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 * PCH Target:: Validity checking for precompiled headers.
54 * Misc:: Everything else.
57 @node Target Structure
58 @section The Global @code{targetm} Variable
60 @cindex target functions
62 @deftypevar {struct gcc_target} targetm
63 The target @file{.c} file must define the global @code{targetm} variable
64 which contains pointers to functions and data relating to the target
65 machine. The variable is declared in @file{target.h};
66 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
67 used to initialize the variable, and macros for the default initializers
68 for elements of the structure. The @file{.c} file should override those
69 macros for which the default definition is inappropriate. For example:
72 #include "target-def.h"
74 /* @r{Initialize the GCC target structure.} */
76 #undef TARGET_COMP_TYPE_ATTRIBUTES
77 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
79 struct gcc_target targetm = TARGET_INITIALIZER;
83 Where a macro should be defined in the @file{.c} file in this manner to
84 form part of the @code{targetm} structure, it is documented below as a
85 ``Target Hook'' with a prototype. Many macros will change in future
86 from being defined in the @file{.h} file to being part of the
87 @code{targetm} structure.
90 @section Controlling the Compilation Driver, @file{gcc}
92 @cindex controlling the compilation driver
94 @c prevent bad page break with this line
95 You can control the compilation driver.
97 @defmac SWITCH_TAKES_ARG (@var{char})
98 A C expression which determines whether the option @option{-@var{char}}
99 takes arguments. The value should be the number of arguments that
100 option takes--zero, for many options.
102 By default, this macro is defined as
103 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
104 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
105 wish to add additional options which take arguments. Any redefinition
106 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
110 @defmac WORD_SWITCH_TAKES_ARG (@var{name})
111 A C expression which determines whether the option @option{-@var{name}}
112 takes arguments. The value should be the number of arguments that
113 option takes--zero, for many options. This macro rather than
114 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
116 By default, this macro is defined as
117 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
118 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
119 wish to add additional options which take arguments. Any redefinition
120 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
124 @defmac SWITCH_CURTAILS_COMPILATION (@var{char})
125 A C expression which determines whether the option @option{-@var{char}}
126 stops compilation before the generation of an executable. The value is
127 boolean, nonzero if the option does stop an executable from being
128 generated, zero otherwise.
130 By default, this macro is defined as
131 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
132 options properly. You need not define
133 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
134 options which affect the generation of an executable. Any redefinition
135 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
136 for additional options.
139 @defmac SWITCHES_NEED_SPACES
140 A string-valued C expression which enumerates the options for which
141 the linker needs a space between the option and its argument.
143 If this macro is not defined, the default value is @code{""}.
146 @defmac TARGET_OPTION_TRANSLATE_TABLE
147 If defined, a list of pairs of strings, the first of which is a
148 potential command line target to the @file{gcc} driver program, and the
149 second of which is a space-separated (tabs and other whitespace are not
150 supported) list of options with which to replace the first option. The
151 target defining this list is responsible for assuring that the results
152 are valid. Replacement options may not be the @code{--opt} style, they
153 must be the @code{-opt} style. It is the intention of this macro to
154 provide a mechanism for substitution that affects the multilibs chosen,
155 such as one option that enables many options, some of which select
156 multilibs. Example nonsensical definition, where @code{-malt-abi},
157 @code{-EB}, and @code{-mspoo} cause different multilibs to be chosen:
160 #define TARGET_OPTION_TRANSLATE_TABLE \
161 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
162 @{ "-compat", "-EB -malign=4 -mspoo" @}
166 @defmac DRIVER_SELF_SPECS
167 A list of specs for the driver itself. It should be a suitable
168 initializer for an array of strings, with no surrounding braces.
170 The driver applies these specs to its own command line between loading
171 default @file{specs} files (but not command-line specified ones) and
172 choosing the multilib directory or running any subcommands. It
173 applies them in the order given, so each spec can depend on the
174 options added by earlier ones. It is also possible to remove options
175 using @samp{%<@var{option}} in the usual way.
177 This macro can be useful when a port has several interdependent target
178 options. It provides a way of standardizing the command line so
179 that the other specs are easier to write.
181 Do not define this macro if it does not need to do anything.
184 @defmac OPTION_DEFAULT_SPECS
185 A list of specs used to support configure-time default options (i.e.@:
186 @option{--with} options) in the driver. It should be a suitable initializer
187 for an array of structures, each containing two strings, without the
188 outermost pair of surrounding braces.
190 The first item in the pair is the name of the default. This must match
191 the code in @file{config.gcc} for the target. The second item is a spec
192 to apply if a default with this name was specified. The string
193 @samp{%(VALUE)} in the spec will be replaced by the value of the default
194 everywhere it occurs.
196 The driver will apply these specs to its own command line between loading
197 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
198 the same mechanism as @code{DRIVER_SELF_SPECS}.
200 Do not define this macro if it does not need to do anything.
204 A C string constant that tells the GCC driver program options to
205 pass to CPP@. It can also specify how to translate options you
206 give to GCC into options for GCC to pass to the CPP@.
208 Do not define this macro if it does not need to do anything.
211 @defmac CPLUSPLUS_CPP_SPEC
212 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
213 than C@. If you do not define this macro, then the value of
214 @code{CPP_SPEC} (if any) will be used instead.
218 A C string constant that tells the GCC driver program options to
219 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
221 It can also specify how to translate options you give to GCC into options
222 for GCC to pass to front ends.
224 Do not define this macro if it does not need to do anything.
228 A C string constant that tells the GCC driver program options to
229 pass to @code{cc1plus}. It can also specify how to translate options you
230 give to GCC into options for GCC to pass to the @code{cc1plus}.
232 Do not define this macro if it does not need to do anything.
233 Note that everything defined in CC1_SPEC is already passed to
234 @code{cc1plus} so there is no need to duplicate the contents of
235 CC1_SPEC in CC1PLUS_SPEC@.
239 A C string constant that tells the GCC driver program options to
240 pass to the assembler. It can also specify how to translate options
241 you give to GCC into options for GCC to pass to the assembler.
242 See the file @file{sun3.h} for an example of this.
244 Do not define this macro if it does not need to do anything.
247 @defmac ASM_FINAL_SPEC
248 A C string constant that tells the GCC driver program how to
249 run any programs which cleanup after the normal assembler.
250 Normally, this is not needed. See the file @file{mips.h} for
253 Do not define this macro if it does not need to do anything.
256 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
257 Define this macro, with no value, if the driver should give the assembler
258 an argument consisting of a single dash, @option{-}, to instruct it to
259 read from its standard input (which will be a pipe connected to the
260 output of the compiler proper). This argument is given after any
261 @option{-o} option specifying the name of the output file.
263 If you do not define this macro, the assembler is assumed to read its
264 standard input if given no non-option arguments. If your assembler
265 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
266 see @file{mips.h} for instance.
270 A C string constant that tells the GCC driver program options to
271 pass to the linker. It can also specify how to translate options you
272 give to GCC into options for GCC to pass to the linker.
274 Do not define this macro if it does not need to do anything.
278 Another C string constant used much like @code{LINK_SPEC}. The difference
279 between the two is that @code{LIB_SPEC} is used at the end of the
280 command given to the linker.
282 If this macro is not defined, a default is provided that
283 loads the standard C library from the usual place. See @file{gcc.c}.
287 Another C string constant that tells the GCC driver program
288 how and when to place a reference to @file{libgcc.a} into the
289 linker command line. This constant is placed both before and after
290 the value of @code{LIB_SPEC}.
292 If this macro is not defined, the GCC driver provides a default that
293 passes the string @option{-lgcc} to the linker.
296 @defmac STARTFILE_SPEC
297 Another C string constant used much like @code{LINK_SPEC}. The
298 difference between the two is that @code{STARTFILE_SPEC} is used at
299 the very beginning of the command given to the linker.
301 If this macro is not defined, a default is provided that loads the
302 standard C startup file from the usual place. See @file{gcc.c}.
306 Another C string constant used much like @code{LINK_SPEC}. The
307 difference between the two is that @code{ENDFILE_SPEC} is used at
308 the very end of the command given to the linker.
310 Do not define this macro if it does not need to do anything.
313 @defmac THREAD_MODEL_SPEC
314 GCC @code{-v} will print the thread model GCC was configured to use.
315 However, this doesn't work on platforms that are multilibbed on thread
316 models, such as AIX 4.3. On such platforms, define
317 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
318 blanks that names one of the recognized thread models. @code{%*}, the
319 default value of this macro, will expand to the value of
320 @code{thread_file} set in @file{config.gcc}.
323 @defmac SYSROOT_SUFFIX_SPEC
324 Define this macro to add a suffix to the target sysroot when GCC is
325 configured with a sysroot. This will cause GCC to search for usr/lib,
326 et al, within sysroot+suffix.
329 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
330 Define this macro to add a headers_suffix to the target sysroot when
331 GCC is configured with a sysroot. This will cause GCC to pass the
332 updated sysroot+headers_suffix to CPP@, causing it to search for
333 usr/include, et al, within sysroot+headers_suffix.
337 Define this macro to provide additional specifications to put in the
338 @file{specs} file that can be used in various specifications like
341 The definition should be an initializer for an array of structures,
342 containing a string constant, that defines the specification name, and a
343 string constant that provides the specification.
345 Do not define this macro if it does not need to do anything.
347 @code{EXTRA_SPECS} is useful when an architecture contains several
348 related targets, which have various @code{@dots{}_SPECS} which are similar
349 to each other, and the maintainer would like one central place to keep
352 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
353 define either @code{_CALL_SYSV} when the System V calling sequence is
354 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
357 The @file{config/rs6000/rs6000.h} target file defines:
360 #define EXTRA_SPECS \
361 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
363 #define CPP_SYS_DEFAULT ""
366 The @file{config/rs6000/sysv.h} target file defines:
370 "%@{posix: -D_POSIX_SOURCE @} \
371 %@{mcall-sysv: -D_CALL_SYSV @} \
372 %@{!mcall-sysv: %(cpp_sysv_default) @} \
373 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
375 #undef CPP_SYSV_DEFAULT
376 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
379 while the @file{config/rs6000/eabiaix.h} target file defines
380 @code{CPP_SYSV_DEFAULT} as:
383 #undef CPP_SYSV_DEFAULT
384 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
388 @defmac LINK_LIBGCC_SPECIAL
389 Define this macro if the driver program should find the library
390 @file{libgcc.a} itself and should not pass @option{-L} options to the
391 linker. If you do not define this macro, the driver program will pass
392 the argument @option{-lgcc} to tell the linker to do the search and will
393 pass @option{-L} options to it.
396 @defmac LINK_LIBGCC_SPECIAL_1
397 Define this macro if the driver program should find the library
398 @file{libgcc.a}. If you do not define this macro, the driver program will pass
399 the argument @option{-lgcc} to tell the linker to do the search.
400 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
401 not affect @option{-L} options.
404 @defmac LINK_GCC_C_SEQUENCE_SPEC
405 The sequence in which libgcc and libc are specified to the linker.
406 By default this is @code{%G %L %G}.
409 @defmac LINK_COMMAND_SPEC
410 A C string constant giving the complete command line need to execute the
411 linker. When you do this, you will need to update your port each time a
412 change is made to the link command line within @file{gcc.c}. Therefore,
413 define this macro only if you need to completely redefine the command
414 line for invoking the linker and there is no other way to accomplish
415 the effect you need. Overriding this macro may be avoidable by overriding
416 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
419 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
420 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
421 directories from linking commands. Do not give it a nonzero value if
422 removing duplicate search directories changes the linker's semantics.
425 @defmac MULTILIB_DEFAULTS
426 Define this macro as a C expression for the initializer of an array of
427 string to tell the driver program which options are defaults for this
428 target and thus do not need to be handled specially when using
429 @code{MULTILIB_OPTIONS}.
431 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
432 the target makefile fragment or if none of the options listed in
433 @code{MULTILIB_OPTIONS} are set by default.
434 @xref{Target Fragment}.
437 @defmac RELATIVE_PREFIX_NOT_LINKDIR
438 Define this macro to tell @command{gcc} that it should only translate
439 a @option{-B} prefix into a @option{-L} linker option if the prefix
440 indicates an absolute file name.
443 @defmac MD_EXEC_PREFIX
444 If defined, this macro is an additional prefix to try after
445 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
446 when the @option{-b} option is used, or the compiler is built as a cross
447 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
448 to the list of directories used to find the assembler in @file{configure.in}.
451 @defmac STANDARD_STARTFILE_PREFIX
452 Define this macro as a C string constant if you wish to override the
453 standard choice of @code{libdir} as the default prefix to
454 try when searching for startup files such as @file{crt0.o}.
455 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
456 is built as a cross compiler.
459 @defmac MD_STARTFILE_PREFIX
460 If defined, this macro supplies an additional prefix to try after the
461 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
462 @option{-b} option is used, or when the compiler is built as a cross
466 @defmac MD_STARTFILE_PREFIX_1
467 If defined, this macro supplies yet another prefix to try after the
468 standard prefixes. It is not searched when the @option{-b} option is
469 used, or when the compiler is built as a cross compiler.
472 @defmac INIT_ENVIRONMENT
473 Define this macro as a C string constant if you wish to set environment
474 variables for programs called by the driver, such as the assembler and
475 loader. The driver passes the value of this macro to @code{putenv} to
476 initialize the necessary environment variables.
479 @defmac LOCAL_INCLUDE_DIR
480 Define this macro as a C string constant if you wish to override the
481 standard choice of @file{/usr/local/include} as the default prefix to
482 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
483 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
485 Cross compilers do not search either @file{/usr/local/include} or its
489 @defmac MODIFY_TARGET_NAME
490 Define this macro if you wish to define command-line switches that
491 modify the default target name.
493 For each switch, you can include a string to be appended to the first
494 part of the configuration name or a string to be deleted from the
495 configuration name, if present. The definition should be an initializer
496 for an array of structures. Each array element should have three
497 elements: the switch name (a string constant, including the initial
498 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
499 indicate whether the string should be inserted or deleted, and the string
500 to be inserted or deleted (a string constant).
502 For example, on a machine where @samp{64} at the end of the
503 configuration name denotes a 64-bit target and you want the @option{-32}
504 and @option{-64} switches to select between 32- and 64-bit targets, you would
508 #define MODIFY_TARGET_NAME \
509 @{ @{ "-32", DELETE, "64"@}, \
510 @{"-64", ADD, "64"@}@}
514 @defmac SYSTEM_INCLUDE_DIR
515 Define this macro as a C string constant if you wish to specify a
516 system-specific directory to search for header files before the standard
517 directory. @code{SYSTEM_INCLUDE_DIR} comes before
518 @code{STANDARD_INCLUDE_DIR} in the search order.
520 Cross compilers do not use this macro and do not search the directory
524 @defmac STANDARD_INCLUDE_DIR
525 Define this macro as a C string constant if you wish to override the
526 standard choice of @file{/usr/include} as the default prefix to
527 try when searching for header files.
529 Cross compilers ignore this macro and do not search either
530 @file{/usr/include} or its replacement.
533 @defmac STANDARD_INCLUDE_COMPONENT
534 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
535 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
536 If you do not define this macro, no component is used.
539 @defmac INCLUDE_DEFAULTS
540 Define this macro if you wish to override the entire default search path
541 for include files. For a native compiler, the default search path
542 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
543 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
544 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
545 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
546 and specify private search areas for GCC@. The directory
547 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
549 The definition should be an initializer for an array of structures.
550 Each array element should have four elements: the directory name (a
551 string constant), the component name (also a string constant), a flag
552 for C++-only directories,
553 and a flag showing that the includes in the directory don't need to be
554 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
555 the array with a null element.
557 The component name denotes what GNU package the include file is part of,
558 if any, in all uppercase letters. For example, it might be @samp{GCC}
559 or @samp{BINUTILS}. If the package is part of a vendor-supplied
560 operating system, code the component name as @samp{0}.
562 For example, here is the definition used for VAX/VMS:
565 #define INCLUDE_DEFAULTS \
567 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
568 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
569 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
576 Here is the order of prefixes tried for exec files:
580 Any prefixes specified by the user with @option{-B}.
583 The environment variable @code{GCC_EXEC_PREFIX}, if any.
586 The directories specified by the environment variable @code{COMPILER_PATH}.
589 The macro @code{STANDARD_EXEC_PREFIX}.
592 @file{/usr/lib/gcc/}.
595 The macro @code{MD_EXEC_PREFIX}, if any.
598 Here is the order of prefixes tried for startfiles:
602 Any prefixes specified by the user with @option{-B}.
605 The environment variable @code{GCC_EXEC_PREFIX}, if any.
608 The directories specified by the environment variable @code{LIBRARY_PATH}
609 (or port-specific name; native only, cross compilers do not use this).
612 The macro @code{STANDARD_EXEC_PREFIX}.
615 @file{/usr/lib/gcc/}.
618 The macro @code{MD_EXEC_PREFIX}, if any.
621 The macro @code{MD_STARTFILE_PREFIX}, if any.
624 The macro @code{STANDARD_STARTFILE_PREFIX}.
633 @node Run-time Target
634 @section Run-time Target Specification
635 @cindex run-time target specification
636 @cindex predefined macros
637 @cindex target specifications
639 @c prevent bad page break with this line
640 Here are run-time target specifications.
642 @defmac TARGET_CPU_CPP_BUILTINS ()
643 This function-like macro expands to a block of code that defines
644 built-in preprocessor macros and assertions for the target cpu, using
645 the functions @code{builtin_define}, @code{builtin_define_std} and
646 @code{builtin_assert}. When the front end
647 calls this macro it provides a trailing semicolon, and since it has
648 finished command line option processing your code can use those
651 @code{builtin_assert} takes a string in the form you pass to the
652 command-line option @option{-A}, such as @code{cpu=mips}, and creates
653 the assertion. @code{builtin_define} takes a string in the form
654 accepted by option @option{-D} and unconditionally defines the macro.
656 @code{builtin_define_std} takes a string representing the name of an
657 object-like macro. If it doesn't lie in the user's namespace,
658 @code{builtin_define_std} defines it unconditionally. Otherwise, it
659 defines a version with two leading underscores, and another version
660 with two leading and trailing underscores, and defines the original
661 only if an ISO standard was not requested on the command line. For
662 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
663 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
664 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
665 defines only @code{_ABI64}.
667 You can also test for the C dialect being compiled. The variable
668 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
669 or @code{clk_objective_c}. Note that if we are preprocessing
670 assembler, this variable will be @code{clk_c} but the function-like
671 macro @code{preprocessing_asm_p()} will return true, so you might want
672 to check for that first. If you need to check for strict ANSI, the
673 variable @code{flag_iso} can be used. The function-like macro
674 @code{preprocessing_trad_p()} can be used to check for traditional
678 @defmac TARGET_OS_CPP_BUILTINS ()
679 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
680 and is used for the target operating system instead.
683 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
684 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
685 and is used for the target object format. @file{elfos.h} uses this
686 macro to define @code{__ELF__}, so you probably do not need to define
690 @deftypevar {extern int} target_flags
691 This declaration should be present.
694 @cindex optional hardware or system features
695 @cindex features, optional, in system conventions
697 @defmac TARGET_@var{featurename}
698 This series of macros is to allow compiler command arguments to
699 enable or disable the use of optional features of the target machine.
700 For example, one machine description serves both the 68000 and
701 the 68020; a command argument tells the compiler whether it should
702 use 68020-only instructions or not. This command argument works
703 by means of a macro @code{TARGET_68020} that tests a bit in
706 Define a macro @code{TARGET_@var{featurename}} for each such option.
707 Its definition should test a bit in @code{target_flags}. It is
708 recommended that a helper macro @code{MASK_@var{featurename}}
709 is defined for each bit-value to test, and used in
710 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
714 #define TARGET_MASK_68020 1
715 #define TARGET_68020 (target_flags & MASK_68020)
718 One place where these macros are used is in the condition-expressions
719 of instruction patterns. Note how @code{TARGET_68020} appears
720 frequently in the 68000 machine description file, @file{m68k.md}.
721 Another place they are used is in the definitions of the other
722 macros in the @file{@var{machine}.h} file.
725 @defmac TARGET_SWITCHES
726 This macro defines names of command options to set and clear
727 bits in @code{target_flags}. Its definition is an initializer
728 with a subgrouping for each command option.
730 Each subgrouping contains a string constant, that defines the option
731 name, a number, which contains the bits to set in
732 @code{target_flags}, and a second string which is the description
733 displayed by @option{--help}. If the number is negative then the bits specified
734 by the number are cleared instead of being set. If the description
735 string is present but empty, then no help information will be displayed
736 for that option, but it will not count as an undocumented option. The
737 actual option name is made by appending @samp{-m} to the specified name.
738 Non-empty description strings should be marked with @code{N_(@dots{})} for
739 @command{xgettext}. Please do not mark empty strings because the empty
740 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
741 of the message catalog with meta information, not the empty string.
743 In addition to the description for @option{--help},
744 more detailed documentation for each option should be added to
747 One of the subgroupings should have a null string. The number in
748 this grouping is the default value for @code{target_flags}. Any
749 target options act starting with that value.
