAIN
3216 @defmacx STATIC_CHAIN_INCOMING
3217 If the static chain is passed in memory, these macros provide rtx giving
3218 @code{mem} expressions that denote where they are stored.
3219 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3220 as seen by the calling and called functions, respectively. Often the former
3221 will be at an offset from the stack pointer and the latter at an offset from
3224 @findex stack_pointer_rtx
3225 @findex frame_pointer_rtx
3226 @findex arg_pointer_rtx
3227 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3228 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3229 macros and should be used to refer to those items.
3231 If the static chain is passed in a register, the two previous macros should
3235 @defmac DWARF_FRAME_REGISTERS
3236 This macro specifies the maximum number of hard registers that can be
3237 saved in a call frame. This is used to size data structures used in
3238 DWARF2 exception handling.
3240 Prior to GCC 3.0, this macro was needed in order to establish a stable
3241 exception handling ABI in the face of adding new hard registers for ISA
3242 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3243 in the number of hard registers. Nevertheless, this macro can still be
3244 used to reduce the runtime memory requirements of the exception handling
3245 routines, which can be substantial if the ISA contains a lot of
3246 registers that are not call-saved.
3248 If this macro is not defined, it defaults to
3249 @code{FIRST_PSEUDO_REGISTER}.
3252 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3254 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3255 for backward compatibility in pre GCC 3.0 compiled code.
3257 If this macro is not defined, it defaults to
3258 @code{DWARF_FRAME_REGISTERS}.
3261 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3263 Define this macro if the target's representation for dwarf registers
3264 is different than the internal representation for unwind column.
3265 Given a dwarf register, this macro should return the internal unwind
3266 column number to use instead.
3268 See the PowerPC's SPE target for an example.
3272 @subsection Eliminating Frame Pointer and Arg Pointer
3274 @c prevent bad page break with this line
3275 This is about eliminating the frame pointer and arg pointer.
3277 @defmac FRAME_POINTER_REQUIRED
3278 A C expression which is nonzero if a function must have and use a frame
3279 pointer. This expression is evaluated in the reload pass. If its value is
3280 nonzero the function will have a frame pointer.
3282 The expression can in principle examine the current function and decide
3283 according to the facts, but on most machines the constant 0 or the
3284 constant 1 suffices. Use 0 when the machine allows code to be generated
3285 with no frame pointer, and doing so saves some time or space. Use 1
3286 when there is no possible advantage to avoiding a frame pointer.
3288 In certain cases, the compiler does not know how to produce valid code
3289 without a frame pointer. The compiler recognizes those cases and
3290 automatically gives the function a frame pointer regardless of what
3291 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3294 In a function that does not require a frame pointer, the frame pointer
3295 register can be allocated for ordinary usage, unless you mark it as a
3296 fixed register. See @code{FIXED_REGISTERS} for more information.
3299 @findex get_frame_size
3300 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3301 A C statement to store in the variable @var{depth-var} the difference
3302 between the frame pointer and the stack pointer values immediately after
3303 the function prologue. The value would be computed from information
3304 such as the result of @code{get_frame_size ()} and the tables of
3305 registers @code{regs_ever_live} and @code{call_used_regs}.
3307 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3308 need not be defined. Otherwise, it must be defined even if
3309 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3310 case, you may set @var{depth-var} to anything.
3313 @defmac ELIMINABLE_REGS
3314 If defined, this macro specifies a table of register pairs used to
3315 eliminate unneeded registers that point into the stack frame. If it is not
3316 defined, the only elimination attempted by the compiler is to replace
3317 references to the frame pointer with references to the stack pointer.
3319 The definition of this macro is a list of structure initializations, each
3320 of which specifies an original and replacement register.
3322 On some machines, the position of the argument pointer is not known until
3323 the compilation is completed. In such a case, a separate hard register
3324 must be used for the argument pointer. This register can be eliminated by
3325 replacing it with either the frame pointer or the argument pointer,
3326 depending on whether or not the frame pointer has been eliminated.
3328 In this case, you might specify:
3330 #define ELIMINABLE_REGS \
3331 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3332 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3333 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3336 Note that the elimination of the argument pointer with the stack pointer is
3337 specified first since that is the preferred elimination.
3340 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3341 A C expression that returns nonzero if the compiler is allowed to try
3342 to replace register number @var{from-reg} with register number
3343 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3344 is defined, and will usually be the constant 1, since most of the cases
3345 preventing register elimination are things that the compiler already
3349 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3350 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3351 specifies the initial difference between the specified pair of
3352 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3356 @node Stack Arguments
3357 @subsection Passing Function Arguments on the Stack
3358 @cindex arguments on stack
3359 @cindex stack arguments
3361 The macros in this section control how arguments are passed
3362 on the stack. See the following section for other macros that
3363 control passing certain arguments in registers.
3365 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3366 This target hook returns @code{true} if an argument declared in a
3367 prototype as an integral type smaller than @code{int} should actually be
3368 passed as an @code{int}. In addition to avoiding errors in certain
3369 cases of mismatch, it also makes for better code on certain machines.
3370 The default is to not promote prototypes.
3374 A C expression. If nonzero, push insns will be used to pass
3376 If the target machine does not have a push instruction, set it to zero.
3377 That directs GCC to use an alternate strategy: to
3378 allocate the entire argument block and then store the arguments into
3379 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3382 @defmac PUSH_ARGS_REVERSED
3383 A C expression. If nonzero, function arguments will be evaluated from
3384 last to first, rather than from first to last. If this macro is not
3385 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3386 and args grow in opposite directions, and 0 otherwise.
3389 @defmac PUSH_ROUNDING (@var{npushed})
3390 A C expression that is the number of bytes actually pushed onto the
3391 stack when an instruction attempts to push @var{npushed} bytes.
3393 On some machines, the definition
3396 #define PUSH_ROUNDING(BYTES) (BYTES)
3400 will suffice. But on other machines, instructions that appear
3401 to push one byte actually push two bytes in an attempt to maintain
3402 alignment. Then the definition should be
3405 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3409 @findex current_function_outgoing_args_size
3410 @defmac ACCUMULATE_OUTGOING_ARGS
3411 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3412 will be computed and placed into the variable
3413 @code{current_function_outgoing_args_size}. No space will be pushed
3414 onto the stack for each call; instead, the function prologue should
3415 increase the stack frame size by this amount.
3417 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3421 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3422 Define this macro if functions should assume that stack space has been
3423 allocated for arguments even when their values are passed in
3426 The value of this macro is the size, in bytes, of the area reserved for
3427 arguments passed in registers for the function represented by @var{fndecl},
3428 which can be zero if GCC is calling a library function.
3430 This space can be allocated by the caller, or be a part of the
3431 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3434 @c above is overfull. not sure what to do. --mew 5feb93 did
3435 @c something, not sure if it looks good. --mew 10feb93
3437 @defmac MAYBE_REG_PARM_STACK_SPACE
3438 @defmacx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
3439 Define these macros in addition to the one above if functions might
3440 allocate stack space for arguments even when their values are passed
3441 in registers. These should be used when the stack space allocated
3442 for arguments in registers is not a simple constant independent of the
3443 function declaration.
3445 The value of the first macro is the size, in bytes, of the area that
3446 we should initially assume would be reserved for arguments passed in registers.
3448 The value of the second macro is the actual size, in bytes, of the area
3449 that will be reserved for arguments passed in registers. This takes two
3450 arguments: an integer representing the number of bytes of fixed sized
3451 arguments on the stack, and a tree representing the number of bytes of
3452 variable sized arguments on the stack.
3454 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
3455 called for libcall functions, the current function, or for a function
3456 being called when it is known that such stack space must be allocated.
3457 In each case this value can be easily computed.
3459 When deciding whether a called function needs such stack space, and how
3460 much space to reserve, GCC uses these two macros instead of
3461 @code{REG_PARM_STACK_SPACE}.
3464 @defmac OUTGOING_REG_PARM_STACK_SPACE
3465 Define this if it is the responsibility of the caller to allocate the area
3466 reserved for arguments passed in registers.
3468 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3469 whether the space for these arguments counts in the value of
3470 @code{current_function_outgoing_args_size}.
3473 @defmac STACK_PARMS_IN_REG_PARM_AREA
3474 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3475 stack parameters don't skip the area specified by it.
3476 @c i changed this, makes more sens and it should have taken care of the
3477 @c overfull.. not as specific, tho. --mew 5feb93
3479 Normally, when a parameter is not passed in registers, it is placed on the
3480 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3481 suppresses this behavior and causes the parameter to be passed on the
3482 stack in its natural location.
3485 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3486 A C expression that should indicate the number of bytes of its own
3487 arguments that a function pops on returning, or 0 if the
3488 function pops no arguments and the caller must therefore pop them all
3489 after the function returns.
3491 @var{fundecl} is a C variable whose value is a tree node that describes
3492 the function in question. Normally it is a node of type
3493 @code{FUNCTION_DECL} that describes the declaration of the function.
3494 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3496 @var{funtype} is a C variable whose value is a tree node that
3497 describes the function in question. Normally it is a node of type
3498 @code{FUNCTION_TYPE} that describes the data type of the function.
3499 From this it is possible to obtain the data types of the value and
3500 arguments (if known).
3502 When a call to a library function is being considered, @var{fundecl}
3503 will contain an identifier node for the library function. Thus, if
3504 you need to distinguish among various library functions, you can do so
3505 by their names. Note that ``library function'' in this context means
3506 a function used to perform arithmetic, whose name is known specially
3507 in the compiler and was not mentioned in the C code being compiled.
3509 @var{stack-size} is the number of bytes of arguments passed on the
3510 stack. If a variable number of bytes is passed, it is zero, and
3511 argument popping will always be the responsibility of the calling function.
3513 On the VAX, all functions always pop their arguments, so the definition
3514 of this macro is @var{stack-size}. On the 68000, using the standard
3515 calling convention, no functions pop their arguments, so the value of
3516 the macro is always 0 in this case. But an alternative calling
3517 convention is available in which functions that take a fixed number of
3518 arguments pop them but other functions (such as @code{printf}) pop
3519 nothing (the caller pops all). When this convention is in use,
3520 @var{funtype} is examined to determine whether a function takes a fixed
3521 number of arguments.
3524 @defmac CALL_POPS_ARGS (@var{cum})
3525 A C expression that should indicate the number of bytes a call sequence
3526 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3527 when compiling a function call.
3529 @var{cum} is the variable in which all arguments to the called function
3530 have been accumulated.
3532 On certain architectures, such as the SH5, a call trampoline is used
3533 that pops certain registers off the stack, depending on the arguments
3534 that have been passed to the function. Since this is a property of the
3535 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3539 @node Register Arguments
3540 @subsection Passing Arguments in Registers
3541 @cindex arguments in registers
3542 @cindex registers arguments
3544 This section describes the macros which let you control how various
3545 types of arguments are passed in registers or how they are arranged in
3548 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3549 A C expression that controls whether a function argument is passed
3550 in a register, and which register.
3552 The arguments are @var{cum}, which summarizes all the previous
3553 arguments; @var{mode}, the machine mode of the argument; @var{type},
3554 the data type of the argument as a tree node or 0 if that is not known
3555 (which happens for C support library functions); and @var{named},
3556 which is 1 for an ordinary argument and 0 for nameless arguments that
3557 correspond to @samp{@dots{}} in the called function's prototype.
3558 @var{type} can be an incomplete type if a syntax error has previously
3561 The value of the expression is usually either a @code{reg} RTX for the
3562 hard register in which to pass the argument, or zero to pass the
3563 argument on the stack.
3565 For machines like the VAX and 68000, where normally all arguments are
3566 pushed, zero suffices as a definition.
3568 The value of the expression can also be a @code{parallel} RTX@. This is
3569 used when an argument is passed in multiple locations. The mode of the
3570 @code{parallel} should be the mode of the entire argument. The
3571 @code{parallel} holds any number of @code{expr_list} pairs; each one
3572 describes where part of the argument is passed. In each
3573 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3574 register in which to pass this part of the argument, and the mode of the
3575 register RTX indicates how large this part of the argument is. The
3576 second operand of the @code{expr_list} is a @code{const_int} which gives
3577 the offset in bytes into the entire argument of where this part starts.
3578 As a special exception the first @code{expr_list} in the @code{parallel}
3579 RTX may have a first operand of zero. This indicates that the entire
3580 argument is also stored on the stack.
3582 The last time this macro is called, it is called with @code{MODE ==
3583 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3584 pattern as operands 2 and 3 respectively.
3586 @cindex @file{stdarg.h} and register arguments
3587 The usual way to make the ISO library @file{stdarg.h} work on a machine
3588 where some arguments are usually passed in registers, is to cause
3589 nameless arguments to be passed on the stack instead. This is done
3590 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3592 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3593 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3594 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3595 in the definition of this macro to determine if this argument is of a
3596 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3597 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3598 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3599 defined, the argument will be computed in the stack and then loaded into
3603 @defmac MUST_PASS_IN_STACK (@var{mode}, @var{type})
3604 Define as a C expression that evaluates to nonzero if we do not know how
3605 to pass TYPE solely in registers. The file @file{expr.h} defines a
3606 definition that is usually appropriate, refer to @file{expr.h} for additional
3610 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3611 Define this macro if the target machine has ``register windows'', so
3612 that the register in which a function sees an arguments is not
3613 necessarily the same as the one in which the caller passed the
3616 For such machines, @code{FUNCTION_ARG} computes the register in which
3617 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3618 be defined in a similar fashion to tell the function being called
3619 where the arguments will arrive.
3621 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3622 serves both purposes.
3625 @defmac FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3626 A C expression for the number of words, at the beginning of an
3627 argument, that must be put in registers. The value must be zero for
3628 arguments that are passed entirely in registers or that are entirely
3629 pushed on the stack.
3631 On some machines, certain arguments must be passed partially in
3632 registers and partially in memory. On these machines, typically the
3633 first @var{n} words of arguments are passed in registers, and the rest
3634 on the stack. If a multi-word argument (a @code{double} or a
3635 structure) crosses that boundary, its first few words must be passed
3636 in registers and the rest must be pushed. This macro tells the
3637 compiler when this occurs, and how many of the words should go in
3640 @code{FUNCTION_ARG} for these arguments should return the first
3641 register to be used by the caller for this argument; likewise
3642 @code{FUNCTION_INCOMING_ARG}, for the called function.
3645 @defmac FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3646 A C expression that indicates when an argument must be passed by reference.
3647 If nonzero for an argument, a copy of that argument is made in memory and a
3648 pointer to the argument is passed instead of the argument itself.
3649 The pointer is passed in whatever way is appropriate for passing a pointer
3652 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3653 definition of this macro might be
3655 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3656 (CUM, MODE, TYPE, NAMED) \
3657 MUST_PASS_IN_STACK (MODE, TYPE)
3659 @c this is *still* too long. --mew 5feb93
3662 @defmac FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3663 If defined, a C expression that indicates when it is the called function's
3664 responsibility to make a copy of arguments passed by invisible reference.
3665 Normally, the caller makes a copy and passes the address of the copy to the
3666 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3667 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3668 ``live'' value. The called function must not modify this value. If it can be
3669 determined that the value won't be modified, it need not make a copy;
3670 otherwise a copy must be made.
3673 @defmac CUMULATIVE_ARGS
3674 A C type for declaring a variable that is used as the first argument of
3675 @code{FUNCTION_ARG} and other related values. For some target machines,
3676 the type @code{int} suffices and can hold the number of bytes of
3679 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3680 arguments that have been passed on the stack. The compiler has other
3681 variables to keep track of that. For target machines on which all
3682 arguments are passed on the stack, there is no need to store anything in
3683 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3684 should not be empty, so use @code{int}.
3687 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl})
3688 A C statement (sans semicolon) for initializing the variable
3689 @var{cum} for the state at the beginning of the argument list. The
3690 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3691 is the tree node for the data type of the function which will receive
3692 the args, or 0 if the args are to a compiler support library function.
3693 For direct calls that are not libcalls, @var{fndecl} contain the
3694 declaration node of the function. @var{fndecl} is also set when
3695 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3698 When processing a call to a compiler support library function,
3699 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3700 contains the name of the function, as a string. @var{libname} is 0 when
3701 an ordinary C function call is being processed. Thus, each time this
3702 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3703 never both of them at once.
3706 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3707 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3708 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3709 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3710 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3711 0)} is used instead.
3714 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3715 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3716 finding the arguments for the function being compiled. If this macro is
3717 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3719 The value passed for @var{libname} is always 0, since library routines
3720 with special calling conventions are never compiled with GCC@. The
3721 argument @var{libname} exists for symmetry with
3722 @code{INIT_CUMULATIVE_ARGS}.
3723 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3724 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3727 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3728 A C statement (sans semicolon) to update the summarizer variable
3729 @var{cum} to advance past an argument in the argument list. The
3730 values @var{mode}, @var{type} and @var{named} describe that argument.
3731 Once this is done, the variable @var{cum} is suitable for analyzing
3732 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3734 This macro need not do anything if the argument in question was passed
3735 on the stack. The compiler knows how to track the amount of stack space
3736 used for arguments without any special help.
3739 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3740 If defined, a C expression which determines whether, and in which direction,
3741 to pad out an argument with extra space. The value should be of type
3742 @code{enum direction}: either @code{upward} to pad above the argument,
3743 @code{downward} to pad below, or @code{none} to inhibit padding.
3745 The @emph{amount} of padding is always just enough to reach the next
3746 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3749 This macro has a default definition which is right for most systems.
3750 For little-endian machines, the default is to pad upward. For
3751 big-endian machines, the default is to pad downward for an argument of
3752 constant size shorter than an @code{int}, and upward otherwise.
3755 @defmac PAD_VARARGS_DOWN
3756 If defined, a C expression which determines whether the default
3757 implementation of va_arg will attempt to pad down before reading the
3758 next argument, if that argument is smaller than its aligned space as
3759 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3760 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3763 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3764 Specify padding for the last element of a block move between registers and
3765 memory. @var{first} is nonzero if this is the only element. Defining this
3766 macro allows better control of register function parameters on big-endian
3767 machines, without using @code{PARALLEL} rtl. In particular,
3768 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3769 registers, as there is no longer a "wrong" part of a register; For example,
3770 a three byte aggregate may be passed in the high part of a register if so
3774 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3775 If defined, a C expression that gives the alignment boundary, in bits,
3776 of an argument with the specified mode and type. If it is not defined,
3777 @code{PARM_BOUNDARY} is used for all arguments.
3780 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3781 A C expression that is nonzero if @var{regno} is the number of a hard
3782 register in which function arguments are sometimes passed. This does
3783 @emph{not} include implicit arguments such as the static chain and
3784 the structure-value address. On many machines, no registers can be
3785 used for this purpose since all function arguments are pushed on the
3789 @defmac SPLIT_COMPLEX_ARGS
3791 Define this macro to a nonzero value if complex function arguments
3792 should be split into their corresponding components. By default, GCC
3793 will attempt to pack complex arguments into the target's word size.
3794 Some ABIs require complex arguments to be split and treated as their
3795 individual components. For example, on AIX64, complex floats should
3796 be passed in a pair of floating point registers, even though a complex
3797 float would fit in one 64-bit floating point register.
3800 @defmac LOAD_ARGS_REVERSED
3801 If defined, the order in which arguments are loaded into their
3802 respective argument registers is reversed so that the last
3803 argument is loaded first. This macro only affects arguments
3804 passed in registers.
3808 @subsection How Scalar Function Values Are Returned
3809 @cindex return values in registers
3810 @cindex values, returned by functions
3811 @cindex scalars, returned as values
3813 This section discusses the macros that control returning scalars as
3814 values---values that can fit in registers.
3816 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3817 A C expression to create an RTX representing the place where a
3818 function returns a value of data type @var{valtype}. @var{valtype} is
3819 a tree node representing a data type. Write @code{TYPE_MODE
3820 (@var{valtype})} to get the machine mode used to represent that type.
3821 On many machines, only the mode is relevant. (Actually, on most
3822 machines, scalar values are returned in the same place regardless of
3825 The value of the expression is usually a @code{reg} RTX for the hard
3826 register where the return value is stored. The value can also be a
3827 @code{parallel} RTX, if the return value is in multiple places. See
3828 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3830 If @code{TARGET_PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3831 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3834 If the precise function being called is known, @var{func} is a tree
3835 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3836 pointer. This makes it possible to use a different value-returning
3837 convention for specific functions when all their calls are
3840 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3841 types, because these are returned in another way. See
3842 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3845 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3846 Define this macro if the target machine has ``register windows''
3847 so that the register in which a function returns its value is not
3848 the same as the one in which the caller sees the value.
3850 For such machines, @code{FUNCTION_VALUE} computes the register in which
3851 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3852 defined in a similar fashion to tell the function where to put the
3855 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3856 @code{FUNCTION_VALUE} serves both purposes.
3858 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3859 aggregate data types, because these are returned in another way. See
3860 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3863 @defmac LIBCALL_VALUE (@var{mode})
3864 A C expression to create an RTX representing the place where a library
3865 function returns a value of mode @var{mode}. If the precise function
3866 being called is known, @var{func} is a tree node
3867 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3868 pointer. This makes it possible to use a different value-returning
3869 convention for specific functions when all their calls are
3872 Note that ``library function'' in this context means a compiler
3873 support routine, used to perform arithmetic, whose name is known
3874 specially by the compiler and was not mentioned in the C code being
3877 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3878 data types, because none of the library functions returns such types.
3881 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3882 A C expression that is nonzero if @var{regno} is the number of a hard
3883 register in which the values of called function may come back.
3885 A register whose use for returning values is limited to serving as the
3886 second of a pair (for a value of type @code{double}, say) need not be
3887 recognized by this macro. So for most machines, this definition
3891 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3894 If the machine has register windows, so that the caller and the called
3895 function use different registers for the return value, this macro
3896 should recognize only the caller's register numbers.
3899 @defmac APPLY_RESULT_SIZE
3900 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3901 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3902 saving and restoring an arbitrary return value.
3905 @node Aggregate Return
3906 @subsection How Large Values Are Returned
3907 @cindex aggregates as return values
3908 @cindex large return values
3909 @cindex returning aggregate values
3910 @cindex structure value address
3912 When a function value's mode is @code{BLKmode} (and in some other
3913 cases), the value is not returned according to @code{FUNCTION_VALUE}
3914 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3915 block of memory in which the value should be stored. This address
3916 is called the @dfn{structure value address}.
3918 This section describes how to control returning structure values in
3921 @deftypefn {Target Hook} bool RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
3922 This target hook should return a nonzero value to say to return the
3923 function value in memory, just as large structures are always returned.
3924 Here @var{type} will be the data type of the value, and @var{fntype}
3925 will be the type of the function doing the returning, or @code{NULL} for
3928 Note that values of mode @code{BLKmode} must be explicitly handled
3929 by this function. Also, the option @option{-fpcc-struct-return}
3930 takes effect regardless of this macro. On most systems, it is
3931 possible to leave the hook undefined; this causes a default
3932 definition to be used, whose value is the constant 1 for @code{BLKmode}
3933 values, and 0 otherwise.
3935 Do not use this hook to indicate that structures and unions should always
3936 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3940 @defmac DEFAULT_PCC_STRUCT_RETURN
3941 Define this macro to be 1 if all structure and union return values must be
3942 in memory. Since this results in slower code, this should be defined
3943 only if needed for compatibility with other compilers or with an ABI@.
3944 If you define this macro to be 0, then the conventions used for structure
3945 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3947 If not defined, this defaults to the value 1.
3950 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
3951 This target hook should return the location of the structure value
3952 address (normally a @code{mem} or @code{reg}), or 0 if the address is
3953 passed as an ``invisible'' first argument. Note that @var{fndecl} may
3954 be @code{NULL}, for libcalls.
3956 On some architectures the place where the structure value address
3957 is found by the called function is not the same place that the
3958 caller put it. This can be due to register windows, or it could
3959 be because the function prologue moves it to a different place.