751 Here is an example which defines @option{-m68000} and @option{-m68020}
752 with opposite meanings, and picks the latter as the default:
755 #define TARGET_SWITCHES \
756 @{ @{ "68020", MASK_68020, "" @}, \
757 @{ "68000", -MASK_68020, \
758 N_("Compile for the 68000") @}, \
759 @{ "", MASK_68020, "" @}, \
764 @defmac TARGET_OPTIONS
765 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
766 options that have values. Its definition is an initializer with a
767 subgrouping for each command option.
769 Each subgrouping contains a string constant, that defines the option
770 name, the address of a variable, a description string, and a value.
771 Non-empty description strings should be marked with @code{N_(@dots{})}
772 for @command{xgettext}. Please do not mark empty strings because the
773 empty string is reserved by GNU gettext. @code{gettext("")} returns the
774 header entry of the message catalog with meta information, not the empty
777 If the value listed in the table is @code{NULL}, then the variable, type
778 @code{char *}, is set to the variable part of the given option if the
779 fixed part matches. In other words, if the first part of the option
780 matches what's in the table, the variable will be set to point to the
781 rest of the option. This allows the user to specify a value for that
782 option. The actual option name is made by appending @samp{-m} to the
783 specified name. Again, each option should also be documented in
786 If the value listed in the table is non-@code{NULL}, then the option
787 must match the option in the table exactly (with @samp{-m}), and the
788 variable is set to point to the value listed in the table.
790 Here is an example which defines @option{-mshort-data-@var{number}}. If the
791 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
792 will be set to the string @code{"512"}.
795 extern char *m88k_short_data;
796 #define TARGET_OPTIONS \
797 @{ @{ "short-data-", &m88k_short_data, \
798 N_("Specify the size of the short data section"), 0 @} @}
801 Here is a variant of the above that allows the user to also specify
802 just @option{-mshort-data} where a default of @code{"64"} is used.
805 extern char *m88k_short_data;
806 #define TARGET_OPTIONS \
807 @{ @{ "short-data-", &m88k_short_data, \
808 N_("Specify the size of the short data section"), 0 @} \
809 @{ "short-data", &m88k_short_data, "", "64" @},
813 Here is an example which defines @option{-mno-alu}, @option{-malu1}, and
814 @option{-malu2} as a three-state switch, along with suitable macros for
815 checking the state of the option (documentation is elided for brevity).
819 char *chip_alu = ""; /* Specify default here. */
822 extern char *chip_alu;
823 #define TARGET_OPTIONS \
824 @{ @{ "no-alu", &chip_alu, "", "" @}, \
825 @{ "alu1", &chip_alu, "", "1" @}, \
826 @{ "alu2", &chip_alu, "", "2" @}, @}
827 #define TARGET_ALU (chip_alu[0] != '\0')
828 #define TARGET_ALU1 (chip_alu[0] == '1')
829 #define TARGET_ALU2 (chip_alu[0] == '2')
833 @defmac TARGET_VERSION
834 This macro is a C statement to print on @code{stderr} a string
835 describing the particular machine description choice. Every machine
836 description should define @code{TARGET_VERSION}. For example:
840 #define TARGET_VERSION \
841 fprintf (stderr, " (68k, Motorola syntax)");
843 #define TARGET_VERSION \
844 fprintf (stderr, " (68k, MIT syntax)");
849 @defmac OVERRIDE_OPTIONS
850 Sometimes certain combinations of command options do not make sense on
851 a particular target machine. You can define a macro
852 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
853 defined, is executed once just after all the command options have been
856 Don't use this macro to turn on various extra optimizations for
857 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
860 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
861 Some machines may desire to change what optimizations are performed for
862 various optimization levels. This macro, if defined, is executed once
863 just after the optimization level is determined and before the remainder
864 of the command options have been parsed. Values set in this macro are
865 used as the default values for the other command line options.
867 @var{level} is the optimization level specified; 2 if @option{-O2} is
868 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
870 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
872 You should not use this macro to change options that are not
873 machine-specific. These should uniformly selected by the same
874 optimization level on all supported machines. Use this macro to enable
875 machine-specific optimizations.
877 @strong{Do not examine @code{write_symbols} in
878 this macro!} The debugging options are not supposed to alter the
882 @defmac CAN_DEBUG_WITHOUT_FP
883 Define this macro if debugging can be performed even without a frame
884 pointer. If this macro is defined, GCC will turn on the
885 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
888 @node Per-Function Data
889 @section Defining data structures for per-function information.
890 @cindex per-function data
891 @cindex data structures
893 If the target needs to store information on a per-function basis, GCC
894 provides a macro and a couple of variables to allow this. Note, just
895 using statics to store the information is a bad idea, since GCC supports
896 nested functions, so you can be halfway through encoding one function
897 when another one comes along.
899 GCC defines a data structure called @code{struct function} which
900 contains all of the data specific to an individual function. This
901 structure contains a field called @code{machine} whose type is
902 @code{struct machine_function *}, which can be used by targets to point
903 to their own specific data.
905 If a target needs per-function specific data it should define the type
906 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
907 This macro should be used to initialize the function pointer
908 @code{init_machine_status}. This pointer is explained below.
910 One typical use of per-function, target specific data is to create an
911 RTX to hold the register containing the function's return address. This
912 RTX can then be used to implement the @code{__builtin_return_address}
913 function, for level 0.
915 Note---earlier implementations of GCC used a single data area to hold
916 all of the per-function information. Thus when processing of a nested
917 function began the old per-function data had to be pushed onto a
918 stack, and when the processing was finished, it had to be popped off the
919 stack. GCC used to provide function pointers called
920 @code{save_machine_status} and @code{restore_machine_status} to handle
921 the saving and restoring of the target specific information. Since the
922 single data area approach is no longer used, these pointers are no
925 @defmac INIT_EXPANDERS
926 Macro called to initialize any target specific information. This macro
927 is called once per function, before generation of any RTL has begun.
928 The intention of this macro is to allow the initialization of the
929 function pointer @code{init_machine_status}.
932 @deftypevar {void (*)(struct function *)} init_machine_status
933 If this function pointer is non-@code{NULL} it will be called once per
934 function, before function compilation starts, in order to allow the
935 target to perform any target specific initialization of the
936 @code{struct function} structure. It is intended that this would be
937 used to initialize the @code{machine} of that structure.
939 @code{struct machine_function} structures are expected to be freed by GC.
940 Generally, any memory that they reference must be allocated by using
941 @code{ggc_alloc}, including the structure itself.
945 @section Storage Layout
946 @cindex storage layout
948 Note that the definitions of the macros in this table which are sizes or
949 alignments measured in bits do not need to be constant. They can be C
950 expressions that refer to static variables, such as the @code{target_flags}.
951 @xref{Run-time Target}.
953 @defmac BITS_BIG_ENDIAN
954 Define this macro to have the value 1 if the most significant bit in a
955 byte has the lowest number; otherwise define it to have the value zero.
956 This means that bit-field instructions count from the most significant
957 bit. If the machine has no bit-field instructions, then this must still
958 be defined, but it doesn't matter which value it is defined to. This
959 macro need not be a constant.
961 This macro does not affect the way structure fields are packed into
962 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
965 @defmac BYTES_BIG_ENDIAN
966 Define this macro to have the value 1 if the most significant byte in a
967 word has the lowest number. This macro need not be a constant.
970 @defmac WORDS_BIG_ENDIAN
971 Define this macro to have the value 1 if, in a multiword object, the
972 most significant word has the lowest number. This applies to both
973 memory locations and registers; GCC fundamentally assumes that the
974 order of words in memory is the same as the order in registers. This
975 macro need not be a constant.
978 @defmac LIBGCC2_WORDS_BIG_ENDIAN
979 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
980 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
981 used only when compiling @file{libgcc2.c}. Typically the value will be set
982 based on preprocessor defines.
985 @defmac FLOAT_WORDS_BIG_ENDIAN
986 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
987 @code{TFmode} floating point numbers are stored in memory with the word
988 containing the sign bit at the lowest address; otherwise define it to
989 have the value 0. This macro need not be a constant.
991 You need not define this macro if the ordering is the same as for
995 @defmac BITS_PER_UNIT
996 Define this macro to be the number of bits in an addressable storage
997 unit (byte). If you do not define this macro the default is 8.
1000 @defmac BITS_PER_WORD
1001 Number of bits in a word. If you do not define this macro, the default
1002 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
1005 @defmac MAX_BITS_PER_WORD
1006 Maximum number of bits in a word. If this is undefined, the default is
1007 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
1008 largest value that @code{BITS_PER_WORD} can have at run-time.
1011 @defmac UNITS_PER_WORD
1012 Number of storage units in a word; normally 4.
1015 @defmac MIN_UNITS_PER_WORD
1016 Minimum number of units in a word. If this is undefined, the default is
1017 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
1018 smallest value that @code{UNITS_PER_WORD} can have at run-time.
1021 @defmac POINTER_SIZE
1022 Width of a pointer, in bits. You must specify a value no wider than the
1023 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1024 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1025 a value the default is @code{BITS_PER_WORD}.
1028 @defmac POINTERS_EXTEND_UNSIGNED
1029 A C expression whose value is greater than zero if pointers that need to be
1030 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
1031 be zero-extended and zero if they are to be sign-extended. If the value
1032 is less then zero then there must be an "ptr_extend" instruction that
1033 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
1035 You need not define this macro if the @code{POINTER_SIZE} is equal
1036 to the width of @code{Pmode}.
1039 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1040 A macro to update @var{m} and @var{unsignedp} when an object whose type
1041 is @var{type} and which has the specified mode and signedness is to be
1042 stored in a register. This macro is only called when @var{type} is a
1045 On most RISC machines, which only have operations that operate on a full
1046 register, define this macro to set @var{m} to @code{word_mode} if
1047 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1048 cases, only integer modes should be widened because wider-precision
1049 floating-point operations are usually more expensive than their narrower
1052 For most machines, the macro definition does not change @var{unsignedp}.
1053 However, some machines, have instructions that preferentially handle
1054 either signed or unsigned quantities of certain modes. For example, on
1055 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1056 sign-extend the result to 64 bits. On such machines, set
1057 @var{unsignedp} according to which kind of extension is more efficient.
1059 Do not define this macro if it would never modify @var{m}.
1062 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1063 This target hook should return @code{true} if the promotion described by
1064 @code{PROMOTE_MODE} should also be done for outgoing function arguments.
1067 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1068 This target hook should return @code{true} if the promotion described by
1069 @code{PROMOTE_MODE} should also be done for the return value of
1072 If this target hook returns @code{true}, @code{FUNCTION_VALUE} must
1073 perform the same promotions done by @code{PROMOTE_MODE}.
1076 @defmac PROMOTE_FOR_CALL_ONLY
1077 Define this macro if the promotion described by @code{PROMOTE_MODE}
1078 should @emph{only} be performed for outgoing function arguments or
1079 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1080 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1083 @defmac PARM_BOUNDARY
1084 Normal alignment required for function parameters on the stack, in
1085 bits. All stack parameters receive at least this much alignment
1086 regardless of data type. On most machines, this is the same as the
1090 @defmac STACK_BOUNDARY
1091 Define this macro to the minimum alignment enforced by hardware for the
1092 stack pointer on this machine. The definition is a C expression for the
1093 desired alignment (measured in bits). This value is used as a default
1094 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1095 this should be the same as @code{PARM_BOUNDARY}.
1098 @defmac PREFERRED_STACK_BOUNDARY
1099 Define this macro if you wish to preserve a certain alignment for the
1100 stack pointer, greater than what the hardware enforces. The definition
1101 is a C expression for the desired alignment (measured in bits). This
1102 macro must evaluate to a value equal to or larger than
1103 @code{STACK_BOUNDARY}.
1106 @defmac FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1107 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1108 not guaranteed by the runtime and we should emit code to align the stack
1109 at the beginning of @code{main}.
1111 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1112 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1113 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1114 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1115 be momentarily unaligned while pushing arguments.
1118 @defmac FUNCTION_BOUNDARY
1119 Alignment required for a function entry point, in bits.
1122 @defmac BIGGEST_ALIGNMENT
1123 Biggest alignment that any data type can require on this machine, in bits.
1126 @defmac MINIMUM_ATOMIC_ALIGNMENT
1127 If defined, the smallest alignment, in bits, that can be given to an
1128 object that can be referenced in one operation, without disturbing any
1129 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1130 on machines that don't have byte or half-word store operations.
1133 @defmac BIGGEST_FIELD_ALIGNMENT
1134 Biggest alignment that any structure or union field can require on this
1135 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1136 structure and union fields only, unless the field alignment has been set
1137 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1140 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1141 An expression for the alignment of a structure field @var{field} if the
1142 alignment computed in the usual way (including applying of
1143 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1144 alignment) is @var{computed}. It overrides alignment only if the
1145 field alignment has not been set by the
1146 @code{__attribute__ ((aligned (@var{n})))} construct.
1149 @defmac MAX_OFILE_ALIGNMENT
1150 Biggest alignment supported by the object file format of this machine.
1151 Use this macro to limit the alignment which can be specified using the
1152 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1153 the default value is @code{BIGGEST_ALIGNMENT}.
1156 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1157 If defined, a C expression to compute the alignment for a variable in
1158 the static store. @var{type} is the data type, and @var{basic-align} is
1159 the alignment that the object would ordinarily have. The value of this
1160 macro is used instead of that alignment to align the object.
1162 If this macro is not defined, then @var{basic-align} is used.
1165 One use of this macro is to increase alignment of medium-size data to
1166 make it all fit in fewer cache lines. Another is to cause character
1167 arrays to be word-aligned so that @code{strcpy} calls that copy
1168 constants to character arrays can be done inline.
1171 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1172 If defined, a C expression to compute the alignment given to a constant
1173 that is being placed in memory. @var{constant} is the constant and
1174 @var{basic-align} is the alignment that the object would ordinarily
1175 have. The value of this macro is used instead of that alignment to
1178 If this macro is not defined, then @var{basic-align} is used.
1180 The typical use of this macro is to increase alignment for string
1181 constants to be word aligned so that @code{strcpy} calls that copy
1182 constants can be done inline.
1185 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1186 If defined, a C expression to compute the alignment for a variable in
1187 the local store. @var{type} is the data type, and @var{basic-align} is
1188 the alignment that the object would ordinarily have. The value of this
1189 macro is used instead of that alignment to align the object.
1191 If this macro is not defined, then @var{basic-align} is used.
1193 One use of this macro is to increase alignment of medium-size data to
1194 make it all fit in fewer cache lines.
1197 @defmac EMPTY_FIELD_BOUNDARY
1198 Alignment in bits to be given to a structure bit-field that follows an
1199 empty field such as @code{int : 0;}.
1201 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1204 @defmac STRUCTURE_SIZE_BOUNDARY
1205 Number of bits which any structure or union's size must be a multiple of.
1206 Each structure or union's size is rounded up to a multiple of this.
1208 If you do not define this macro, the default is the same as
1209 @code{BITS_PER_UNIT}.
1212 @defmac STRICT_ALIGNMENT
1213 Define this macro to be the value 1 if instructions will fail to work
1214 if given data not on the nominal alignment. If instructions will merely
1215 go slower in that case, define this macro as 0.
1218 @defmac PCC_BITFIELD_TYPE_MATTERS
1219 Define this if you wish to imitate the way many other C compilers handle
1220 alignment of bit-fields and the structures that contain them.
1222 The behavior is that the type written for a named bit-field (@code{int},
1223 @code{short}, or other integer type) imposes an alignment for the entire
1224 structure, as if the structure really did contain an ordinary field of
1225 that type. In addition, the bit-field is placed within the structure so
1226 that it would fit within such a field, not crossing a boundary for it.
1228 Thus, on most machines, a named bit-field whose type is written as
1229 @code{int} would not cross a four-byte boundary, and would force
1230 four-byte alignment for the whole structure. (The alignment used may
1231 not be four bytes; it is controlled by the other alignment parameters.)
1233 An unnamed bit-field will not affect the alignment of the containing
1236 If the macro is defined, its definition should be a C expression;
1237 a nonzero value for the expression enables this behavior.
1239 Note that if this macro is not defined, or its value is zero, some
1240 bit-fields may cross more than one alignment boundary. The compiler can
1241 support such references if there are @samp{insv}, @samp{extv}, and
1242 @samp{extzv} insns that can directly reference memory.
1244 The other known way of making bit-fields work is to define
1245 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1246 Then every structure can be accessed with fullwords.
1248 Unless the machine has bit-field instructions or you define
1249 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1250 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1252 If your aim is to make GCC use the same conventions for laying out
1253 bit-fields as are used by another compiler, here is how to investigate
1254 what the other compiler does. Compile and run this program:
1273 printf ("Size of foo1 is %d\n",
1274 sizeof (struct foo1));
1275 printf ("Size of foo2 is %d\n",
1276 sizeof (struct foo2));
1281 If this prints 2 and 5, then the compiler's behavior is what you would
1282 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1285 @defmac BITFIELD_NBYTES_LIMITED
1286 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1287 to aligning a bit-field within the structure.
1290 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1291 Return 1 if a structure or array containing @var{field} should be accessed using
1294 If @var{field} is the only field in the structure, @var{mode} is its
1295 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1296 case where structures of one field would require the structure's mode to
1297 retain the field's mode.
1299 Normally, this is not needed. See the file @file{c4x.h} for an example
1300 of how to use this macro to prevent a structure having a floating point
1301 field from being accessed in an integer mode.
1304 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1305 Define this macro as an expression for the alignment of a type (given
1306 by @var{type} as a tree node) if the alignment computed in the usual
1307 way is @var{computed} and the alignment explicitly specified was
1310 The default is to use @var{specified} if it is larger; otherwise, use
1311 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1314 @defmac MAX_FIXED_MODE_SIZE
1315 An integer expression for the size in bits of the largest integer
1316 machine mode that should actually be used. All integer machine modes of
1317 this size or smaller can be used for structures and unions with the
1318 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1319 (DImode)} is assumed.
1322 @defmac VECTOR_MODE_SUPPORTED_P (@var{mode})
1323 Define this macro to be nonzero if the port is prepared to handle insns
1324 involving vector mode @var{mode}. At the very least, it must have move
1325 patterns for this mode.
1328 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1329 If defined, an expression of type @code{enum machine_mode} that
1330 specifies the mode of the save area operand of a
1331 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1332 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1333 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1334 having its mode specified.
1336 You need not define this macro if it always returns @code{Pmode}. You
1337 would most commonly define this macro if the
1338 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1342 @defmac STACK_SIZE_MODE
1343 If defined, an expression of type @code{enum machine_mode} that
1344 specifies the mode of the size increment operand of an
1345 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1347 You need not define this macro if it always returns @code{word_mode}.
1348 You would most commonly define this macro if the @code{allocate_stack}
1349 pattern needs to support both a 32- and a 64-bit mode.
1352 @defmac TARGET_FLOAT_FORMAT
1353 A code distinguishing the floating point format of the target machine.
1354 There are four defined values:
1357 @item IEEE_FLOAT_FORMAT
1358 This code indicates IEEE floating point. It is the default; there is no
1359 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1361 @item VAX_FLOAT_FORMAT
1362 This code indicates the ``F float'' (for @code{float}) and ``D float''
1363 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1365 @item IBM_FLOAT_FORMAT
1366 This code indicates the format used on the IBM System/370.
1368 @item C4X_FLOAT_FORMAT
1369 This code indicates the format used on the TMS320C3x/C4x.
1372 If your target uses a floating point format other than these, you must
1373 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1374 it to @file{real.c}.
1376 The ordering of the component words of floating point values stored in
1377 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1380 @defmac MODE_HAS_NANS (@var{mode})
1381 When defined, this macro should be true if @var{mode} has a NaN
1382 representation. The compiler assumes that NaNs are not equal to
1383 anything (including themselves) and that addition, subtraction,
1384 multiplication and division all return NaNs when one operand is
1387 By default, this macro is true if @var{mode} is a floating-point
1388 mode and the target floating-point format is IEEE@.
1391 @defmac MODE_HAS_INFINITIES (@var{mode})
1392 This macro should be true if @var{mode} can represent infinity. At
1393 present, the compiler uses this macro to decide whether @samp{x - x}
1394 is always defined. By default, the macro is true when @var{mode}
1395 is a floating-point mode and the target format is IEEE@.
1398 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1399 True if @var{mode} distinguishes between positive and negative zero.
1400 The rules are expected to follow the IEEE standard:
1404 @samp{x + x} has the same sign as @samp{x}.
1407 If the sum of two values with opposite sign is zero, the result is
1408 positive for all rounding modes expect towards @minus{}infinity, for
1409 which it is negative.
1412 The sign of a product or quotient is negative when exactly one
1413 of the operands is negative.
1416 The default definition is true if @var{mode} is a floating-point
1417 mode and the target format is IEEE@.
1420 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1421 If defined, this macro should be true for @var{mode} if it has at
1422 least one rounding mode in which @samp{x} and @samp{-x} can be
1423 rounded to numbers of different magnitude. Two such modes are
1424 towards @minus{}infinity and towards +infinity.
1426 The default definition of this macro is true if @var{mode} is
1427 a floating-point mode and the target format is IEEE@.
1430 @defmac ROUND_TOWARDS_ZERO
1431 If defined, this macro should be true if the prevailing rounding
1432 mode is towards zero. A true value has the following effects:
1436 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1439 @file{libgcc.a}'s floating-point emulator will round towards zero
1440 rather than towards nearest.
1443 The compiler's floating-point emulator will round towards zero after
1444 doing arithmetic, and when converting from the internal float format to
1448 The macro does not affect the parsing of string literals. When the
1449 primary rounding mode is towards zero, library functions like
1450 @code{strtod} might still round towards nearest, and the compiler's
1451 parser should behave like the target's @code{strtod} where possible.
1453 Not defining this macro is equivalent to returning zero.
1456 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1457 This macro should return true if floats with @var{size}
1458 bits do not have a NaN or infinity representation, but use the largest
1459 exponent for normal numbers instead.
1461 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1462 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1463 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1464 floating-point arithmetic.
1466 The default definition of this macro returns false for all sizes.
1469 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1470 This target hook should return @code{true} a vector is opaque. That
1471 is, if no cast is needed when copying a vector value of type
1472 @var{type} into another vector lvalue of the same size. Vector opaque
1473 types cannot be initialized. The default is that there are no such
1477 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1478 This target hook returns @code{true} if bit-fields in the given
1479 @var{record_type} are to be laid out following the rules of Microsoft
1480 Visual C/C++, namely: (i) a bit-field won't share the same storage
1481 unit with the previous bit-field if their underlying types have
1482 different sizes, and the bit-field will be aligned to the highest
1483 alignment of the underlying types of itself and of the previous
1484 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1485 the whole enclosing structure, even if it is unnamed; except that
1486 (iii) a zero-sized bit-field will be disregarded unless it follows
1487 another bit-field of nonzero size. If this hook returns @code{true},
1488 other macros that control bit-field layout are ignored.
1490 When a bit-field is inserted into a packed record, the whole size
1491 of the underlying type is used by one or more same-size adjacent
1492 bit-fields (that is, if its long:3, 32 bits is used in the record,
1493 and any additional adjacent long bit-fields are packed into the same
1494 chunk of 32 bits. However, if the size changes, a new field of that
1495 size is allocated). In an unpacked record, this is the same as using
1496 alignment, but not equivalent when packing.
1498 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1499 the latter will take precedence. If @samp{__attribute__((packed))} is
1500 used on a single field when MS bit-fields are in use, it will take
1501 precedence for that field, but the alignment of the rest of the structure
1502 may affect its placement.
1506 @section Layout of Source Language Data Types
1508 These macros define the sizes and other characteristics of the standard
1509 basic data types used in programs being compiled. Unlike the macros in
1510 the previous section, these apply to specific features of C and related
1511 languages, rather than to fundamental aspects of storage layout.
1513 @defmac INT_TYPE_SIZE
1514 A C expression for the size in bits of the type @code{int} on the
1515 target machine. If you don't define this, the default is one word.
1518 @defmac SHORT_TYPE_SIZE
1519 A C expression for the size in bits of the type @code{short} on the
1520 target machine. If you don't define this, the default is half a word.
1521 (If this would be less than one storage unit, it is rounded up to one
1525 @defmac LONG_TYPE_SIZE
1526 A C expression for the size in bits of the type @code{long} on the
1527 target machine. If you don't define this, the default is one word.
1530 @defmac ADA_LONG_TYPE_SIZE
1531 On some machines, the size used for the Ada equivalent of the type
1532 @code{long} by a native Ada compiler differs from that used by C. In
1533 that situation, define this macro to be a C expression to be used for
1534 the size of that type. If you don't define this, the default is the
1535 value of @code{LONG_TYPE_SIZE}.
1538 @defmac MAX_LONG_TYPE_SIZE
1539 Maximum number for the size in bits of the type @code{long} on the
1540 target machine. If this is undefined, the default is
1541 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1542 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1546 @defmac LONG_LONG_TYPE_SIZE
1547 A C expression for the size in bits of the type @code{long long} on the
1548 target machine. If you don't define this, the default is two
1549 words. If you want to support GNU Ada on your machine, the value of this
1550 macro must be at least 64.
1553 @defmac CHAR_TYPE_SIZE
1554 A C expression for the size in bits of the type @code{char} on the
1555 target machine. If you don't define this, the default is
1556 @code{BITS_PER_UNIT}.
1559 @defmac BOOL_TYPE_SIZE
1560 A C expression for the size in bits of the C++ type @code{bool} and
1561 C99 type @code{_Bool} on the target machine. If you don't define
1562 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1565 @defmac FLOAT_TYPE_SIZE
1566 A C expression for the size in bits of the type @code{float} on the
1567 target machine. If you don't define this, the default is one word.
1570 @defmac DOUBLE_TYPE_SIZE
1571 A C expression for the size in bits of the type @code{double} on the
1572 target machine. If you don't define this, the default is two
1576 @defmac LONG_DOUBLE_TYPE_SIZE
1577 A C expression for the size in bits of the type @code{long double} on
1578 the target machine. If you don't define this, the default is two
1582 @defmac MAX_LONG_DOUBLE_TYPE_SIZE
1583 Maximum number for the size in bits of the type @code{long double} on the
1584 target machine. If this is undefined, the default is
1585 @code{LONG_DOUBLE_TYPE_SIZE}. Otherwise, it is the constant value that is
1586 the largest value that @code{LONG_DOUBLE_TYPE_SIZE} can have at run-time.
1587 This is used in @code{cpp}.
1590 @defmac TARGET_FLT_EVAL_METHOD
1591 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1592 assuming, if applicable, that the floating-point control word is in its
1593 default state. If you do not define this macro the value of
1594 @code{FLT_EVAL_METHOD} will be zero.
1597 @defmac WIDEST_HARDWARE_FP_SIZE
1598 A C expression for the size in bits of the widest floating-point format
1599 supported by the hardware. If you define this macro, you must specify a
1600 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1601 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1605 @defmac DEFAULT_SIGNED_CHAR
1606 An expression whose value is 1 or 0, according to whether the type
1607 @code{char} should be signed or unsigned by default. The user can
1608 always override this default with the options @option{-fsigned-char}
1609 and @option{-funsigned-char}.
1612 @defmac DEFAULT_SHORT_ENUMS
1613 A C expression to determine whether to give an @code{enum} type
1614 only as many bytes as it takes to represent the range of possible values
1615 of that type. A nonzero value means to do that; a zero value means all
1616 @code{enum} types should be allocated like @code{int}.
1618 If you don't define the macro, the default is 0.
1622 A C expression for a string describing the name of the data type to use
1623 for size values. The typedef name @code{size_t} is defined using the
1624 contents of the string.
1626 The string can contain more than one keyword. If so, separate them with
1627 spaces, and write first any length keyword, then @code{unsigned} if
1628 appropriate, and finally @code{int}. The string must exactly match one
1629 of the data type names defined in the function
1630 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1631 omit @code{int} or change the order---that would cause the compiler to
1634 If you don't define this macro, the default is @code{"long unsigned
1638 @defmac PTRDIFF_TYPE
1639 A C expression for a string describing the name of the data type to use
1640 for the result of subtracting two pointers. The typedef name
1641 @code{ptrdiff_t} is defined using the contents of the string. See
1642 @code{SIZE_TYPE} above for more information.
1644 If you don't define this macro, the default is @code{"long int"}.
1648 A C expression for a string describing the name of the data type to use
1649 for wide characters. The typedef name @code{wchar_t} is defined using
1650 the contents of the string. See @code{SIZE_TYPE} above for more
1653 If you don't define this macro, the default is @code{"int"}.
1656 @defmac WCHAR_TYPE_SIZE
1657 A C expression for the size in bits of the data type for wide
1658 characters. This is used in @code{cpp}, which cannot make use of
1662 @defmac MAX_WCHAR_TYPE_SIZE
1663 Maximum number for the size in bits of the data type for wide
1664 characters. If this is undefined, the default is
1665 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1666 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1670 @defmac GCOV_TYPE_SIZE
1671 A C expression for the size in bits of the type used for gcov counters on the
1672 target machine. If you don't define this, the default is one
1673 @code{LONG_TYPE_SIZE} in case it is greater or equal to 64-bit and
1674 @code{LONG_LONG_TYPE_SIZE} otherwise. You may want to re-define the type to
1675 ensure atomicity for counters in multithreaded programs.
1679 A C expression for a string describing the name of the data type to
1680 use for wide characters passed to @code{printf} and returned from
1681 @code{getwc}. The typedef name @code{wint_t} is defined using the
1682 contents of the string. See @code{SIZE_TYPE} above for more
1685 If you don't define this macro, the default is @code{"unsigned int"}.
1689 A C expression for a string describing the name of the data type that
1690 can represent any value of any standard or extended signed integer type.
1691 The typedef name @code{intmax_t} is defined using the contents of the
1692 string. See @code{SIZE_TYPE} above for more information.
1694 If you don't define this macro, the default is the first of
1695 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1696 much precision as @code{long long int}.
1699 @defmac UINTMAX_TYPE
1700 A C expression for a string describing the name of the data type that
1701 can represent any value of any standard or extended unsigned integer
1702 type. The typedef name @code{uintmax_t} is defined using the contents
1703 of the string. See @code{SIZE_TYPE} above for more information.
1705 If you don't define this macro, the default is the first of
1706 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1707 unsigned int"} that has as much precision as @code{long long unsigned
1711 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1712 The C++ compiler represents a pointer-to-member-function with a struct
1719 ptrdiff_t vtable_index;
1726 The C++ compiler must use one bit to indicate whether the function that
1727 will be called through a pointer-to-member-function is virtual.
1728 Normally, we assume that the low-order bit of a function pointer must
1729 always be zero. Then, by ensuring that the vtable_index is odd, we can
1730 distinguish which variant of the union is in use. But, on some
1731 platforms function pointers can be odd, and so this doesn't work. In
1732 that case, we use the low-order bit of the @code{delta} field, and shift
1733 the remainder of the @code{delta} field to the left.
1735 GCC will automatically make the right selection about where to store
1736 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1737 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1738 set such that functions always start at even addresses, but the lowest
1739 bit of pointers to functions indicate whether the function at that
1740 address is in ARM or Thumb mode. If this is the case of your
1741 architecture, you should define this macro to
1742 @code{ptrmemfunc_vbit_in_delta}.
1744 In general, you should not have to define this macro. On architectures
1745 in which function addresses are always even, according to
1746 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1747 @code{ptrmemfunc_vbit_in_pfn}.
1750 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1751 Normally, the C++ compiler uses function pointers in vtables. This
1752 macro allows the target to change to use ``function descriptors''
1753 instead. Function descriptors are found on targets for whom a
1754 function pointer is actually a small data structure. Normally the
1755 data structure consists of the actual code address plus a data
1756 pointer to which the function's data is relative.
1758 If vtables are used, the value of this macro should be the number
1759 of words that the function descriptor occupies.
1762 @defmac TARGET_VTABLE_ENTRY_ALIGN
1763 By default, the vtable entries are void pointers, the so the alignment
1764 is the same as pointer alignment. The value of this macro specifies
1765 the alignment of the vtable entry in bits. It should be defined only
1766 when special alignment is necessary. */
1769 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1770 There are a few non-descriptor entries in the vtable at offsets below
1771 zero. If these entries must be padded (say, to preserve the alignment
1772 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1773 of words in each data entry.
1776 @node Escape Sequences
1777 @section Target Character Escape Sequences
1778 @cindex escape sequences
1780 By default, GCC assumes that the C character escape sequences take on
1781 their ASCII values for the target. If this is not correct, you must
1782 explicitly define all of the macros below. All of them must evaluate
1783 to constants; they are used in @code{case} statements.
1789 @findex TARGET_NEWLINE
1792 @multitable {@code{TARGET_NEWLINE}} {Escape} {ASCII character}
1793 @item Macro @tab Escape @tab ASCII character
1794 @item @code{TARGET_BELL} @tab @kbd{\a} @tab @code{07}, @code{BEL}
1795 @item @code{TARGET_CR} @tab @kbd{\r} @tab @code{0D}, @code{CR}
1796 @item @code{TARGET_ESC} @tab @kbd{\e}, @kbd{\E} @tab @code{1B}, @code{ESC}
1797 @item @code{TARGET_FF} @tab @kbd{\f} @tab @code{0C}, @code{FF}
1798 @item @code{TARGET_NEWLINE} @tab @kbd{\n} @tab @code{0A}, @code{LF}
1799 @item @code{TARGET_TAB} @tab @kbd{\t} @tab @code{09}, @code{HT}
1800 @item @code{TARGET_VT} @tab @kbd{\v} @tab @code{0B}, @code{VT}
1804 Note that the @kbd{\e} and @kbd{\E} escapes are GNU extensions, not
1805 part of the C standard.
1808 @section Register Usage
1809 @cindex register usage
1811 This section explains how to describe what registers the target machine
1812 has, and how (in general) they can be used.
1814 The description of which registers a specific instruction can use is
1815 done with register classes; see @ref{Register Classes}. For information
1816 on using registers to access a stack frame, see @ref{Frame Registers}.
1817 For passing values in registers, see @ref{Register Arguments}.
1818 For returning values in registers, see @ref{Scalar Return}.
1821 * Register Basics:: Number and kinds of registers.
1822 * Allocation Order:: Order in which registers are allocated.
1823 * Values in Registers:: What kinds of values each reg can hold.
1824 * Leaf Functions:: Renumbering registers for leaf functions.
1825 * Stack Registers:: Handling a register stack such as 80387.
1828 @node Register Basics
1829 @subsection Basic Characteristics of Registers
1831 @c prevent bad page break with this line
1832 Registers have various characteristics.
1834 @defmac FIRST_PSEUDO_REGISTER
1835 Number of hardware registers known to the compiler. They receive
1836 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1837 pseudo register's number really is assigned the number
1838 @code{FIRST_PSEUDO_REGISTER}.
1841 @defmac FIXED_REGISTERS
1842 @cindex fixed register
1843 An initializer that says which registers are used for fixed purposes
1844 all throughout the compiled code and are therefore not available for
1845 general allocation. These would include the stack pointer, the frame
1846 pointer (except on machines where that can be used as a general
1847 register when no frame pointer is needed), the program counter on
1848 machines where that is considered one of the addressable registers,
1849 and any other numbered register with a standard use.
1851 This information is expressed as a sequence of numbers, separated by
1852 commas and surrounded by braces. The @var{n}th number is 1 if
1853 register @var{n} is fixed, 0 otherwise.
1855 The table initialized from this macro, and the table initialized by
1856 the following one, may be overridden at run time either automatically,
1857 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1858 the user with the command options @option{-ffixed-@var{reg}},
1859 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1862 @defmac CALL_USED_REGISTERS
1863 @cindex call-used register
1864 @cindex call-clobbered register
1865 @cindex call-saved register
1866 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1867 clobbered (in general) by function calls as well as for fixed
1868 registers. This macro therefore identifies the registers that are not
1869 available for general allocation of values that must live across
1872 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1873 automatically saves it on function entry and restores it on function
1874 exit, if the register is used within the function.
1877 @defmac CALL_REALLY_USED_REGISTERS
1878 @cindex call-used register
1879 @cindex call-clobbered register
1880 @cindex call-saved register
1881 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1882 that the entire set of @code{FIXED_REGISTERS} be included.
1883 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1884 This macro is optional. If not specified, it defaults to the value
1885 of @code{CALL_USED_REGISTERS}.
1888 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1889 @cindex call-used register
1890 @cindex call-clobbered register
1891 @cindex call-saved register
1892 A C expression that is nonzero if it is not permissible to store a
1893 value of mode @var{mode} in hard register number @var{regno} across a
1894 call without some part of it being clobbered. For most machines this
1895 macro need not be defined. It is only required for machines that do not
1896 preserve the entire contents of a register across a call.
1900 @findex call_used_regs
1903 @findex reg_class_contents
1904 @defmac CONDITIONAL_REGISTER_USAGE
1905 Zero or more C statements that may conditionally modify five variables
1906 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1907 @code{reg_names}, and @code{reg_class_contents}, to take into account
1908 any dependence of these register sets on target flags. The first three
1909 of these are of type @code{char []} (interpreted as Boolean vectors).
1910 @code{global_regs} is a @code{const char *[]}, and
1911 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1912 called, @code{fixed_regs}, @code{call_used_regs},
1913 @code{reg_class_contents}, and @code{reg_names} have been initialized
1914 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1915 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1916 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1917 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1918 command options have been applied.
1920 You need not define this macro if it has no work to do.
1922 @cindex disabling certain registers
1923 @cindex controlling register usage
1924 If the usage of an entire class of registers depends on the target
1925 flags, you may indicate this to GCC by using this macro to modify
1926 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1927 registers in the classes which should not be used by GCC@. Also define
1928 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1929 to return @code{NO_REGS} if it
1930 is called with a letter for a class that shouldn't be used.
1932 (However, if this class is not included in @code{GENERAL_REGS} and all
1933 of the insn patterns whose constraints permit this class are
1934 controlled by target switches, then GCC will automatically avoid using
1935 these registers when the target switches are opposed to them.)
1938 @defmac NON_SAVING_SETJMP
1939 If this macro is defined and has a nonzero value, it means that
1940 @code{setjmp} and related functions fail to save the registers, or that
1941 @code{longjmp} fails to restore them. To compensate, the compiler
1942 avoids putting variables in registers in functions that use
1946 @defmac INCOMING_REGNO (@var{out})
1947 Define this macro if the target machine has register windows. This C
1948 expression returns the register number as seen by the called function
1949 corresponding to the register number @var{out} as seen by the calling
1950 function. Return @var{out} if register number @var{out} is not an
1954 @defmac OUTGOING_REGNO (@var{in})
1955 Define this macro if the target machine has register windows. This C
1956 expression returns the register number as seen by the calling function
1957 corresponding to the register number @var{in} as seen by the called
1958 function. Return @var{in} if register number @var{in} is not an inbound
1962 @defmac LOCAL_REGNO (@var{regno})
1963 Define this macro if the target machine has register windows. This C
1964 expression returns true if the register is call-saved but is in the
1965 register window. Unlike most call-saved registers, such registers
1966 need not be explicitly restored on function exit or during non-local
1971 If the program counter has a register number, define this as that
1972 register number. Otherwise, do not define it.
1975 @node Allocation Order
1976 @subsection Order of Allocation of Registers
1977 @cindex order of register allocation
1978 @cindex register allocation order
1980 @c prevent bad page break with this line
1981 Registers are allocated in order.
1983 @defmac REG_ALLOC_ORDER
1984 If defined, an initializer for a vector of integers, containing the
1985 numbers of hard registers in the order in which GCC should prefer
1986 to use them (from most preferred to least).
1988 If this macro is not defined, registers are used lowest numbered first
1989 (all else being equal).
1991 One use of this macro is on machines where the highest numbered
1992 registers must always be saved and the save-multiple-registers
1993 instruction supports only sequences of consecutive registers. On such
1994 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1995 the highest numbered allocable register first.
1998 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
1999 A C statement (sans semicolon) to choose the order in which to allocate
2000 hard registers for pseudo-registers local to a basic block.
2002 Store the desired register order in the array @code{reg_alloc_order}.
2003 Element 0 should be the register to allocate first; element 1, the next
2004 register; and so on.
2006 The macro body should not assume anything about the contents of
2007 @code{reg_alloc_order} before execution of the macro.
2009 On most machines, it is not necessary to define this macro.
2012 @node Values in Registers
2013 @subsection How Values Fit in Registers
2015 This section discusses the macros that describe which kinds of values
2016 (specifically, which machine modes) each register can hold, and how many
2017 consecutive registers are needed for a given mode.
2019 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2020 A C expression for the number of consecutive hard registers, starting
2021 at register number @var{regno}, required to hold a value of mode
2024 On a machine where all registers are exactly one word, a suitable
2025 definition of this macro is
2028 #define HARD_REGNO_NREGS(REGNO, MODE) \
2029 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2034 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2035 A C expression that is nonzero if it is permissible to store a value
2036 of mode @var{mode} in hard register number @var{regno} (or in several
2037 registers starting with that one). For a machine where all registers
2038 are equivalent, a suitable definition is
2041 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2044 You need not include code to check for the numbers of fixed registers,
2045 because the allocation mechanism considers them to be always occupied.
2047 @cindex register pairs
2048 On some machines, double-precision values must be kept in even/odd
2049 register pairs. You can implement that by defining this macro to reject
2050 odd register numbers for such modes.
2052 The minimum requirement for a mode to be OK in a register is that the
2053 @samp{mov@var{mode}} instruction pattern support moves between the
2054 register and other hard register in the same class and that moving a
2055 value into the register and back out not alter it.