3960 @var{incoming} is @code{true} when the location is needed in
3961 the context of the called function, and @code{false} in the context of
3964 If @var{incoming} is @code{true} and the address is to be found on the
3965 stack, return a @code{mem} which refers to the frame pointer.
3968 @defmac PCC_STATIC_STRUCT_RETURN
3969 Define this macro if the usual system convention on the target machine
3970 for returning structures and unions is for the called function to return
3971 the address of a static variable containing the value.
3973 Do not define this if the usual system convention is for the caller to
3974 pass an address to the subroutine.
3976 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3977 nothing when you use @option{-freg-struct-return} mode.
3981 @subsection Caller-Saves Register Allocation
3983 If you enable it, GCC can save registers around function calls. This
3984 makes it possible to use call-clobbered registers to hold variables that
3985 must live across calls.
3987 @defmac DEFAULT_CALLER_SAVES
3988 Define this macro if function calls on the target machine do not preserve
3989 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3990 for all registers. When defined, this macro enables @option{-fcaller-saves}
3991 by default for all optimization levels. It has no effect for optimization
3992 levels 2 and higher, where @option{-fcaller-saves} is the default.
3995 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3996 A C expression to determine whether it is worthwhile to consider placing
3997 a pseudo-register in a call-clobbered hard register and saving and
3998 restoring it around each function call. The expression should be 1 when
3999 this is worth doing, and 0 otherwise.
4001 If you don't define this macro, a default is used which is good on most
4002 machines: @code{4 * @var{calls} < @var{refs}}.
4005 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4006 A C expression specifying which mode is required for saving @var{nregs}
4007 of a pseudo-register in call-clobbered hard register @var{regno}. If
4008 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4009 returned. For most machines this macro need not be defined since GCC
4010 will select the smallest suitable mode.
4013 @node Function Entry
4014 @subsection Function Entry and Exit
4015 @cindex function entry and exit
4019 This section describes the macros that output function entry
4020 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4022 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4023 If defined, a function that outputs the assembler code for entry to a
4024 function. The prologue is responsible for setting up the stack frame,
4025 initializing the frame pointer register, saving registers that must be
4026 saved, and allocating @var{size} additional bytes of storage for the
4027 local variables. @var{size} is an integer. @var{file} is a stdio
4028 stream to which the assembler code should be output.
4030 The label for the beginning of the function need not be output by this
4031 macro. That has already been done when the macro is run.
4033 @findex regs_ever_live
4034 To determine which registers to save, the macro can refer to the array
4035 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4036 @var{r} is used anywhere within the function. This implies the function
4037 prologue should save register @var{r}, provided it is not one of the
4038 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4039 @code{regs_ever_live}.)
4041 On machines that have ``register windows'', the function entry code does
4042 not save on the stack the registers that are in the windows, even if
4043 they are supposed to be preserved by function calls; instead it takes
4044 appropriate steps to ``push'' the register stack, if any non-call-used
4045 registers are used in the function.
4047 @findex frame_pointer_needed
4048 On machines where functions may or may not have frame-pointers, the
4049 function entry code must vary accordingly; it must set up the frame
4050 pointer if one is wanted, and not otherwise. To determine whether a
4051 frame pointer is in wanted, the macro can refer to the variable
4052 @code{frame_pointer_needed}. The variable's value will be 1 at run
4053 time in a function that needs a frame pointer. @xref{Elimination}.
4055 The function entry code is responsible for allocating any stack space
4056 required for the function. This stack space consists of the regions
4057 listed below. In most cases, these regions are allocated in the
4058 order listed, with the last listed region closest to the top of the
4059 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4060 the highest address if it is not defined). You can use a different order
4061 for a machine if doing so is more convenient or required for
4062 compatibility reasons. Except in cases where required by standard
4063 or by a debugger, there is no reason why the stack layout used by GCC
4064 need agree with that used by other compilers for a machine.
4067 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4068 If defined, a function that outputs assembler code at the end of a
4069 prologue. This should be used when the function prologue is being
4070 emitted as RTL, and you have some extra assembler that needs to be
4071 emitted. @xref{prologue instruction pattern}.
4074 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4075 If defined, a function that outputs assembler code at the start of an
4076 epilogue. This should be used when the function epilogue is being
4077 emitted as RTL, and you have some extra assembler that needs to be
4078 emitted. @xref{epilogue instruction pattern}.
4081 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4082 If defined, a function that outputs the assembler code for exit from a
4083 function. The epilogue is responsible for restoring the saved
4084 registers and stack pointer to their values when the function was
4085 called, and returning control to the caller. This macro takes the
4086 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4087 registers to restore are determined from @code{regs_ever_live} and
4088 @code{CALL_USED_REGISTERS} in the same way.
4090 On some machines, there is a single instruction that does all the work
4091 of returning from the function. On these machines, give that
4092 instruction the name @samp{return} and do not define the macro
4093 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4095 Do not define a pattern named @samp{return} if you want the
4096 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4097 switches to control whether return instructions or epilogues are used,
4098 define a @samp{return} pattern with a validity condition that tests the
4099 target switches appropriately. If the @samp{return} pattern's validity
4100 condition is false, epilogues will be used.
4102 On machines where functions may or may not have frame-pointers, the
4103 function exit code must vary accordingly. Sometimes the code for these
4104 two cases is completely different. To determine whether a frame pointer
4105 is wanted, the macro can refer to the variable
4106 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4107 a function that needs a frame pointer.
4109 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4110 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4111 The C variable @code{current_function_is_leaf} is nonzero for such a
4112 function. @xref{Leaf Functions}.
4114 On some machines, some functions pop their arguments on exit while
4115 others leave that for the caller to do. For example, the 68020 when
4116 given @option{-mrtd} pops arguments in functions that take a fixed
4117 number of arguments.
4119 @findex current_function_pops_args
4120 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4121 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4122 needs to know what was decided. The variable that is called
4123 @code{current_function_pops_args} is the number of bytes of its
4124 arguments that a function should pop. @xref{Scalar Return}.
4125 @c what is the "its arguments" in the above sentence referring to, pray
4126 @c tell? --mew 5feb93
4131 @findex current_function_pretend_args_size
4132 A region of @code{current_function_pretend_args_size} bytes of
4133 uninitialized space just underneath the first argument arriving on the
4134 stack. (This may not be at the very start of the allocated stack region
4135 if the calling sequence has pushed anything else since pushing the stack
4136 arguments. But usually, on such machines, nothing else has been pushed
4137 yet, because the function prologue itself does all the pushing.) This
4138 region is used on machines where an argument may be passed partly in
4139 registers and partly in memory, and, in some cases to support the
4140 features in @code{<stdarg.h>}.
4143 An area of memory used to save certain registers used by the function.
4144 The size of this area, which may also include space for such things as
4145 the return address and pointers to previous stack frames, is
4146 machine-specific and usually depends on which registers have been used
4147 in the function. Machines with register windows often do not require
4151 A region of at least @var{size} bytes, possibly rounded up to an allocation
4152 boundary, to contain the local variables of the function. On some machines,
4153 this region and the save area may occur in the opposite order, with the
4154 save area closer to the top of the stack.
4157 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4158 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4159 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4160 argument lists of the function. @xref{Stack Arguments}.
4163 Normally, it is necessary for the macros
4164 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4165 @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially.
4166 The C variable @code{current_function_is_leaf} is nonzero for such a
4169 @defmac EXIT_IGNORE_STACK
4170 Define this macro as a C expression that is nonzero if the return
4171 instruction or the function epilogue ignores the value of the stack
4172 pointer; in other words, if it is safe to delete an instruction to
4173 adjust the stack pointer before a return from the function.
4175 Note that this macro's value is relevant only for functions for which
4176 frame pointers are maintained. It is never safe to delete a final
4177 stack adjustment in a function that has no frame pointer, and the
4178 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4181 @defmac EPILOGUE_USES (@var{regno})
4182 Define this macro as a C expression that is nonzero for registers that are
4183 used by the epilogue or the @samp{return} pattern. The stack and frame
4184 pointer registers are already be assumed to be used as needed.
4187 @defmac EH_USES (@var{regno})
4188 Define this macro as a C expression that is nonzero for registers that are
4189 used by the exception handling mechanism, and so should be considered live
4190 on entry to an exception edge.
4193 @defmac DELAY_SLOTS_FOR_EPILOGUE
4194 Define this macro if the function epilogue contains delay slots to which
4195 instructions from the rest of the function can be ``moved''. The
4196 definition should be a C expression whose value is an integer
4197 representing the number of delay slots there.
4200 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4201 A C expression that returns 1 if @var{insn} can be placed in delay
4202 slot number @var{n} of the epilogue.
4204 The argument @var{n} is an integer which identifies the delay slot now
4205 being considered (since different slots may have different rules of
4206 eligibility). It is never negative and is always less than the number
4207 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4208 If you reject a particular insn for a given delay slot, in principle, it
4209 may be reconsidered for a subsequent delay slot. Also, other insns may
4210 (at least in principle) be considered for the so far unfilled delay
4213 @findex current_function_epilogue_delay_list
4214 @findex final_scan_insn
4215 The insns accepted to fill the epilogue delay slots are put in an RTL
4216 list made with @code{insn_list} objects, stored in the variable
4217 @code{current_function_epilogue_delay_list}. The insn for the first
4218 delay slot comes first in the list. Your definition of the macro
4219 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4220 outputting the insns in this list, usually by calling
4221 @code{final_scan_insn}.
4223 You need not define this macro if you did not define
4224 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4227 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, tree @var{function})
4228 A function that outputs the assembler code for a thunk
4229 function, used to implement C++ virtual function calls with multiple
4230 inheritance. The thunk acts as a wrapper around a virtual function,
4231 adjusting the implicit object parameter before handing control off to
4234 First, emit code to add the integer @var{delta} to the location that
4235 contains the incoming first argument. Assume that this argument
4236 contains a pointer, and is the one used to pass the @code{this} pointer
4237 in C++. This is the incoming argument @emph{before} the function prologue,
4238 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4239 all other incoming arguments.
4241 After the addition, emit code to jump to @var{function}, which is a
4242 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4243 not touch the return address. Hence returning from @var{FUNCTION} will
4244 return to whoever called the current @samp{thunk}.
4246 The effect must be as if @var{function} had been called directly with
4247 the adjusted first argument. This macro is responsible for emitting all
4248 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4249 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4251 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4252 have already been extracted from it.) It might possibly be useful on
4253 some targets, but probably not.
4255 If you do not define this macro, the target-independent code in the C++
4256 front end will generate a less efficient heavyweight thunk that calls
4257 @var{function} instead of jumping to it. The generic approach does
4258 not support varargs.
4261 @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})
4262 A function like @code{TARGET_ASM_OUTPUT_MI_THUNK}, except that if
4263 @var{vcall_offset} is nonzero, an additional adjustment should be made
4264 after adding @code{delta}. In particular, if @var{p} is the
4265 adjusted pointer, the following adjustment should be made:
4268 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4272 If this function is defined, it will always be used in place of
4273 @code{TARGET_ASM_OUTPUT_MI_THUNK}.
4277 @subsection Generating Code for Profiling
4278 @cindex profiling, code generation
4280 These macros will help you generate code for profiling.
4282 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4283 A C statement or compound statement to output to @var{file} some
4284 assembler code to call the profiling subroutine @code{mcount}.
4287 The details of how @code{mcount} expects to be called are determined by
4288 your operating system environment, not by GCC@. To figure them out,
4289 compile a small program for profiling using the system's installed C
4290 compiler and look at the assembler code that results.
4292 Older implementations of @code{mcount} expect the address of a counter
4293 variable to be loaded into some register. The name of this variable is
4294 @samp{LP} followed by the number @var{labelno}, so you would generate
4295 the name using @samp{LP%d} in a @code{fprintf}.
4298 @defmac PROFILE_HOOK
4299 A C statement or compound statement to output to @var{file} some assembly
4300 code to call the profiling subroutine @code{mcount} even the target does
4301 not support profiling.
4304 @defmac NO_PROFILE_COUNTERS
4305 Define this macro if the @code{mcount} subroutine on your system does
4306 not need a counter variable allocated for each function. This is true
4307 for almost all modern implementations. If you define this macro, you
4308 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4311 @defmac PROFILE_BEFORE_PROLOGUE
4312 Define this macro if the code for function profiling should come before
4313 the function prologue. Normally, the profiling code comes after.
4317 @subsection Permitting tail calls
4320 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4321 True if it is ok to do sibling call optimization for the specified
4322 call expression @var{exp}. @var{decl} will be the called function,
4323 or @code{NULL} if this is an indirect call.
4325 It is not uncommon for limitations of calling conventions to prevent
4326 tail calls to functions outside the current unit of translation, or
4327 during PIC compilation. The hook is used to enforce these restrictions,
4328 as the @code{sibcall} md pattern can not fail, or fall over to a
4329 ``normal'' call. The criteria for successful sibling call optimization
4330 may vary greatly between different architectures.
4334 @section Implementing the Varargs Macros
4335 @cindex varargs implementation
4337 GCC comes with an implementation of @code{<varargs.h>} and
4338 @code{<stdarg.h>} that work without change on machines that pass arguments
4339 on the stack. Other machines require their own implementations of
4340 varargs, and the two machine independent header files must have
4341 conditionals to include it.
4343 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4344 the calling convention for @code{va_start}. The traditional
4345 implementation takes just one argument, which is the variable in which
4346 to store the argument pointer. The ISO implementation of
4347 @code{va_start} takes an additional second argument. The user is
4348 supposed to write the last named argument of the function here.
4350 However, @code{va_start} should not use this argument. The way to find
4351 the end of the named arguments is with the built-in functions described
4354 @defmac __builtin_saveregs ()
4355 Use this built-in function to save the argument registers in memory so
4356 that the varargs mechanism can access them. Both ISO and traditional
4357 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4358 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
4360 On some machines, @code{__builtin_saveregs} is open-coded under the
4361 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
4362 it calls a routine written in assembler language, found in
4365 Code generated for the call to @code{__builtin_saveregs} appears at the
4366 beginning of the function, as opposed to where the call to
4367 @code{__builtin_saveregs} is written, regardless of what the code is.
4368 This is because the registers must be saved before the function starts
4369 to use them for its own purposes.
4370 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4374 @defmac __builtin_args_info (@var{category})
4375 Use this built-in function to find the first anonymous arguments in
4378 In general, a machine may have several categories of registers used for
4379 arguments, each for a particular category of data types. (For example,
4380 on some machines, floating-point registers are used for floating-point
4381 arguments while other arguments are passed in the general registers.)
4382 To make non-varargs functions use the proper calling convention, you
4383 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4384 registers in each category have been used so far
4386 @code{__builtin_args_info} accesses the same data structure of type
4387 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4388 with it, with @var{category} specifying which word to access. Thus, the
4389 value indicates the first unused register in a given category.
4391 Normally, you would use @code{__builtin_args_info} in the implementation
4392 of @code{va_start}, accessing each category just once and storing the
4393 value in the @code{va_list} object. This is because @code{va_list} will
4394 have to update the values, and there is no way to alter the
4395 values accessed by @code{__builtin_args_info}.
4398 @defmac __builtin_next_arg (@var{lastarg})
4399 This is the equivalent of @code{__builtin_args_info}, for stack
4400 arguments. It returns the address of the first anonymous stack
4401 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4402 returns the address of the location above the first anonymous stack
4403 argument. Use it in @code{va_start} to initialize the pointer for
4404 fetching arguments from the stack. Also use it in @code{va_start} to
4405 verify that the second parameter @var{lastarg} is the last named argument
4406 of the current function.
4409 @defmac __builtin_classify_type (@var{object})
4410 Since each machine has its own conventions for which data types are
4411 passed in which kind of register, your implementation of @code{va_arg}
4412 has to embody these conventions. The easiest way to categorize the
4413 specified data type is to use @code{__builtin_classify_type} together
4414 with @code{sizeof} and @code{__alignof__}.
4416 @code{__builtin_classify_type} ignores the value of @var{object},
4417 considering only its data type. It returns an integer describing what
4418 kind of type that is---integer, floating, pointer, structure, and so on.
4420 The file @file{typeclass.h} defines an enumeration that you can use to
4421 interpret the values of @code{__builtin_classify_type}.
4424 These machine description macros help implement varargs:
4426 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4427 If defined, this hook produces the machine-specific code for a call to
4428 @code{__builtin_saveregs}. This code will be moved to the very
4429 beginning of the function, before any parameter access are made. The
4430 return value of this function should be an RTX that contains the value
4431 to use as the return of @code{__builtin_saveregs}.
4434 @deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARGS (CUMULATIVE_ARGS *@var{args_so_far}, enum machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time})
4435 This target hook offers an alternative to using
4436 @code{__builtin_saveregs} and defining the hook
4437 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4438 register arguments into the stack so that all the arguments appear to
4439 have been passed consecutively on the stack. Once this is done, you can
4440 use the standard implementation of varargs that works for machines that
4441 pass all their arguments on the stack.
4443 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4444 structure, containing the values that are obtained after processing the
4445 named arguments. The arguments @var{mode} and @var{type} describe the
4446 last named argument---its machine mode and its data type as a tree node.
4448 The target hook should do two things: first, push onto the stack all the
4449 argument registers @emph{not} used for the named arguments, and second,
4450 store the size of the data thus pushed into the @code{int}-valued
4451 variable pointed to by @var{pretend_args_size}. The value that you
4452 store here will serve as additional offset for setting up the stack
4455 Because you must generate code to push the anonymous arguments at
4456 compile time without knowing their data types,
4457 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4458 have just a single category of argument register and use it uniformly
4461 If the argument @var{second_time} is nonzero, it means that the
4462 arguments of the function are being analyzed for the second time. This
4463 happens for an inline function, which is not actually compiled until the
4464 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4465 not generate any instructions in this case.
4468 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4469 Define this hook to return @code{true} if the location where a function
4470 argument is passed depends on whether or not it is a named argument.
4472 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4473 is set for varargs and stdarg functions. If this hook returns
4474 @code{true}, the @var{named} argument is always true for named
4475 arguments, and false for unnamed arguments. If it returns @code{false},
4476 but @code{TARGET_PRETEND_OUTOGOING_VARARGS_NAMED} returns @code{true},
4477 then all arguments are treated as named. Otherwise, all named arguments
4478 except the last are treated as named.
4480 You need not define this hook if it always returns zero.
4483 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4484 If you need to conditionally change ABIs so that one works with
4485 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4486 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4487 defined, then define this hook to return @code{true} if
4488 @code{SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4489 Otherwise, you should not define this hook.
4493 @section Trampolines for Nested Functions
4494 @cindex trampolines for nested functions
4495 @cindex nested functions, trampolines for
4497 A @dfn{trampoline} is a small piece of code that is created at run time
4498 when the address of a nested function is taken. It normally resides on
4499 the stack, in the stack frame of the containing function. These macros
4500 tell GCC how to generate code to allocate and initialize a
4503 The instructions in the trampoline must do two things: load a constant
4504 address into the static chain register, and jump to the real address of
4505 the nested function. On CISC machines such as the m68k, this requires
4506 two instructions, a move immediate and a jump. Then the two addresses
4507 exist in the trampoline as word-long immediate operands. On RISC
4508 machines, it is often necessary to load each address into a register in
4509 two parts. Then pieces of each address form separate immediate
4512 The code generated to initialize the trampoline must store the variable
4513 parts---the static chain value and the function address---into the
4514 immediate operands of the instructions. On a CISC machine, this is
4515 simply a matter of copying each address to a memory reference at the
4516 proper offset from the start of the trampoline. On a RISC machine, it
4517 may be necessary to take out pieces of the address and store them
4520 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4521 A C statement to output, on the stream @var{file}, assembler code for a
4522 block of data that contains the constant parts of a trampoline. This
4523 code should not include a label---the label is taken care of
4526 If you do not define this macro, it means no template is needed
4527 for the target. Do not define this macro on systems where the block move
4528 code to copy the trampoline into place would be larger than the code
4529 to generate it on the spot.
4532 @defmac TRAMPOLINE_SECTION
4533 The name of a subroutine to switch to the section in which the
4534 trampoline template is to be placed (@pxref{Sections}). The default is
4535 a value of @samp{readonly_data_section}, which places the trampoline in
4536 the section containing read-only data.
4539 @defmac TRAMPOLINE_SIZE
4540 A C expression for the size in bytes of the trampoline, as an integer.
4543 @defmac TRAMPOLINE_ALIGNMENT
4544 Alignment required for trampolines, in bits.
4546 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4547 is used for aligning trampolines.
4550 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4551 A C statement to initialize the variable parts of a trampoline.
4552 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4553 an RTX for the address of the nested function; @var{static_chain} is an
4554 RTX for the static chain value that should be passed to the function
4558 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4559 A C statement that should perform any machine-specific adjustment in
4560 the address of the trampoline. Its argument contains the address that
4561 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4562 used for a function call should be different from the address in which
4563 the template was stored, the different address should be assigned to
4564 @var{addr}. If this macro is not defined, @var{addr} will be used for
4567 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4568 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4569 If this macro is not defined, by default the trampoline is allocated as
4570 a stack slot. This default is right for most machines. The exceptions
4571 are machines where it is impossible to execute instructions in the stack
4572 area. On such machines, you may have to implement a separate stack,
4573 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4574 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4576 @var{fp} points to a data structure, a @code{struct function}, which
4577 describes the compilation status of the immediate containing function of
4578 the function which the trampoline is for. The stack slot for the
4579 trampoline is in the stack frame of this containing function. Other
4580 allocation strategies probably must do something analogous with this
4584 Implementing trampolines is difficult on many machines because they have
4585 separate instruction and data caches. Writing into a stack location
4586 fails to clear the memory in the instruction cache, so when the program
4587 jumps to that location, it executes the old contents.
4589 Here are two possible solutions. One is to clear the relevant parts of
4590 the instruction cache whenever a trampoline is set up. The other is to
4591 make all trampolines identical, by having them jump to a standard
4592 subroutine. The former technique makes trampoline execution faster; the
4593 latter makes initialization faster.
4595 To clear the instruction cache when a trampoline is initialized, define
4596 the following macro.
4598 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4599 If defined, expands to a C expression clearing the @emph{instruction
4600 cache} in the specified interval. The definition of this macro would
4601 typically be a series of @code{asm} statements. Both @var{beg} and
4602 @var{end} are both pointer expressions.
4605 To use a standard subroutine, define the following macro. In addition,
4606 you must make sure that the instructions in a trampoline fill an entire
4607 cache line with identical instructions, or else ensure that the
4608 beginning of the trampoline code is always aligned at the same point in
4609 its cache line. Look in @file{m68k.h} as a guide.
4611 @defmac TRANSFER_FROM_TRAMPOLINE
4612 Define this macro if trampolines need a special subroutine to do their
4613 work. The macro should expand to a series of @code{asm} statements
4614 which will be compiled with GCC@. They go in a library function named
4615 @code{__transfer_from_trampoline}.
4617 If you need to avoid executing the ordinary prologue code of a compiled
4618 C function when you jump to the subroutine, you can do so by placing a
4619 special label of your own in the assembler code. Use one @code{asm}
4620 statement to generate an assembler label, and another to make the label
4621 global. Then trampolines can use that label to jump directly to your
4622 special assembler code.
4626 @section Implicit Calls to Library Routines
4627 @cindex library subroutine names
4628 @cindex @file{libgcc.a}
4630 @c prevent bad page break with this line
4631 Here is an explanation of implicit calls to library routines.
4633 @defmac DECLARE_LIBRARY_RENAMES
4634 This macro, if defined, should expand to a piece of C code that will get
4635 expanded when compiling functions for libgcc.a. It can be used to
4636 provide alternate names for gcc's internal library functions if there
4637 are ABI-mandated names that the compiler should provide.
4640 @findex init_one_libfunc
4641 @findex set_optab_libfunc
4642 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
4643 This hook should declare additional library routines or rename
4644 existing ones, using the functions @code{set_optab_libfunc} and
4645 @code{init_one_libfunc} defined in @file{optabs.c}.