2057 Since the same instruction used to move @code{word_mode} will work for
2058 all narrower integer modes, it is not necessary on any machine for
2059 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2060 you define patterns @samp{movhi}, etc., to take advantage of this. This
2061 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2062 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2065 Many machines have special registers for floating point arithmetic.
2066 Often people assume that floating point machine modes are allowed only
2067 in floating point registers. This is not true. Any registers that
2068 can hold integers can safely @emph{hold} a floating point machine
2069 mode, whether or not floating arithmetic can be done on it in those
2070 registers. Integer move instructions can be used to move the values.
2072 On some machines, though, the converse is true: fixed-point machine
2073 modes may not go in floating registers. This is true if the floating
2074 registers normalize any value stored in them, because storing a
2075 non-floating value there would garble it. In this case,
2076 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2077 floating registers. But if the floating registers do not automatically
2078 normalize, if you can store any bit pattern in one and retrieve it
2079 unchanged without a trap, then any machine mode may go in a floating
2080 register, so you can define this macro to say so.
2082 The primary significance of special floating registers is rather that
2083 they are the registers acceptable in floating point arithmetic
2084 instructions. However, this is of no concern to
2085 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2086 constraints for those instructions.
2088 On some machines, the floating registers are especially slow to access,
2089 so that it is better to store a value in a stack frame than in such a
2090 register if floating point arithmetic is not being done. As long as the
2091 floating registers are not in class @code{GENERAL_REGS}, they will not
2092 be used unless some pattern's constraint asks for one.
2095 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2096 A C expression that is nonzero if a value of mode
2097 @var{mode1} is accessible in mode @var{mode2} without copying.
2099 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2100 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2101 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2102 should be nonzero. If they differ for any @var{r}, you should define
2103 this macro to return zero unless some other mechanism ensures the
2104 accessibility of the value in a narrower mode.
2106 You should define this macro to return nonzero in as many cases as
2107 possible since doing so will allow GCC to perform better register
2111 @defmac AVOID_CCMODE_COPIES
2112 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2113 registers. You should only define this macro if support for copying to/from
2114 @code{CCmode} is incomplete.
2117 @node Leaf Functions
2118 @subsection Handling Leaf Functions
2120 @cindex leaf functions
2121 @cindex functions, leaf
2122 On some machines, a leaf function (i.e., one which makes no calls) can run
2123 more efficiently if it does not make its own register window. Often this
2124 means it is required to receive its arguments in the registers where they
2125 are passed by the caller, instead of the registers where they would
2128 The special treatment for leaf functions generally applies only when
2129 other conditions are met; for example, often they may use only those
2130 registers for its own variables and temporaries. We use the term ``leaf
2131 function'' to mean a function that is suitable for this special
2132 handling, so that functions with no calls are not necessarily ``leaf
2135 GCC assigns register numbers before it knows whether the function is
2136 suitable for leaf function treatment. So it needs to renumber the
2137 registers in order to output a leaf function. The following macros
2140 @defmac LEAF_REGISTERS
2141 Name of a char vector, indexed by hard register number, which
2142 contains 1 for a register that is allowable in a candidate for leaf
2145 If leaf function treatment involves renumbering the registers, then the
2146 registers marked here should be the ones before renumbering---those that
2147 GCC would ordinarily allocate. The registers which will actually be
2148 used in the assembler code, after renumbering, should not be marked with 1
2151 Define this macro only if the target machine offers a way to optimize
2152 the treatment of leaf functions.
2155 @defmac LEAF_REG_REMAP (@var{regno})
2156 A C expression whose value is the register number to which @var{regno}
2157 should be renumbered, when a function is treated as a leaf function.
2159 If @var{regno} is a register number which should not appear in a leaf
2160 function before renumbering, then the expression should yield @minus{}1, which
2161 will cause the compiler to abort.
2163 Define this macro only if the target machine offers a way to optimize the
2164 treatment of leaf functions, and registers need to be renumbered to do
2168 @findex current_function_is_leaf
2169 @findex current_function_uses_only_leaf_regs
2170 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2171 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2172 specially. They can test the C variable @code{current_function_is_leaf}
2173 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2174 set prior to local register allocation and is valid for the remaining
2175 compiler passes. They can also test the C variable
2176 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2177 functions which only use leaf registers.
2178 @code{current_function_uses_only_leaf_regs} is valid after reload and is
2179 only useful if @code{LEAF_REGISTERS} is defined.
2180 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2181 @c of the next paragraph?! --mew 2feb93
2183 @node Stack Registers
2184 @subsection Registers That Form a Stack
2186 There are special features to handle computers where some of the
2187 ``registers'' form a stack. Stack registers are normally written by
2188 pushing onto the stack, and are numbered relative to the top of the
2191 Currently, GCC can only handle one group of stack-like registers, and
2192 they must be consecutively numbered. Furthermore, the existing
2193 support for stack-like registers is specific to the 80387 floating
2194 point coprocessor. If you have a new architecture that uses
2195 stack-like registers, you will need to do substantial work on
2196 @file{reg-stack.c} and write your machine description to cooperate
2197 with it, as well as defining these macros.
2200 Define this if the machine has any stack-like registers.
2203 @defmac FIRST_STACK_REG
2204 The number of the first stack-like register. This one is the top
2208 @defmac LAST_STACK_REG
2209 The number of the last stack-like register. This one is the bottom of
2213 @node Register Classes
2214 @section Register Classes
2215 @cindex register class definitions
2216 @cindex class definitions, register
2218 On many machines, the numbered registers are not all equivalent.
2219 For example, certain registers may not be allowed for indexed addressing;
2220 certain registers may not be allowed in some instructions. These machine
2221 restrictions are described to the compiler using @dfn{register classes}.
2223 You define a number of register classes, giving each one a name and saying
2224 which of the registers belong to it. Then you can specify register classes
2225 that are allowed as operands to particular instruction patterns.
2229 In general, each register will belong to several classes. In fact, one
2230 class must be named @code{ALL_REGS} and contain all the registers. Another
2231 class must be named @code{NO_REGS} and contain no registers. Often the
2232 union of two classes will be another class; however, this is not required.
2234 @findex GENERAL_REGS
2235 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2236 terribly special about the name, but the operand constraint letters
2237 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2238 the same as @code{ALL_REGS}, just define it as a macro which expands
2241 Order the classes so that if class @var{x} is contained in class @var{y}
2242 then @var{x} has a lower class number than @var{y}.
2244 The way classes other than @code{GENERAL_REGS} are specified in operand
2245 constraints is through machine-dependent operand constraint letters.
2246 You can define such letters to correspond to various classes, then use
2247 them in operand constraints.
2249 You should define a class for the union of two classes whenever some
2250 instruction allows both classes. For example, if an instruction allows
2251 either a floating point (coprocessor) register or a general register for a
2252 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2253 which includes both of them. Otherwise you will get suboptimal code.
2255 You must also specify certain redundant information about the register
2256 classes: for each class, which classes contain it and which ones are
2257 contained in it; for each pair of classes, the largest class contained
2260 When a value occupying several consecutive registers is expected in a
2261 certain class, all the registers used must belong to that class.
2262 Therefore, register classes cannot be used to enforce a requirement for
2263 a register pair to start with an even-numbered register. The way to
2264 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2266 Register classes used for input-operands of bitwise-and or shift
2267 instructions have a special requirement: each such class must have, for
2268 each fixed-point machine mode, a subclass whose registers can transfer that
2269 mode to or from memory. For example, on some machines, the operations for
2270 single-byte values (@code{QImode}) are limited to certain registers. When
2271 this is so, each register class that is used in a bitwise-and or shift
2272 instruction must have a subclass consisting of registers from which
2273 single-byte values can be loaded or stored. This is so that
2274 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2276 @deftp {Data type} {enum reg_class}
2277 An enumeral type that must be defined with all the register class names
2278 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
2279 must be the last register class, followed by one more enumeral value,
2280 @code{LIM_REG_CLASSES}, which is not a register class but rather
2281 tells how many classes there are.
2283 Each register class has a number, which is the value of casting
2284 the class name to type @code{int}. The number serves as an index
2285 in many of the tables described below.
2288 @defmac N_REG_CLASSES
2289 The number of distinct register classes, defined as follows:
2292 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2296 @defmac REG_CLASS_NAMES
2297 An initializer containing the names of the register classes as C string
2298 constants. These names are used in writing some of the debugging dumps.
2301 @defmac REG_CLASS_CONTENTS
2302 An initializer containing the contents of the register classes, as integers
2303 which are bit masks. The @var{n}th integer specifies the contents of class
2304 @var{n}. The way the integer @var{mask} is interpreted is that
2305 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2307 When the machine has more than 32 registers, an integer does not suffice.
2308 Then the integers are replaced by sub-initializers, braced groupings containing
2309 several integers. Each sub-initializer must be suitable as an initializer
2310 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2311 In this situation, the first integer in each sub-initializer corresponds to
2312 registers 0 through 31, the second integer to registers 32 through 63, and
2316 @defmac REGNO_REG_CLASS (@var{regno})
2317 A C expression whose value is a register class containing hard register
2318 @var{regno}. In general there is more than one such class; choose a class
2319 which is @dfn{minimal}, meaning that no smaller class also contains the
2323 @defmac BASE_REG_CLASS
2324 A macro whose definition is the name of the class to which a valid
2325 base register must belong. A base register is one used in an address
2326 which is the register value plus a displacement.
2329 @defmac MODE_BASE_REG_CLASS (@var{mode})
2330 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2331 the selection of a base register in a mode dependent manner. If
2332 @var{mode} is VOIDmode then it should return the same value as
2333 @code{BASE_REG_CLASS}.
2336 @defmac INDEX_REG_CLASS
2337 A macro whose definition is the name of the class to which a valid
2338 index register must belong. An index register is one used in an
2339 address where its value is either multiplied by a scale factor or
2340 added to another register (as well as added to a displacement).
2343 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2344 For the constraint at the start of @var{str}, which starts with the letter
2345 @var{c}, return the length. This allows you to have register class /
2346 constant / extra constraints that are longer than a single letter;
2347 you don't need to define this macro if you can do with single-letter
2348 constraints only. The definition of this macro should use
2349 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2350 to handle specially.
2351 There are some sanity checks in genoutput.c that check the constraint lengths
2352 for the md file, so you can also use this macro to help you while you are
2353 transitioning from a byzantine single-letter-constraint scheme: when you
2354 return a negative length for a constraint you want to re-use, genoutput
2355 will complain about every instance where it is used in the md file.
2358 @defmac REG_CLASS_FROM_LETTER (@var{char})
2359 A C expression which defines the machine-dependent operand constraint
2360 letters for register classes. If @var{char} is such a letter, the
2361 value should be the register class corresponding to it. Otherwise,
2362 the value should be @code{NO_REGS}. The register letter @samp{r},
2363 corresponding to class @code{GENERAL_REGS}, will not be passed
2364 to this macro; you do not need to handle it.
2367 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2368 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2369 passed in @var{str}, so that you can use suffixes to distinguish between
2373 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2374 A C expression which is nonzero if register number @var{num} is
2375 suitable for use as a base register in operand addresses. It may be
2376 either a suitable hard register or a pseudo register that has been
2377 allocated such a hard register.
2380 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2381 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2382 that expression may examine the mode of the memory reference in
2383 @var{mode}. You should define this macro if the mode of the memory
2384 reference affects whether a register may be used as a base register. If
2385 you define this macro, the compiler will use it instead of
2386 @code{REGNO_OK_FOR_BASE_P}.
2389 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2390 A C expression which is nonzero if register number @var{num} is
2391 suitable for use as an index register in operand addresses. It may be
2392 either a suitable hard register or a pseudo register that has been
2393 allocated such a hard register.
2395 The difference between an index register and a base register is that
2396 the index register may be scaled. If an address involves the sum of
2397 two registers, neither one of them scaled, then either one may be
2398 labeled the ``base'' and the other the ``index''; but whichever
2399 labeling is used must fit the machine's constraints of which registers
2400 may serve in each capacity. The compiler will try both labelings,
2401 looking for one that is valid, and will reload one or both registers
2402 only if neither labeling works.
2405 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2406 A C expression that places additional restrictions on the register class
2407 to use when it is necessary to copy value @var{x} into a register in class
2408 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2409 another, smaller class. On many machines, the following definition is
2413 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2416 Sometimes returning a more restrictive class makes better code. For
2417 example, on the 68000, when @var{x} is an integer constant that is in range
2418 for a @samp{moveq} instruction, the value of this macro is always
2419 @code{DATA_REGS} as long as @var{class} includes the data registers.
2420 Requiring a data register guarantees that a @samp{moveq} will be used.
2422 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2423 you can force @var{x} into a memory constant. This is useful on
2424 certain machines where immediate floating values cannot be loaded into
2425 certain kinds of registers.
2428 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2429 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2430 input reloads. If you don't define this macro, the default is to use
2431 @var{class}, unchanged.
2434 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2435 A C expression that places additional restrictions on the register class
2436 to use when it is necessary to be able to hold a value of mode
2437 @var{mode} in a reload register for which class @var{class} would
2440 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2441 there are certain modes that simply can't go in certain reload classes.
2443 The value is a register class; perhaps @var{class}, or perhaps another,
2446 Don't define this macro unless the target machine has limitations which
2447 require the macro to do something nontrivial.
2450 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2451 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2452 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2453 Many machines have some registers that cannot be copied directly to or
2454 from memory or even from other types of registers. An example is the
2455 @samp{MQ} register, which on most machines, can only be copied to or
2456 from general registers, but not memory. Some machines allow copying all
2457 registers to and from memory, but require a scratch register for stores
2458 to some memory locations (e.g., those with symbolic address on the RT,
2459 and those with certain symbolic address on the SPARC when compiling
2460 PIC)@. In some cases, both an intermediate and a scratch register are
2463 You should define these macros to indicate to the reload phase that it may
2464 need to allocate at least one register for a reload in addition to the
2465 register to contain the data. Specifically, if copying @var{x} to a
2466 register @var{class} in @var{mode} requires an intermediate register,
2467 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2468 largest register class all of whose registers can be used as
2469 intermediate registers or scratch registers.
2471 If copying a register @var{class} in @var{mode} to @var{x} requires an
2472 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2473 should be defined to return the largest register class required. If the
2474 requirements for input and output reloads are the same, the macro
2475 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2478 The values returned by these macros are often @code{GENERAL_REGS}.
2479 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2480 can be directly copied to or from a register of @var{class} in
2481 @var{mode} without requiring a scratch register. Do not define this
2482 macro if it would always return @code{NO_REGS}.
2484 If a scratch register is required (either with or without an
2485 intermediate register), you should define patterns for
2486 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2487 (@pxref{Standard Names}. These patterns, which will normally be
2488 implemented with a @code{define_expand}, should be similar to the
2489 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2492 Define constraints for the reload register and scratch register that
2493 contain a single register class. If the original reload register (whose
2494 class is @var{class}) can meet the constraint given in the pattern, the
2495 value returned by these macros is used for the class of the scratch
2496 register. Otherwise, two additional reload registers are required.
2497 Their classes are obtained from the constraints in the insn pattern.
2499 @var{x} might be a pseudo-register or a @code{subreg} of a
2500 pseudo-register, which could either be in a hard register or in memory.
2501 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2502 in memory and the hard register number if it is in a register.
2504 These macros should not be used in the case where a particular class of
2505 registers can only be copied to memory and not to another class of
2506 registers. In that case, secondary reload registers are not needed and
2507 would not be helpful. Instead, a stack location must be used to perform
2508 the copy and the @code{mov@var{m}} pattern should use memory as an
2509 intermediate storage. This case often occurs between floating-point and
2513 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2514 Certain machines have the property that some registers cannot be copied
2515 to some other registers without using memory. Define this macro on
2516 those machines to be a C expression that is nonzero if objects of mode
2517 @var{m} in registers of @var{class1} can only be copied to registers of
2518 class @var{class2} by storing a register of @var{class1} into memory
2519 and loading that memory location into a register of @var{class2}.
2521 Do not define this macro if its value would always be zero.
2524 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2525 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2526 allocates a stack slot for a memory location needed for register copies.
2527 If this macro is defined, the compiler instead uses the memory location
2528 defined by this macro.
2530 Do not define this macro if you do not define
2531 @code{SECONDARY_MEMORY_NEEDED}.
2534 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2535 When the compiler needs a secondary memory location to copy between two
2536 registers of mode @var{mode}, it normally allocates sufficient memory to
2537 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2538 load operations in a mode that many bits wide and whose class is the
2539 same as that of @var{mode}.
2541 This is right thing to do on most machines because it ensures that all
2542 bits of the register are copied and prevents accesses to the registers
2543 in a narrower mode, which some machines prohibit for floating-point
2546 However, this default behavior is not correct on some machines, such as
2547 the DEC Alpha, that store short integers in floating-point registers
2548 differently than in integer registers. On those machines, the default
2549 widening will not work correctly and you must define this macro to
2550 suppress that widening in some cases. See the file @file{alpha.h} for
2553 Do not define this macro if you do not define
2554 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2555 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2558 @defmac SMALL_REGISTER_CLASSES
2559 On some machines, it is risky to let hard registers live across arbitrary
2560 insns. Typically, these machines have instructions that require values
2561 to be in specific registers (like an accumulator), and reload will fail
2562 if the required hard register is used for another purpose across such an
2565 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2566 value on these machines. When this macro has a nonzero value, the
2567 compiler will try to minimize the lifetime of hard registers.
2569 It is always safe to define this macro with a nonzero value, but if you
2570 unnecessarily define it, you will reduce the amount of optimizations
2571 that can be performed in some cases. If you do not define this macro
2572 with a nonzero value when it is required, the compiler will run out of
2573 spill registers and print a fatal error message. For most machines, you
2574 should not define this macro at all.
2577 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2578 A C expression whose value is nonzero if pseudos that have been assigned
2579 to registers of class @var{class} would likely be spilled because
2580 registers of @var{class} are needed for spill registers.
2582 The default value of this macro returns 1 if @var{class} has exactly one
2583 register and zero otherwise. On most machines, this default should be
2584 used. Only define this macro to some other expression if pseudos
2585 allocated by @file{local-alloc.c} end up in memory because their hard
2586 registers were needed for spill registers. If this macro returns nonzero
2587 for those classes, those pseudos will only be allocated by
2588 @file{global.c}, which knows how to reallocate the pseudo to another
2589 register. If there would not be another register available for
2590 reallocation, you should not change the definition of this macro since
2591 the only effect of such a definition would be to slow down register
2595 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2596 A C expression for the maximum number of consecutive registers
2597 of class @var{class} needed to hold a value of mode @var{mode}.
2599 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2600 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2601 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2602 @var{mode})} for all @var{regno} values in the class @var{class}.
2604 This macro helps control the handling of multiple-word values
2608 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2609 If defined, a C expression that returns nonzero for a @var{class} for which
2610 a change from mode @var{from} to mode @var{to} is invalid.
2612 For the example, loading 32-bit integer or floating-point objects into
2613 floating-point registers on the Alpha extends them to 64 bits.
2614 Therefore loading a 64-bit object and then storing it as a 32-bit object
2615 does not store the low-order 32 bits, as would be the case for a normal
2616 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2620 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2621 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2622 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2626 Three other special macros describe which operands fit which constraint
2629 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2630 A C expression that defines the machine-dependent operand constraint
2631 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2632 particular ranges of integer values. If @var{c} is one of those
2633 letters, the expression should check that @var{value}, an integer, is in
2634 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2635 not one of those letters, the value should be 0 regardless of
2639 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2640 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2641 string passed in @var{str}, so that you can use suffixes to distinguish
2642 between different variants.
2645 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2646 A C expression that defines the machine-dependent operand constraint
2647 letters that specify particular ranges of @code{const_double} values
2648 (@samp{G} or @samp{H}).
2650 If @var{c} is one of those letters, the expression should check that
2651 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2652 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2653 letters, the value should be 0 regardless of @var{value}.
2655 @code{const_double} is used for all floating-point constants and for
2656 @code{DImode} fixed-point constants. A given letter can accept either
2657 or both kinds of values. It can use @code{GET_MODE} to distinguish
2658 between these kinds.
2661 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2662 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2663 string passed in @var{str}, so that you can use suffixes to distinguish
2664 between different variants.
2667 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2668 A C expression that defines the optional machine-dependent constraint
2669 letters that can be used to segregate specific types of operands, usually
2670 memory references, for the target machine. Any letter that is not
2671 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2672 @code{REG_CLASS_FROM_CONSTRAINT}
2673 may be used. Normally this macro will not be defined.
2675 If it is required for a particular target machine, it should return 1
2676 if @var{value} corresponds to the operand type represented by the
2677 constraint letter @var{c}. If @var{c} is not defined as an extra
2678 constraint, the value returned should be 0 regardless of @var{value}.
2680 For example, on the ROMP, load instructions cannot have their output
2681 in r0 if the memory reference contains a symbolic address. Constraint
2682 letter @samp{Q} is defined as representing a memory address that does
2683 @emph{not} contain a symbolic address. An alternative is specified with
2684 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2685 alternative specifies @samp{m} on the input and a register class that
2686 does not include r0 on the output.
2689 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2690 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2691 in @var{str}, so that you can use suffixes to distinguish between different
2695 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2696 A C expression that defines the optional machine-dependent constraint
2697 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2698 be treated like memory constraints by the reload pass.
2700 It should return 1 if the operand type represented by the constraint
2701 at the start of @var{str}, the first letter of which is the letter @var{c},
2702 comprises a subset of all memory references including
2703 all those whose address is simply a base register. This allows the reload
2704 pass to reload an operand, if it does not directly correspond to the operand
2705 type of @var{c}, by copying its address into a base register.
2707 For example, on the S/390, some instructions do not accept arbitrary
2708 memory references, but only those that do not make use of an index
2709 register. The constraint letter @samp{Q} is defined via
2710 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2711 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2712 a @samp{Q} constraint can handle any memory operand, because the
2713 reload pass knows it can be reloaded by copying the memory address
2714 into a base register if required. This is analogous to the way
2715 a @samp{o} constraint can handle any memory operand.
2718 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2719 A C expression that defines the optional machine-dependent constraint
2720 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2721 @code{EXTRA_CONSTRAINT_STR}, that should
2722 be treated like address constraints by the reload pass.
2724 It should return 1 if the operand type represented by the constraint
2725 at the start of @var{str}, which starts with the letter @var{c}, comprises
2726 a subset of all memory addresses including
2727 all those that consist of just a base register. This allows the reload
2728 pass to reload an operand, if it does not directly correspond to the operand
2729 type of @var{str}, by copying it into a base register.
2731 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2732 be used with the @code{address_operand} predicate. It is treated
2733 analogously to the @samp{p} constraint.
2736 @node Stack and Calling
2737 @section Stack Layout and Calling Conventions
2738 @cindex calling conventions
2740 @c prevent bad page break with this line
2741 This describes the stack layout and calling conventions.
2745 * Exception Handling::
2750 * Register Arguments::
2752 * Aggregate Return::
2760 @subsection Basic Stack Layout
2761 @cindex stack frame layout
2762 @cindex frame layout
2764 @c prevent bad page break with this line
2765 Here is the basic stack layout.
2767 @defmac STACK_GROWS_DOWNWARD
2768 Define this macro if pushing a word onto the stack moves the stack
2769 pointer to a smaller address.
2771 When we say, ``define this macro if @dots{},'' it means that the
2772 compiler checks this macro only with @code{#ifdef} so the precise
2773 definition used does not matter.
2776 @defmac STACK_PUSH_CODE
2777 This macro defines the operation used when something is pushed
2778 on the stack. In RTL, a push operation will be
2779 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2781 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2782 and @code{POST_INC}. Which of these is correct depends on
2783 the stack direction and on whether the stack pointer points
2784 to the last item on the stack or whether it points to the
2785 space for the next item on the stack.
2787 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2788 defined, which is almost always right, and @code{PRE_INC} otherwise,
2789 which is often wrong.
2792 @defmac FRAME_GROWS_DOWNWARD
2793 Define this macro if the addresses of local variable slots are at negative
2794 offsets from the frame pointer.
2797 @defmac ARGS_GROW_DOWNWARD
2798 Define this macro if successive arguments to a function occupy decreasing
2799 addresses on the stack.
2802 @defmac STARTING_FRAME_OFFSET
2803 Offset from the frame pointer to the first local variable slot to be allocated.
2805 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2806 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2807 Otherwise, it is found by adding the length of the first slot to the
2808 value @code{STARTING_FRAME_OFFSET}.
2809 @c i'm not sure if the above is still correct.. had to change it to get
2810 @c rid of an overfull. --mew 2feb93
2813 @defmac STACK_ALIGNMENT_NEEDED
2814 Define to zero to disable final alignment of the stack during reload.
2815 The nonzero default for this macro is suitable for most ports.
2817 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2818 is a register save block following the local block that doesn't require
2819 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2820 stack alignment and do it in the backend.
2823 @defmac STACK_POINTER_OFFSET
2824 Offset from the stack pointer register to the first location at which
2825 outgoing arguments are placed. If not specified, the default value of
2826 zero is used. This is the proper value for most machines.
2828 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2829 the first location at which outgoing arguments are placed.
2832 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2833 Offset from the argument pointer register to the first argument's
2834 address. On some machines it may depend on the data type of the
2837 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2838 the first argument's address.
2841 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2842 Offset from the stack pointer register to an item dynamically allocated
2843 on the stack, e.g., by @code{alloca}.
2845 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2846 length of the outgoing arguments. The default is correct for most
2847 machines. See @file{function.c} for details.
2850 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2851 A C expression whose value is RTL representing the address in a stack
2852 frame where the pointer to the caller's frame is stored. Assume that
2853 @var{frameaddr} is an RTL expression for the address of the stack frame
2856 If you don't define this macro, the default is to return the value
2857 of @var{frameaddr}---that is, the stack frame address is also the
2858 address of the stack word that points to the previous frame.
2861 @defmac SETUP_FRAME_ADDRESSES
2862 If defined, a C expression that produces the machine-specific code to
2863 setup the stack so that arbitrary frames can be accessed. For example,
2864 on the SPARC, we must flush all of the register windows to the stack
2865 before we can access arbitrary stack frames. You will seldom need to
2869 @defmac BUILTIN_SETJMP_FRAME_VALUE
2870 If defined, a C expression that contains an rtx that is used to store
2871 the address of the current frame into the built in @code{setjmp} buffer.
2872 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2873 machines. One reason you may need to define this macro is if
2874 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2877 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2878 A C expression whose value is RTL representing the value of the return
2879 address for the frame @var{count} steps up from the current frame, after
2880 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2881 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2882 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2884 The value of the expression must always be the correct address when
2885 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2886 determine the return address of other frames.
2889 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2890 Define this if the return address of a particular stack frame is accessed
2891 from the frame pointer of the previous stack frame.
2894 @defmac INCOMING_RETURN_ADDR_RTX
2895 A C expression whose value is RTL representing the location of the
2896 incoming return address at the beginning of any function, before the
2897 prologue. This RTL is either a @code{REG}, indicating that the return
2898 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2901 You only need to define this macro if you want to support call frame
2902 debugging information like that provided by DWARF 2.
2904 If this RTL is a @code{REG}, you should also define
2905 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2908 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2909 A C expression whose value is an integer giving a DWARF 2 column
2910 number that may be used as an alternate return column. This should
2911 be defined only if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2912 general register, but an alternate column needs to be used for
2916 @defmac INCOMING_FRAME_SP_OFFSET
2917 A C expression whose value is an integer giving the offset, in bytes,
2918 from the value of the stack pointer register to the top of the stack
2919 frame at the beginning of any function, before the prologue. The top of
2920 the frame is defined to be the value of the stack pointer in the
2921 previous frame, just before the call instruction.
2923 You only need to define this macro if you want to support call frame
2924 debugging information like that provided by DWARF 2.
2927 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2928 A C expression whose value is an integer giving the offset, in bytes,
2929 from the argument pointer to the canonical frame address (cfa). The
2930 final value should coincide with that calculated by
2931 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2932 during virtual register instantiation.
2934 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2935 which is correct for most machines; in general, the arguments are found
2936 immediately before the stack frame. Note that this is not the case on
2937 some targets that save registers into the caller's frame, such as SPARC
2938 and rs6000, and so such targets need to define this macro.
2940 You only need to define this macro if the default is incorrect, and you
2941 want to support call frame debugging information like that provided by
2945 @node Exception Handling
2946 @subsection Exception Handling Support
2947 @cindex exception handling
2949 @defmac EH_RETURN_DATA_REGNO (@var{N})
2950 A C expression whose value is the @var{N}th register number used for
2951 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2952 @var{N} registers are usable.
2954 The exception handling library routines communicate with the exception
2955 handlers via a set of agreed upon registers. Ideally these registers
2956 should be call-clobbered; it is possible to use call-saved registers,
2957 but may negatively impact code size. The target must support at least
2958 2 data registers, but should define 4 if there are enough free registers.
2960 You must define this macro if you want to support call frame exception
2961 handling like that provided by DWARF 2.
2964 @defmac EH_RETURN_STACKADJ_RTX
2965 A C expression whose value is RTL representing a location in which
2966 to store a stack adjustment to be applied before function return.
2967 This is used to unwind the stack to an exception handler's call frame.
2968 It will be assigned zero on code paths that return normally.
2970 Typically this is a call-clobbered hard register that is otherwise
2971 untouched by the epilogue, but could also be a stack slot.
2973 Do not define this macro if the stack pointer is saved and restored
2974 by the regular prolog and epilog code in the call frame itself; in
2975 this case, the exception handling library routines will update the
2976 stack location to be restored in place. Otherwise, you must define
2977 this macro if you want to support call frame exception handling like
2978 that provided by DWARF 2.
2981 @defmac EH_RETURN_HANDLER_RTX
2982 A C expression whose value is RTL representing a location in which
2983 to store the address of an exception handler to which we should
2984 return. It will not be assigned on code paths that return normally.
2986 Typically this is the location in the call frame at which the normal
2987 return address is stored. For targets that return by popping an
2988 address off the stack, this might be a memory address just below
2989 the @emph{target} call frame rather than inside the current call
2990 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
2991 been assigned, so it may be used to calculate the location of the
2994 Some targets have more complex requirements than storing to an
2995 address calculable during initial code generation. In that case
2996 the @code{eh_return} instruction pattern should be used instead.
2998 If you want to support call frame exception handling, you must
2999 define either this macro or the @code{eh_return} instruction pattern.
3002 @defmac RETURN_ADDR_OFFSET
3003 If defined, an integer-valued C expression for which rtl will be generated
3004 to add it to the exception handler address before it is searched in the
3005 exception handling tables, and to subtract it again from the address before
3006 using it to return to the exception handler.
3009 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3010 This macro chooses the encoding of pointers embedded in the exception
3011 handling sections. If at all possible, this should be defined such
3012 that the exception handling section will not require dynamic relocations,
3013 and so may be read-only.
3015 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3016 @var{global} is true if the symbol may be affected by dynamic relocations.
3017 The macro should return a combination of the @code{DW_EH_PE_*} defines
3018 as found in @file{dwarf2.h}.
3020 If this macro is not defined, pointers will not be encoded but
3021 represented directly.
3024 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3025 This macro allows the target to emit whatever special magic is required
3026 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3027 Generic code takes care of pc-relative and indirect encodings; this must
3028 be defined if the target uses text-relative or data-relative encodings.
3030 This is a C statement that branches to @var{done} if the format was
3031 handled. @var{encoding} is the format chosen, @var{size} is the number
3032 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3036 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs}, @var{success})
3037 This macro allows the target to add cpu and operating system specific
3038 code to the call-frame unwinder for use when there is no unwind data
3039 available. The most common reason to implement this macro is to unwind
3040 through signal frames.
3042 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3043 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3044 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3045 for the address of the code being executed and @code{context->cfa} for
3046 the stack pointer value. If the frame can be decoded, the register save
3047 addresses should be updated in @var{fs} and the macro should branch to
3048 @var{success}. If the frame cannot be decoded, the macro should do
3051 For proper signal handling in Java this macro is accompanied by
3052 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3055 @node Stack Checking
3056 @subsection Specifying How Stack Checking is Done
3058 GCC will check that stack references are within the boundaries of
3059 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3063 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3064 will assume that you have arranged for stack checking to be done at
3065 appropriate places in the configuration files, e.g., in
3066 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3070 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3071 called @code{check_stack} in your @file{md} file, GCC will call that
3072 pattern with one argument which is the address to compare the stack
3073 value against. You must arrange for this pattern to report an error if
3074 the stack pointer is out of range.
3077 If neither of the above are true, GCC will generate code to periodically
3078 ``probe'' the stack pointer using the values of the macros defined below.
3081 Normally, you will use the default values of these macros, so GCC
3082 will use the third approach.
3084 @defmac STACK_CHECK_BUILTIN
3085 A nonzero value if stack checking is done by the configuration files in a
3086 machine-dependent manner. You should define this macro if stack checking
3087 is require by the ABI of your machine or if you would like to have to stack
3088 checking in some more efficient way than GCC's portable approach.
3089 The default value of this macro is zero.
3092 @defmac STACK_CHECK_PROBE_INTERVAL
3093 An integer representing the interval at which GCC must generate stack
3094 probe instructions. You will normally define this macro to be no larger
3095 than the size of the ``guard pages'' at the end of a stack area. The
3096 default value of 4096 is suitable for most systems.
3099 @defmac STACK_CHECK_PROBE_LOAD
3100 A integer which is nonzero if GCC should perform the stack probe
3101 as a load instruction and zero if GCC should use a store instruction.
3102 The default is zero, which is the most efficient choice on most systems.
3105 @defmac STACK_CHECK_PROTECT
3106 The number of bytes of stack needed to recover from a stack overflow,
3107 for languages where such a recovery is supported. The default value of
3108 75 words should be adequate for most machines.
3111 @defmac STACK_CHECK_MAX_FRAME_SIZE
3112 The maximum size of a stack frame, in bytes. GCC will generate probe
3113 instructions in non-leaf functions to ensure at least this many bytes of
3114 stack are available. If a stack frame is larger than this size, stack
3115 checking will not be reliable and GCC will issue a warning. The
3116 default is chosen so that GCC only generates one instruction on most
3117 systems. You should normally not change the default value of this macro.
3120 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3121 GCC uses this value to generate the above warning message. It
3122 represents the amount of fixed frame used by a function, not including
3123 space for any callee-saved registers, temporaries and user variables.
3124 You need only specify an upper bound for this amount and will normally
3125 use the default of four words.
3128 @defmac STACK_CHECK_MAX_VAR_SIZE
3129 The maximum size, in bytes, of an object that GCC will place in the
3130 fixed area of the stack frame when the user specifies
3131 @option{-fstack-check}.
3132 GCC computed the default from the values of the above macros and you will
3133 normally not need to override that default.
3137 @node Frame Registers
3138 @subsection Registers That Address the Stack Frame
3140 @c prevent bad page break with this line
3141 This discusses registers that address the stack frame.
3143 @defmac STACK_POINTER_REGNUM
3144 The register number of the stack pointer register, which must also be a
3145 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3146 the hardware determines which register this is.
3149 @defmac FRAME_POINTER_REGNUM
3150 The register number of the frame pointer register, which is used to
3151 access automatic variables in the stack frame. On some machines, the
3152 hardware determines which register this is. On other machines, you can
3153 choose any register you wish for this purpose.
3156 @defmac HARD_FRAME_POINTER_REGNUM
3157 On some machines the offset between the frame pointer and starting
3158 offset of the automatic variables is not known until after register
3159 allocation has been done (for example, because the saved registers are
3160 between these two locations). On those machines, define
3161 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3162 be used internally until the offset is known, and define
3163 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3164 used for the frame pointer.
3166 You should define this macro only in the very rare circumstances when it
3167 is not possible to calculate the offset between the frame pointer and
3168 the automatic variables until after register allocation has been
3169 completed. When this macro is defined, you must also indicate in your
3170 definition of @code{ELIMINABLE_REGS} how to eliminate
3171 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3172 or @code{STACK_POINTER_REGNUM}.
3174 Do not define this macro if it would be the same as
3175 @code{FRAME_POINTER_REGNUM}.
3178 @defmac ARG_POINTER_REGNUM
3179 The register number of the arg pointer register, which is used to access
3180 the function's argument list. On some machines, this is the same as the
3181 frame pointer register. On some machines, the hardware determines which
3182 register this is. On other machines, you can choose any register you
3183 wish for this purpose. If this is not the same register as the frame
3184 pointer register, then you must mark it as a fixed register according to
3185 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3186 (@pxref{Elimination}).
3189 @defmac RETURN_ADDRESS_POINTER_REGNUM
3190 The register number of the return address pointer register, which is used to
3191 access the current function's return address from the stack. On some
3192 machines, the return address is not at a fixed offset from the frame
3193 pointer or stack pointer or argument pointer. This register can be defined
3194 to point to the return address on the stack, and then be converted by
3195 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3197 Do not define this macro unless there is no other way to get the return
3198 address from the stack.
3201 @defmac STATIC_CHAIN_REGNUM
3202 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3203 Register numbers used for passing a function's static chain pointer. If
3204 register windows are used, the register number as seen by the called
3205 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3206 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3207 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3210 The static chain register need not be a fixed register.
3212 If the static chain is passed in memory, these macros should not be
3213 defined; instead, the next two macros should be defined.
3216 @defmac STATIC_CHAIN
3217 @defmacx STATIC_CHAIN_INCOMING
3218 If the static chain is passed in memory, these macros provide rtx giving
3219 @code{mem} expressions that denote where they are stored.
3220 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3221 as seen by the calling and called functions, respectively. Often the former
3222 will be at an offset from the stack pointer and the latter at an offset from
3225 @findex stack_pointer_rtx
3226 @findex frame_pointer_rtx
3227 @findex arg_pointer_rtx
3228 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3229 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3230 macros and should be used to refer to those items.
3232 If the static chain is passed in a register, the two previous macros should
3236 @defmac DWARF_FRAME_REGISTERS
3237 This macro specifies the maximum number of hard registers that can be
3238 saved in a call frame. This is used to size data structures used in
3239 DWARF2 exception handling.
3241 Prior to GCC 3.0, this macro was needed in order to establish a stable
3242 exception handling ABI in the face of adding new hard registers for ISA
3243 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3244 in the number of hard registers. Nevertheless, this macro can still be
3245 used to reduce the runtime memory requirements of the exception handling
3246 routines, which can be substantial if the ISA contains a lot of
3247 registers that are not call-saved.
3249 If this macro is not defined, it defaults to
3250 @code{FIRST_PSEUDO_REGISTER}.
3253 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3255 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3256 for backward compatibility in pre GCC 3.0 compiled code.
3258 If this macro is not defined, it defaults to
3259 @code{DWARF_FRAME_REGISTERS}.
3262 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3264 Define this macro if the target's representation for dwarf registers
3265 is different than the internal representation for unwind column.
3266 Given a dwarf register, this macro should return the internal unwind
3267 column number to use instead.
3269 See the PowerPC's SPE target for an example.
3273 @subsection Eliminating Frame Pointer and Arg Pointer
3275 @c prevent bad page break with this line
3276 This is about eliminating the frame pointer and arg pointer.
3278 @defmac FRAME_POINTER_REQUIRED
3279 A C expression which is nonzero if a function must have and use a frame
3280 pointer. This expression is evaluated in the reload pass. If its value is
3281 nonzero the function will have a frame pointer.
3283 The expression can in principle examine the current function and decide
3284 according to the facts, but on most machines the constant 0 or the
3285 constant 1 suffices. Use 0 when the machine allows code to be generated
3286 with no frame pointer, and doing so saves some time or space. Use 1
3287 when there is no possible advantage to avoiding a frame pointer.
3289 In certain cases, the compiler does not know how to produce valid code
3290 without a frame pointer. The compiler recognizes those cases and
3291 automatically gives the function a frame pointer regardless of what
3292 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3295 In a function that does not require a frame pointer, the frame pointer
3296 register can be allocated for ordinary usage, unless you mark it as a
3297 fixed register. See @code{FIXED_REGISTERS} for more information.
3300 @findex get_frame_size
3301 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3302 A C statement to store in the variable @var{depth-var} the difference
3303 between the frame pointer and the stack pointer values immediately after
3304 the function prologue. The value would be computed from information
3305 such as the result of @code{get_frame_size ()} and the tables of
3306 registers @code{regs_ever_live} and @code{call_used_regs}.
3308 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3309 need not be defined. Otherwise, it must be defined even if
3310 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3311 case, you may set @var{depth-var} to anything.
3314 @defmac ELIMINABLE_REGS
3315 If defined, this macro specifies a table of register pairs used to
3316 eliminate unneeded registers that point into the stack frame. If it is not
3317 defined, the only elimination attempted by the compiler is to replace
3318 references to the frame pointer with references to the stack pointer.
3320 The definition of this macro is a list of structure initializations, each
3321 of which specifies an original and replacement register.
3323 On some machines, the position of the argument pointer is not known until
3324 the compilation is completed. In such a case, a separate hard register
3325 must be used for the argument pointer. This register can be eliminated by
3326 replacing it with either the frame pointer or the argument pointer,
3327 depending on whether or not the frame pointer has been eliminated.
3329 In this case, you might specify:
3331 #define ELIMINABLE_REGS \
3332 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3333 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3334 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3337 Note that the elimination of the argument pointer with the stack pointer is
3338 specified first since that is the preferred elimination.
3341 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3342 A C expression that returns nonzero if the compiler is allowed to try
3343 to replace register number @var{from-reg} with register number
3344 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3345 is defined, and will usually be the constant 1, since most of the cases
3346 preventing register elimination are things that the compiler already
3350 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3351 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3352 specifies the initial difference between the specified pair of
3353 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3357 @node Stack Arguments
3358 @subsection Passing Function Arguments on the Stack
3359 @cindex arguments on stack
3360 @cindex stack arguments
3362 The macros in this section control how arguments are passed
3363 on the stack. See the following section for other macros that
3364 control passing certain arguments in registers.