4646 @code{init_optabs} calls this macro after initializing all the normal
4649 The default is to do nothing. Most ports don't need to define this hook.
4652 @defmac TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4653 This macro should return @code{true} if the library routine that
4654 implements the floating point comparison operator @var{comparison} in
4655 mode @var{mode} will return a boolean, and @var{false} if it will
4658 GCC's own floating point libraries return tristates from the
4659 comparison operators, so the default returns false always. Most ports
4660 don't need to define this macro.
4663 @cindex US Software GOFAST, floating point emulation library
4664 @cindex floating point emulation library, US Software GOFAST
4665 @cindex GOFAST, floating point emulation library
4666 @findex gofast_maybe_init_libfuncs
4667 @defmac US_SOFTWARE_GOFAST
4668 Define this macro if your system C library uses the US Software GOFAST
4669 library to provide floating point emulation.
4671 In addition to defining this macro, your architecture must set
4672 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
4673 else call that function from its version of that hook. It is defined
4674 in @file{config/gofast.h}, which must be included by your
4675 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
4678 If this macro is defined, the
4679 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
4680 false for @code{SFmode} and @code{DFmode} comparisons.
4683 @cindex @code{EDOM}, implicit usage
4686 The value of @code{EDOM} on the target machine, as a C integer constant
4687 expression. If you don't define this macro, GCC does not attempt to
4688 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4689 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4692 If you do not define @code{TARGET_EDOM}, then compiled code reports
4693 domain errors by calling the library function and letting it report the
4694 error. If mathematical functions on your system use @code{matherr} when
4695 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4696 that @code{matherr} is used normally.
4699 @cindex @code{errno}, implicit usage
4700 @defmac GEN_ERRNO_RTX
4701 Define this macro as a C expression to create an rtl expression that
4702 refers to the global ``variable'' @code{errno}. (On certain systems,
4703 @code{errno} may not actually be a variable.) If you don't define this
4704 macro, a reasonable default is used.
4707 @cindex @code{bcopy}, implicit usage
4708 @cindex @code{memcpy}, implicit usage
4709 @cindex @code{memmove}, implicit usage
4710 @cindex @code{bzero}, implicit usage
4711 @cindex @code{memset}, implicit usage
4712 @defmac TARGET_MEM_FUNCTIONS
4713 Define this macro if GCC should generate calls to the ISO C
4714 (and System V) library functions @code{memcpy}, @code{memmove} and
4715 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4718 @cindex C99 math functions, implicit usage
4719 @defmac TARGET_C99_FUNCTIONS
4720 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4721 @code{sinf} and similarly for other functions defined by C99 standard. The
4722 default is nonzero that should be proper value for most modern systems, however
4723 number of existing systems lacks support for these functions in the runtime so
4724 they needs this macro to be redefined to 0.
4727 @defmac NEXT_OBJC_RUNTIME
4728 Define this macro to generate code for Objective-C message sending using
4729 the calling convention of the NeXT system. This calling convention
4730 involves passing the object, the selector and the method arguments all
4731 at once to the method-lookup library function.
4733 The default calling convention passes just the object and the selector
4734 to the lookup function, which returns a pointer to the method.
4737 @node Addressing Modes
4738 @section Addressing Modes
4739 @cindex addressing modes
4741 @c prevent bad page break with this line
4742 This is about addressing modes.
4744 @defmac HAVE_PRE_INCREMENT
4745 @defmacx HAVE_PRE_DECREMENT
4746 @defmacx HAVE_POST_INCREMENT
4747 @defmacx HAVE_POST_DECREMENT
4748 A C expression that is nonzero if the machine supports pre-increment,
4749 pre-decrement, post-increment, or post-decrement addressing respectively.
4752 @defmac HAVE_PRE_MODIFY_DISP
4753 @defmacx HAVE_POST_MODIFY_DISP
4754 A C expression that is nonzero if the machine supports pre- or
4755 post-address side-effect generation involving constants other than
4756 the size of the memory operand.
4759 @defmac HAVE_PRE_MODIFY_REG
4760 @defmacx HAVE_POST_MODIFY_REG
4761 A C expression that is nonzero if the machine supports pre- or
4762 post-address side-effect generation involving a register displacement.
4765 @defmac CONSTANT_ADDRESS_P (@var{x})
4766 A C expression that is 1 if the RTX @var{x} is a constant which
4767 is a valid address. On most machines, this can be defined as
4768 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4769 in which constant addresses are supported.
4772 @defmac CONSTANT_P (@var{x})
4773 @code{CONSTANT_P}, which is defined by target-independent code,
4774 accepts integer-values expressions whose values are not explicitly
4775 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4776 expressions and @code{const} arithmetic expressions, in addition to
4777 @code{const_int} and @code{const_double} expressions.
4780 @defmac MAX_REGS_PER_ADDRESS
4781 A number, the maximum number of registers that can appear in a valid
4782 memory address. Note that it is up to you to specify a value equal to
4783 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4787 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4788 A C compound statement with a conditional @code{goto @var{label};}
4789 executed if @var{x} (an RTX) is a legitimate memory address on the
4790 target machine for a memory operand of mode @var{mode}.
4792 It usually pays to define several simpler macros to serve as
4793 subroutines for this one. Otherwise it may be too complicated to
4796 This macro must exist in two variants: a strict variant and a
4797 non-strict one. The strict variant is used in the reload pass. It
4798 must be defined so that any pseudo-register that has not been
4799 allocated a hard register is considered a memory reference. In
4800 contexts where some kind of register is required, a pseudo-register
4801 with no hard register must be rejected.
4803 The non-strict variant is used in other passes. It must be defined to
4804 accept all pseudo-registers in every context where some kind of
4805 register is required.
4807 @findex REG_OK_STRICT
4808 Compiler source files that want to use the strict variant of this
4809 macro define the macro @code{REG_OK_STRICT}. You should use an
4810 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4811 in that case and the non-strict variant otherwise.
4813 Subroutines to check for acceptable registers for various purposes (one
4814 for base registers, one for index registers, and so on) are typically
4815 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4816 Then only these subroutine macros need have two variants; the higher
4817 levels of macros may be the same whether strict or not.
4819 Normally, constant addresses which are the sum of a @code{symbol_ref}
4820 and an integer are stored inside a @code{const} RTX to mark them as
4821 constant. Therefore, there is no need to recognize such sums
4822 specifically as legitimate addresses. Normally you would simply
4823 recognize any @code{const} as legitimate.
4825 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4826 sums that are not marked with @code{const}. It assumes that a naked
4827 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4828 naked constant sums as illegitimate addresses, so that none of them will
4829 be given to @code{PRINT_OPERAND_ADDRESS}.
4831 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4832 On some machines, whether a symbolic address is legitimate depends on
4833 the section that the address refers to. On these machines, define the
4834 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4835 into the @code{symbol_ref}, and then check for it here. When you see a
4836 @code{const}, you will have to look inside it to find the
4837 @code{symbol_ref} in order to determine the section. @xref{Assembler
4841 @defmac REG_OK_FOR_BASE_P (@var{x})
4842 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4843 RTX) is valid for use as a base register. For hard registers, it
4844 should always accept those which the hardware permits and reject the
4845 others. Whether the macro accepts or rejects pseudo registers must be
4846 controlled by @code{REG_OK_STRICT} as described above. This usually
4847 requires two variant definitions, of which @code{REG_OK_STRICT}
4848 controls the one actually used.
4851 @defmac REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4852 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4853 that expression may examine the mode of the memory reference in
4854 @var{mode}. You should define this macro if the mode of the memory
4855 reference affects whether a register may be used as a base register. If
4856 you define this macro, the compiler will use it instead of
4857 @code{REG_OK_FOR_BASE_P}.
4860 @defmac REG_OK_FOR_INDEX_P (@var{x})
4861 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4862 RTX) is valid for use as an index register.
4864 The difference between an index register and a base register is that
4865 the index register may be scaled. If an address involves the sum of
4866 two registers, neither one of them scaled, then either one may be
4867 labeled the ``base'' and the other the ``index''; but whichever
4868 labeling is used must fit the machine's constraints of which registers
4869 may serve in each capacity. The compiler will try both labelings,
4870 looking for one that is valid, and will reload one or both registers
4871 only if neither labeling works.
4874 @defmac FIND_BASE_TERM (@var{x})
4875 A C expression to determine the base term of address @var{x}.
4876 This macro is used in only one place: `find_base_term' in alias.c.
4878 It is always safe for this macro to not be defined. It exists so
4879 that alias analysis can understand machine-dependent addresses.
4881 The typical use of this macro is to handle addresses containing
4882 a label_ref or symbol_ref within an UNSPEC@.
4885 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4886 A C compound statement that attempts to replace @var{x} with a valid
4887 memory address for an operand of mode @var{mode}. @var{win} will be a
4888 C statement label elsewhere in the code; the macro definition may use
4891 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4895 to avoid further processing if the address has become legitimate.
4897 @findex break_out_memory_refs
4898 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4899 and @var{oldx} will be the operand that was given to that function to produce
4902 The code generated by this macro should not alter the substructure of
4903 @var{x}. If it transforms @var{x} into a more legitimate form, it
4904 should assign @var{x} (which will always be a C variable) a new value.
4906 It is not necessary for this macro to come up with a legitimate
4907 address. The compiler has standard ways of doing so in all cases. In
4908 fact, it is safe for this macro to do nothing. But often a
4909 machine-dependent strategy can generate better code.
4912 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4913 A C compound statement that attempts to replace @var{x}, which is an address
4914 that needs reloading, with a valid memory address for an operand of mode
4915 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4916 It is not necessary to define this macro, but it might be useful for
4917 performance reasons.
4919 For example, on the i386, it is sometimes possible to use a single
4920 reload register instead of two by reloading a sum of two pseudo
4921 registers into a register. On the other hand, for number of RISC
4922 processors offsets are limited so that often an intermediate address
4923 needs to be generated in order to address a stack slot. By defining
4924 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4925 generated for adjacent some stack slots can be made identical, and thus
4928 @emph{Note}: This macro should be used with caution. It is necessary
4929 to know something of how reload works in order to effectively use this,
4930 and it is quite easy to produce macros that build in too much knowledge
4931 of reload internals.
4933 @emph{Note}: This macro must be able to reload an address created by a
4934 previous invocation of this macro. If it fails to handle such addresses
4935 then the compiler may generate incorrect code or abort.
4938 The macro definition should use @code{push_reload} to indicate parts that
4939 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4940 suitable to be passed unaltered to @code{push_reload}.
4942 The code generated by this macro must not alter the substructure of
4943 @var{x}. If it transforms @var{x} into a more legitimate form, it
4944 should assign @var{x} (which will always be a C variable) a new value.
4945 This also applies to parts that you change indirectly by calling
4948 @findex strict_memory_address_p
4949 The macro definition may use @code{strict_memory_address_p} to test if
4950 the address has become legitimate.
4953 If you want to change only a part of @var{x}, one standard way of doing
4954 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4955 single level of rtl. Thus, if the part to be changed is not at the
4956 top level, you'll need to replace first the top level.
4957 It is not necessary for this macro to come up with a legitimate
4958 address; but often a machine-dependent strategy can generate better code.
4961 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4962 A C statement or compound statement with a conditional @code{goto
4963 @var{label};} executed if memory address @var{x} (an RTX) can have
4964 different meanings depending on the machine mode of the memory
4965 reference it is used for or if the address is valid for some modes
4968 Autoincrement and autodecrement addresses typically have mode-dependent
4969 effects because the amount of the increment or decrement is the size
4970 of the operand being addressed. Some machines have other mode-dependent
4971 addresses. Many RISC machines have no mode-dependent addresses.
4973 You may assume that @var{addr} is a valid address for the machine.
4976 @defmac LEGITIMATE_CONSTANT_P (@var{x})
4977 A C expression that is nonzero if @var{x} is a legitimate constant for
4978 an immediate operand on the target machine. You can assume that
4979 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4980 @samp{1} is a suitable definition for this macro on machines where
4981 anything @code{CONSTANT_P} is valid.
4984 @node Condition Code
4985 @section Condition Code Status
4986 @cindex condition code status
4988 @c prevent bad page break with this line
4989 This describes the condition code status.
4992 The file @file{conditions.h} defines a variable @code{cc_status} to
4993 describe how the condition code was computed (in case the interpretation of
4994 the condition code depends on the instruction that it was set by). This
4995 variable contains the RTL expressions on which the condition code is
4996 currently based, and several standard flags.
4998 Sometimes additional machine-specific flags must be defined in the machine
4999 description header file. It can also add additional machine-specific
5000 information by defining @code{CC_STATUS_MDEP}.
5002 @defmac CC_STATUS_MDEP
5003 C code for a data type which is used for declaring the @code{mdep}
5004 component of @code{cc_status}. It defaults to @code{int}.
5006 This macro is not used on machines that do not use @code{cc0}.
5009 @defmac CC_STATUS_MDEP_INIT
5010 A C expression to initialize the @code{mdep} field to ``empty''.
5011 The default definition does nothing, since most machines don't use
5012 the field anyway. If you want to use the field, you should probably
5013 define this macro to initialize it.
5015 This macro is not used on machines that do not use @code{cc0}.
5018 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5019 A C compound statement to set the components of @code{cc_status}
5020 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5021 this macro's responsibility to recognize insns that set the condition
5022 code as a byproduct of other activity as well as those that explicitly
5025 This macro is not used on machines that do not use @code{cc0}.
5027 If there are insns that do not set the condition code but do alter
5028 other machine registers, this macro must check to see whether they
5029 invalidate the expressions that the condition code is recorded as
5030 reflecting. For example, on the 68000, insns that store in address
5031 registers do not set the condition code, which means that usually
5032 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5033 insns. But suppose that the previous insn set the condition code
5034 based on location @samp{a4@@(102)} and the current insn stores a new
5035 value in @samp{a4}. Although the condition code is not changed by
5036 this, it will no longer be true that it reflects the contents of
5037 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5038 @code{cc_status} in this case to say that nothing is known about the
5039 condition code value.
5041 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5042 with the results of peephole optimization: insns whose patterns are
5043 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5044 constants which are just the operands. The RTL structure of these
5045 insns is not sufficient to indicate what the insns actually do. What
5046 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5047 @code{CC_STATUS_INIT}.
5049 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5050 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5051 @samp{cc}. This avoids having detailed information about patterns in
5052 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5055 @defmac EXTRA_CC_MODES
5056 Condition codes are represented in registers by machine modes of class
5057 @code{MODE_CC}. By default, there is just one mode, @code{CCmode}, with
5058 this class. If you need more such modes, create a file named
5059 @file{@var{machine}-modes.def} in your @file{config/@var{machine}}
5060 directory (@pxref{Back End, , Anatomy of a Target Back End}), containing
5061 a list of these modes. Each entry in the list should be a call to the
5062 macro @code{CC}. This macro takes one argument, which is the name of
5063 the mode: it should begin with @samp{CC}. Do not put quotation marks
5064 around the name, or include the trailing @samp{mode}; these are
5065 automatically added. There should not be anything else in the file
5068 A sample @file{@var{machine}-modes.def} file might look like this:
5071 CC (CC_NOOV) /* @r{Comparison only valid if there was no overflow.} */
5072 CC (CCFP) /* @r{Floating point comparison that cannot trap.} */
5073 CC (CCFPE) /* @r{Floating point comparison that may trap.} */
5076 When you create this file, the macro @code{EXTRA_CC_MODES} is
5077 automatically defined by @command{configure}, with value @samp{1}.
5080 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5081 Returns a mode from class @code{MODE_CC} to be used when comparison
5082 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5083 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5084 @pxref{Jump Patterns} for a description of the reason for this
5088 #define SELECT_CC_MODE(OP,X,Y) \
5089 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5090 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5091 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5092 || GET_CODE (X) == NEG) \
5093 ? CC_NOOVmode : CCmode))
5096 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
5099 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5100 On some machines not all possible comparisons are defined, but you can
5101 convert an invalid comparison into a valid one. For example, the Alpha
5102 does not have a @code{GT} comparison, but you can use an @code{LT}
5103 comparison instead and swap the order of the operands.
5105 On such machines, define this macro to be a C statement to do any
5106 required conversions. @var{code} is the initial comparison code
5107 and @var{op0} and @var{op1} are the left and right operands of the
5108 comparison, respectively. You should modify @var{code}, @var{op0}, and
5109 @var{op1} as required.
5111 GCC will not assume that the comparison resulting from this macro is
5112 valid but will see if the resulting insn matches a pattern in the
5115 You need not define this macro if it would never change the comparison
5119 @defmac REVERSIBLE_CC_MODE (@var{mode})
5120 A C expression whose value is one if it is always safe to reverse a
5121 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5122 can ever return @var{mode} for a floating-point inequality comparison,
5123 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5125 You need not define this macro if it would always returns zero or if the
5126 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5127 For example, here is the definition used on the SPARC, where floating-point
5128 inequality comparisons are always given @code{CCFPEmode}:
5131 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5135 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5136 A C expression whose value is reversed condition code of the @var{code} for
5137 comparison done in CC_MODE @var{mode}. The macro is used only in case
5138 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5139 machine has some non-standard way how to reverse certain conditionals. For
5140 instance in case all floating point conditions are non-trapping, compiler may
5141 freely convert unordered compares to ordered one. Then definition may look
5145 #define REVERSE_CONDITION(CODE, MODE) \
5146 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5147 : reverse_condition_maybe_unordered (CODE))
5151 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
5152 A C expression that returns true if the conditional execution predicate
5153 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
5154 return 0 if the target has conditional execution predicates that cannot be
5155 reversed safely. If no expansion is specified, this macro is defined as
5159 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5160 ((x) == reverse_condition (y))
5165 @section Describing Relative Costs of Operations
5166 @cindex costs of instructions
5167 @cindex relative costs
5168 @cindex speed of instructions
5170 These macros let you describe the relative speed of various operations
5171 on the target machine.
5173 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5174 A C expression for the cost of moving data of mode @var{mode} from a
5175 register in class @var{from} to one in class @var{to}. The classes are
5176 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5177 value of 2 is the default; other values are interpreted relative to
5180 It is not required that the cost always equal 2 when @var{from} is the
5181 same as @var{to}; on some machines it is expensive to move between
5182 registers if they are not general registers.
5184 If reload sees an insn consisting of a single @code{set} between two
5185 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5186 classes returns a value of 2, reload does not check to ensure that the
5187 constraints of the insn are met. Setting a cost of other than 2 will
5188 allow reload to verify that the constraints are met. You should do this
5189 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5192 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5193 A C expression for the cost of moving data of mode @var{mode} between a
5194 register of class @var{class} and memory; @var{in} is zero if the value
5195 is to be written to memory, nonzero if it is to be read in. This cost
5196 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5197 registers and memory is more expensive than between two registers, you
5198 should define this macro to express the relative cost.
5200 If you do not define this macro, GCC uses a default cost of 4 plus
5201 the cost of copying via a secondary reload register, if one is
5202 needed. If your machine requires a secondary reload register to copy
5203 between memory and a register of @var{class} but the reload mechanism is
5204 more complex than copying via an intermediate, define this macro to
5205 reflect the actual cost of the move.
5207 GCC defines the function @code{memory_move_secondary_cost} if
5208 secondary reloads are needed. It computes the costs due to copying via
5209 a secondary register. If your machine copies from memory using a
5210 secondary register in the conventional way but the default base value of
5211 4 is not correct for your machine, define this macro to add some other
5212 value to the result of that function. The arguments to that function
5213 are the same as to this macro.
5217 A C expression for the cost of a branch instruction. A value of 1 is
5218 the default; other values are interpreted relative to that.
5221 Here are additional macros which do not specify precise relative costs,
5222 but only that certain actions are more expensive than GCC would
5225 @defmac SLOW_BYTE_ACCESS
5226 Define this macro as a C expression which is nonzero if accessing less
5227 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5228 faster than accessing a word of memory, i.e., if such access
5229 require more than one instruction or if there is no difference in cost
5230 between byte and (aligned) word loads.
5232 When this macro is not defined, the compiler will access a field by
5233 finding the smallest containing object; when it is defined, a fullword
5234 load will be used if alignment permits. Unless bytes accesses are
5235 faster than word accesses, using word accesses is preferable since it
5236 may eliminate subsequent memory access if subsequent accesses occur to
5237 other fields in the same word of the structure, but to different bytes.
5240 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5241 Define this macro to be the value 1 if memory accesses described by the
5242 @var{mode} and @var{alignment} parameters have a cost many times greater
5243 than aligned accesses, for example if they are emulated in a trap
5246 When this macro is nonzero, the compiler will act as if
5247 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5248 moves. This can cause significantly more instructions to be produced.
5249 Therefore, do not set this macro nonzero if unaligned accesses only add a
5250 cycle or two to the time for a memory access.
5252 If the value of this macro is always zero, it need not be defined. If
5253 this macro is defined, it should produce a nonzero value when
5254 @code{STRICT_ALIGNMENT} is nonzero.
5258 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5259 which a sequence of insns should be generated instead of a
5260 string move insn or a library call. Increasing the value will always
5261 make code faster, but eventually incurs high cost in increased code size.
5263 Note that on machines where the corresponding move insn is a
5264 @code{define_expand} that emits a sequence of insns, this macro counts
5265 the number of such sequences.
5267 If you don't define this, a reasonable default is used.
5270 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5271 A C expression used to determine whether @code{move_by_pieces} will be used to
5272 copy a chunk of memory, or whether some other block move mechanism
5273 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5274 than @code{MOVE_RATIO}.
5277 @defmac MOVE_MAX_PIECES
5278 A C expression used by @code{move_by_pieces} to determine the largest unit
5279 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5283 The threshold of number of scalar move insns, @emph{below} which a sequence
5284 of insns should be generated to clear memory instead of a string clear insn
5285 or a library call. Increasing the value will always make code faster, but
5286 eventually incurs high cost in increased code size.
5288 If you don't define this, a reasonable default is used.
5291 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5292 A C expression used to determine whether @code{clear_by_pieces} will be used
5293 to clear a chunk of memory, or whether some other block clear mechanism
5294 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5295 than @code{CLEAR_RATIO}.
5298 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5299 A C expression used to determine whether @code{store_by_pieces} will be
5300 used to set a chunk of memory to a constant value, or whether some other
5301 mechanism will be used. Used by @code{__builtin_memset} when storing
5302 values other than constant zero and by @code{__builtin_strcpy} when
5303 when called with a constant source string.
5304 Defaults to @code{MOVE_BY_PIECES_P}.
5307 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5308 A C expression used to determine whether a load postincrement is a good
5309 thing to use for a given mode. Defaults to the value of
5310 @code{HAVE_POST_INCREMENT}.
5313 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5314 A C expression used to determine whether a load postdecrement is a good
5315 thing to use for a given mode. Defaults to the value of
5316 @code{HAVE_POST_DECREMENT}.
5319 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5320 A C expression used to determine whether a load preincrement is a good
5321 thing to use for a given mode. Defaults to the value of
5322 @code{HAVE_PRE_INCREMENT}.
5325 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5326 A C expression used to determine whether a load predecrement is a good
5327 thing to use for a given mode. Defaults to the value of
5328 @code{HAVE_PRE_DECREMENT}.
5331 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5332 A C expression used to determine whether a store postincrement is a good
5333 thing to use for a given mode. Defaults to the value of
5334 @code{HAVE_POST_INCREMENT}.
5337 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5338 A C expression used to determine whether a store postdecrement is a good
5339 thing to use for a given mode. Defaults to the value of
5340 @code{HAVE_POST_DECREMENT}.
5343 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5344 This macro is used to determine whether a store preincrement is a good
5345 thing to use for a given mode. Defaults to the value of
5346 @code{HAVE_PRE_INCREMENT}.
5349 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5350 This macro is used to determine whether a store predecrement is a good
5351 thing to use for a given mode. Defaults to the value of
5352 @code{HAVE_PRE_DECREMENT}.