3366 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3367 This target hook returns @code{true} if an argument declared in a
3368 prototype as an integral type smaller than @code{int} should actually be
3369 passed as an @code{int}. In addition to avoiding errors in certain
3370 cases of mismatch, it also makes for better code on certain machines.
3371 The default is to not promote prototypes.
3375 A C expression. If nonzero, push insns will be used to pass
3377 If the target machine does not have a push instruction, set it to zero.
3378 That directs GCC to use an alternate strategy: to
3379 allocate the entire argument block and then store the arguments into
3380 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3383 @defmac PUSH_ARGS_REVERSED
3384 A C expression. If nonzero, function arguments will be evaluated from
3385 last to first, rather than from first to last. If this macro is not
3386 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3387 and args grow in opposite directions, and 0 otherwise.
3390 @defmac PUSH_ROUNDING (@var{npushed})
3391 A C expression that is the number of bytes actually pushed onto the
3392 stack when an instruction attempts to push @var{npushed} bytes.
3394 On some machines, the definition
3397 #define PUSH_ROUNDING(BYTES) (BYTES)
3401 will suffice. But on other machines, instructions that appear
3402 to push one byte actually push two bytes in an attempt to maintain
3403 alignment. Then the definition should be
3406 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3410 @findex current_function_outgoing_args_size
3411 @defmac ACCUMULATE_OUTGOING_ARGS
3412 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3413 will be computed and placed into the variable
3414 @code{current_function_outgoing_args_size}. No space will be pushed
3415 onto the stack for each call; instead, the function prologue should
3416 increase the stack frame size by this amount.
3418 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3422 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3423 Define this macro if functions should assume that stack space has been
3424 allocated for arguments even when their values are passed in
3427 The value of this macro is the size, in bytes, of the area reserved for
3428 arguments passed in registers for the function represented by @var{fndecl},
3429 which can be zero if GCC is calling a library function.
3431 This space can be allocated by the caller, or be a part of the
3432 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3435 @c above is overfull. not sure what to do. --mew 5feb93 did
3436 @c something, not sure if it looks good. --mew 10feb93
3438 @defmac MAYBE_REG_PARM_STACK_SPACE
3439 @defmacx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
3440 Define these macros in addition to the one above if functions might
3441 allocate stack space for arguments even when their values are passed
3442 in registers. These should be used when the stack space allocated
3443 for arguments in registers is not a simple constant independent of the
3444 function declaration.
3446 The value of the first macro is the size, in bytes, of the area that
3447 we should initially assume would be reserved for arguments passed in registers.
3449 The value of the second macro is the actual size, in bytes, of the area
3450 that will be reserved for arguments passed in registers. This takes two
3451 arguments: an integer representing the number of bytes of fixed sized
3452 arguments on the stack, and a tree representing the number of bytes of
3453 variable sized arguments on the stack.
3455 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
3456 called for libcall functions, the current function, or for a function
3457 being called when it is known that such stack space must be allocated.
3458 In each case this value can be easily computed.
3460 When deciding whether a called function needs such stack space, and how
3461 much space to reserve, GCC uses these two macros instead of
3462 @code{REG_PARM_STACK_SPACE}.
3465 @defmac OUTGOING_REG_PARM_STACK_SPACE
3466 Define this if it is the responsibility of the caller to allocate the area
3467 reserved for arguments passed in registers.
3469 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3470 whether the space for these arguments counts in the value of
3471 @code{current_function_outgoing_args_size}.
3474 @defmac STACK_PARMS_IN_REG_PARM_AREA
3475 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3476 stack parameters don't skip the area specified by it.
3477 @c i changed this, makes more sens and it should have taken care of the
3478 @c overfull.. not as specific, tho. --mew 5feb93
3480 Normally, when a parameter is not passed in registers, it is placed on the
3481 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3482 suppresses this behavior and causes the parameter to be passed on the
3483 stack in its natural location.
3486 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3487 A C expression that should indicate the number of bytes of its own
3488 arguments that a function pops on returning, or 0 if the
3489 function pops no arguments and the caller must therefore pop them all
3490 after the function returns.
3492 @var{fundecl} is a C variable whose value is a tree node that describes
3493 the function in question. Normally it is a node of type
3494 @code{FUNCTION_DECL} that describes the declaration of the function.
3495 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3497 @var{funtype} is a C variable whose value is a tree node that
3498 describes the function in question. Normally it is a node of type
3499 @code{FUNCTION_TYPE} that describes the data type of the function.
3500 From this it is possible to obtain the data types of the value and
3501 arguments (if known).
3503 When a call to a library function is being considered, @var{fundecl}
3504 will contain an identifier node for the library function. Thus, if
3505 you need to distinguish among various library functions, you can do so
3506 by their names. Note that ``library function'' in this context means
3507 a function used to perform arithmetic, whose name is known specially
3508 in the compiler and was not mentioned in the C code being compiled.
3510 @var{stack-size} is the number of bytes of arguments passed on the
3511 stack. If a variable number of bytes is passed, it is zero, and
3512 argument popping will always be the responsibility of the calling function.
3514 On the VAX, all functions always pop their arguments, so the definition
3515 of this macro is @var{stack-size}. On the 68000, using the standard
3516 calling convention, no functions pop their arguments, so the value of
3517 the macro is always 0 in this case. But an alternative calling
3518 convention is available in which functions that take a fixed number of
3519 arguments pop them but other functions (such as @code{printf}) pop
3520 nothing (the caller pops all). When this convention is in use,
3521 @var{funtype} is examined to determine whether a function takes a fixed
3522 number of arguments.
3525 @defmac CALL_POPS_ARGS (@var{cum})
3526 A C expression that should indicate the number of bytes a call sequence
3527 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3528 when compiling a function call.
3530 @var{cum} is the variable in which all arguments to the called function
3531 have been accumulated.
3533 On certain architectures, such as the SH5, a call trampoline is used
3534 that pops certain registers off the stack, depending on the arguments
3535 that have been passed to the function. Since this is a property of the
3536 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3540 @node Register Arguments
3541 @subsection Passing Arguments in Registers
3542 @cindex arguments in registers
3543 @cindex registers arguments
3545 This section describes the macros which let you control how various
3546 types of arguments are passed in registers or how they are arranged in
3549 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3550 A C expression that controls whether a function argument is passed
3551 in a register, and which register.
3553 The arguments are @var{cum}, which summarizes all the previous
3554 arguments; @var{mode}, the machine mode of the argument; @var{type},
3555 the data type of the argument as a tree node or 0 if that is not known
3556 (which happens for C support library functions); and @var{named},
3557 which is 1 for an ordinary argument and 0 for nameless arguments that
3558 correspond to @samp{@dots{}} in the called function's prototype.
3559 @var{type} can be an incomplete type if a syntax error has previously
3562 The value of the expression is usually either a @code{reg} RTX for the
3563 hard register in which to pass the argument, or zero to pass the
3564 argument on the stack.
3566 For machines like the VAX and 68000, where normally all arguments are
3567 pushed, zero suffices as a definition.
3569 The value of the expression can also be a @code{parallel} RTX@. This is
3570 used when an argument is passed in multiple locations. The mode of the
3571 @code{parallel} should be the mode of the entire argument. The
3572 @code{parallel} holds any number of @code{expr_list} pairs; each one
3573 describes where part of the argument is passed. In each
3574 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3575 register in which to pass this part of the argument, and the mode of the
3576 register RTX indicates how large this part of the argument is. The
3577 second operand of the @code{expr_list} is a @code{const_int} which gives
3578 the offset in bytes into the entire argument of where this part starts.
3579 As a special exception the first @code{expr_list} in the @code{parallel}
3580 RTX may have a first operand of zero. This indicates that the entire
3581 argument is also stored on the stack.
3583 The last time this macro is called, it is called with @code{MODE ==
3584 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3585 pattern as operands 2 and 3 respectively.
3587 @cindex @file{stdarg.h} and register arguments
3588 The usual way to make the ISO library @file{stdarg.h} work on a machine
3589 where some arguments are usually passed in registers, is to cause
3590 nameless arguments to be passed on the stack instead. This is done
3591 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3593 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3594 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3595 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3596 in the definition of this macro to determine if this argument is of a
3597 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3598 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3599 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3600 defined, the argument will be computed in the stack and then loaded into
3604 @defmac MUST_PASS_IN_STACK (@var{mode}, @var{type})
3605 Define as a C expression that evaluates to nonzero if we do not know how
3606 to pass TYPE solely in registers. The file @file{expr.h} defines a
3607 definition that is usually appropriate, refer to @file{expr.h} for additional
3611 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3612 Define this macro if the target machine has ``register windows'', so
3613 that the register in which a function sees an arguments is not
3614 necessarily the same as the one in which the caller passed the
3617 For such machines, @code{FUNCTION_ARG} computes the register in which
3618 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3619 be defined in a similar fashion to tell the function being called
3620 where the arguments will arrive.
3622 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3623 serves both purposes.
3626 @defmac FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3627 A C expression for the number of words, at the beginning of an
3628 argument, that must be put in registers. The value must be zero for
3629 arguments that are passed entirely in registers or that are entirely
3630 pushed on the stack.
3632 On some machines, certain arguments must be passed partially in
3633 registers and partially in memory. On these machines, typically the
3634 first @var{n} words of arguments are passed in registers, and the rest
3635 on the stack. If a multi-word argument (a @code{double} or a
3636 structure) crosses that boundary, its first few words must be passed
3637 in registers and the rest must be pushed. This macro tells the
3638 compiler when this occurs, and how many of the words should go in
3641 @code{FUNCTION_ARG} for these arguments should return the first
3642 register to be used by the caller for this argument; likewise
3643 @code{FUNCTION_INCOMING_ARG}, for the called function.
3646 @defmac FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3647 A C expression that indicates when an argument must be passed by reference.
3648 If nonzero for an argument, a copy of that argument is made in memory and a
3649 pointer to the argument is passed instead of the argument itself.
3650 The pointer is passed in whatever way is appropriate for passing a pointer
3653 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3654 definition of this macro might be
3656 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3657 (CUM, MODE, TYPE, NAMED) \
3658 MUST_PASS_IN_STACK (MODE, TYPE)
3660 @c this is *still* too long. --mew 5feb93
3663 @defmac FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3664 If defined, a C expression that indicates when it is the called function's
3665 responsibility to make a copy of arguments passed by invisible reference.
3666 Normally, the caller makes a copy and passes the address of the copy to the
3667 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3668 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3669 ``live'' value. The called function must not modify this value. If it can be
3670 determined that the value won't be modified, it need not make a copy;
3671 otherwise a copy must be made.
3674 @defmac CUMULATIVE_ARGS
3675 A C type for declaring a variable that is used as the first argument of
3676 @code{FUNCTION_ARG} and other related values. For some target machines,
3677 the type @code{int} suffices and can hold the number of bytes of
3680 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3681 arguments that have been passed on the stack. The compiler has other
3682 variables to keep track of that. For target machines on which all
3683 arguments are passed on the stack, there is no need to store anything in
3684 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3685 should not be empty, so use @code{int}.
3688 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl})
3689 A C statement (sans semicolon) for initializing the variable
3690 @var{cum} for the state at the beginning of the argument list. The
3691 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3692 is the tree node for the data type of the function which will receive
3693 the args, or 0 if the args are to a compiler support library function.
3694 For direct calls that are not libcalls, @var{fndecl} contain the
3695 declaration node of the function. @var{fndecl} is also set when
3696 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3699 When processing a call to a compiler support library function,
3700 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3701 contains the name of the function, as a string. @var{libname} is 0 when
3702 an ordinary C function call is being processed. Thus, each time this
3703 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3704 never both of them at once.
3707 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3708 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3709 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3710 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3711 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3712 0)} is used instead.
3715 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3716 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3717 finding the arguments for the function being compiled. If this macro is
3718 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3720 The value passed for @var{libname} is always 0, since library routines
3721 with special calling conventions are never compiled with GCC@. The
3722 argument @var{libname} exists for symmetry with
3723 @code{INIT_CUMULATIVE_ARGS}.
3724 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3725 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3728 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3729 A C statement (sans semicolon) to update the summarizer variable
3730 @var{cum} to advance past an argument in the argument list. The
3731 values @var{mode}, @var{type} and @var{named} describe that argument.
3732 Once this is done, the variable @var{cum} is suitable for analyzing
3733 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3735 This macro need not do anything if the argument in question was passed
3736 on the stack. The compiler knows how to track the amount of stack space
3737 used for arguments without any special help.
3740 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3741 If defined, a C expression which determines whether, and in which direction,
3742 to pad out an argument with extra space. The value should be of type
3743 @code{enum direction}: either @code{upward} to pad above the argument,
3744 @code{downward} to pad below, or @code{none} to inhibit padding.
3746 The @emph{amount} of padding is always just enough to reach the next
3747 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3750 This macro has a default definition which is right for most systems.
3751 For little-endian machines, the default is to pad upward. For
3752 big-endian machines, the default is to pad downward for an argument of
3753 constant size shorter than an @code{int}, and upward otherwise.
3756 @defmac PAD_VARARGS_DOWN
3757 If defined, a C expression which determines whether the default
3758 implementation of va_arg will attempt to pad down before reading the
3759 next argument, if that argument is smaller than its aligned space as
3760 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3761 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3764 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3765 Specify padding for the last element of a block move between registers and
3766 memory. @var{first} is nonzero if this is the only element. Defining this
3767 macro allows better control of register function parameters on big-endian
3768 machines, without using @code{PARALLEL} rtl. In particular,
3769 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3770 registers, as there is no longer a "wrong" part of a register; For example,
3771 a three byte aggregate may be passed in the high part of a register if so
3775 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3776 If defined, a C expression that gives the alignment boundary, in bits,
3777 of an argument with the specified mode and type. If it is not defined,
3778 @code{PARM_BOUNDARY} is used for all arguments.
3781 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3782 A C expression that is nonzero if @var{regno} is the number of a hard
3783 register in which function arguments are sometimes passed. This does
3784 @emph{not} include implicit arguments such as the static chain and
3785 the structure-value address. On many machines, no registers can be
3786 used for this purpose since all function arguments are pushed on the
3790 @defmac SPLIT_COMPLEX_ARGS
3792 Define this macro to a nonzero value if complex function arguments
3793 should be split into their corresponding components. By default, GCC
3794 will attempt to pack complex arguments into the target's word size.
3795 Some ABIs require complex arguments to be split and treated as their
3796 individual components. For example, on AIX64, complex floats should
3797 be passed in a pair of floating point registers, even though a complex
3798 float would fit in one 64-bit floating point register.
3801 @defmac LOAD_ARGS_REVERSED
3802 If defined, the order in which arguments are loaded into their
3803 respective argument registers is reversed so that the last
3804 argument is loaded first. This macro only affects arguments
3805 passed in registers.
3809 @subsection How Scalar Function Values Are Returned
3810 @cindex return values in registers
3811 @cindex values, returned by functions
3812 @cindex scalars, returned as values
3814 This section discusses the macros that control returning scalars as
3815 values---values that can fit in registers.
3817 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3818 A C expression to create an RTX representing the place where a
3819 function returns a value of data type @var{valtype}. @var{valtype} is
3820 a tree node representing a data type. Write @code{TYPE_MODE
3821 (@var{valtype})} to get the machine mode used to represent that type.
3822 On many machines, only the mode is relevant. (Actually, on most
3823 machines, scalar values are returned in the same place regardless of
3826 The value of the expression is usually a @code{reg} RTX for the hard
3827 register where the return value is stored. The value can also be a
3828 @code{parallel} RTX, if the return value is in multiple places. See
3829 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3831 If @code{TARGET_PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3832 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3835 If the precise function being called is known, @var{func} is a tree
3836 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3837 pointer. This makes it possible to use a different value-returning
3838 convention for specific functions when all their calls are
3841 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3842 types, because these are returned in another way. See
3843 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3846 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3847 Define this macro if the target machine has ``register windows''
3848 so that the register in which a function returns its value is not
3849 the same as the one in which the caller sees the value.
3851 For such machines, @code{FUNCTION_VALUE} computes the register in which
3852 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3853 defined in a similar fashion to tell the function where to put the
3856 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3857 @code{FUNCTION_VALUE} serves both purposes.
3859 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3860 aggregate data types, because these are returned in another way. See
3861 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3864 @defmac LIBCALL_VALUE (@var{mode})
3865 A C expression to create an RTX representing the place where a library
3866 function returns a value of mode @var{mode}. If the precise function
3867 being called is known, @var{func} is a tree node
3868 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3869 pointer. This makes it possible to use a different value-returning
3870 convention for specific functions when all their calls are
3873 Note that ``library function'' in this context means a compiler
3874 support routine, used to perform arithmetic, whose name is known
3875 specially by the compiler and was not mentioned in the C code being
3878 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3879 data types, because none of the library functions returns such types.
3882 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3883 A C expression that is nonzero if @var{regno} is the number of a hard
3884 register in which the values of called function may come back.
3886 A register whose use for returning values is limited to serving as the
3887 second of a pair (for a value of type @code{double}, say) need not be
3888 recognized by this macro. So for most machines, this definition
3892 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3895 If the machine has register windows, so that the caller and the called
3896 function use different registers for the return value, this macro
3897 should recognize only the caller's register numbers.
3900 @defmac APPLY_RESULT_SIZE
3901 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3902 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3903 saving and restoring an arbitrary return value.
3906 @node Aggregate Return
3907 @subsection How Large Values Are Returned
3908 @cindex aggregates as return values
3909 @cindex large return values
3910 @cindex returning aggregate values
3911 @cindex structure value address
3913 When a function value's mode is @code{BLKmode} (and in some other
3914 cases), the value is not returned according to @code{FUNCTION_VALUE}
3915 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3916 block of memory in which the value should be stored. This address
3917 is called the @dfn{structure value address}.
3919 This section describes how to control returning structure values in
3922 @deftypefn {Target Hook} bool RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
3923 This target hook should return a nonzero value to say to return the
3924 function value in memory, just as large structures are always returned.
3925 Here @var{type} will be the data type of the value, and @var{fntype}
3926 will be the type of the function doing the returning, or @code{NULL} for
3929 Note that values of mode @code{BLKmode} must be explicitly handled
3930 by this function. Also, the option @option{-fpcc-struct-return}
3931 takes effect regardless of this macro. On most systems, it is
3932 possible to leave the hook undefined; this causes a default
3933 definition to be used, whose value is the constant 1 for @code{BLKmode}
3934 values, and 0 otherwise.
3936 Do not use this hook to indicate that structures and unions should always
3937 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3941 @defmac DEFAULT_PCC_STRUCT_RETURN
3942 Define this macro to be 1 if all structure and union return values must be
3943 in memory. Since this results in slower code, this should be defined
3944 only if needed for compatibility with other compilers or with an ABI@.
3945 If you define this macro to be 0, then the conventions used for structure
3946 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3948 If not defined, this defaults to the value 1.
3951 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
3952 This target hook should return the location of the structure value
3953 address (normally a @code{mem} or @code{reg}), or 0 if the address is
3954 passed as an ``invisible'' first argument. Note that @var{fndecl} may
3955 be @code{NULL}, for libcalls.
3957 On some architectures the place where the structure value address
3958 is found by the called function is not the same place that the
3959 caller put it. This can be due to register windows, or it could
3960 be because the function prologue moves it to a different place.
3961 @var{incoming} is @code{true} when the location is needed in
3962 the context of the called function, and @code{false} in the context of
3965 If @var{incoming} is @code{true} and the address is to be found on the
3966 stack, return a @code{mem} which refers to the frame pointer.
3969 @defmac PCC_STATIC_STRUCT_RETURN
3970 Define this macro if the usual system convention on the target machine
3971 for returning structures and unions is for the called function to return
3972 the address of a static variable containing the value.
3974 Do not define this if the usual system convention is for the caller to
3975 pass an address to the subroutine.
3977 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3978 nothing when you use @option{-freg-struct-return} mode.
3982 @subsection Caller-Saves Register Allocation
3984 If you enable it, GCC can save registers around function calls. This
3985 makes it possible to use call-clobbered registers to hold variables that
3986 must live across calls.
3988 @defmac DEFAULT_CALLER_SAVES
3989 Define this macro if function calls on the target machine do not preserve
3990 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3991 for all registers. When defined, this macro enables @option{-fcaller-saves}
3992 by default for all optimization levels. It has no effect for optimization
3993 levels 2 and higher, where @option{-fcaller-saves} is the default.
3996 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3997 A C expression to determine whether it is worthwhile to consider placing
3998 a pseudo-register in a call-clobbered hard register and saving and
3999 restoring it around each function call. The expression should be 1 when
4000 this is worth doing, and 0 otherwise.
4002 If you don't define this macro, a default is used which is good on most
4003 machines: @code{4 * @var{calls} < @var{refs}}.
4006 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4007 A C expression specifying which mode is required for saving @var{nregs}
4008 of a pseudo-register in call-clobbered hard register @var{regno}. If
4009 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4010 returned. For most machines this macro need not be defined since GCC
4011 will select the smallest suitable mode.
4014 @node Function Entry
4015 @subsection Function Entry and Exit
4016 @cindex function entry and exit
4020 This section describes the macros that output function entry
4021 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4023 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4024 If defined, a function that outputs the assembler code for entry to a
4025 function. The prologue is responsible for setting up the stack frame,
4026 initializing the frame pointer register, saving registers that must be
4027 saved, and allocating @var{size} additional bytes of storage for the
4028 local variables. @var{size} is an integer. @var{file} is a stdio
4029 stream to which the assembler code should be output.
4031 The label for the beginning of the function need not be output by this
4032 macro. That has already been done when the macro is run.
4034 @findex regs_ever_live
4035 To determine which registers to save, the macro can refer to the array
4036 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4037 @var{r} is used anywhere within the function. This implies the function
4038 prologue should save register @var{r}, provided it is not one of the
4039 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4040 @code{regs_ever_live}.)
4042 On machines that have ``register windows'', the function entry code does
4043 not save on the stack the registers that are in the windows, even if
4044 they are supposed to be preserved by function calls; instead it takes
4045 appropriate steps to ``push'' the register stack, if any non-call-used
4046 registers are used in the function.
4048 @findex frame_pointer_needed
4049 On machines where functions may or may not have frame-pointers, the
4050 function entry code must vary accordingly; it must set up the frame
4051 pointer if one is wanted, and not otherwise. To determine whether a
4052 frame pointer is in wanted, the macro can refer to the variable
4053 @code{frame_pointer_needed}. The variable's value will be 1 at run
4054 time in a function that needs a frame pointer. @xref{Elimination}.
4056 The function entry code is responsible for allocating any stack space
4057 required for the function. This stack space consists of the regions
4058 listed below. In most cases, these regions are allocated in the
4059 order listed, with the last listed region closest to the top of the
4060 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4061 the highest address if it is not defined). You can use a different order
4062 for a machine if doing so is more convenient or required for
4063 compatibility reasons. Except in cases where required by standard
4064 or by a debugger, there is no reason why the stack layout used by GCC
4065 need agree with that used by other compilers for a machine.
4068 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4069 If defined, a function that outputs assembler code at the end of a
4070 prologue. This should be used when the function prologue is being
4071 emitted as RTL, and you have some extra assembler that needs to be
4072 emitted. @xref{prologue instruction pattern}.
4075 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4076 If defined, a function that outputs assembler code at the start of an
4077 epilogue. This should be used when the function epilogue is being
4078 emitted as RTL, and you have some extra assembler that needs to be
4079 emitted. @xref{epilogue instruction pattern}.
4082 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4083 If defined, a function that outputs the assembler code for exit from a
4084 function. The epilogue is responsible for restoring the saved
4085 registers and stack pointer to their values when the function was
4086 called, and returning control to the caller. This macro takes the
4087 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4088 registers to restore are determined from @code{regs_ever_live} and
4089 @code{CALL_USED_REGISTERS} in the same way.
4091 On some machines, there is a single instruction that does all the work
4092 of returning from the function. On these machines, give that
4093 instruction the name @samp{return} and do not define the macro
4094 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4096 Do not define a pattern named @samp{return} if you want the
4097 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4098 switches to control whether return instructions or epilogues are used,
4099 define a @samp{return} pattern with a validity condition that tests the
4100 target switches appropriately. If the @samp{return} pattern's validity
4101 condition is false, epilogues will be used.
4103 On machines where functions may or may not have frame-pointers, the
4104 function exit code must vary accordingly. Sometimes the code for these
4105 two cases is completely different. To determine whether a frame pointer
4106 is wanted, the macro can refer to the variable
4107 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4108 a function that needs a frame pointer.
4110 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4111 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4112 The C variable @code{current_function_is_leaf} is nonzero for such a
4113 function. @xref{Leaf Functions}.
4115 On some machines, some functions pop their arguments on exit while
4116 others leave that for the caller to do. For example, the 68020 when
4117 given @option{-mrtd} pops arguments in functions that take a fixed
4118 number of arguments.
4120 @findex current_function_pops_args
4121 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4122 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4123 needs to know what was decided. The variable that is called
4124 @code{current_function_pops_args} is the number of bytes of its
4125 arguments that a function should pop. @xref{Scalar Return}.
4126 @c what is the "its arguments" in the above sentence referring to, pray
4127 @c tell? --mew 5feb93
4132 @findex current_function_pretend_args_size
4133 A region of @code{current_function_pretend_args_size} bytes of
4134 uninitialized space just underneath the first argument arriving on the
4135 stack. (This may not be at the very start of the allocated stack region
4136 if the calling sequence has pushed anything else since pushing the stack
4137 arguments. But usually, on such machines, nothing else has been pushed
4138 yet, because the function prologue itself does all the pushing.) This
4139 region is used on machines where an argument may be passed partly in
4140 registers and partly in memory, and, in some cases to support the
4141 features in @code{<stdarg.h>}.
4144 An area of memory used to save certain registers used by the function.
4145 The size of this area, which may also include space for such things as
4146 the return address and pointers to previous stack frames, is
4147 machine-specific and usually depends on which registers have been used
4148 in the function. Machines with register windows often do not require
4152 A region of at least @var{size} bytes, possibly rounded up to an allocation
4153 boundary, to contain the local variables of the function. On some machines,
4154 this region and the save area may occur in the opposite order, with the
4155 save area closer to the top of the stack.
4158 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4159 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4160 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4161 argument lists of the function. @xref{Stack Arguments}.
4164 Normally, it is necessary for the macros
4165 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4166 @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially.
4167 The C variable @code{current_function_is_leaf} is nonzero for such a
4170 @defmac EXIT_IGNORE_STACK
4171 Define this macro as a C expression that is nonzero if the return
4172 instruction or the function epilogue ignores the value of the stack
4173 pointer; in other words, if it is safe to delete an instruction to
4174 adjust the stack pointer before a return from the function.
4176 Note that this macro's value is relevant only for functions for which
4177 frame pointers are maintained. It is never safe to delete a final
4178 stack adjustment in a function that has no frame pointer, and the
4179 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4182 @defmac EPILOGUE_USES (@var{regno})
4183 Define this macro as a C expression that is nonzero for registers that are
4184 used by the epilogue or the @samp{return} pattern. The stack and frame
4185 pointer registers are already be assumed to be used as needed.
4188 @defmac EH_USES (@var{regno})
4189 Define this macro as a C expression that is nonzero for registers that are
4190 used by the exception handling mechanism, and so should be considered live
4191 on entry to an exception edge.
4194 @defmac DELAY_SLOTS_FOR_EPILOGUE
4195 Define this macro if the function epilogue contains delay slots to which
4196 instructions from the rest of the function can be ``moved''. The
4197 definition should be a C expression whose value is an integer
4198 representing the number of delay slots there.
4201 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4202 A C expression that returns 1 if @var{insn} can be placed in delay
4203 slot number @var{n} of the epilogue.
4205 The argument @var{n} is an integer which identifies the delay slot now
4206 being considered (since different slots may have different rules of
4207 eligibility). It is never negative and is always less than the number
4208 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4209 If you reject a particular insn for a given delay slot, in principle, it
4210 may be reconsidered for a subsequent delay slot. Also, other insns may
4211 (at least in principle) be considered for the so far unfilled delay
4214 @findex current_function_epilogue_delay_list
4215 @findex final_scan_insn
4216 The insns accepted to fill the epilogue delay slots are put in an RTL
4217 list made with @code{insn_list} objects, stored in the variable
4218 @code{current_function_epilogue_delay_list}. The insn for the first
4219 delay slot comes first in the list. Your definition of the macro
4220 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4221 outputting the insns in this list, usually by calling
4222 @code{final_scan_insn}.
4224 You need not define this macro if you did not define
4225 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4228 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, tree @var{function})
4229 A function that outputs the assembler code for a thunk
4230 function, used to implement C++ virtual function calls with multiple
4231 inheritance. The thunk acts as a wrapper around a virtual function,
4232 adjusting the implicit object parameter before handing control off to
4235 First, emit code to add the integer @var{delta} to the location that
4236 contains the incoming first argument. Assume that this argument
4237 contains a pointer, and is the one used to pass the @code{this} pointer
4238 in C++. This is the incoming argument @emph{before} the function prologue,
4239 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4240 all other incoming arguments.
4242 After the addition, emit code to jump to @var{function}, which is a
4243 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4244 not touch the return address. Hence returning from @var{FUNCTION} will
4245 return to whoever called the current @samp{thunk}.
4247 The effect must be as if @var{function} had been called directly with
4248 the adjusted first argument. This macro is responsible for emitting all
4249 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4250 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4252 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4253 have already been extracted from it.) It might possibly be useful on
4254 some targets, but probably not.
4256 If you do not define this macro, the target-independent code in the C++
4257 front end will generate a less efficient heavyweight thunk that calls
4258 @var{function} instead of jumping to it. The generic approach does
4259 not support varargs.
4262 @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})
4263 A function like @code{TARGET_ASM_OUTPUT_MI_THUNK}, except that if
4264 @var{vcall_offset} is nonzero, an additional adjustment should be made
4265 after adding @code{delta}. In particular, if @var{p} is the
4266 adjusted pointer, the following adjustment should be made:
4269 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4273 If this function is defined, it will always be used in place of
4274 @code{TARGET_ASM_OUTPUT_MI_THUNK}.
4278 @subsection Generating Code for Profiling
4279 @cindex profiling, code generation
4281 These macros will help you generate code for profiling.
4283 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4284 A C statement or compound statement to output to @var{file} some
4285 assembler code to call the profiling subroutine @code{mcount}.
4288 The details of how @code{mcount} expects to be called are determined by
4289 your operating system environment, not by GCC@. To figure them out,
4290 compile a small program for profiling using the system's installed C
4291 compiler and look at the assembler code that results.
4293 Older implementations of @code{mcount} expect the address of a counter
4294 variable to be loaded into some register. The name of this variable is
4295 @samp{LP} followed by the number @var{labelno}, so you would generate
4296 the name using @samp{LP%d} in a @code{fprintf}.
4299 @defmac PROFILE_HOOK
4300 A C statement or compound statement to output to @var{file} some assembly
4301 code to call the profiling subroutine @code{mcount} even the target does
4302 not support profiling.
4305 @defmac NO_PROFILE_COUNTERS
4306 Define this macro if the @code{mcount} subroutine on your system does
4307 not need a counter variable allocated for each function. This is true
4308 for almost all modern implementations. If you define this macro, you
4309 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4312 @defmac PROFILE_BEFORE_PROLOGUE
4313 Define this macro if the code for function profiling should come before
4314 the function prologue. Normally, the profiling code comes after.
4318 @subsection Permitting tail calls
4321 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4322 True if it is ok to do sibling call optimization for the specified
4323 call expression @var{exp}. @var{decl} will be the called function,
4324 or @code{NULL} if this is an indirect call.
4326 It is not uncommon for limitations of calling conventions to prevent
4327 tail calls to functions outside the current unit of translation, or
4328 during PIC compilation. The hook is used to enforce these restrictions,
4329 as the @code{sibcall} md pattern can not fail, or fall over to a
4330 ``normal'' call. The criteria for successful sibling call optimization
4331 may vary greatly between different architectures.
4335 @section Implementing the Varargs Macros
4336 @cindex varargs implementation
4338 GCC comes with an implementation of @code{<varargs.h>} and
4339 @code{<stdarg.h>} that work without change on machines that pass arguments
4340 on the stack. Other machines require their own implementations of
4341 varargs, and the two machine independent header files must have
4342 conditionals to include it.
4344 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4345 the calling convention for @code{va_start}. The traditional
4346 implementation takes just one argument, which is the variable in which
4347 to store the argument pointer. The ISO implementation of
4348 @code{va_start} takes an additional second argument. The user is
4349 supposed to write the last named argument of the function here.
4351 However, @code{va_start} should not use this argument. The way to find
4352 the end of the named arguments is with the built-in functions described
4355 @defmac __builtin_saveregs ()
4356 Use this built-in function to save the argument registers in memory so
4357 that the varargs mechanism can access them. Both ISO and traditional
4358 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4359 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
4361 On some machines, @code{__builtin_saveregs} is open-coded under the
4362 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
4363 it calls a routine written in assembler language, found in
4366 Code generated for the call to @code{__builtin_saveregs} appears at the
4367 beginning of the function, as opposed to where the call to
4368 @code{__builtin_saveregs} is written, regardless of what the code is.
4369 This is because the registers must be saved before the function starts
4370 to use them for its own purposes.
4371 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4375 @defmac __builtin_args_info (@var{category})
4376 Use this built-in function to find the first anonymous arguments in
4379 In general, a machine may have several categories of registers used for
4380 arguments, each for a particular category of data types. (For example,
4381 on some machines, floating-point registers are used for floating-point
4382 arguments while other arguments are passed in the general registers.)
4383 To make non-varargs functions use the proper calling convention, you
4384 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4385 registers in each category have been used so far
4387 @code{__builtin_args_info} accesses the same data structure of type
4388 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4389 with it, with @var{category} specifying which word to access. Thus, the
4390 value indicates the first unused register in a given category.
4392 Normally, you would use @code{__builtin_args_info} in the implementation
4393 of @code{va_start}, accessing each category just once and storing the
4394 value in the @code{va_list} object. This is because @code{va_list} will
4395 have to update the values, and there is no way to alter the
4396 values accessed by @code{__builtin_args_info}.
4399 @defmac __builtin_next_arg (@var{lastarg})
4400 This is the equivalent of @code{__builtin_args_info}, for stack
4401 arguments. It returns the address of the first anonymous stack
4402 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4403 returns the address of the location above the first anonymous stack
4404 argument. Use it in @code{va_start} to initialize the pointer for
4405 fetching arguments from the stack. Also use it in @code{va_start} to
4406 verify that the second parameter @var{lastarg} is the last named argument
4407 of the current function.
4410 @defmac __builtin_classify_type (@var{object})
4411 Since each machine has its own conventions for which data types are
4412 passed in which kind of register, your implementation of @code{va_arg}
4413 has to embody these conventions. The easiest way to categorize the
4414 specified data type is to use @code{__builtin_classify_type} together
4415 with @code{sizeof} and @code{__alignof__}.
4417 @code{__builtin_classify_type} ignores the value of @var{object},
4418 considering only its data type. It returns an integer describing what
4419 kind of type that is---integer, floating, pointer, structure, and so on.
4421 The file @file{typeclass.h} defines an enumeration that you can use to
4422 interpret the values of @code{__builtin_classify_type}.
4425 These machine description macros help implement varargs:
4427 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4428 If defined, this hook produces the machine-specific code for a call to
4429 @code{__builtin_saveregs}. This code will be moved to the very
4430 beginning of the function, before any parameter access are made. The
4431 return value of this function should be an RTX that contains the value
4432 to use as the return of @code{__builtin_saveregs}.
4435 @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})
4436 This target hook offers an alternative to using
4437 @code{__builtin_saveregs} and defining the hook
4438 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4439 register arguments into the stack so that all the arguments appear to
4440 have been passed consecutively on the stack. Once this is done, you can
4441 use the standard implementation of varargs that works for machines that
4442 pass all their arguments on the stack.
4444 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4445 structure, containing the values that are obtained after processing the
4446 named arguments. The arguments @var{mode} and @var{type} describe the
4447 last named argument---its machine mode and its data type as a tree node.
4449 The target hook should do two things: first, push onto the stack all the
4450 argument registers @emph{not} used for the named arguments, and second,
4451 store the size of the data thus pushed into the @code{int}-valued
4452 variable pointed to by @var{pretend_args_size}. The value that you
4453 store here will serve as additional offset for setting up the stack
4456 Because you must generate code to push the anonymous arguments at
4457 compile time without knowing their data types,
4458 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4459 have just a single category of argument register and use it uniformly
4462 If the argument @var{second_time} is nonzero, it means that the
4463 arguments of the function are being analyzed for the second time. This
4464 happens for an inline function, which is not actually compiled until the
4465 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4466 not generate any instructions in this case.
4469 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4470 Define this hook to return @code{true} if the location where a function
4471 argument is passed depends on whether or not it is a named argument.
4473 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4474 is set for varargs and stdarg functions. If this hook returns
4475 @code{true}, the @var{named} argument is always true for named
4476 arguments, and false for unnamed arguments. If it returns @code{false},
4477 but @code{TARGET_PRETEND_OUTOGOING_VARARGS_NAMED} returns @code{true},
4478 then all arguments are treated as named. Otherwise, all named arguments
4479 except the last are treated as named.
4481 You need not define this hook if it always returns zero.
4484 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4485 If you need to conditionally change ABIs so that one works with
4486 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4487 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4488 defined, then define this hook to return @code{true} if
4489 @code{SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4490 Otherwise, you should not define this hook.
4494 @section Trampolines for Nested Functions
4495 @cindex trampolines for nested functions
4496 @cindex nested functions, trampolines for
4498 A @dfn{trampoline} is a small piece of code that is created at run time
4499 when the address of a nested function is taken. It normally resides on
4500 the stack, in the stack frame of the containing function. These macros
4501 tell GCC how to generate code to allocate and initialize a
4504 The instructions in the trampoline must do two things: load a constant
4505 address into the static chain register, and jump to the real address of
4506 the nested function. On CISC machines such as the m68k, this requires
4507 two instructions, a move immediate and a jump. Then the two addresses
4508 exist in the trampoline as word-long immediate operands. On RISC
4509 machines, it is often necessary to load each address into a register in
4510 two parts. Then pieces of each address form separate immediate
4513 The code generated to initialize the trampoline must store the variable
4514 parts---the static chain value and the function address---into the
4515 immediate operands of the instructions. On a CISC machine, this is
4516 simply a matter of copying each address to a memory reference at the
4517 proper offset from the start of the trampoline. On a RISC machine, it
4518 may be necessary to take out pieces of the address and store them
4521 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4522 A C statement to output, on the stream @var{file}, assembler code for a
4523 block of data that contains the constant parts of a trampoline. This
4524 code should not include a label---the label is taken care of
4527 If you do not define this macro, it means no template is needed
4528 for the target. Do not define this macro on systems where the block move
4529 code to copy the trampoline into place would be larger than the code
4530 to generate it on the spot.
4533 @defmac TRAMPOLINE_SECTION
4534 The name of a subroutine to switch to the section in which the
4535 trampoline template is to be placed (@pxref{Sections}). The default is
4536 a value of @samp{readonly_data_section}, which places the trampoline in
4537 the section containing read-only data.
4540 @defmac TRAMPOLINE_SIZE
4541 A C expression for the size in bytes of the trampoline, as an integer.
4544 @defmac TRAMPOLINE_ALIGNMENT
4545 Alignment required for trampolines, in bits.
4547 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4548 is used for aligning trampolines.
4551 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4552 A C statement to initialize the variable parts of a trampoline.
4553 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4554 an RTX for the address of the nested function; @var{static_chain} is an
4555 RTX for the static chain value that should be passed to the function
4559 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4560 A C statement that should perform any machine-specific adjustment in
4561 the address of the trampoline. Its argument contains the address that
4562 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4563 used for a function call should be different from the address in which
4564 the template was stored, the different address should be assigned to
4565 @var{addr}. If this macro is not defined, @var{addr} will be used for
4568 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4569 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4570 If this macro is not defined, by default the trampoline is allocated as
4571 a stack slot. This default is right for most machines. The exceptions
4572 are machines where it is impossible to execute instructions in the stack
4573 area. On such machines, you may have to implement a separate stack,
4574 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4575 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4577 @var{fp} points to a data structure, a @code{struct function}, which
4578 describes the compilation status of the immediate containing function of
4579 the function which the trampoline is for. The stack slot for the
4580 trampoline is in the stack frame of this containing function. Other
4581 allocation strategies probably must do something analogous with this
4585 Implementing trampolines is difficult on many machines because they have
4586 separate instruction and data caches. Writing into a stack location
4587 fails to clear the memory in the instruction cache, so when the program
4588 jumps to that location, it executes the old contents.
4590 Here are two possible solutions. One is to clear the relevant parts of
4591 the instruction cache whenever a trampoline is set up. The other is to
4592 make all trampolines identical, by having them jump to a standard
4593 subroutine. The former technique makes trampoline execution faster; the
4594 latter makes initialization faster.
4596 To clear the instruction cache when a trampoline is initialized, define
4597 the following macro.
4599 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4600 If defined, expands to a C expression clearing the @emph{instruction
4601 cache} in the specified interval. The definition of this macro would
4602 typically be a series of @code{asm} statements. Both @var{beg} and
4603 @var{end} are both pointer expressions.