5355 @defmac NO_FUNCTION_CSE
5356 Define this macro if it is as good or better to call a constant
5357 function address than to call an address kept in a register.
5360 @defmac NO_RECURSIVE_FUNCTION_CSE
5361 Define this macro if it is as good or better for a function to call
5362 itself with an explicit address than to call an address kept in a
5366 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5367 Define this macro if a non-short-circuit operation produced by
5368 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5369 @code{BRANCH_COST} is greater than or equal to the value 2.
5372 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5373 This target hook describes the relative costs of RTL expressions.
5375 The cost may depend on the precise form of the expression, which is
5376 available for examination in @var{x}, and the rtx code of the expression
5377 in which it is contained, found in @var{outer_code}. @var{code} is the
5378 expression code---redundant, since it can be obtained with
5379 @code{GET_CODE (@var{x})}.
5381 In implementing this hook, you can use the construct
5382 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5385 On entry to the hook, @code{*@var{total}} contains a default estimate
5386 for the cost of the expression. The hook should modify this value as
5389 The hook returns true when all subexpressions of @var{x} have been
5390 processed, and false when @code{rtx_cost} should recurse.
5393 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5394 This hook computes the cost of an addressing mode that contains
5395 @var{address}. If not defined, the cost is computed from
5396 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5398 For most CISC machines, the default cost is a good approximation of the
5399 true cost of the addressing mode. However, on RISC machines, all
5400 instructions normally have the same length and execution time. Hence
5401 all addresses will have equal costs.
5403 In cases where more than one form of an address is known, the form with
5404 the lowest cost will be used. If multiple forms have the same, lowest,
5405 cost, the one that is the most complex will be used.
5407 For example, suppose an address that is equal to the sum of a register
5408 and a constant is used twice in the same basic block. When this macro
5409 is not defined, the address will be computed in a register and memory
5410 references will be indirect through that register. On machines where
5411 the cost of the addressing mode containing the sum is no higher than
5412 that of a simple indirect reference, this will produce an additional
5413 instruction and possibly require an additional register. Proper
5414 specification of this macro eliminates this overhead for such machines.
5416 This hook is never called with an invalid address.
5418 On machines where an address involving more than one register is as
5419 cheap as an address computation involving only one register, defining
5420 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5421 be live over a region of code where only one would have been if
5422 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5423 should be considered in the definition of this macro. Equivalent costs
5424 should probably only be given to addresses with different numbers of
5425 registers on machines with lots of registers.
5429 @section Adjusting the Instruction Scheduler
5431 The instruction scheduler may need a fair amount of machine-specific
5432 adjustment in order to produce good code. GCC provides several target
5433 hooks for this purpose. It is usually enough to define just a few of
5434 them: try the first ones in this list first.
5436 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5437 This hook returns the maximum number of instructions that can ever
5438 issue at the same time on the target machine. The default is one.
5439 Although the insn scheduler can define itself the possibility of issue
5440 an insn on the same cycle, the value can serve as an additional
5441 constraint to issue insns on the same simulated processor cycle (see
5442 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5443 This value must be constant over the entire compilation. If you need
5444 it to vary depending on what the instructions are, you must use
5445 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5447 For the automaton based pipeline interface, you could define this hook
5448 to return the value of the macro @code{MAX_DFA_ISSUE_RATE}.
5451 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5452 This hook is executed by the scheduler after it has scheduled an insn
5453 from the ready list. It should return the number of insns which can
5454 still be issued in the current cycle. The default is
5455 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5456 @code{USE}, which normally are not counted against the issue rate.
5457 You should define this hook if some insns take more machine resources
5458 than others, so that fewer insns can follow them in the same cycle.
5459 @var{file} is either a null pointer, or a stdio stream to write any
5460 debug output to. @var{verbose} is the verbose level provided by
5461 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5465 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5466 This function corrects the value of @var{cost} based on the
5467 relationship between @var{insn} and @var{dep_insn} through the
5468 dependence @var{link}. It should return the new value. The default
5469 is to make no adjustment to @var{cost}. This can be used for example
5470 to specify to the scheduler using the traditional pipeline description
5471 that an output- or anti-dependence does not incur the same cost as a
5472 data-dependence. If the scheduler using the automaton based pipeline
5473 description, the cost of anti-dependence is zero and the cost of
5474 output-dependence is maximum of one and the difference of latency
5475 times of the first and the second insns. If these values are not
5476 acceptable, you could use the hook to modify them too. See also
5477 @pxref{Automaton pipeline description}.
5480 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5481 This hook adjusts the integer scheduling priority @var{priority} of
5482 @var{insn}. It should return the new priority. Reduce the priority to
5483 execute @var{insn} earlier, increase the priority to execute @var{insn}
5484 later. Do not define this hook if you do not need to adjust the
5485 scheduling priorities of insns.
5488 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5489 This hook is executed by the scheduler after it has scheduled the ready
5490 list, to allow the machine description to reorder it (for example to
5491 combine two small instructions together on @samp{VLIW} machines).
5492 @var{file} is either a null pointer, or a stdio stream to write any
5493 debug output to. @var{verbose} is the verbose level provided by
5494 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5495 list of instructions that are ready to be scheduled. @var{n_readyp} is
5496 a pointer to the number of elements in the ready list. The scheduler
5497 reads the ready list in reverse order, starting with
5498 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5499 is the timer tick of the scheduler. You may modify the ready list and
5500 the number of ready insns. The return value is the number of insns that
5501 can issue this cycle; normally this is just @code{issue_rate}. See also
5502 @samp{TARGET_SCHED_REORDER2}.
5505 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5506 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5507 function is called whenever the scheduler starts a new cycle. This one
5508 is called once per iteration over a cycle, immediately after
5509 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5510 return the number of insns to be scheduled in the same cycle. Defining
5511 this hook can be useful if there are frequent situations where
5512 scheduling one insn causes other insns to become ready in the same
5513 cycle. These other insns can then be taken into account properly.
5516 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5517 This hook is called after evaluation forward dependencies of insns in
5518 chain given by two parameter values (@var{head} and @var{tail}
5519 correspondingly) but before insns scheduling of the insn chain. For
5520 example, it can be used for better insn classification if it requires
5521 analysis of dependencies. This hook can use backward and forward
5522 dependencies of the insn scheduler because they are already
5526 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5527 This hook is executed by the scheduler at the beginning of each block of
5528 instructions that are to be scheduled. @var{file} is either a null
5529 pointer, or a stdio stream to write any debug output to. @var{verbose}
5530 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5531 @var{max_ready} is the maximum number of insns in the current scheduling
5532 region that can be live at the same time. This can be used to allocate
5533 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5536 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5537 This hook is executed by the scheduler at the end of each block of
5538 instructions that are to be scheduled. It can be used to perform
5539 cleanup of any actions done by the other scheduling hooks. @var{file}
5540 is either a null pointer, or a stdio stream to write any debug output
5541 to. @var{verbose} is the verbose level provided by
5542 @option{-fsched-verbose-@var{n}}.
5545 @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void)
5546 This hook is called many times during insn scheduling. If the hook
5547 returns nonzero, the automaton based pipeline description is used for
5548 insn scheduling. Otherwise the traditional pipeline description is
5549 used. The default is usage of the traditional pipeline description.
5551 You should also remember that to simplify the insn scheduler sources
5552 an empty traditional pipeline description interface is generated even
5553 if there is no a traditional pipeline description in the @file{.md}
5554 file. The same is true for the automaton based pipeline description.
5555 That means that you should be accurate in defining the hook.
5558 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5559 The hook returns an RTL insn. The automaton state used in the
5560 pipeline hazard recognizer is changed as if the insn were scheduled
5561 when the new simulated processor cycle starts. Usage of the hook may
5562 simplify the automaton pipeline description for some @acronym{VLIW}
5563 processors. If the hook is defined, it is used only for the automaton
5564 based pipeline description. The default is not to change the state
5565 when the new simulated processor cycle starts.
5568 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5569 The hook can be used to initialize data used by the previous hook.
5572 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5573 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5574 to changed the state as if the insn were scheduled when the new
5575 simulated processor cycle finishes.
5578 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5579 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5580 used to initialize data used by the previous hook.
5583 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5584 This hook controls better choosing an insn from the ready insn queue
5585 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5586 chooses the first insn from the queue. If the hook returns a positive
5587 value, an additional scheduler code tries all permutations of
5588 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5589 subsequent ready insns to choose an insn whose issue will result in
5590 maximal number of issued insns on the same cycle. For the
5591 @acronym{VLIW} processor, the code could actually solve the problem of
5592 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5593 rules of @acronym{VLIW} packing are described in the automaton.
5595 This code also could be used for superscalar @acronym{RISC}
5596 processors. Let us consider a superscalar @acronym{RISC} processor
5597 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5598 @var{B}, some insns can be executed only in pipelines @var{B} or
5599 @var{C}, and one insn can be executed in pipeline @var{B}. The
5600 processor may issue the 1st insn into @var{A} and the 2nd one into
5601 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5602 until the next cycle. If the scheduler issues the 3rd insn the first,
5603 the processor could issue all 3 insns per cycle.
5605 Actually this code demonstrates advantages of the automaton based
5606 pipeline hazard recognizer. We try quickly and easy many insn
5607 schedules to choose the best one.
5609 The default is no multipass scheduling.
5612 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5614 This hook controls what insns from the ready insn queue will be
5615 considered for the multipass insn scheduling. If the hook returns
5616 zero for insn passed as the parameter, the insn will be not chosen to
5619 The default is that any ready insns can be chosen to be issued.
5622 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5624 This hook is called by the insn scheduler before issuing insn passed
5625 as the third parameter on given cycle. If the hook returns nonzero,
5626 the insn is not issued on given processors cycle. Instead of that,
5627 the processor cycle is advanced. If the value passed through the last
5628 parameter is zero, the insn ready queue is not sorted on the new cycle
5629 start as usually. The first parameter passes file for debugging
5630 output. The second one passes the scheduler verbose level of the
5631 debugging output. The forth and the fifth parameter values are
5632 correspondingly processor cycle on which the previous insn has been
5633 issued and the current processor cycle.
5636 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void)
5637 The @acronym{DFA}-based scheduler could take the insertion of nop
5638 operations for better insn scheduling into account. It can be done
5639 only if the multi-pass insn scheduling works (see hook
5640 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}).
5642 Let us consider a @acronym{VLIW} processor insn with 3 slots. Each
5643 insn can be placed only in one of the three slots. We have 3 ready
5644 insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be
5645 placed only in the 1st slot, @var{B} can be placed only in the 3rd
5646 slot. We described the automaton which does not permit empty slot
5647 gaps between insns (usually such description is simpler). Without
5648 this code the scheduler would place each insn in 3 separate
5649 @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd
5650 slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is
5651 the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook
5652 @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or
5653 create the nop insns.
5655 You should remember that the scheduler does not insert the nop insns.
5656 It is not wise because of the following optimizations. The scheduler
5657 only considers such possibility to improve the result schedule. The
5658 nop insns should be inserted lately, e.g. on the final phase.
5661 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index})
5662 This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert
5663 nop operations for better insn scheduling when @acronym{DFA}-based
5664 scheduler makes multipass insn scheduling (see also description of
5665 hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook
5666 returns a nop insn with given @var{index}. The indexes start with
5667 zero. The hook should return @code{NULL} if there are no more nop
5668 insns with indexes greater than given index.
5671 @deftypefn {Target Hook} bool IS_COSTLY_DEPENDENCE (rtx @var{insn1}, rtx @var{insn2}, rtx @var{dep_link}, int @var{dep_cost}, int @var{distance})
5672 This hook is used to define which dependences are considered costly by
5673 the target, so costly that it is not advisable to schedule the insns that
5674 are involved in the dependence too close to one another. The parameters
5675 to this hook are as follows: The second parameter @var{insn2} is dependent
5676 upon the first parameter @var{insn1}. The dependence between @var{insn1}
5677 and @var{insn2} is represented by the third parameter @var{dep_link}. The
5678 fourth parameter @var{cost} is the cost of the dependence, and the fifth
5679 parameter @var{distance} is the distance in cycles between the two insns.
5680 The hook returns @code{true} if considering the distance between the two
5681 insns the dependence between them is considered costly by the target,
5682 and @code{false} otherwise.
5684 Defining this hook can be useful in multiple-issue out-of-order machines,
5685 where (a) it's practically hopeless to predict the actual data/resource
5686 delays, however: (b) there's a better chance to predict the actual grouping
5687 that will be formed, and (c) correctly emulating the grouping can be very
5688 important. In such targets one may want to allow issuing dependent insns
5689 closer to one another - i.e, closer than the dependence distance; however,
5690 not in cases of "costly dependences", which this hooks allows to define.
5693 Macros in the following table are generated by the program
5694 @file{genattr} and can be useful for writing the hooks.
5696 @defmac TRADITIONAL_PIPELINE_INTERFACE
5697 The macro definition is generated if there is a traditional pipeline
5698 description in @file{.md} file. You should also remember that to
5699 simplify the insn scheduler sources an empty traditional pipeline
5700 description interface is generated even if there is no a traditional
5701 pipeline description in the @file{.md} file. The macro can be used to
5702 distinguish the two types of the traditional interface.
5705 @defmac DFA_PIPELINE_INTERFACE
5706 The macro definition is generated if there is an automaton pipeline
5707 description in @file{.md} file. You should also remember that to
5708 simplify the insn scheduler sources an empty automaton pipeline
5709 description interface is generated even if there is no an automaton
5710 pipeline description in the @file{.md} file. The macro can be used to
5711 distinguish the two types of the automaton interface.
5714 @defmac MAX_DFA_ISSUE_RATE
5715 The macro definition is generated in the automaton based pipeline
5716 description interface. Its value is calculated from the automaton
5717 based pipeline description and is equal to maximal number of all insns
5718 described in constructions @samp{define_insn_reservation} which can be
5719 issued on the same processor cycle.
5723 @section Dividing the Output into Sections (Texts, Data, @dots{})
5724 @c the above section title is WAY too long. maybe cut the part between
5725 @c the (...)? --mew 10feb93
5727 An object file is divided into sections containing different types of
5728 data. In the most common case, there are three sections: the @dfn{text
5729 section}, which holds instructions and read-only data; the @dfn{data
5730 section}, which holds initialized writable data; and the @dfn{bss
5731 section}, which holds uninitialized data. Some systems have other kinds
5734 The compiler must tell the assembler when to switch sections. These
5735 macros control what commands to output to tell the assembler this. You
5736 can also define additional sections.
5738 @defmac TEXT_SECTION_ASM_OP
5739 A C expression whose value is a string, including spacing, containing the
5740 assembler operation that should precede instructions and read-only data.
5741 Normally @code{"\t.text"} is right.
5744 @defmac TEXT_SECTION
5745 A C statement that switches to the default section containing instructions.
5746 Normally this is not needed, as simply defining @code{TEXT_SECTION_ASM_OP}
5747 is enough. The MIPS port uses this to sort all functions after all data
5751 @defmac HOT_TEXT_SECTION_NAME
5752 If defined, a C string constant for the name of the section containing most
5753 frequently executed functions of the program. If not defined, GCC will provide
5754 a default definition if the target supports named sections.
5757 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5758 If defined, a C string constant for the name of the section containing unlikely
5759 executed functions in the program.
5762 @defmac DATA_SECTION_ASM_OP
5763 A C expression whose value is a string, including spacing, containing the
5764 assembler operation to identify the following data as writable initialized
5765 data. Normally @code{"\t.data"} is right.
5768 @defmac READONLY_DATA_SECTION_ASM_OP
5769 A C expression whose value is a string, including spacing, containing the
5770 assembler operation to identify the following data as read-only initialized
5774 @defmac READONLY_DATA_SECTION
5775 A macro naming a function to call to switch to the proper section for
5776 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5777 if defined, else fall back to @code{text_section}.
5779 The most common definition will be @code{data_section}, if the target
5780 does not have a special read-only data section, and does not put data
5781 in the text section.
5784 @defmac SHARED_SECTION_ASM_OP
5785 If defined, a C expression whose value is a string, including spacing,
5786 containing the assembler operation to identify the following data as
5787 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5790 @defmac BSS_SECTION_ASM_OP
5791 If defined, a C expression whose value is a string, including spacing,
5792 containing the assembler operation to identify the following data as
5793 uninitialized global data. If not defined, and neither
5794 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5795 uninitialized global data will be output in the data section if
5796 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5800 @defmac SHARED_BSS_SECTION_ASM_OP
5801 If defined, a C expression whose value is a string, including spacing,
5802 containing the assembler operation to identify the following data as
5803 uninitialized global shared data. If not defined, and
5804 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5807 @defmac INIT_SECTION_ASM_OP
5808 If defined, a C expression whose value is a string, including spacing,
5809 containing the assembler operation to identify the following data as
5810 initialization code. If not defined, GCC will assume such a section does
5814 @defmac FINI_SECTION_ASM_OP
5815 If defined, a C expression whose value is a string, including spacing,
5816 containing the assembler operation to identify the following data as
5817 finalization code. If not defined, GCC will assume such a section does
5821 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5822 If defined, an ASM statement that switches to a different section
5823 via @var{section_op}, calls @var{function}, and switches back to
5824 the text section. This is used in @file{crtstuff.c} if
5825 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5826 to initialization and finalization functions from the init and fini
5827 sections. By default, this macro uses a simple function call. Some
5828 ports need hand-crafted assembly code to avoid dependencies on
5829 registers initialized in the function prologue or to ensure that
5830 constant pools don't end up too far way in the text section.
5833 @defmac FORCE_CODE_SECTION_ALIGN
5834 If defined, an ASM statement that aligns a code section to some
5835 arbitrary boundary. This is used to force all fragments of the
5836 @code{.init} and @code{.fini} sections to have to same alignment
5837 and thus prevent the linker from having to add any padding.
5842 @defmac EXTRA_SECTIONS
5843 A list of names for sections other than the standard two, which are
5844 @code{in_text} and @code{in_data}. You need not define this macro
5845 on a system with no other sections (that GCC needs to use).
5848 @findex text_section
5849 @findex data_section
5850 @defmac EXTRA_SECTION_FUNCTIONS
5851 One or more functions to be defined in @file{varasm.c}. These
5852 functions should do jobs analogous to those of @code{text_section} and
5853 @code{data_section}, for your additional sections. Do not define this
5854 macro if you do not define @code{EXTRA_SECTIONS}.
5857 @defmac JUMP_TABLES_IN_TEXT_SECTION
5858 Define this macro to be an expression with a nonzero value if jump
5859 tables (for @code{tablejump} insns) should be output in the text
5860 section, along with the assembler instructions. Otherwise, the
5861 readonly data section is used.
5863 This macro is irrelevant if there is no separate readonly data section.
5866 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5867 Switches to the appropriate section for output of @var{exp}. You can
5868 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5869 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5870 requires link-time relocations. Bit 0 is set when variable contains
5871 local relocations only, while bit 1 is set for global relocations.
5872 Select the section by calling @code{data_section} or one of the
5873 alternatives for other sections. @var{align} is the constant alignment
5876 The default version of this function takes care of putting read-only
5877 variables in @code{readonly_data_section}.
5880 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5881 Build up a unique section name, expressed as a @code{STRING_CST} node,
5882 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5883 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5884 the initial value of @var{exp} requires link-time relocations.
5886 The default version of this function appends the symbol name to the
5887 ELF section name that would normally be used for the symbol. For
5888 example, the function @code{foo} would be placed in @code{.text.foo}.
5889 Whatever the actual target object format, this is often good enough.
5892 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
5893 Switches to the appropriate section for output of constant pool entry
5894 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
5895 constant in RTL@. The argument @var{mode} is redundant except in the
5896 case of a @code{const_int} rtx. Select the section by calling
5897 @code{readonly_data_section} or one of the alternatives for other
5898 sections. @var{align} is the constant alignment in bits.
5900 The default version of this function takes care of putting symbolic
5901 constants in @code{flag_pic} mode in @code{data_section} and everything
5902 else in @code{readonly_data_section}.
5905 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
5906 Define this hook if references to a symbol or a constant must be
5907 treated differently depending on something about the variable or
5908 function named by the symbol (such as what section it is in).
5910 The hook is executed immediately after rtl has been created for
5911 @var{decl}, which may be a variable or function declaration or
5912 an entry in the constant pool. In either case, @var{rtl} is the
5913 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
5914 in this hook; that field may not have been initialized yet.
5916 In the case of a constant, it is safe to assume that the rtl is
5917 a @code{mem} whose address is a @code{symbol_ref}. Most decls
5918 will also have this form, but that is not guaranteed. Global
5919 register variables, for instance, will have a @code{reg} for their
5920 rtl. (Normally the right thing to do with such unusual rtl is
5923 The @var{new_decl_p} argument will be true if this is the first time
5924 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
5925 be false for subsequent invocations, which will happen for duplicate
5926 declarations. Whether or not anything must be done for the duplicate
5927 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
5928 @var{new_decl_p} is always true when the hook is called for a constant.
5930 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
5931 The usual thing for this hook to do is to record flags in the
5932 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
5933 Historically, the name string was modified if it was necessary to
5934 encode more than one bit of information, but this practice is now
5935 discouraged; use @code{SYMBOL_REF_FLAGS}.
5937 The default definition of this hook, @code{default_encode_section_info}
5938 in @file{varasm.c}, sets a number of commonly-useful bits in
5939 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
5940 before overriding it.
5943 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
5944 Decode @var{name} and return the real name part, sans
5945 the characters that @code{TARGET_ENCODE_SECTION_INFO}
5949 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
5950 Returns true if @var{exp} should be placed into a ``small data'' section.
5951 The default version of this hook always returns false.
5954 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
5955 Contains the value true if the target places read-only
5956 ``small data'' into a separate section. The default value is false.
5959 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
5960 Returns true if @var{exp} names an object for which name resolution
5961 rules must resolve to the current ``module'' (dynamic shared library
5962 or executable image).
5964 The default version of this hook implements the name resolution rules
5965 for ELF, which has a looser model of global name binding than other
5966 currently supported object file formats.
5969 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
5970 Contains the value true if the target supports thread-local storage.
5971 The default value is false.
5976 @section Position Independent Code
5977 @cindex position independent code
5980 This section describes macros that help implement generation of position
5981 independent code. Simply defining these macros is not enough to
5982 generate valid PIC; you must also add support to the macros
5983 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5984 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5985 @samp{movsi} to do something appropriate when the source operand
5986 contains a symbolic address. You may also need to alter the handling of
5987 switch statements so that they use relative addresses.
5988 @c i rearranged the order of the macros above to try to force one of
5989 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5991 @defmac PIC_OFFSET_TABLE_REGNUM
5992 The register number of the register used to address a table of static
5993 data addresses in memory. In some cases this register is defined by a
5994 processor's ``application binary interface'' (ABI)@. When this macro
5995 is defined, RTL is generated for this register once, as with the stack
5996 pointer and frame pointer registers. If this macro is not defined, it
5997 is up to the machine-dependent files to allocate such a register (if
5998 necessary). Note that this register must be fixed when in use (e.g.@:
5999 when @code{flag_pic} is true).
6002 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6003 Define this macro if the register defined by
6004 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6005 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6008 @defmac FINALIZE_PIC
6009 By generating position-independent code, when two different programs (A
6010 and B) share a common library (libC.a), the text of the library can be
6011 shared whether or not the library is linked at the same address for both
6012 programs. In some of these environments, position-independent code
6013 requires not only the use of different addressing modes, but also
6014 special code to enable the use of these addressing modes.
6016 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
6017 codes once the function is being compiled into assembly code, but not
6018 before. (It is not done before, because in the case of compiling an
6019 inline function, it would lead to multiple PIC prologues being
6020 included in functions which used inline functions and were compiled to
6024 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6025 A C expression that is nonzero if @var{x} is a legitimate immediate
6026 operand on the target machine when generating position independent code.
6027 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6028 check this. You can also assume @var{flag_pic} is true, so you need not
6029 check it either. You need not define this macro if all constants
6030 (including @code{SYMBOL_REF}) can be immediate operands when generating
6031 position independent code.
6034 @node Assembler Format
6035 @section Defining the Output Assembler Language
6037 This section describes macros whose principal purpose is to describe how
6038 to write instructions in assembler language---rather than what the
6042 * File Framework:: Structural information for the assembler file.
6043 * Data Output:: Output of constants (numbers, strings, addresses).
6044 * Uninitialized Data:: Output of uninitialized variables.
6045 * Label Output:: Output and generation of labels.
6046 * Initialization:: General principles of initialization
6047 and termination routines.
6048 * Macros for Initialization::
6049 Specific macros that control the handling of
6050 initialization and termination routines.
6051 * Instruction Output:: Output of actual instructions.
6052 * Dispatch Tables:: Output of jump tables.
6053 * Exception Region Output:: Output of exception region code.
6054 * Alignment Output:: Pseudo ops for alignment and skipping data.
6057 @node File Framework
6058 @subsection The Overall Framework of an Assembler File
6059 @cindex assembler format
6060 @cindex output of assembler code
6062 @c prevent bad page break with this line
6063 This describes the overall framework of an assembly file.
6065 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6066 @findex default_file_start
6067 Output to @code{asm_out_file} any text which the assembler expects to
6068 find at the beginning of a file. The default behavior is controlled
6069 by two flags, documented below. Unless your target's assembler is
6070 quite unusual, if you override the default, you should call
6071 @code{default_file_start} at some point in your target hook. This
6072 lets other target files rely on these variables.
6075 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6076 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6077 printed as the very first line in the assembly file, unless
6078 @option{-fverbose-asm} is in effect. (If that macro has been defined
6079 to the empty string, this variable has no effect.) With the normal
6080 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6081 assembler that it need not bother stripping comments or extra
6082 whitespace from its input. This allows it to work a bit faster.
6084 The default is false. You should not set it to true unless you have
6085 verified that your port does not generate any extra whitespace or
6086 comments that will cause GAS to issue errors in NO_APP mode.
6089 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6090 If this flag is true, @code{output_file_directive} will be called
6091 for the primary source file, immediately after printing
6092 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6093 this to be done. The default is false.
6096 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6097 Output to @code{asm_out_file} any text which the assembler expects
6098 to find at the end of a file. The default is to output nothing.
6101 @deftypefun void file_end_indicate_exec_stack ()
6102 Some systems use a common convention, the @samp{.note.GNU-stack}
6103 special section, to indicate whether or not an object file relies on
6104 the stack being executable. If your system uses this convention, you
6105 should define @code{TARGET_ASM_FILE_END} to this function. If you
6106 need to do other things in that hook, have your hook function call
6110 @defmac ASM_COMMENT_START
6111 A C string constant describing how to begin a comment in the target
6112 assembler language. The compiler assumes that the comment will end at
6113 the end of the line.
6117 A C string constant for text to be output before each @code{asm}
6118 statement or group of consecutive ones. Normally this is
6119 @code{"#APP"}, which is a comment that has no effect on most
6120 assemblers but tells the GNU assembler that it must check the lines
6121 that follow for all valid assembler constructs.
6125 A C string constant for text to be output after each @code{asm}
6126 statement or group of consecutive ones. Normally this is
6127 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6128 time-saving assumptions that are valid for ordinary compiler output.
6131 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6132 A C statement to output COFF information or DWARF debugging information
6133 which indicates that filename @var{name} is the current source file to
6134 the stdio stream @var{stream}.
6136 This macro need not be defined if the standard form of output
6137 for the file format in use is appropriate.
6140 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6141 A C statement to output the string @var{string} to the stdio stream
6142 @var{stream}. If you do not call the function @code{output_quoted_string}
6143 in your config files, GCC will only call it to output filenames to
6144 the assembler source. So you can use it to canonicalize the format
6145 of the filename using this macro.
6148 @defmac ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
6149 A C statement to output DBX or SDB debugging information before code
6150 for line number @var{line} of the current source file to the
6151 stdio stream @var{stream}. @var{counter} is the number of time the
6152 macro was invoked, including the current invocation; it is intended
6153 to generate unique labels in the assembly output.
6155 This macro need not be defined if the standard form of debugging
6156 information for the debugger in use is appropriate.
6159 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6160 A C statement to output something to the assembler file to handle a
6161 @samp{#ident} directive containing the text @var{string}. If this
6162 macro is not defined, nothing is output for a @samp{#ident} directive.
6165 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6166 Output assembly directives to switch to section @var{name}. The section
6167 should have attributes as specified by @var{flags}, which is a bit mask
6168 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6169 is nonzero, it contains an alignment in bytes to be used for the section,
6170 otherwise some target default should be used. Only targets that must
6171 specify an alignment within the section directive need pay attention to
6172 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6175 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6176 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6179 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6180 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6181 based on a variable or function decl, a section name, and whether or not the
6182 declaration's initializer may contain runtime relocations. @var{decl} may be
6183 null, in which case read-write data should be assumed.
6185 The default version if this function handles choosing code vs data,
6186 read-only vs read-write data, and @code{flag_pic}. You should only
6187 need to override this if your target has special flags that might be
6188 set via @code{__attribute__}.
6193 @subsection Output of Data
6196 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6197 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6198 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6199 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6200 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6201 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6202 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6203 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6204 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6205 These hooks specify assembly directives for creating certain kinds
6206 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6207 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6208 aligned two-byte object, and so on. Any of the hooks may be
6209 @code{NULL}, indicating that no suitable directive is available.
6211 The compiler will print these strings at the start of a new line,
6212 followed immediately by the object's initial value. In most cases,
6213 the string should contain a tab, a pseudo-op, and then another tab.
6216 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6217 The @code{assemble_integer} function uses this hook to output an
6218 integer object. @var{x} is the object's value, @var{size} is its size
6219 in bytes and @var{aligned_p} indicates whether it is aligned. The
6220 function should return @code{true} if it was able to output the
6221 object. If it returns false, @code{assemble_integer} will try to
6222 split the object into smaller parts.
6224 The default implementation of this hook will use the
6225 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6226 when the relevant string is @code{NULL}.
6229 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6230 A C statement to recognize @var{rtx} patterns that
6231 @code{output_addr_const} can't deal with, and output assembly code to
6232 @var{stream} corresponding to the pattern @var{x}. This may be used to
6233 allow machine-dependent @code{UNSPEC}s to appear within constants.
6235 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6236 @code{goto fail}, so that a standard error message is printed. If it
6237 prints an error message itself, by calling, for example,
6238 @code{output_operand_lossage}, it may just complete normally.
6241 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6242 A C statement to output to the stdio stream @var{stream} an assembler
6243 instruction to assemble a string constant containing the @var{len}
6244 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6245 @code{char *} and @var{len} a C expression of type @code{int}.
6247 If the assembler has a @code{.ascii} pseudo-op as found in the
6248 Berkeley Unix assembler, do not define the macro
6249 @code{ASM_OUTPUT_ASCII}.
6252 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6253 A C statement to output word @var{n} of a function descriptor for
6254 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6255 is defined, and is otherwise unused.
6258 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6259 You may define this macro as a C expression. You should define the
6260 expression to have a nonzero value if GCC should output the constant
6261 pool for a function before the code for the function, or a zero value if
6262 GCC should output the constant pool after the function. If you do
6263 not define this macro, the usual case, GCC will output the constant
6264 pool before the function.
6267 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6268 A C statement to output assembler commands to define the start of the
6269 constant pool for a function. @var{funname} is a string giving
6270 the name of the function. Should the return type of the function
6271 be required, it can be obtained via @var{fundecl}. @var{size}
6272 is the size, in bytes, of the constant pool that will be written
6273 immediately after this call.
6275 If no constant-pool prefix is required, the usual case, this macro need
6279 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6280 A C statement (with or without semicolon) to output a constant in the
6281 constant pool, if it needs special treatment. (This macro need not do
6282 anything for RTL expressions that can be output normally.)
6284 The argument @var{file} is the standard I/O stream to output the
6285 assembler code on. @var{x} is the RTL expression for the constant to
6286 output, and @var{mode} is the machine mode (in case @var{x} is a
6287 @samp{const_int}). @var{align} is the required alignment for the value
6288 @var{x}; you should output an assembler directive to force this much
6291 The argument @var{labelno} is a number to use in an internal label for
6292 the address of this pool entry. The definition of this macro is
6293 responsible for outputting the label definition at the proper place.
6294 Here is how to do this:
6297 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6300 When you output a pool entry specially, you should end with a
6301 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6302 entry from being output a second time in the usual manner.
6304 You need not define this macro if it would do nothing.
6307 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6308 A C statement to output assembler commands to at the end of the constant
6309 pool for a function. @var{funname} is a string giving the name of the
6310 function. Should the return type of the function be required, you can
6311 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6312 constant pool that GCC wrote immediately before this call.
6314 If no constant-pool epilogue is required, the usual case, you need not
6318 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6319 Define this macro as a C expression which is nonzero if @var{C} is
6320 used as a logical line separator by the assembler.
6322 If you do not define this macro, the default is that only
6323 the character @samp{;} is treated as a logical line separator.
6326 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6327 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6328 These target hooks are C string constants, describing the syntax in the
6329 assembler for grouping arithmetic expressions. If not overridden, they
6330 default to normal parentheses, which is correct for most assemblers.
6333 These macros are provided by @file{real.h} for writing the definitions
6334 of @code{ASM_OUTPUT_DOUBLE} and the like:
6336 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6337 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6338 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6339 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6340 floating point representation, and store its bit pattern in the variable
6341 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6342 be a simple @code{long int}. For the others, it should be an array of
6343 @code{long int}. The number of elements in this array is determined by
6344 the size of the desired target floating point data type: 32 bits of it
6345 go in each @code{long int} array element. Each array element holds 32
6346 bits of the result, even if @code{long int} is wider than 32 bits on the
6349 The array element values are designed so that you can print them out
6350 using @code{fprintf} in the order they should appear in the target
6354 @node Uninitialized Data
6355 @subsection Output of Uninitialized Variables
6357 Each of the macros in this section is used to do the whole job of
6358 outputting a single uninitialized variable.
6360 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6361 A C statement (sans semicolon) to output to the stdio stream
6362 @var{stream} the assembler definition of a common-label named
6363 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6364 is the size rounded up to whatever alignment the caller wants.
6366 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6367 output the name itself; before and after that, output the additional
6368 assembler syntax for defining the name, and a newline.
6370 This macro controls how the assembler definitions of uninitialized
6371 common global variables are output.
6374 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6375 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6376 separate, explicit argument. If you define this macro, it is used in
6377 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6378 handling the required alignment of the variable. The alignment is specified
6379 as the number of bits.
6382 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6383 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6384 variable to be output, if there is one, or @code{NULL_TREE} if there
6385 is no corresponding variable. If you define this macro, GCC will use it
6386 in place of both @code{ASM_OUTPUT_COMMON} and
6387 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6388 the variable's decl in order to chose what to output.
6391 @defmac ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6392 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6393 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6397 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6398 A C statement (sans semicolon) to output to the stdio stream
6399 @var{stream} the assembler definition of uninitialized global @var{decl} named
6400 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6401 is the size rounded up to whatever alignment the caller wants.
6403 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6404 defining this macro. If unable, use the expression
6405 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6406 before and after that, output the additional assembler syntax for defining
6407 the name, and a newline.
6409 This macro controls how the assembler definitions of uninitialized global
6410 variables are output. This macro exists to properly support languages like
6411 C++ which do not have @code{common} data. However, this macro currently
6412 is not defined for all targets. If this macro and
6413 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6414 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6415 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6418 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6419 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6420 separate, explicit argument. If you define this macro, it is used in
6421 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6422 handling the required alignment of the variable. The alignment is specified
6423 as the number of bits.
6425 Try to use function @code{asm_output_aligned_bss} defined in file
6426 @file{varasm.c} when defining this macro.
6429 @defmac ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6430 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6431 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6435 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6436 A C statement (sans semicolon) to output to the stdio stream
6437 @var{stream} the assembler definition of a local-common-label named
6438 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6439 is the size rounded up to whatever alignment the caller wants.
6441 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6442 output the name itself; before and after that, output the additional
6443 assembler syntax for defining the name, and a newline.
6445 This macro controls how the assembler definitions of uninitialized
6446 static variables are output.
6449 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6450 Like @code{ASM_OUTPUT_LOCAL} 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_LOCAL}, and gives you more flexibility in
6453 handling the required alignment of the variable. The alignment is specified
6454 as the number of bits.
6457 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6458 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6459 variable to be output, if there is one, or @code{NULL_TREE} if there
6460 is no corresponding variable. If you define this macro, GCC will use it
6461 in place of both @code{ASM_OUTPUT_DECL} and
6462 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6463 the variable's decl in order to chose what to output.
6466 @defmac ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6467 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6468 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6473 @subsection Output and Generation of Labels
6475 @c prevent bad page break with this line
6476 This is about outputting labels.
6478 @findex assemble_name
6479 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6480 A C statement (sans semicolon) to output to the stdio stream
6481 @var{stream} the assembler definition of a label named @var{name}.
6482 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6483 output the name itself; before and after that, output the additional
6484 assembler syntax for defining the name, and a newline. A default
6485 definition of this macro is provided which is correct for most systems.
6489 A C string containing the appropriate assembler directive to specify the
6490 size of a symbol, without any arguments. On systems that use ELF, the
6491 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6492 systems, the default is not to define this macro.
6494 Define this macro only if it is correct to use the default definitions
6495 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6496 for your system. If you need your own custom definitions of those
6497 macros, or if you do not need explicit symbol sizes at all, do not
6501 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6502 A C statement (sans semicolon) to output to the stdio stream
6503 @var{stream} a directive telling the assembler that the size of the
6504 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6505 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6509 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6510 A C statement (sans semicolon) to output to the stdio stream
6511 @var{stream} a directive telling the assembler to calculate the size of
6512 the symbol @var{name} by subtracting its address from the current
6515 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6516 provided. The default assumes that the assembler recognizes a special
6517 @samp{.} symbol as referring to the current address, and can calculate
6518 the difference between this and another symbol. If your assembler does
6519 not recognize @samp{.} or cannot do calculations with it, you will need
6520 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6524 A C string containing the appropriate assembler directive to specify the
6525 type of a symbol, without any arguments. On systems that use ELF, the
6526 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6527 systems, the default is not to define this macro.
6529 Define this macro only if it is correct to use the default definition of
6530 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6531 custom definition of this macro, or if you do not need explicit symbol
6532 types at all, do not define this macro.
6535 @defmac TYPE_OPERAND_FMT
6536 A C string which specifies (using @code{printf} syntax) the format of
6537 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6538 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6539 the default is not to define this macro.
6541 Define this macro only if it is correct to use the default definition of
6542 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6543 custom definition of this macro, or if you do not need explicit symbol
6544 types at all, do not define this macro.
6547 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6548 A C statement (sans semicolon) to output to the stdio stream
6549 @var{stream} a directive telling the assembler that the type of the
6550 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6551 that string is always either @samp{"function"} or @samp{"object"}, but
6552 you should not count on this.
6554 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6555 definition of this macro is provided.
6558 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6559 A C statement (sans semicolon) to output to the stdio stream
6560 @var{stream} any text necessary for declaring the name @var{name} of a
6561 function which is being defined. This macro is responsible for
6562 outputting the label definition (perhaps using
6563 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6564 @code{FUNCTION_DECL} tree node representing the function.
6566 If this macro is not defined, then the function name is defined in the
6567 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6569 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6573 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6574 A C statement (sans semicolon) to output to the stdio stream
6575 @var{stream} any text necessary for declaring the size of a function
6576 which is being defined. The argument @var{name} is the name of the
6577 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6578 representing the function.
6580 If this macro is not defined, then the function size is not defined.
6582 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6586 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6587 A C statement (sans semicolon) to output to the stdio stream
6588 @var{stream} any text necessary for declaring the name @var{name} of an
6589 initialized variable which is being defined. This macro must output the
6590 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6591 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6593 If this macro is not defined, then the variable name is defined in the
6594 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6596 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6597 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6600 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
6601 A C statement (sans semicolon) to output to the stdio stream
6602 @var{stream} any text necessary for declaring the name @var{name} of a
6603 constant which is being defined. This macro is responsible for
6604 outputting the label definition (perhaps using
6605 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
6606 value of the constant, and @var{size} is the size of the constant
6607 in bytes. @var{name} will be an internal label.
6609 If this macro is not defined, then the @var{name} is defined in the
6610 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6612 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6616 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6617 A C statement (sans semicolon) to output to the stdio stream
6618 @var{stream} any text necessary for claiming a register @var{regno}
6619 for a global variable @var{decl} with name @var{name}.
6621 If you don't define this macro, that is equivalent to defining it to do
6625 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6626 A C statement (sans semicolon) to finish up declaring a variable name
6627 once the compiler has processed its initializer fully and thus has had a
6628 chance to determine the size of an array when controlled by an
6629 initializer. This is used on systems where it's necessary to declare
6630 something about the size of the object.
6632 If you don't define this macro, that is equivalent to defining it to do
6635 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6636 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6639 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6640 This target hook is a function to output to the stdio stream
6641 @var{stream} some commands that will make the label @var{name} global;
6642 that is, available for reference from other files.
6644 The default implementation relies on a proper definition of
6645 @code{GLOBAL_ASM_OP}.
6648 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6649 A C statement (sans semicolon) to output to the stdio stream
6650 @var{stream} some commands that will make the label @var{name} weak;
6651 that is, available for reference from other files but only used if
6652 no other definition is available. Use the expression
6653 @code{assemble_name (@var{stream}, @var{name})} to output the name
6654 itself; before and after that, output the additional assembler syntax
6655 for making that name weak, and a newline.
6657 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6658 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6662 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6663 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6664 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6665 or variable decl. If @var{value} is not @code{NULL}, this C statement
6666 should output to the stdio stream @var{stream} assembler code which
6667 defines (equates) the weak symbol @var{name} to have the value
6668 @var{value}. If @var{value} is @code{NULL}, it should output commands
6669 to make @var{name} weak.
6672 @defmac SUPPORTS_WEAK
6673 A C expression which evaluates to true if the target supports weak symbols.
6675 If you don't define this macro, @file{defaults.h} provides a default
6676 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6677 is defined, the default definition is @samp{1}; otherwise, it is
6678 @samp{0}. Define this macro if you want to control weak symbol support
6679 with a compiler flag such as @option{-melf}.
6682 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
6683 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6684 public symbol such that extra copies in multiple translation units will
6685 be discarded by the linker. Define this macro if your object file
6686 format provides support for this concept, such as the @samp{COMDAT}
6687 section flags in the Microsoft Windows PE/COFF format, and this support
6688 requires changes to @var{decl}, such as putting it in a separate section.
6691 @defmac SUPPORTS_ONE_ONLY
6692 A C expression which evaluates to true if the target supports one-only
6695 If you don't define this macro, @file{varasm.c} provides a default
6696 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6697 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6698 you want to control one-only symbol support with a compiler flag, or if
6699 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6700 be emitted as one-only.
6703 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6704 This target hook is a function to output to @var{asm_out_file} some
6705 commands that will make the symbol(s) associated with @var{decl} have
6706 hidden, protected or internal visibility as specified by @var{visibility}.
6709 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6710 A C statement (sans semicolon) to output to the stdio stream
6711 @var{stream} any text necessary for declaring the name of an external
6712 symbol named @var{name} which is referenced in this compilation but
6713 not defined. The value of @var{decl} is the tree node for the
6716 This macro need not be defined if it does not need to output anything.
6717 The GNU assembler and most Unix assemblers don't require anything.
6720 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
6721 This target hook is a function to output to @var{asm_out_file} an assembler
6722 pseudo-op to declare a library function name external. The name of the
6723 library function is given by @var{symref}, which is a @code{symbol_ref}.
6726 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6727 A C statement (sans semicolon) to output to the stdio stream
6728 @var{stream} a reference in assembler syntax to a label named
6729 @var{name}. This should add @samp{_} to the front of the name, if that
6730 is customary on your operating system, as it is in most Berkeley Unix
6731 systems. This macro is used in @code{assemble_name}.
6734 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6735 A C statement (sans semicolon) to output a reference to
6736 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6737 will be used to output the name of the symbol. This macro may be used
6738 to modify the way a symbol is referenced depending on information
6739 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6742 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6743 A C statement (sans semicolon) to output a reference to @var{buf}, the
6744 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6745 @code{assemble_name} will be used to output the name of the symbol.
6746 This macro is not used by @code{output_asm_label}, or the @code{%l}
6747 specifier that calls it; the intention is that this macro should be set
6748 when it is necessary to output a label differently when its address is
6752 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6753 A function to output to the stdio stream @var{stream} a label whose
6754 name is made from the string @var{prefix} and the number @var{labelno}.
6756 It is absolutely essential that these labels be distinct from the labels
6757 used for user-level functions and variables. Otherwise, certain programs
6758 will have name conflicts with internal labels.
6760 It is desirable to exclude internal labels from the symbol table of the
6761 object file. Most assemblers have a naming convention for labels that
6762 should be excluded; on many systems, the letter @samp{L} at the
6763 beginning of a label has this effect. You should find out what
6764 convention your system uses, and follow it.
6766 The default version of this function utilizes ASM_GENERATE_INTERNAL_LABEL.
6769 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6770 A C statement to output to the stdio stream @var{stream} a debug info
6771 label whose name is made from the string @var{prefix} and the number
6772 @var{num}. This is useful for VLIW targets, where debug info labels
6773 may need to be treated differently than branch target labels. On some
6774 systems, branch target labels must be at the beginning of instruction
6775 bundles, but debug info labels can occur in the middle of instruction
6778 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
6782 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6783 A C statement to store into the string @var{string} a label whose name
6784 is made from the string @var{prefix} and the number @var{num}.
6786 This string, when output subsequently by @code{assemble_name}, should
6787 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
6788 with the same @var{prefix} and @var{num}.
6790 If the string begins with @samp{*}, then @code{assemble_name} will
6791 output the rest of the string unchanged. It is often convenient for
6792 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6793 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6794 to output the string, and may change it. (Of course,
6795 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6796 you should know what it does on your machine.)
6799 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6800 A C expression to assign to @var{outvar} (which is a variable of type
6801 @code{char *}) a newly allocated string made from the string
6802 @var{name} and the number @var{number}, with some suitable punctuation
6803 added. Use @code{alloca} to get space for the string.
6805 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6806 produce an assembler label for an internal static variable whose name is
6807 @var{name}. Therefore, the string must be such as to result in valid
6808 assembler code. The argument @var{number} is different each time this
6809 macro is executed; it prevents conflicts between similarly-named
6810 internal static variables in different scopes.
6812 Ideally this string should not be a valid C identifier, to prevent any
6813 conflict with the user's own symbols. Most assemblers allow periods
6814 or percent signs in assembler symbols; putting at least one of these
6815 between the name and the number will suffice.
6817 If this macro is not defined, a default definition will be provided
6818 which is correct for most systems.
6821 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6822 A C statement to output to the stdio stream @var{stream} assembler code
6823 which defines (equates) the symbol @var{name} to have the value @var{value}.
6826 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6827 correct for most systems.
6830 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6831 A C statement to output to the stdio stream @var{stream} assembler code
6832 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6833 to have the value of the tree node @var{decl_of_value}. This macro will
6834 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6835 the tree nodes are available.
6838 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6839 correct for most systems.
6842 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6843 A C statement to output to the stdio stream @var{stream} assembler code
6844 which defines (equates) the weak symbol @var{name} to have the value
6845 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6846 an undefined weak symbol.
6848 Define this macro if the target only supports weak aliases; define
6849 @code{ASM_OUTPUT_DEF} instead if possible.
6852 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6853 Define this macro to override the default assembler names used for
6854 Objective-C methods.
6856 The default name is a unique method number followed by the name of the
6857 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6858 the category is also included in the assembler name (e.g.@:
6861 These names are safe on most systems, but make debugging difficult since
6862 the method's selector is not present in the name. Therefore, particular
6863 systems define other ways of computing names.
6865 @var{buf} is an expression of type @code{char *} which gives you a
6866 buffer in which to store the name; its length is as long as
6867 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6868 50 characters extra.
6870 The argument @var{is_inst} specifies whether the method is an instance
6871 method or a class method; @var{class_name} is the name of the class;
6872 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6873 in a category); and @var{sel_name} is the name of the selector.
6875 On systems where the assembler can handle quoted names, you can use this
6876 macro to provide more human-readable names.
6879 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
6880 A C statement (sans semicolon) to output to the stdio stream
6881 @var{stream} commands to declare that the label @var{name} is an
6882 Objective-C class reference. This is only needed for targets whose
6883 linkers have special support for NeXT-style runtimes.
6886 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6887 A C statement (sans semicolon) to output to the stdio stream
6888 @var{stream} commands to declare that the label @var{name} is an
6889 unresolved Objective-C class reference. This is only needed for targets
6890 whose linkers have special support for NeXT-style runtimes.
6893 @node Initialization
6894 @subsection How Initialization Functions Are Handled
6895 @cindex initialization routines
6896 @cindex termination routines
6897 @cindex constructors, output of
6898 @cindex destructors, output of
6900 The compiled code for certain languages includes @dfn{constructors}
6901 (also called @dfn{initialization routines})---functions to initialize
6902 data in the program when the program is started. These functions need
6903 to be called before the program is ``started''---that is to say, before
6904 @code{main} is called.
6906 Compiling some languages generates @dfn{destructors} (also called
6907 @dfn{termination routines}) that should be called when the program
6910 To make the initialization and termination functions work, the compiler
6911 must output something in the assembler code to cause those functions to
6912 be called at the appropriate time. When you port the compiler to a new
6913 system, you need to specify how to do this.
6915 There are two major ways that GCC currently supports the execution of
6916 initialization and termination functions. Each way has two variants.
6917 Much of the structure is common to all four variations.
6919 @findex __CTOR_LIST__
6920 @findex __DTOR_LIST__
6921 The linker must build two lists of these functions---a list of
6922 initialization functions, called @code{__CTOR_LIST__}, and a list of
6923 termination functions, called @code{__DTOR_LIST__}.
6925 Each list always begins with an ignored function pointer (which may hold
6926 0, @minus{}1, or a count of the function pointers after it, depending on
6927 the environment). This is followed by a series of zero or more function
6928 pointers to constructors (or destructors), followed by a function
6929 pointer containing zero.
6931 Depending on the operating system and its executable file format, either
6932 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6933 time and exit time. Constructors are called in reverse order of the
6934 list; destructors in forward order.
6936 The best way to handle static constructors works only for object file
6937 formats which provide arbitrarily-named sections. A section is set
6938 aside for a list of constructors, and another for a list of destructors.
6939 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6940 object file that defines an initialization function also puts a word in
6941 the constructor section to point to that function. The linker
6942 accumulates all these words into one contiguous @samp{.ctors} section.
6943 Termination functions are handled similarly.
6945 This method will be chosen as the default by @file{target-def.h} if
6946 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
6947 support arbitrary sections, but does support special designated
6948 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
6949 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
6951 When arbitrary sections are available, there are two variants, depending
6952 upon how the code in @file{crtstuff.c} is called. On systems that
6953 support a @dfn{.init} section which is executed at program startup,
6954 parts of @file{crtstuff.c} are compiled into that section. The
6955 program is linked by the @command{gcc} driver like this:
6958 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
6961 The prologue of a function (@code{__init}) appears in the @code{.init}
6962 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
6963 for the function @code{__fini} in the @dfn{.fini} section. Normally these
6964 files are provided by the operating system or by the GNU C library, but
6965 are provided by GCC for a few targets.
6967 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
6968 compiled from @file{crtstuff.c}. They contain, among other things, code
6969 fragments within the @code{.init} and @code{.fini} sections that branch
6970 to routines in the @code{.text} section. The linker will pull all parts
6971 of a section together, which results in a complete @code{__init} function
6972 that invokes the routines we need at startup.
6974 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
6977 If no init section is available, when GCC compiles any function called
6978 @code{main} (or more accurately, any function designated as a program
6979 entry point by the language front end calling @code{expand_main_function}),
6980 it inserts a procedure call to @code{__main} as the first executable code
6981 after the function prologue. The @code{__main} function is defined
6982 in @file{libgcc2.c} and runs the global constructors.
6984 In file formats that don't support arbitrary sections, there are again
6985 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
6986 and an `a.out' format must be used. In this case,
6987 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
6988 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
6989 and with the address of the void function containing the initialization
6990 code as its value. The GNU linker recognizes this as a request to add
6991 the value to a @dfn{set}; the values are accumulated, and are eventually
6992 placed in the executable as a vector in the format described above, with
6993 a leading (ignored) count and a trailing zero element.
6994 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
6995 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
6996 the compilation of @code{main} to call @code{__main} as above, starting
6997 the initialization process.
6999 The last variant uses neither arbitrary sections nor the GNU linker.
7000 This is preferable when you want to do dynamic linking and when using
7001 file formats which the GNU linker does not support, such as `ECOFF'@. In
7002 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7003 termination functions are recognized simply by their names. This requires
7004 an extra program in the linkage step, called @command{collect2}. This program
7005 pretends to be the linker, for use with GCC; it does its job by running
7006 the ordinary linker, but also arranges to include the vectors of
7007 initialization and termination functions. These functions are called
7008 via @code{__main} as described above. In order to use this method,
7009 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7012 The following section describes the specific macros that control and
7013 customize the handling of initialization and termination functions.
7016 @node Macros for Initialization
7017 @subsection Macros Controlling Initialization Routines
7019 Here are the macros that control how the compiler handles initialization
7020 and termination functions:
7022 @defmac INIT_SECTION_ASM_OP
7023 If defined, a C string constant, including spacing, for the assembler
7024 operation to identify the following data as initialization code. If not
7025 defined, GCC will assume such a section does not exist. When you are
7026 using special sections for initialization and termination functions, this
7027 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7028 run the initialization functions.
7031 @defmac HAS_INIT_SECTION
7032 If defined, @code{main} will not call @code{__main} as described above.
7033 This macro should be defined for systems that control start-up code
7034 on a symbol-by-symbol basis, such as OSF/1, and should not
7035 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7038 @defmac LD_INIT_SWITCH
7039 If defined, a C string constant for a switch that tells the linker that
7040 the following symbol is an initialization routine.
7043 @defmac LD_FINI_SWITCH
7044 If defined, a C string constant for a switch that tells the linker that
7045 the following symbol is a finalization routine.
7048 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7049 If defined, a C statement that will write a function that can be
7050 automatically called when a shared library is loaded. The function
7051 should call @var{func}, which takes no arguments. If not defined, and
7052 the object format requires an explicit initialization function, then a
7053 function called @code{_GLOBAL__DI} will be generated.
7055 This function and the following one are used by collect2 when linking a
7056 shared library that needs constructors or destructors, or has DWARF2
7057 exception tables embedded in the code.
7060 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7061 If defined, a C statement that will write a function that can be
7062 automatically called when a shared library is unloaded. The function
7063 should call @var{func}, which takes no arguments. If not defined, and
7064 the object format requires an explicit finalization function, then a
7065 function called @code{_GLOBAL__DD} will be generated.
7068 @defmac INVOKE__main
7069 If defined, @code{main} will call @code{__main} despite the presence of
7070 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7071 where the init section is not actually run automatically, but is still
7072 useful for collecting the lists of constructors and destructors.
7075 @defmac SUPPORTS_INIT_PRIORITY
7076 If nonzero, the C++ @code{init_priority} attribute is supported and the
7077 compiler should emit instructions to control the order of initialization
7078 of objects. If zero, the compiler will issue an error message upon
7079 encountering an @code{init_priority} attribute.
7082 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7083 This value is true if the target supports some ``native'' method of
7084 collecting constructors and destructors to be run at startup and exit.
7085 It is false if we must use @command{collect2}.
7088 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7089 If defined, a function that outputs assembler code to arrange to call
7090 the function referenced by @var{symbol} at initialization time.
7092 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7093 no arguments and with no return value. If the target supports initialization
7094 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7095 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7097 If this macro is not defined by the target, a suitable default will
7098 be chosen if (1) the target supports arbitrary section names, (2) the
7099 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7103 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7104 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7105 functions rather than initialization functions.
7108 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7109 generated for the generated object file will have static linkage.
7111 If your system uses @command{collect2} as the means of processing
7112 constructors, then that program normally uses @command{nm} to scan
7113 an object file for constructor functions to be called.
7115 On certain kinds of systems, you can define this macro to make
7116 @command{collect2} work faster (and, in some cases, make it work at all):
7118 @defmac OBJECT_FORMAT_COFF
7119 Define this macro if the system uses COFF (Common Object File Format)
7120 object files, so that @command{collect2} can assume this format and scan
7121 object files directly for dynamic constructor/destructor functions.
7123 This macro is effective only in a native compiler; @command{collect2} as
7124 part of a cross compiler always uses @command{nm} for the target machine.
7127 @defmac REAL_NM_FILE_NAME
7128 Define this macro as a C string constant containing the file name to use
7129 to execute @command{nm}. The default is to search the path normally for
7132 If your system supports shared libraries and has a program to list the
7133 dynamic dependencies of a given library or executable, you can define
7134 these macros to enable support for running initialization and
7135 termination functions in shared libraries:
7139 Define this macro to a C string constant containing the name of the program
7140 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7143 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7144 Define this macro to be C code that extracts filenames from the output
7145 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7146 of type @code{char *} that points to the beginning of a line of output
7147 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7148 code must advance @var{ptr} to the beginning of the filename on that
7149 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7152 @node Instruction Output
7153 @subsection Output of Assembler Instructions
7155 @c prevent bad page break with this line
7156 This describes assembler instruction output.
7158 @defmac REGISTER_NAMES
7159 A C initializer containing the assembler's names for the machine
7160 registers, each one as a C string constant. This is what translates
7161 register numbers in the compiler into assembler language.
7164 @defmac ADDITIONAL_REGISTER_NAMES
7165 If defined, a C initializer for an array of structures containing a name
7166 and a register number. This macro defines additional names for hard
7167 registers, thus allowing the @code{asm} option in declarations to refer
7168 to registers using alternate names.
7171 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7172 Define this macro if you are using an unusual assembler that
7173 requires different names for the machine instructions.
7175 The definition is a C statement or statements which output an
7176 assembler instruction opcode to the stdio stream @var{stream}. The
7177 macro-operand @var{ptr} is a variable of type @code{char *} which
7178 points to the opcode name in its ``internal'' form---the form that is
7179 written in the machine description. The definition should output the
7180 opcode name to @var{stream}, performing any translation you desire, and
7181 increment the variable @var{ptr} to point at the end of the opcode
7182 so that it will not be output twice.
7184 In fact, your macro definition may process less than the entire opcode
7185 name, or more than the opcode name; but if you want to process text
7186 that includes @samp{%}-sequences to substitute operands, you must take
7187 care of the substitution yourself. Just be sure to increment
7188 @var{ptr} over whatever text should not be output normally.
7190 @findex recog_data.operand
7191 If you need to look at the operand values, they can be found as the
7192 elements of @code{recog_data.operand}.
7194 If the macro definition does nothing, the instruction is output
7198 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7199 If defined, a C statement to be executed just prior to the output of
7200 assembler code for @var{insn}, to modify the extracted operands so
7201 they will be output differently.
7203 Here the argument @var{opvec} is the vector containing the operands
7204 extracted from @var{insn}, and @var{noperands} is the number of
7205 elements of the vector which contain meaningful data for this insn.
7206 The contents of this vector are what will be used to convert the insn
7207 template into assembler code, so you can change the assembler output
7208 by changing the contents of the vector.
7210 This macro is useful when various assembler syntaxes share a single
7211 file of instruction patterns; by defining this macro differently, you
7212 can cause a large class of instructions to be output differently (such
7213 as with rearranged operands). Naturally, variations in assembler
7214 syntax affecting individual insn patterns ought to be handled by
7215 writing conditional output routines in those patterns.
7217 If this macro is not defined, it is equivalent to a null statement.
7220 @defmac FINAL_PRESCAN_LABEL
7221 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
7222 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
7223 @var{noperands} will be zero.
7226 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7227 A C compound statement to output to stdio stream @var{stream} the
7228 assembler syntax for an instruction operand @var{x}. @var{x} is an
7231 @var{code} is a value that can be used to specify one of several ways
7232 of printing the operand. It is used when identical operands must be
7233 printed differently depending on the context. @var{code} comes from
7234 the @samp{%} specification that was used to request printing of the
7235 operand. If the specification was just @samp{%@var{digit}} then
7236 @var{code} is 0; if the specification was @samp{%@var{ltr}
7237 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7240 If @var{x} is a register, this macro should print the register's name.
7241 The names can be found in an array @code{reg_names} whose type is
7242 @code{char *[]}. @code{reg_names} is initialized from
7243 @code{REGISTER_NAMES}.
7245 When the machine description has a specification @samp{%@var{punct}}
7246 (a @samp{%} followed by a punctuation character), this macro is called
7247 with a null pointer for @var{x} and the punctuation character for
7251 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7252 A C expression which evaluates to true if @var{code} is a valid
7253 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7254 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7255 punctuation characters (except for the standard one, @samp{%}) are used
7259 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7260 A C compound statement to output to stdio stream @var{stream} the
7261 assembler syntax for an instruction operand that is a memory reference
7262 whose address is @var{x}. @var{x} is an RTL expression.
7264 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7265 On some machines, the syntax for a symbolic address depends on the
7266 section that the address refers to. On these machines, define the hook
7267 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7268 @code{symbol_ref}, and then check for it here. @xref{Assembler
7272 @findex dbr_sequence_length
7273 @defmac DBR_OUTPUT_SEQEND (@var{file})
7274 A C statement, to be executed after all slot-filler instructions have
7275 been output. If necessary, call @code{dbr_sequence_length} to
7276 determine the number of slots filled in a sequence (zero if not
7277 currently outputting a sequence), to decide how many no-ops to output,
7280 Don't define this macro if it has nothing to do, but it is helpful in
7281 reading assembly output if the extent of the delay sequence is made
7282 explicit (e.g.@: with white space).
7285 @findex final_sequence
7286 Note that output routines for instructions with delay slots must be
7287 prepared to deal with not being output as part of a sequence
7288 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7289 found.) The variable @code{final_sequence} is null when not
7290 processing a sequence, otherwise it contains the @code{sequence} rtx
7294 @defmac REGISTER_PREFIX
7295 @defmacx LOCAL_LABEL_PREFIX
7296 @defmacx USER_LABEL_PREFIX
7297 @defmacx IMMEDIATE_PREFIX
7298 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7299 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7300 @file{final.c}). These are useful when a single @file{md} file must
7301 support multiple assembler formats. In that case, the various @file{tm.h}
7302 files can define these macros differently.
7305 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
7306 If defined this macro should expand to a series of @code{case}
7307 statements which will be parsed inside the @code{switch} statement of
7308 the @code{asm_fprintf} function. This allows targets to define extra
7309 printf formats which may useful when generating their assembler
7310 statements. Note that uppercase letters are reserved for future
7311 generic extensions to asm_fprintf, and so are not available to target
7312 specific code. The output file is given by the parameter @var{file}.
7313 The varargs input pointer is @var{argptr} and the rest of the format
7314 string, starting the character after the one that is being switched
7315 upon, is pointed to by @var{format}.
7318 @defmac ASSEMBLER_DIALECT
7319 If your target supports multiple dialects of assembler language (such as
7320 different opcodes), define this macro as a C expression that gives the
7321 numeric index of the assembler language dialect to use, with zero as the
7324 If this macro is defined, you may use constructs of the form
7326 @samp{@{option0|option1|option2@dots{}@}}
7329 in the output templates of patterns (@pxref{Output Template}) or in the
7330 first argument of @code{asm_fprintf}. This construct outputs
7331 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7332 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7333 within these strings retain their usual meaning. If there are fewer
7334 alternatives within the braces than the value of
7335 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7337 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7338 @samp{@}} do not have any special meaning when used in templates or
7339 operands to @code{asm_fprintf}.
7341 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7342 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7343 the variations in assembler language syntax with that mechanism. Define
7344 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7345 if the syntax variant are larger and involve such things as different
7346 opcodes or operand order.
7349 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7350 A C expression to output to @var{stream} some assembler code
7351 which will push hard register number @var{regno} onto the stack.
7352 The code need not be optimal, since this macro is used only when
7356 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7357 A C expression to output to @var{stream} some assembler code
7358 which will pop hard register number @var{regno} off of the stack.
7359 The code need not be optimal, since this macro is used only when
7363 @node Dispatch Tables
7364 @subsection Output of Dispatch Tables
7366 @c prevent bad page break with this line
7367 This concerns dispatch tables.
7369 @cindex dispatch table
7370 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7371 A C statement to output to the stdio stream @var{stream} an assembler
7372 pseudo-instruction to generate a difference between two labels.
7373 @var{value} and @var{rel} are the numbers of two internal labels. The
7374 definitions of these labels are output using
7375 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7376 way here. For example,
7379 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7380 @var{value}, @var{rel})
7383 You must provide this macro on machines where the addresses in a
7384 dispatch table are relative to the table's own address. If defined, GCC
7385 will also use this macro on all machines when producing PIC@.
7386 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7387 mode and flags can be read.
7390 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7391 This macro should be provided on machines where the addresses
7392 in a dispatch table are absolute.
7394 The definition should be a C statement to output to the stdio stream
7395 @var{stream} an assembler pseudo-instruction to generate a reference to
7396 a label. @var{value} is the number of an internal label whose
7397 definition is output using @code{(*targetm.asm_out.internal_label)}.
7401 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7405 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7406 Define this if the label before a jump-table needs to be output
7407 specially. The first three arguments are the same as for
7408 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7409 jump-table which follows (a @code{jump_insn} containing an
7410 @code{addr_vec} or @code{addr_diff_vec}).
7412 This feature is used on system V to output a @code{swbeg} statement
7415 If this macro is not defined, these labels are output with
7416 @code{(*targetm.asm_out.internal_label)}.
7419 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7420 Define this if something special must be output at the end of a
7421 jump-table. The definition should be a C statement to be executed
7422 after the assembler code for the table is written. It should write
7423 the appropriate code to stdio stream @var{stream}. The argument
7424 @var{table} is the jump-table insn, and @var{num} is the label-number
7425 of the preceding label.
7427 If this macro is not defined, nothing special is output at the end of
7431 @node Exception Region Output
7432 @subsection Assembler Commands for Exception Regions
7434 @c prevent bad page break with this line
7436 This describes commands marking the start and the end of an exception
7439 @defmac EH_FRAME_SECTION_NAME
7440 If defined, a C string constant for the name of the section containing
7441 exception handling frame unwind information. If not defined, GCC will
7442 provide a default definition if the target supports named sections.
7443 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7445 You should define this symbol if your target supports DWARF 2 frame
7446 unwind information and the default definition does not work.
7449 @defmac EH_FRAME_IN_DATA_SECTION
7450 If defined, DWARF 2 frame unwind information will be placed in the
7451 data section even though the target supports named sections. This
7452 might be necessary, for instance, if the system linker does garbage
7453 collection and sections cannot be marked as not to be collected.
7455 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7459 @defmac MASK_RETURN_ADDR
7460 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7461 that it does not contain any extraneous set bits in it.
7464 @defmac DWARF2_UNWIND_INFO
7465 Define this macro to 0 if your target supports DWARF 2 frame unwind
7466 information, but it does not yet work with exception handling.
7467 Otherwise, if your target supports this information (if it defines
7468 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7469 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7472 If this macro is defined to 1, the DWARF 2 unwinder will be the default
7473 exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by
7476 If this macro is defined to anything, the DWARF 2 unwinder will be used
7477 instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case.
7480 @defmac MUST_USE_SJLJ_EXCEPTIONS
7481 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
7482 runtime-variable. In that case, @file{except.h} cannot correctly
7483 determine the corresponding definition of
7484 @code{MUST_USE_SJLJ_EXCEPTIONS}, so the target must provide it directly.
7487 @defmac DWARF_CIE_DATA_ALIGNMENT
7488 This macro need only be defined if the target might save registers in the
7489 function prologue at an offset to the stack pointer that is not aligned to
7490 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7491 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7492 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7493 the target supports DWARF 2 frame unwind information.
7496 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7497 If defined, a function that switches to the section in which the main
7498 exception table is to be placed (@pxref{Sections}). The default is a
7499 function that switches to a section named @code{.gcc_except_table} on
7500 machines that support named sections via
7501 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7502 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7503 @code{readonly_data_section}.
7506 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7507 If defined, a function that switches to the section in which the DWARF 2
7508 frame unwind information to be placed (@pxref{Sections}). The default
7509 is a function that outputs a standard GAS section directive, if
7510 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7511 directive followed by a synthetic label.
7514 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7515 Contains the value true if the target should add a zero word onto the
7516 end of a Dwarf-2 frame info section when used for exception handling.
7517 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7521 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
7522 Given a register, this hook should return a parallel of registers to
7523 represent where to find the register pieces. Define this hook if the
7524 register and its mode are represented in Dwarf in non-contiguous
7525 locations, or if the register should be represented in more than one
7526 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
7527 If not defined, the default is to return @code{NULL_RTX}.
7530 @node Alignment Output
7531 @subsection Assembler Commands for Alignment
7533 @c prevent bad page break with this line
7534 This describes commands for alignment.
7536 @defmac JUMP_ALIGN (@var{label})
7537 The alignment (log base 2) to put in front of @var{label}, which is
7538 a common destination of jumps and has no fallthru incoming edge.
7540 This macro need not be defined if you don't want any special alignment
7541 to be done at such a time. Most machine descriptions do not currently
7544 Unless it's necessary to inspect the @var{label} parameter, it is better
7545 to set the variable @var{align_jumps} in the target's
7546 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7547 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7550 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
7551 The alignment (log base 2) to put in front of @var{label}, which follows
7554 This macro need not be defined if you don't want any special alignment
7555 to be done at such a time. Most machine descriptions do not currently
7559 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7560 The maximum number of bytes to skip when applying
7561 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7562 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7565 @defmac LOOP_ALIGN (@var{label})
7566 The alignment (log base 2) to put in front of @var{label}, which follows
7567 a @code{NOTE_INSN_LOOP_BEG} note.
7569 This macro need not be defined if you don't want any special alignment
7570 to be done at such a time. Most machine descriptions do not currently
7573 Unless it's necessary to inspect the @var{label} parameter, it is better
7574 to set the variable @code{align_loops} in the target's
7575 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7576 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7579 @defmac LOOP_ALIGN_MAX_SKIP
7580 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7581 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7584 @defmac LABEL_ALIGN (@var{label})
7585 The alignment (log base 2) to put in front of @var{label}.
7586 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7587 the maximum of the specified values is used.
7589 Unless it's necessary to inspect the @var{label} parameter, it is better
7590 to set the variable @code{align_labels} in the target's
7591 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7592 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7595 @defmac LABEL_ALIGN_MAX_SKIP
7596 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7597 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7600 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7601 A C statement to output to the stdio stream @var{stream} an assembler
7602 instruction to advance the location counter by @var{nbytes} bytes.
7603 Those bytes should be zero when loaded. @var{nbytes} will be a C
7604 expression of type @code{int}.
7607 @defmac ASM_NO_SKIP_IN_TEXT
7608 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7609 text section because it fails to put zeros in the bytes that are skipped.
7610 This is true on many Unix systems, where the pseudo--op to skip bytes
7611 produces no-op instructions rather than zeros when used in the text
7615 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7616 A C statement to output to the stdio stream @var{stream} an assembler
7617 command to advance the location counter to a multiple of 2 to the
7618 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7621 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7622 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7623 for padding, if necessary.
7626 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7627 A C statement to output to the stdio stream @var{stream} an assembler
7628 command to advance the location counter to a multiple of 2 to the
7629 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7630 satisfy the alignment request. @var{power} and @var{max_skip} will be
7631 a C expression of type @code{int}.
7635 @node Debugging Info
7636 @section Controlling Debugging Information Format
7638 @c prevent bad page break with this line
7639 This describes how to specify debugging information.
7642 * All Debuggers:: Macros that affect all debugging formats uniformly.
7643 * DBX Options:: Macros enabling specific options in DBX format.
7644 * DBX Hooks:: Hook macros for varying DBX format.
7645 * File Names and DBX:: Macros controlling output of file names in DBX format.
7646 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7647 * VMS Debug:: Macros for VMS debug format.
7651 @subsection Macros Affecting All Debugging Formats
7653 @c prevent bad page break with this line
7654 These macros affect all debugging formats.
7656 @defmac DBX_REGISTER_NUMBER (@var{regno})
7657 A C expression that returns the DBX register number for the compiler
7658 register number @var{regno}. In the default macro provided, the value
7659 of this expression will be @var{regno} itself. But sometimes there are
7660 some registers that the compiler knows about and DBX does not, or vice
7661 versa. In such cases, some register may need to have one number in the
7662 compiler and another for DBX@.
7664 If two registers have consecutive numbers inside GCC, and they can be
7665 used as a pair to hold a multiword value, then they @emph{must} have
7666 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7667 Otherwise, debuggers will be unable to access such a pair, because they
7668 expect register pairs to be consecutive in their own numbering scheme.
7670 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7671 does not preserve register pairs, then what you must do instead is
7672 redefine the actual register numbering scheme.
7675 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
7676 A C expression that returns the integer offset value for an automatic
7677 variable having address @var{x} (an RTL expression). The default
7678 computation assumes that @var{x} is based on the frame-pointer and
7679 gives the offset from the frame-pointer. This is required for targets
7680 that produce debugging output for DBX or COFF-style debugging output
7681 for SDB and allow the frame-pointer to be eliminated when the
7682 @option{-g} options is used.
7685 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7686 A C expression that returns the integer offset value for an argument
7687 having address @var{x} (an RTL expression). The nominal offset is
7691 @defmac PREFERRED_DEBUGGING_TYPE
7692 A C expression that returns the type of debugging output GCC should
7693 produce when the user specifies just @option{-g}. Define
7694 this if you have arranged for GCC to support more than one format of
7695 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7696 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7697 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7699 When the user specifies @option{-ggdb}, GCC normally also uses the
7700 value of this macro to select the debugging output format, but with two
7701 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7702 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7703 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7704 defined, GCC uses @code{DBX_DEBUG}.
7706 The value of this macro only affects the default debugging output; the
7707 user can always get a specific type of output by using @option{-gstabs},
7708 @option{-gcoff}, @option{-gdwarf-1}, @option{-gdwarf-2}, @option{-gxcoff},
7713 @subsection Specific Options for DBX Output
7715 @c prevent bad page break with this line
7716 These are specific options for DBX output.
7718 @defmac DBX_DEBUGGING_INFO
7719 Define this macro if GCC should produce debugging output for DBX
7720 in response to the @option{-g} option.
7723 @defmac XCOFF_DEBUGGING_INFO
7724 Define this macro if GCC should produce XCOFF format debugging output
7725 in response to the @option{-g} option. This is a variant of DBX format.
7728 @defmac DEFAULT_GDB_EXTENSIONS
7729 Define this macro to control whether GCC should by default generate
7730 GDB's extended version of DBX debugging information (assuming DBX-format
7731 debugging information is enabled at all). If you don't define the
7732 macro, the default is 1: always generate the extended information
7733 if there is any occasion to.
7736 @defmac DEBUG_SYMS_TEXT
7737 Define this macro if all @code{.stabs} commands should be output while
7738 in the text section.
7741 @defmac ASM_STABS_OP
7742 A C string constant, including spacing, naming the assembler pseudo op to
7743 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7744 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7745 applies only to DBX debugging information format.
7748 @defmac ASM_STABD_OP
7749 A C string constant, including spacing, naming the assembler pseudo op to
7750 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7751 value is the current location. If you don't define this macro,
7752 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7756 @defmac ASM_STABN_OP
7757 A C string constant, including spacing, naming the assembler pseudo op to
7758 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7759 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7760 macro applies only to DBX debugging information format.
7763 @defmac DBX_NO_XREFS
7764 Define this macro if DBX on your system does not support the construct
7765 @samp{xs@var{tagname}}. On some systems, this construct is used to
7766 describe a forward reference to a structure named @var{tagname}.
7767 On other systems, this construct is not supported at all.
7770 @defmac DBX_CONTIN_LENGTH
7771 A symbol name in DBX-format debugging information is normally
7772 continued (split into two separate @code{.stabs} directives) when it
7773 exceeds a certain length (by default, 80 characters). On some
7774 operating systems, DBX requires this splitting; on others, splitting
7775 must not be done. You can inhibit splitting by defining this macro
7776 with the value zero. You can override the default splitting-length by
7777 defining this macro as an expression for the length you desire.
7780 @defmac DBX_CONTIN_CHAR
7781 Normally continuation is indicated by adding a @samp{\} character to
7782 the end of a @code{.stabs} string when a continuation follows. To use
7783 a different character instead, define this macro as a character
7784 constant for the character you want to use. Do not define this macro
7785 if backslash is correct for your system.
7788 @defmac DBX_STATIC_STAB_DATA_SECTION
7789 Define this macro if it is necessary to go to the data section before
7790 outputting the @samp{.stabs} pseudo-op for a non-global static
7794 @defmac DBX_TYPE_DECL_STABS_CODE
7795 The value to use in the ``code'' field of the @code{.stabs} directive
7796 for a typedef. The default is @code{N_LSYM}.
7799 @defmac DBX_STATIC_CONST_VAR_CODE
7800 The value to use in the ``code'' field of the @code{.stabs} directive
7801 for a static variable located in the text section. DBX format does not
7802 provide any ``right'' way to do this. The default is @code{N_FUN}.
7805 @defmac DBX_REGPARM_STABS_CODE
7806 The value to use in the ``code'' field of the @code{.stabs} directive
7807 for a parameter passed in registers. DBX format does not provide any
7808 ``right'' way to do this. The default is @code{N_RSYM}.
7811 @defmac DBX_REGPARM_STABS_LETTER
7812 The letter to use in DBX symbol data to identify a symbol as a parameter
7813 passed in registers. DBX format does not customarily provide any way to
7814 do this. The default is @code{'P'}.
7817 @defmac DBX_MEMPARM_STABS_LETTER
7818 The letter to use in DBX symbol data to identify a symbol as a stack
7819 parameter. The default is @code{'p'}.
7822 @defmac DBX_FUNCTION_FIRST
7823 Define this macro if the DBX information for a function and its
7824 arguments should precede the assembler code for the function. Normally,
7825 in DBX format, the debugging information entirely follows the assembler
7829 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
7830 Define this macro if the value of a symbol describing the scope of a
7831 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7832 of the enclosing function. Normally, GCC uses an absolute address.
7835 @defmac DBX_USE_BINCL
7836 Define this macro if GCC should generate @code{N_BINCL} and
7837 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7838 macro also directs GCC to output a type number as a pair of a file
7839 number and a type number within the file. Normally, GCC does not
7840 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7841 number for a type number.
7845 @subsection Open-Ended Hooks for DBX Format
7847 @c prevent bad page break with this line
7848 These are hooks for DBX format.
7850 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7851 Define this macro to say how to output to @var{stream} the debugging
7852 information for the start of a scope level for variable names. The
7853 argument @var{name} is the name of an assembler symbol (for use with
7854 @code{assemble_name}) whose value is the address where the scope begins.
7857 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7858 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7861 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
7862 Define this macro if the target machine requires special handling to
7863 output an @code{N_FUN} entry for the function @var{decl}.
7866 @defmac DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7867 Define this macro if the target machine requires special output at the
7868 end of the debugging information for a function. The definition should
7869 be a C statement (sans semicolon) to output the appropriate information
7870 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7874 @defmac DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7875 Define this macro if you need to control the order of output of the
7876 standard data types at the beginning of compilation. The argument
7877 @var{syms} is a @code{tree} which is a chain of all the predefined
7878 global symbols, including names of data types.
7880 Normally, DBX output starts with definitions of the types for integers
7881 and characters, followed by all the other predefined types of the
7882 particular language in no particular order.
7884 On some machines, it is necessary to output different particular types
7885 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
7886 those symbols in the necessary order. Any predefined types that you
7887 don't explicitly output will be output afterward in no particular order.
7889 Be careful not to define this macro so that it works only for C@. There
7890 are no global variables to access most of the built-in types, because
7891 another language may have another set of types. The way to output a
7892 particular type is to look through @var{syms} to see if you can find it.
7898 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7899 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
7901 dbxout_symbol (decl);
7907 This does nothing if the expected type does not exist.
7909 See the function @code{init_decl_processing} in @file{c-decl.c} to find
7910 the names to use for all the built-in C types.
7912 Here is another way of finding a particular type:
7914 @c this is still overfull. --mew 10feb93
7918 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7919 if (TREE_CODE (decl) == TYPE_DECL
7920 && (TREE_CODE (TREE_TYPE (decl))
7922 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
7923 && TYPE_UNSIGNED (TREE_TYPE (decl)))
7925 /* @r{This must be @code{unsigned short}.} */
7926 dbxout_symbol (decl);
7933 @defmac NO_DBX_FUNCTION_END
7934 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7935 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
7936 On those machines, define this macro to turn this feature off without
7937 disturbing the rest of the gdb extensions.
7940 @node File Names and DBX
7941 @subsection File Names in DBX Format
7943 @c prevent bad page break with this line
7944 This describes file names in DBX format.
7946 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7947 A C statement to output DBX debugging information to the stdio stream
7948 @var{stream} which indicates that file @var{name} is the main source
7949 file---the file specified as the input file for compilation.
7950 This macro is called only once, at the beginning of compilation.
7952 This macro need not be defined if the standard form of output
7953 for DBX debugging information is appropriate.
7956 @defmac DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
7957 A C statement to output DBX debugging information to the stdio stream
7958 @var{stream} which indicates that the current directory during
7959 compilation is named @var{name}.
7961 This macro need not be defined if the standard form of output
7962 for DBX debugging information is appropriate.
7965 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7966 A C statement to output DBX debugging information at the end of
7967 compilation of the main source file @var{name}.
7969 If you don't define this macro, nothing special is output at the end
7970 of compilation, which is correct for most machines.
7975 @subsection Macros for SDB and DWARF Output
7977 @c prevent bad page break with this line
7978 Here are macros for SDB and DWARF output.
7980 @defmac SDB_DEBUGGING_INFO
7981 Define this macro if GCC should produce COFF-style debugging output
7982 for SDB in response to the @option{-g} option.
7985 @defmac DWARF_DEBUGGING_INFO
7986 Define this macro if GCC should produce dwarf format debugging output
7987 in response to the @option{-g} option.
7990 @defmac DWARF2_DEBUGGING_INFO
7991 Define this macro if GCC should produce dwarf version 2 format
7992 debugging output in response to the @option{-g} option.
7994 To support optional call frame debugging information, you must also
7995 define @code{INCOMING_RETURN_ADDR_RTX} and either set
7996 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
7997 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
7998 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8001 @defmac DWARF2_FRAME_INFO
8002 Define this macro to a nonzero value if GCC should always output
8003 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8004 (@pxref{Exception Region Output} is nonzero, GCC will output this
8005 information not matter how you define @code{DWARF2_FRAME_INFO}.
8008 @defmac LINKER_DOES_NOT_WORK_WITH_DWARF2
8009 Define this macro if the linker does not work with Dwarf version 2.
8010 Normally, if the user specifies only @option{-ggdb} GCC will use Dwarf
8011 version 2 if available; this macro disables this. See the description
8012 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
8015 @defmac DWARF2_GENERATE_TEXT_SECTION_LABEL
8016 By default, the Dwarf 2 debugging information generator will generate a
8017 label to mark the beginning of the text section. If it is better simply
8018 to use the name of the text section itself, rather than an explicit label,
8019 to indicate the beginning of the text section, define this macro to zero.
8022 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8023 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8024 line debug info sections. This will result in much more compact line number
8025 tables, and hence is desirable if it works.
8028 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8029 A C statement to issue assembly directives that create a difference
8030 between the two given labels, using an integer of the given size.
8033 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8034 A C statement to issue assembly directives that create a
8035 section-relative reference to the given label, using an integer of the
8039 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8040 A C statement to issue assembly directives that create a self-relative
8041 reference to the given label, using an integer of the given size.
8044 @defmac PUT_SDB_@dots{}
8045 Define these macros to override the assembler syntax for the special
8046 SDB assembler directives. See @file{sdbout.c} for a list of these
8047 macros and their arguments. If the standard syntax is used, you need
8048 not define them yourself.
8052 Some assemblers do not support a semicolon as a delimiter, even between
8053 SDB assembler directives. In that case, define this macro to be the
8054 delimiter to use (usually @samp{\n}). It is not necessary to define
8055 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8059 @defmac SDB_GENERATE_FAKE
8060 Define this macro to override the usual method of constructing a dummy
8061 name for anonymous structure and union types. See @file{sdbout.c} for
8065 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8066 Define this macro to allow references to unknown structure,
8067 union, or enumeration tags to be emitted. Standard COFF does not
8068 allow handling of unknown references, MIPS ECOFF has support for
8072 @defmac SDB_ALLOW_FORWARD_REFERENCES
8073 Define this macro to allow references to structure, union, or
8074 enumeration tags that have not yet been seen to be handled. Some
8075 assemblers choke if forward tags are used, while some require it.
8080 @subsection Macros for VMS Debug Format
8082 @c prevent bad page break with this line
8083 Here are macros for VMS debug format.
8085 @defmac VMS_DEBUGGING_INFO
8086 Define this macro if GCC should produce debugging output for VMS
8087 in response to the @option{-g} option. The default behavior for VMS
8088 is to generate minimal debug info for a traceback in the absence of
8089 @option{-g} unless explicitly overridden with @option{-g0}. This
8090 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8091 @code{OVERRIDE_OPTIONS}.
8094 @node Floating Point
8095 @section Cross Compilation and Floating Point
8096 @cindex cross compilation and floating point
8097 @cindex floating point and cross compilation
8099 While all modern machines use twos-complement representation for integers,
8100 there are a variety of representations for floating point numbers. This
8101 means that in a cross-compiler the representation of floating point numbers
8102 in the compiled program may be different from that used in the machine
8103 doing the compilation.
8105 Because different representation systems may offer different amounts of
8106 range and precision, all floating point constants must be represented in
8107 the target machine's format. Therefore, the cross compiler cannot
8108 safely use the host machine's floating point arithmetic; it must emulate
8109 the target's arithmetic. To ensure consistency, GCC always uses
8110 emulation to work with floating point values, even when the host and
8111 target floating point formats are identical.
8113 The following macros are provided by @file{real.h} for the compiler to
8114 use. All parts of the compiler which generate or optimize
8115 floating-point calculations must use these macros. They may evaluate
8116 their operands more than once, so operands must not have side effects.
8118 @defmac REAL_VALUE_TYPE
8119 The C data type to be used to hold a floating point value in the target
8120 machine's format. Typically this is a @code{struct} containing an
8121 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8125 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8126 Compares for equality the two values, @var{x} and @var{y}. If the target
8127 floating point format supports negative zeroes and/or NaNs,
8128 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8129 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8132 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8133 Tests whether @var{x} is less than @var{y}.
8136 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8137 Truncates @var{x} to a signed integer, rounding toward zero.
8140 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8141 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8142 @var{x} is negative, returns zero.
8145 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8146 Converts @var{string} into a floating point number in the target machine's
8147 representation for mode @var{mode}. This routine can handle both
8148 decimal and hexadecimal floating point constants, using the syntax
8149 defined by the C language for both.
8152 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8153 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8156 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8157 Determines whether @var{x} represents infinity (positive or negative).
8160 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8161 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8164 @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})
8165 Calculates an arithmetic operation on the two floating point values
8166 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8169 The operation to be performed is specified by @var{code}. Only the
8170 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8171 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8173 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8174 target's floating point format cannot represent infinity, it will call
8175 @code{abort}. Callers should check for this situation first, using
8176 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8179 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8180 Returns the negative of the floating point value @var{x}.
8183 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8184 Returns the absolute value of @var{x}.
8187 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8188 Truncates the floating point value @var{x} to fit in @var{mode}. The
8189 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8190 appropriate bit pattern to be output asa floating constant whose
8191 precision accords with mode @var{mode}.
8194 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8195 Converts a floating point value @var{x} into a double-precision integer
8196 which is then stored into @var{low} and @var{high}. If the value is not
8197 integral, it is truncated.
8200 @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})
8201 Converts a double-precision integer found in @var{low} and @var{high},
8202 into a floating point value which is then stored into @var{x}. The
8203 value is truncated to fit in mode @var{mode}.
8206 @node Mode Switching
8207 @section Mode Switching Instructions
8208 @cindex mode switching
8209 The following macros control mode switching optimizations:
8211 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8212 Define this macro if the port needs extra instructions inserted for mode
8213 switching in an optimizing compilation.
8215 For an example, the SH4 can perform both single and double precision
8216 floating point operations, but to perform a single precision operation,
8217 the FPSCR PR bit has to be cleared, while for a double precision
8218 operation, this bit has to be set. Changing the PR bit requires a general
8219 purpose register as a scratch register, hence these FPSCR sets have to
8220 be inserted before reload, i.e.@: you can't put this into instruction emitting
8221 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
8223 You can have multiple entities that are mode-switched, and select at run time
8224 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8225 return nonzero for any @var{entity} that needs mode-switching.
8226 If you define this macro, you also have to define
8227 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8228 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8229 @code{NORMAL_MODE} is optional.
8232 @defmac NUM_MODES_FOR_MODE_SWITCHING
8233 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8234 initializer for an array of integers. Each initializer element
8235 N refers to an entity that needs mode switching, and specifies the number
8236 of different modes that might need to be set for this entity.
8237 The position of the initializer in the initializer - starting counting at
8238 zero - determines the integer that is used to refer to the mode-switched
8240 In macros that take mode arguments / yield a mode result, modes are
8241 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8242 switch is needed / supplied.
8245 @defmac MODE_NEEDED (@var{entity}, @var{insn})
8246 @var{entity} is an integer specifying a mode-switched entity. If
8247 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8248 return an integer value not larger than the corresponding element in
8249 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8250 be switched into prior to the execution of @var{insn}.
8253 @defmac NORMAL_MODE (@var{entity})
8254 If this macro is defined, it is evaluated for every @var{entity} that needs
8255 mode switching. It should evaluate to an integer, which is a mode that
8256 @var{entity} is assumed to be switched to at function entry and exit.
8259 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8260 This macro specifies the order in which modes for @var{entity} are processed.
8261 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8262 lowest. The value of the macro should be an integer designating a mode
8263 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8264 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8265 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8268 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8269 Generate one or more insns to set @var{entity} to @var{mode}.
8270 @var{hard_reg_live} is the set of hard registers live at the point where
8271 the insn(s) are to be inserted.
8274 @node Target Attributes
8275 @section Defining target-specific uses of @code{__attribute__}
8276 @cindex target attributes
8277 @cindex machine attributes
8278 @cindex attributes, target-specific
8280 Target-specific attributes may be defined for functions, data and types.
8281 These are described using the following target hooks; they also need to
8282 be documented in @file{extend.texi}.
8284 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8285 If defined, this target hook points to an array of @samp{struct
8286 attribute_spec} (defined in @file{tree.h}) specifying the machine
8287 specific attributes for this target and some of the restrictions on the
8288 entities to which these attributes are applied and the arguments they
8292 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8293 If defined, this target hook is a function which returns zero if the attributes on
8294 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8295 and two if they are nearly compatible (which causes a warning to be
8296 generated). If this is not defined, machine-specific attributes are
8297 supposed always to be compatible.
8300 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8301 If defined, this target hook is a function which assigns default attributes to
8302 newly defined @var{type}.
8305 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8306 Define this target hook if the merging of type attributes needs special
8307 handling. If defined, the result is a list of the combined
8308 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8309 that @code{comptypes} has already been called and returned 1. This
8310 function may call @code{merge_attributes} to handle machine-independent
8314 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8315 Define this target hook if the merging of decl attributes needs special
8316 handling. If defined, the result is a list of the combined
8317 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8318 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8319 when this is needed are when one attribute overrides another, or when an
8320 attribute is nullified by a subsequent definition. This function may
8321 call @code{merge_attributes} to handle machine-independent merging.
8323 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8324 If the only target-specific handling you require is @samp{dllimport} for
8325 Windows targets, you should define the macro
8326 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function
8327 called @code{merge_dllimport_decl_attributes} which can then be defined
8328 as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done
8329 in @file{i386/cygwin.h} and @file{i386/i386.c}, for example.
8332 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8333 Define this target hook if you want to be able to add attributes to a decl
8334 when it is being created. This is normally useful for back ends which
8335 wish to implement a pragma by using the attributes which correspond to
8336 the pragma's effect. The @var{node} argument is the decl which is being
8337 created. The @var{attr_ptr} argument is a pointer to the attribute list
8338 for this decl. The list itself should not be modified, since it may be
8339 shared with other decls, but attributes may be chained on the head of
8340 the list and @code{*@var{attr_ptr}} modified to point to the new
8341 attributes, or a copy of the list may be made if further changes are
8345 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8347 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8348 into the current function, despite its having target-specific
8349 attributes, @code{false} otherwise. By default, if a function has a
8350 target specific attribute attached to it, it will not be inlined.
8353 @node MIPS Coprocessors
8354 @section Defining coprocessor specifics for MIPS targets.
8355 @cindex MIPS coprocessor-definition macros
8357 The MIPS specification allows MIPS implementations to have as many as 4
8358 coprocessors, each with as many as 32 private registers. gcc supports
8359 accessing these registers and transferring values between the registers
8360 and memory using asm-ized variables. For example:
8363 register unsigned int cp0count asm ("c0r1");
8369 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8370 names may be added as described below, or the default names may be
8371 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8373 Coprocessor registers are assumed to be epilogue-used; sets to them will
8374 be preserved even if it does not appear that the register is used again
8375 later in the function.
8377 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8378 the FPU. One accesses COP1 registers through standard mips
8379 floating-point support; they are not included in this mechanism.
8381 There is one macro used in defining the MIPS coprocessor interface which
8382 you may want to override in subtargets; it is described below.
8384 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
8385 A comma-separated list (with leading comma) of pairs describing the
8386 alternate names of coprocessor registers. The format of each entry should be
8388 @{ @var{alternatename}, @var{register_number}@}
8394 @section Miscellaneous Parameters
8395 @cindex parameters, miscellaneous
8397 @c prevent bad page break with this line
8398 Here are several miscellaneous parameters.
8400 @defmac PREDICATE_CODES
8401 Define this if you have defined special-purpose predicates in the file
8402 @file{@var{machine}.c}. This macro is called within an initializer of an
8403 array of structures. The first field in the structure is the name of a
8404 predicate and the second field is an array of rtl codes. For each
8405 predicate, list all rtl codes that can be in expressions matched by the
8406 predicate. The list should have a trailing comma. Here is an example
8407 of two entries in the list for a typical RISC machine:
8410 #define PREDICATE_CODES \
8411 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8412 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8415 Defining this macro does not affect the generated code (however,
8416 incorrect definitions that omit an rtl code that may be matched by the
8417 predicate can cause the compiler to malfunction). Instead, it allows
8418 the table built by @file{genrecog} to be more compact and efficient,
8419 thus speeding up the compiler. The most important predicates to include
8420 in the list specified by this macro are those used in the most insn
8423 For each predicate function named in @code{PREDICATE_CODES}, a
8424 declaration will be generated in @file{insn-codes.h}.
8427 @defmac SPECIAL_MODE_PREDICATES
8428 Define this if you have special predicates that know special things
8429 about modes. Genrecog will warn about certain forms of
8430 @code{match_operand} without a mode; if the operand predicate is
8431 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8434 Here is an example from the IA-32 port (@code{ext_register_operand}
8435 specially checks for @code{HImode} or @code{SImode} in preparation
8436 for a byte extraction from @code{%ah} etc.).
8439 #define SPECIAL_MODE_PREDICATES \
8440 "ext_register_operand",
8444 @defmac CASE_VECTOR_MODE
8445 An alias for a machine mode name. This is the machine mode that
8446 elements of a jump-table should have.
8449 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8450 Optional: return the preferred mode for an @code{addr_diff_vec}
8451 when the minimum and maximum offset are known. If you define this,
8452 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8453 To make this work, you also have to define @code{INSN_ALIGN} and
8454 make the alignment for @code{addr_diff_vec} explicit.
8455 The @var{body} argument is provided so that the offset_unsigned and scale
8456 flags can be updated.
8459 @defmac CASE_VECTOR_PC_RELATIVE
8460 Define this macro to be a C expression to indicate when jump-tables
8461 should contain relative addresses. If jump-tables never contain
8462 relative addresses, then you need not define this macro.
8465 @defmac CASE_DROPS_THROUGH
8466 Define this if control falls through a @code{case} insn when the index
8467 value is out of range. This means the specified default-label is
8468 actually ignored by the @code{case} insn proper.
8471 @defmac CASE_VALUES_THRESHOLD
8472 Define this to be the smallest number of different values for which it
8473 is best to use a jump-table instead of a tree of conditional branches.
8474 The default is four for machines with a @code{casesi} instruction and
8475 five otherwise. This is best for most machines.
8478 @defmac CASE_USE_BIT_TESTS
8479 Define this macro to be a C expression to indicate whether C switch
8480 statements may be implemented by a sequence of bit tests. This is
8481 advantageous on processors that can efficiently implement left shift
8482 of 1 by the number of bits held in a register, but inappropriate on
8483 targets that would require a loop. By default, this macro returns
8484 @code{true} if the target defines an @code{ashlsi3} pattern, and
8485 @code{false} otherwise.
8488 @defmac WORD_REGISTER_OPERATIONS
8489 Define this macro if operations between registers with integral mode
8490 smaller than a word are always performed on the entire register.
8491 Most RISC machines have this property and most CISC machines do not.
8494 @defmac LOAD_EXTEND_OP (@var{mode})
8495 Define this macro to be a C expression indicating when insns that read
8496 memory in @var{mode}, an integral mode narrower than a word, set the
8497 bits outside of @var{mode} to be either the sign-extension or the
8498 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8499 of @var{mode} for which the
8500 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8501 @code{NIL} for other modes.
8503 This macro is not called with @var{mode} non-integral or with a width
8504 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8505 value in this case. Do not define this macro if it would always return
8506 @code{NIL}. On machines where this macro is defined, you will normally
8507 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8510 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
8511 Define this macro if loading short immediate values into registers sign
8515 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
8516 Define this macro if the same instructions that convert a floating
8517 point number to a signed fixed point number also convert validly to an
8522 The maximum number of bytes that a single instruction can move quickly
8523 between memory and registers or between two memory locations.
8526 @defmac MAX_MOVE_MAX
8527 The maximum number of bytes that a single instruction can move quickly
8528 between memory and registers or between two memory locations. If this
8529 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8530 constant value that is the largest value that @code{MOVE_MAX} can have
8534 @defmac SHIFT_COUNT_TRUNCATED
8535 A C expression that is nonzero if on this machine the number of bits
8536 actually used for the count of a shift operation is equal to the number
8537 of bits needed to represent the size of the object being shifted. When
8538 this macro is nonzero, the compiler will assume that it is safe to omit
8539 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8540 truncates the count of a shift operation. On machines that have
8541 instructions that act on bit-fields at variable positions, which may
8542 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8543 also enables deletion of truncations of the values that serve as
8544 arguments to bit-field instructions.
8546 If both types of instructions truncate the count (for shifts) and
8547 position (for bit-field operations), or if no variable-position bit-field
8548 instructions exist, you should define this macro.
8550 However, on some machines, such as the 80386 and the 680x0, truncation
8551 only applies to shift operations and not the (real or pretended)
8552 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8553 such machines. Instead, add patterns to the @file{md} file that include
8554 the implied truncation of the shift instructions.
8556 You need not define this macro if it would always have the value of zero.
8559 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8560 A C expression which is nonzero if on this machine it is safe to
8561 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8562 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8563 operating on it as if it had only @var{outprec} bits.
8565 On many machines, this expression can be 1.
8567 @c rearranged this, removed the phrase "it is reported that". this was
8568 @c to fix an overfull hbox. --mew 10feb93
8569 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8570 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8571 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8572 such cases may improve things.
8575 @defmac STORE_FLAG_VALUE
8576 A C expression describing the value returned by a comparison operator
8577 with an integral mode and stored by a store-flag instruction
8578 (@samp{s@var{cond}}) when the condition is true. This description must
8579 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8580 comparison operators whose results have a @code{MODE_INT} mode.
8582 A value of 1 or @minus{}1 means that the instruction implementing the
8583 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8584 and 0 when the comparison is false. Otherwise, the value indicates
8585 which bits of the result are guaranteed to be 1 when the comparison is
8586 true. This value is interpreted in the mode of the comparison
8587 operation, which is given by the mode of the first operand in the
8588 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8589 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8592 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8593 generate code that depends only on the specified bits. It can also
8594 replace comparison operators with equivalent operations if they cause
8595 the required bits to be set, even if the remaining bits are undefined.
8596 For example, on a machine whose comparison operators return an
8597 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8598 @samp{0x80000000}, saying that just the sign bit is relevant, the
8602 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8609 (ashift:SI @var{x} (const_int @var{n}))
8613 where @var{n} is the appropriate shift count to move the bit being
8614 tested into the sign bit.
8616 There is no way to describe a machine that always sets the low-order bit
8617 for a true value, but does not guarantee the value of any other bits,
8618 but we do not know of any machine that has such an instruction. If you
8619 are trying to port GCC to such a machine, include an instruction to
8620 perform a logical-and of the result with 1 in the pattern for the
8621 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8623 Often, a machine will have multiple instructions that obtain a value
8624 from a comparison (or the condition codes). Here are rules to guide the
8625 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8630 Use the shortest sequence that yields a valid definition for
8631 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8632 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8633 comparison operators to do so because there may be opportunities to
8634 combine the normalization with other operations.
8637 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8638 slightly preferred on machines with expensive jumps and 1 preferred on
8642 As a second choice, choose a value of @samp{0x80000001} if instructions
8643 exist that set both the sign and low-order bits but do not define the
8647 Otherwise, use a value of @samp{0x80000000}.
8650 Many machines can produce both the value chosen for
8651 @code{STORE_FLAG_VALUE} and its negation in the same number of
8652 instructions. On those machines, you should also define a pattern for
8653 those cases, e.g., one matching
8656 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8659 Some machines can also perform @code{and} or @code{plus} operations on
8660 condition code values with less instructions than the corresponding
8661 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8662 machines, define the appropriate patterns. Use the names @code{incscc}
8663 and @code{decscc}, respectively, for the patterns which perform
8664 @code{plus} or @code{minus} operations on condition code values. See
8665 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8666 find such instruction sequences on other machines.
8668 If this macro is not defined, the default value, 1, is used. You need
8669 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8670 instructions, or if the value generated by these instructions is 1.
8673 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
8674 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8675 returned when comparison operators with floating-point results are true.
8676 Define this macro on machine that have comparison operations that return
8677 floating-point values. If there are no such operations, do not define
8681 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8682 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8683 A C expression that evaluates to true if the architecture defines a value
8684 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
8685 should be set to this value. If this macro is not defined, the value of
8686 @code{clz} or @code{ctz} is assumed to be undefined.
8688 This macro must be defined if the target's expansion for @code{ffs}
8689 relies on a particular value to get correct results. Otherwise it
8690 is not necessary, though it may be used to optimize some corner cases.
8692 Note that regardless of this macro the ``definedness'' of @code{clz}
8693 and @code{ctz} at zero do @emph{not} extend to the builtin functions
8694 visible to the user. Thus one may be free to adjust the value at will
8695 to match the target expansion of these operations without fear of
8700 An alias for the machine mode for pointers. On most machines, define
8701 this to be the integer mode corresponding to the width of a hardware
8702 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8703 On some machines you must define this to be one of the partial integer
8704 modes, such as @code{PSImode}.
8706 The width of @code{Pmode} must be at least as large as the value of
8707 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8708 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8712 @defmac FUNCTION_MODE
8713 An alias for the machine mode used for memory references to functions
8714 being called, in @code{call} RTL expressions. On most machines this
8715 should be @code{QImode}.
8718 @defmac INTEGRATE_THRESHOLD (@var{decl})
8719 A C expression for the maximum number of instructions above which the
8720 function @var{decl} should not be inlined. @var{decl} is a
8721 @code{FUNCTION_DECL} node.
8723 The default definition of this macro is 64 plus 8 times the number of
8724 arguments that the function accepts. Some people think a larger
8725 threshold should be used on RISC machines.
8728 @defmac STDC_0_IN_SYSTEM_HEADERS
8729 In normal operation, the preprocessor expands @code{__STDC__} to the
8730 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8731 hosts, like Solaris, the system compiler uses a different convention,
8732 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8733 strict conformance to the C Standard.
8735 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8736 convention when processing system header files, but when processing user
8737 files @code{__STDC__} will always expand to 1.
8740 @defmac NO_IMPLICIT_EXTERN_C
8741 Define this macro if the system header files support C++ as well as C@.
8742 This macro inhibits the usual method of using system header files in
8743 C++, which is to pretend that the file's contents are enclosed in
8744 @samp{extern "C" @{@dots{}@}}.
8749 @defmac REGISTER_TARGET_PRAGMAS ()
8750 Define this macro if you want to implement any target-specific pragmas.
8751 If defined, it is a C expression which makes a series of calls to
8752 @code{c_register_pragma} for each pragma. The macro may also do any
8753 setup required for the pragmas.
8755 The primary reason to define this macro is to provide compatibility with
8756 other compilers for the same target. In general, we discourage
8757 definition of target-specific pragmas for GCC@.
8759 If the pragma can be implemented by attributes then you should consider
8760 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8762 Preprocessor macros that appear on pragma lines are not expanded. All
8763 @samp{#pragma} directives that do not match any registered pragma are
8764 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8767 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
8769 Each call to @code{c_register_pragma} establishes one pragma. The
8770 @var{callback} routine will be called when the preprocessor encounters a
8774 #pragma [@var{space}] @var{name} @dots{}
8777 @var{space} is the case-sensitive namespace of the pragma, or
8778 @code{NULL} to put the pragma in the global namespace. The callback
8779 routine receives @var{pfile} as its first argument, which can be passed
8780 on to cpplib's functions if necessary. You can lex tokens after the
8781 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8782 callback will be silently ignored. The end of the line is indicated by
8783 a token of type @code{CPP_EOF}
8785 For an example use of this routine, see @file{c4x.h} and the callback
8786 routines defined in @file{c4x-c.c}.
8788 Note that the use of @code{c_lex} is specific to the C and C++
8789 compilers. It will not work in the Java or Fortran compilers, or any
8790 other language compilers for that matter. Thus if @code{c_lex} is going
8791 to be called from target-specific code, it must only be done so when
8792 building the C and C++ compilers. This can be done by defining the
8793 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8794 target entry in the @file{config.gcc} file. These variables should name
8795 the target-specific, language-specific object file which contains the
8796 code that uses @code{c_lex}. Note it will also be necessary to add a
8797 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8798 how to build this object file.
8803 @defmac HANDLE_SYSV_PRAGMA
8804 Define this macro (to a value of 1) if you want the System V style
8805 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8806 [=<value>]} to be supported by gcc.
8808 The pack pragma specifies the maximum alignment (in bytes) of fields
8809 within a structure, in much the same way as the @samp{__aligned__} and
8810 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8811 the behavior to the default.
8813 A subtlety for Microsoft Visual C/C++ style bit-field packing
8814 (e.g. -mms-bitfields) for targets that support it:
8815 When a bit-field is inserted into a packed record, the whole size
8816 of the underlying type is used by one or more same-size adjacent
8817 bit-fields (that is, if its long:3, 32 bits is used in the record,
8818 and any additional adjacent long bit-fields are packed into the same
8819 chunk of 32 bits. However, if the size changes, a new field of that
8822 If both MS bit-fields and @samp{__attribute__((packed))} are used,
8823 the latter will take precedence. If @samp{__attribute__((packed))} is
8824 used on a single field when MS bit-fields are in use, it will take
8825 precedence for that field, but the alignment of the rest of the structure
8826 may affect its placement.
8828 The weak pragma only works if @code{SUPPORTS_WEAK} and
8829 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8830 of specifically named weak labels, optionally with a value.
8835 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
8836 Define this macro (to a value of 1) if you want to support the Win32
8837 style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma
8838 pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment
8839 (in bytes) of fields within a structure, in much the same way as the
8840 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8841 pack value of zero resets the behavior to the default. Successive
8842 invocations of this pragma cause the previous values to be stacked, so
8843 that invocations of @samp{#pragma pack(pop)} will return to the previous
8847 @defmac DOLLARS_IN_IDENTIFIERS
8848 Define this macro to control use of the character @samp{$} in
8849 identifier names for the C family of languages. 0 means @samp{$} is
8850 not allowed by default; 1 means it is allowed. 1 is the default;
8851 there is no need to define this macro in that case.
8854 @defmac NO_DOLLAR_IN_LABEL
8855 Define this macro if the assembler does not accept the character
8856 @samp{$} in label names. By default constructors and destructors in
8857 G++ have @samp{$} in the identifiers. If this macro is defined,
8858 @samp{.} is used instead.
8861 @defmac NO_DOT_IN_LABEL
8862 Define this macro if the assembler does not accept the character
8863 @samp{.} in label names. By default constructors and destructors in G++
8864 have names that use @samp{.}. If this macro is defined, these names
8865 are rewritten to avoid @samp{.}.
8868 @defmac DEFAULT_MAIN_RETURN
8869 Define this macro if the target system expects every program's @code{main}
8870 function to return a standard ``success'' value by default (if no other
8871 value is explicitly returned).
8873 The definition should be a C statement (sans semicolon) to generate the
8874 appropriate rtl instructions. It is used only when compiling the end of
8878 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
8879 Define this macro as a C expression that is nonzero if it is safe for the
8880 delay slot scheduler to place instructions in the delay slot of @var{insn},
8881 even if they appear to use a resource set or clobbered in @var{insn}.
8882 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8883 every @code{call_insn} has this behavior. On machines where some @code{insn}
8884 or @code{jump_insn} is really a function call and hence has this behavior,
8885 you should define this macro.
8887 You need not define this macro if it would always return zero.
8890 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
8891 Define this macro as a C expression that is nonzero if it is safe for the
8892 delay slot scheduler to place instructions in the delay slot of @var{insn},
8893 even if they appear to set or clobber a resource referenced in @var{insn}.
8894 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8895 some @code{insn} or @code{jump_insn} is really a function call and its operands
8896 are registers whose use is actually in the subroutine it calls, you should
8897 define this macro. Doing so allows the delay slot scheduler to move
8898 instructions which copy arguments into the argument registers into the delay
8901 You need not define this macro if it would always return zero.
8904 @defmac MULTIPLE_SYMBOL_SPACES
8905 Define this macro if in some cases global symbols from one translation
8906 unit may not be bound to undefined symbols in another translation unit
8907 without user intervention. For instance, under Microsoft Windows
8908 symbols must be explicitly imported from shared libraries (DLLs).
8911 @defmac MD_ASM_CLOBBERS (@var{clobbers})
8912 A C statement that adds to @var{clobbers} @code{STRING_CST} trees for
8913 any hard regs the port wishes to automatically clobber for all asms.
8916 @defmac MAX_INTEGER_COMPUTATION_MODE
8917 Define this to the largest integer machine mode which can be used for
8918 operations other than load, store and copy operations.
8920 You need only define this macro if the target holds values larger than
8921 @code{word_mode} in general purpose registers. Most targets should not define
8925 @defmac MATH_LIBRARY
8926 Define this macro as a C string constant for the linker argument to link
8927 in the system math library, or @samp{""} if the target does not have a
8928 separate math library.
8930 You need only define this macro if the default of @samp{"-lm"} is wrong.
8933 @defmac LIBRARY_PATH_ENV
8934 Define this macro as a C string constant for the environment variable that
8935 specifies where the linker should look for libraries.
8937 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
8941 @defmac TARGET_HAS_F_SETLKW
8942 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
8943 Note that this functionality is part of POSIX@.
8944 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
8945 to use file locking when exiting a program, which avoids race conditions
8946 if the program has forked.
8949 @defmac MAX_CONDITIONAL_EXECUTE
8951 A C expression for the maximum number of instructions to execute via
8952 conditional execution instructions instead of a branch. A value of
8953 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8954 1 if it does use cc0.
8957 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
8958 Used if the target needs to perform machine-dependent modifications on the
8959 conditionals used for turning basic blocks into conditionally executed code.
8960 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
8961 contains information about the currently processed blocks. @var{true_expr}
8962 and @var{false_expr} are the tests that are used for converting the
8963 then-block and the else-block, respectively. Set either @var{true_expr} or
8964 @var{false_expr} to a null pointer if the tests cannot be converted.
8967 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
8968 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
8969 if-statements into conditions combined by @code{and} and @code{or} operations.
8970 @var{bb} contains the basic block that contains the test that is currently
8971 being processed and about to be turned into a condition.
8974 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
8975 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
8976 be converted to conditional execution format. @var{ce_info} points to
8977 a data structure, @code{struct ce_if_block}, which contains information
8978 about the currently processed blocks.
8981 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
8982 A C expression to perform any final machine dependent modifications in
8983 converting code to conditional execution. The involved basic blocks
8984 can be found in the @code{struct ce_if_block} structure that is pointed
8985 to by @var{ce_info}.
8988 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
8989 A C expression to cancel any machine dependent modifications in
8990 converting code to conditional execution. The involved basic blocks
8991 can be found in the @code{struct ce_if_block} structure that is pointed
8992 to by @var{ce_info}.
8995 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
8996 A C expression to initialize any extra fields in a @code{struct ce_if_block}
8997 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9000 @defmac IFCVT_EXTRA_FIELDS
9001 If defined, it should expand to a set of field declarations that will be
9002 added to the @code{struct ce_if_block} structure. These should be initialized
9003 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9006 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
9007 If non-null, this hook performs a target-specific pass over the
9008 instruction stream. The compiler will run it at all optimization levels,
9009 just before the point at which it normally does delayed-branch scheduling.
9011 The exact purpose of the hook varies from target to target. Some use
9012 it to do transformations that are necessary for correctness, such as
9013 laying out in-function constant pools or avoiding hardware hazards.
9014 Others use it as an opportunity to do some machine-dependent optimizations.
9016 You need not implement the hook if it has nothing to do. The default
9020 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9021 Define this hook if you have any machine-specific built-in functions
9022 that need to be defined. It should be a function that performs the
9025 Machine specific built-in functions can be useful to expand special machine
9026 instructions that would otherwise not normally be generated because
9027 they have no equivalent in the source language (for example, SIMD vector
9028 instructions or prefetch instructions).
9030 To create a built-in function, call the function @code{builtin_function}
9031 which is defined by the language front end. You can use any type nodes set
9032 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9033 only language front ends that use those two functions will call
9034 @samp{TARGET_INIT_BUILTINS}.
9037 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9039 Expand a call to a machine specific built-in function that was set up by
9040 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9041 function call; the result should go to @var{target} if that is
9042 convenient, and have mode @var{mode} if that is convenient.
9043 @var{subtarget} may be used as the target for computing one of
9044 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9045 ignored. This function should return the result of the call to the
9049 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
9051 Take a branch insn in @var{branch1} and another in @var{branch2}.
9052 Return true if redirecting @var{branch1} to the destination of
9053 @var{branch2} is possible.
9055 On some targets, branches may have a limited range. Optimizing the
9056 filling of delay slots can result in branches being redirected, and this
9057 may in turn cause a branch offset to overflow.
9060 @defmac ALLOCATE_INITIAL_VALUE (@var{hard_reg})
9062 When the initial value of a hard register has been copied in a pseudo
9063 register, it is often not necessary to actually allocate another register
9064 to this pseudo register, because the original hard register or a stack slot
9065 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9066 defined, is called at the start of register allocation once for each
9067 hard register that had its initial value copied by using
9068 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9069 Possible values are @code{NULL_RTX}, if you don't want
9070 to do any special allocation, a @code{REG} rtx---that would typically be
9071 the hard register itself, if it is known not to be clobbered---or a
9073 If you are returning a @code{MEM}, this is only a hint for the allocator;
9074 it might decide to use another register anyways.
9075 You may use @code{current_function_leaf_function} in the definition of the
9076 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9077 register in question will not be clobbered.
9080 @defmac TARGET_OBJECT_SUFFIX
9081 Define this macro to be a C string representing the suffix for object
9082 files on your target machine. If you do not define this macro, GCC will
9083 use @samp{.o} as the suffix for object files.
9086 @defmac TARGET_EXECUTABLE_SUFFIX
9087 Define this macro to be a C string representing the suffix to be
9088 automatically added to executable files on your target machine. If you
9089 do not define this macro, GCC will use the null string as the suffix for
9093 @defmac COLLECT_EXPORT_LIST
9094 If defined, @code{collect2} will scan the individual object files
9095 specified on its command line and create an export list for the linker.
9096 Define this macro for systems like AIX, where the linker discards
9097 object files that are not referenced from @code{main} and uses export
9101 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
9102 Define this macro to a C expression representing a variant of the
9103 method call @var{mdecl}, if Java Native Interface (JNI) methods
9104 must be invoked differently from other methods on your target.
9105 For example, on 32-bit Windows, JNI methods must be invoked using
9106 the @code{stdcall} calling convention and this macro is then
9107 defined as this expression:
9110 build_type_attribute_variant (@var{mdecl},
9112 (get_identifier ("stdcall"),
9117 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9118 This target hook returns @code{true} past the point in which new jump
9119 instructions could be created. On machines that require a register for
9120 every jump such as the SHmedia ISA of SH5, this point would typically be
9121 reload, so this target hook should be defined to a function such as:
9125 cannot_modify_jumps_past_reload_p ()
9127 return (reload_completed || reload_in_progress);
9132 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
9133 This target hook returns a register class for which branch target register
9134 optimizations should be applied. All registers in this class should be
9135 usable interchangeably. After reload, registers in this class will be
9136 re-allocated and loads will be hoisted out of loops and be subjected
9137 to inter-block scheduling.
9140 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
9141 Branch target register optimization will by default exclude callee-saved
9143 that are not already live during the current function; if this target hook
9144 returns true, they will be included. The target code must than make sure
9145 that all target registers in the class returned by
9146 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
9147 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
9148 epilogues have already been generated. Note, even if you only return
9149 true when @var{after_prologue_epilogue_gen} is false, you still are likely
9150 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
9151 to reserve space for caller-saved target registers.
9154 @defmac POWI_MAX_MULTS
9155 If defined, this macro is interpreted as a signed integer C expression
9156 that specifies the maximum number of floating point multiplications
9157 that should be emitted when expanding exponentiation by an integer
9158 constant inline. When this value is defined, exponentiation requiring
9159 more than this number of multiplications is implemented by calling the
9160 system library's @code{pow}, @code{powf} or @code{powl} routines.
9161 The default value places no upper bound on the multiplication count.