4606 To use a standard subroutine, define the following macro. In addition,
4607 you must make sure that the instructions in a trampoline fill an entire
4608 cache line with identical instructions, or else ensure that the
4609 beginning of the trampoline code is always aligned at the same point in
4610 its cache line. Look in @file{m68k.h} as a guide.
4612 @defmac TRANSFER_FROM_TRAMPOLINE
4613 Define this macro if trampolines need a special subroutine to do their
4614 work. The macro should expand to a series of @code{asm} statements
4615 which will be compiled with GCC@. They go in a library function named
4616 @code{__transfer_from_trampoline}.
4618 If you need to avoid executing the ordinary prologue code of a compiled
4619 C function when you jump to the subroutine, you can do so by placing a
4620 special label of your own in the assembler code. Use one @code{asm}
4621 statement to generate an assembler label, and another to make the label
4622 global. Then trampolines can use that label to jump directly to your
4623 special assembler code.
4627 @section Implicit Calls to Library Routines
4628 @cindex library subroutine names
4629 @cindex @file{libgcc.a}
4631 @c prevent bad page break with this line
4632 Here is an explanation of implicit calls to library routines.
4634 @defmac DECLARE_LIBRARY_RENAMES
4635 This macro, if defined, should expand to a piece of C code that will get
4636 expanded when compiling functions for libgcc.a. It can be used to
4637 provide alternate names for gcc's internal library functions if there
4638 are ABI-mandated names that the compiler should provide.
4641 @findex init_one_libfunc
4642 @findex set_optab_libfunc
4643 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
4644 This hook should declare additional library routines or rename
4645 existing ones, using the functions @code{set_optab_libfunc} and
4646 @code{init_one_libfunc} defined in @file{optabs.c}.
4647 @code{init_optabs} calls this macro after initializing all the normal
4650 The default is to do nothing. Most ports don't need to define this hook.
4653 @defmac TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4654 This macro should return @code{true} if the library routine that
4655 implements the floating point comparison operator @var{comparison} in
4656 mode @var{mode} will return a boolean, and @var{false} if it will
4659 GCC's own floating point libraries return tristates from the
4660 comparison operators, so the default returns false always. Most ports
4661 don't need to define this macro.
4664 @cindex US Software GOFAST, floating point emulation library
4665 @cindex floating point emulation library, US Software GOFAST
4666 @cindex GOFAST, floating point emulation library
4667 @findex gofast_maybe_init_libfuncs
4668 @defmac US_SOFTWARE_GOFAST
4669 Define this macro if your system C library uses the US Software GOFAST
4670 library to provide floating point emulation.
4672 In addition to defining this macro, your architecture must set
4673 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
4674 else call that function from its version of that hook. It is defined
4675 in @file{config/gofast.h}, which must be included by your
4676 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
4679 If this macro is defined, the
4680 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
4681 false for @code{SFmode} and @code{DFmode} comparisons.
4684 @cindex @code{EDOM}, implicit usage
4687 The value of @code{EDOM} on the target machine, as a C integer constant
4688 expression. If you don't define this macro, GCC does not attempt to
4689 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4690 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4693 If you do not define @code{TARGET_EDOM}, then compiled code reports
4694 domain errors by calling the library function and letting it report the
4695 error. If mathematical functions on your system use @code{matherr} when
4696 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4697 that @code{matherr} is used normally.
4700 @cindex @code{errno}, implicit usage
4701 @defmac GEN_ERRNO_RTX
4702 Define this macro as a C expression to create an rtl expression that
4703 refers to the global ``variable'' @code{errno}. (On certain systems,
4704 @code{errno} may not actually be a variable.) If you don't define this
4705 macro, a reasonable default is used.
4708 @cindex @code{bcopy}, implicit usage
4709 @cindex @code{memcpy}, implicit usage
4710 @cindex @code{memmove}, implicit usage
4711 @cindex @code{bzero}, implicit usage
4712 @cindex @code{memset}, implicit usage
4713 @defmac TARGET_MEM_FUNCTIONS
4714 Define this macro if GCC should generate calls to the ISO C
4715 (and System V) library functions @code{memcpy}, @code{memmove} and
4716 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4719 @cindex C99 math functions, implicit usage
4720 @defmac TARGET_C99_FUNCTIONS
4721 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4722 @code{sinf} and similarly for other functions defined by C99 standard. The
4723 default is nonzero that should be proper value for most modern systems, however
4724 number of existing systems lacks support for these functions in the runtime so
4725 they needs this macro to be redefined to 0.
4728 @defmac NEXT_OBJC_RUNTIME
4729 Define this macro to generate code for Objective-C message sending using
4730 the calling convention of the NeXT system. This calling convention
4731 involves passing the object, the selector and the method arguments all
4732 at once to the method-lookup library function.
4734 The default calling convention passes just the object and the selector
4735 to the lookup function, which returns a pointer to the method.
4738 @node Addressing Modes
4739 @section Addressing Modes
4740 @cindex addressing modes
4742 @c prevent bad page break with this line
4743 This is about addressing modes.
4745 @defmac HAVE_PRE_INCREMENT
4746 @defmacx HAVE_PRE_DECREMENT
4747 @defmacx HAVE_POST_INCREMENT
4748 @defmacx HAVE_POST_DECREMENT
4749 A C expression that is nonzero if the machine supports pre-increment,
4750 pre-decrement, post-increment, or post-decrement addressing respectively.
4753 @defmac HAVE_PRE_MODIFY_DISP
4754 @defmacx HAVE_POST_MODIFY_DISP
4755 A C expression that is nonzero if the machine supports pre- or
4756 post-address side-effect generation involving constants other than
4757 the size of the memory operand.
4760 @defmac HAVE_PRE_MODIFY_REG
4761 @defmacx HAVE_POST_MODIFY_REG
4762 A C expression that is nonzero if the machine supports pre- or
4763 post-address side-effect generation involving a register displacement.
4766 @defmac CONSTANT_ADDRESS_P (@var{x})
4767 A C expression that is 1 if the RTX @var{x} is a constant which
4768 is a valid address. On most machines, this can be defined as
4769 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4770 in which constant addresses are supported.
4773 @defmac CONSTANT_P (@var{x})
4774 @code{CONSTANT_P}, which is defined by target-independent code,
4775 accepts integer-values expressions whose values are not explicitly
4776 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4777 expressions and @code{const} arithmetic expressions, in addition to
4778 @code{const_int} and @code{const_double} expressions.
4781 @defmac MAX_REGS_PER_ADDRESS
4782 A number, the maximum number of registers that can appear in a valid
4783 memory address. Note that it is up to you to specify a value equal to
4784 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4788 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4789 A C compound statement with a conditional @code{goto @var{label};}
4790 executed if @var{x} (an RTX) is a legitimate memory address on the
4791 target machine for a memory operand of mode @var{mode}.
4793 It usually pays to define several simpler macros to serve as
4794 subroutines for this one. Otherwise it may be too complicated to
4797 This macro must exist in two variants: a strict variant and a
4798 non-strict one. The strict variant is used in the reload pass. It
4799 must be defined so that any pseudo-register that has not been
4800 allocated a hard register is considered a memory reference. In
4801 contexts where some kind of register is required, a pseudo-register
4802 with no hard register must be rejected.
4804 The non-strict variant is used in other passes. It must be defined to
4805 accept all pseudo-registers in every context where some kind of
4806 register is required.
4808 @findex REG_OK_STRICT
4809 Compiler source files that want to use the strict variant of this
4810 macro define the macro @code{REG_OK_STRICT}. You should use an
4811 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4812 in that case and the non-strict variant otherwise.
4814 Subroutines to check for acceptable registers for various purposes (one
4815 for base registers, one for index registers, and so on) are typically
4816 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4817 Then only these subroutine macros need have two variants; the higher
4818 levels of macros may be the same whether strict or not.
4820 Normally, constant addresses which are the sum of a @code{symbol_ref}
4821 and an integer are stored inside a @code{const} RTX to mark them as
4822 constant. Therefore, there is no need to recognize such sums
4823 specifically as legitimate addresses. Normally you would simply
4824 recognize any @code{const} as legitimate.
4826 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4827 sums that are not marked with @code{const}. It assumes that a naked
4828 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4829 naked constant sums as illegitimate addresses, so that none of them will
4830 be given to @code{PRINT_OPERAND_ADDRESS}.
4832 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4833 On some machines, whether a symbolic address is legitimate depends on
4834 the section that the address refers to. On these machines, define the
4835 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4836 into the @code{symbol_ref}, and then check for it here. When you see a
4837 @code{const}, you will have to look inside it to find the
4838 @code{symbol_ref} in order to determine the section. @xref{Assembler
4842 @defmac REG_OK_FOR_BASE_P (@var{x})
4843 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4844 RTX) is valid for use as a base register. For hard registers, it
4845 should always accept those which the hardware permits and reject the
4846 others. Whether the macro accepts or rejects pseudo registers must be
4847 controlled by @code{REG_OK_STRICT} as described above. This usually
4848 requires two variant definitions, of which @code{REG_OK_STRICT}
4849 controls the one actually used.
4852 @defmac REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4853 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4854 that expression may examine the mode of the memory reference in
4855 @var{mode}. You should define this macro if the mode of the memory
4856 reference affects whether a register may be used as a base register. If
4857 you define this macro, the compiler will use it instead of
4858 @code{REG_OK_FOR_BASE_P}.
4861 @defmac REG_OK_FOR_INDEX_P (@var{x})
4862 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4863 RTX) is valid for use as an index register.
4865 The difference between an index register and a base register is that
4866 the index register may be scaled. If an address involves the sum of
4867 two registers, neither one of them scaled, then either one may be
4868 labeled the ``base'' and the other the ``index''; but whichever
4869 labeling is used must fit the machine's constraints of which registers
4870 may serve in each capacity. The compiler will try both labelings,
4871 looking for one that is valid, and will reload one or both registers
4872 only if neither labeling works.
4875 @defmac FIND_BASE_TERM (@var{x})
4876 A C expression to determine the base term of address @var{x}.
4877 This macro is used in only one place: `find_base_term' in alias.c.
4879 It is always safe for this macro to not be defined. It exists so
4880 that alias analysis can understand machine-dependent addresses.
4882 The typical use of this macro is to handle addresses containing
4883 a label_ref or symbol_ref within an UNSPEC@.
4886 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4887 A C compound statement that attempts to replace @var{x} with a valid
4888 memory address for an operand of mode @var{mode}. @var{win} will be a
4889 C statement label elsewhere in the code; the macro definition may use
4892 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4896 to avoid further processing if the address has become legitimate.
4898 @findex break_out_memory_refs
4899 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4900 and @var{oldx} will be the operand that was given to that function to produce
4903 The code generated by this macro should not alter the substructure of
4904 @var{x}. If it transforms @var{x} into a more legitimate form, it
4905 should assign @var{x} (which will always be a C variable) a new value.
4907 It is not necessary for this macro to come up with a legitimate
4908 address. The compiler has standard ways of doing so in all cases. In
4909 fact, it is safe for this macro to do nothing. But often a
4910 machine-dependent strategy can generate better code.
4913 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4914 A C compound statement that attempts to replace @var{x}, which is an address
4915 that needs reloading, with a valid memory address for an operand of mode
4916 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4917 It is not necessary to define this macro, but it might be useful for
4918 performance reasons.
4920 For example, on the i386, it is sometimes possible to use a single
4921 reload register instead of two by reloading a sum of two pseudo
4922 registers into a register. On the other hand, for number of RISC
4923 processors offsets are limited so that often an intermediate address
4924 needs to be generated in order to address a stack slot. By defining
4925 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4926 generated for adjacent some stack slots can be made identical, and thus
4929 @emph{Note}: This macro should be used with caution. It is necessary
4930 to know something of how reload works in order to effectively use this,
4931 and it is quite easy to produce macros that build in too much knowledge
4932 of reload internals.
4934 @emph{Note}: This macro must be able to reload an address created by a
4935 previous invocation of this macro. If it fails to handle such addresses
4936 then the compiler may generate incorrect code or abort.
4939 The macro definition should use @code{push_reload} to indicate parts that
4940 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4941 suitable to be passed unaltered to @code{push_reload}.
4943 The code generated by this macro must not alter the substructure of
4944 @var{x}. If it transforms @var{x} into a more legitimate form, it
4945 should assign @var{x} (which will always be a C variable) a new value.
4946 This also applies to parts that you change indirectly by calling
4949 @findex strict_memory_address_p
4950 The macro definition may use @code{strict_memory_address_p} to test if
4951 the address has become legitimate.
4954 If you want to change only a part of @var{x}, one standard way of doing
4955 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4956 single level of rtl. Thus, if the part to be changed is not at the
4957 top level, you'll need to replace first the top level.
4958 It is not necessary for this macro to come up with a legitimate
4959 address; but often a machine-dependent strategy can generate better code.
4962 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4963 A C statement or compound statement with a conditional @code{goto
4964 @var{label};} executed if memory address @var{x} (an RTX) can have
4965 different meanings depending on the machine mode of the memory
4966 reference it is used for or if the address is valid for some modes
4969 Autoincrement and autodecrement addresses typically have mode-dependent
4970 effects because the amount of the increment or decrement is the size
4971 of the operand being addressed. Some machines have other mode-dependent
4972 addresses. Many RISC machines have no mode-dependent addresses.
4974 You may assume that @var{addr} is a valid address for the machine.
4977 @defmac LEGITIMATE_CONSTANT_P (@var{x})
4978 A C expression that is nonzero if @var{x} is a legitimate constant for
4979 an immediate operand on the target machine. You can assume that
4980 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4981 @samp{1} is a suitable definition for this macro on machines where
4982 anything @code{CONSTANT_P} is valid.
4985 @node Condition Code
4986 @section Condition Code Status
4987 @cindex condition code status
4989 @c prevent bad page break with this line
4990 This describes the condition code status.
4993 The file @file{conditions.h} defines a variable @code{cc_status} to
4994 describe how the condition code was computed (in case the interpretation of
4995 the condition code depends on the instruction that it was set by). This
4996 variable contains the RTL expressions on which the condition code is
4997 currently based, and several standard flags.
4999 Sometimes additional machine-specific flags must be defined in the machine
5000 description header file. It can also add additional machine-specific
5001 information by defining @code{CC_STATUS_MDEP}.
5003 @defmac CC_STATUS_MDEP
5004 C code for a data type which is used for declaring the @code{mdep}
5005 component of @code{cc_status}. It defaults to @code{int}.
5007 This macro is not used on machines that do not use @code{cc0}.
5010 @defmac CC_STATUS_MDEP_INIT
5011 A C expression to initialize the @code{mdep} field to ``empty''.
5012 The default definition does nothing, since most machines don't use
5013 the field anyway. If you want to use the field, you should probably
5014 define this macro to initialize it.
5016 This macro is not used on machines that do not use @code{cc0}.
5019 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5020 A C compound statement to set the components of @code{cc_status}
5021 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5022 this macro's responsibility to recognize insns that set the condition
5023 code as a byproduct of other activity as well as those that explicitly
5026 This macro is not used on machines that do not use @code{cc0}.
5028 If there are insns that do not set the condition code but do alter
5029 other machine registers, this macro must check to see whether they
5030 invalidate the expressions that the condition code is recorded as
5031 reflecting. For example, on the 68000, insns that store in address
5032 registers do not set the condition code, which means that usually
5033 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5034 insns. But suppose that the previous insn set the condition code
5035 based on location @samp{a4@@(102)} and the current insn stores a new
5036 value in @samp{a4}. Although the condition code is not changed by
5037 this, it will no longer be true that it reflects the contents of
5038 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5039 @code{cc_status} in this case to say that nothing is known about the
5040 condition code value.
5042 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5043 with the results of peephole optimization: insns whose patterns are
5044 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5045 constants which are just the operands. The RTL structure of these
5046 insns is not sufficient to indicate what the insns actually do. What
5047 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5048 @code{CC_STATUS_INIT}.
5050 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5051 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5052 @samp{cc}. This avoids having detailed information about patterns in
5053 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5056 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5057 Returns a mode from class @code{MODE_CC} to be used when comparison
5058 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5059 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5060 @pxref{Jump Patterns} for a description of the reason for this
5064 #define SELECT_CC_MODE(OP,X,Y) \
5065 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5066 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5067 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5068 || GET_CODE (X) == NEG) \
5069 ? CC_NOOVmode : CCmode))
5072 You should define this macro if and only if you define extra CC modes
5073 in @file{@var{machine}-modes.def}.
5076 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5077 On some machines not all possible comparisons are defined, but you can
5078 convert an invalid comparison into a valid one. For example, the Alpha
5079 does not have a @code{GT} comparison, but you can use an @code{LT}
5080 comparison instead and swap the order of the operands.
5082 On such machines, define this macro to be a C statement to do any
5083 required conversions. @var{code} is the initial comparison code
5084 and @var{op0} and @var{op1} are the left and right operands of the
5085 comparison, respectively. You should modify @var{code}, @var{op0}, and
5086 @var{op1} as required.
5088 GCC will not assume that the comparison resulting from this macro is
5089 valid but will see if the resulting insn matches a pattern in the
5092 You need not define this macro if it would never change the comparison
5096 @defmac REVERSIBLE_CC_MODE (@var{mode})
5097 A C expression whose value is one if it is always safe to reverse a
5098 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5099 can ever return @var{mode} for a floating-point inequality comparison,
5100 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5102 You need not define this macro if it would always returns zero or if the
5103 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5104 For example, here is the definition used on the SPARC, where floating-point
5105 inequality comparisons are always given @code{CCFPEmode}:
5108 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5112 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5113 A C expression whose value is reversed condition code of the @var{code} for
5114 comparison done in CC_MODE @var{mode}. The macro is used only in case
5115 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5116 machine has some non-standard way how to reverse certain conditionals. For
5117 instance in case all floating point conditions are non-trapping, compiler may
5118 freely convert unordered compares to ordered one. Then definition may look
5122 #define REVERSE_CONDITION(CODE, MODE) \
5123 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5124 : reverse_condition_maybe_unordered (CODE))
5128 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
5129 A C expression that returns true if the conditional execution predicate
5130 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
5131 return 0 if the target has conditional execution predicates that cannot be
5132 reversed safely. If no expansion is specified, this macro is defined as
5136 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5137 ((x) == reverse_condition (y))
5142 @section Describing Relative Costs of Operations
5143 @cindex costs of instructions
5144 @cindex relative costs
5145 @cindex speed of instructions
5147 These macros let you describe the relative speed of various operations
5148 on the target machine.
5150 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5151 A C expression for the cost of moving data of mode @var{mode} from a
5152 register in class @var{from} to one in class @var{to}. The classes are
5153 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5154 value of 2 is the default; other values are interpreted relative to
5157 It is not required that the cost always equal 2 when @var{from} is the
5158 same as @var{to}; on some machines it is expensive to move between
5159 registers if they are not general registers.
5161 If reload sees an insn consisting of a single @code{set} between two
5162 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5163 classes returns a value of 2, reload does not check to ensure that the
5164 constraints of the insn are met. Setting a cost of other than 2 will
5165 allow reload to verify that the constraints are met. You should do this
5166 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5169 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5170 A C expression for the cost of moving data of mode @var{mode} between a
5171 register of class @var{class} and memory; @var{in} is zero if the value
5172 is to be written to memory, nonzero if it is to be read in. This cost
5173 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5174 registers and memory is more expensive than between two registers, you
5175 should define this macro to express the relative cost.
5177 If you do not define this macro, GCC uses a default cost of 4 plus
5178 the cost of copying via a secondary reload register, if one is
5179 needed. If your machine requires a secondary reload register to copy
5180 between memory and a register of @var{class} but the reload mechanism is
5181 more complex than copying via an intermediate, define this macro to
5182 reflect the actual cost of the move.
5184 GCC defines the function @code{memory_move_secondary_cost} if
5185 secondary reloads are needed. It computes the costs due to copying via
5186 a secondary register. If your machine copies from memory using a
5187 secondary register in the conventional way but the default base value of
5188 4 is not correct for your machine, define this macro to add some other
5189 value to the result of that function. The arguments to that function
5190 are the same as to this macro.
5194 A C expression for the cost of a branch instruction. A value of 1 is
5195 the default; other values are interpreted relative to that.
5198 Here are additional macros which do not specify precise relative costs,
5199 but only that certain actions are more expensive than GCC would
5202 @defmac SLOW_BYTE_ACCESS
5203 Define this macro as a C expression which is nonzero if accessing less
5204 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5205 faster than accessing a word of memory, i.e., if such access
5206 require more than one instruction or if there is no difference in cost
5207 between byte and (aligned) word loads.
5209 When this macro is not defined, the compiler will access a field by
5210 finding the smallest containing object; when it is defined, a fullword
5211 load will be used if alignment permits. Unless bytes accesses are
5212 faster than word accesses, using word accesses is preferable since it
5213 may eliminate subsequent memory access if subsequent accesses occur to
5214 other fields in the same word of the structure, but to different bytes.
5217 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5218 Define this macro to be the value 1 if memory accesses described by the
5219 @var{mode} and @var{alignment} parameters have a cost many times greater
5220 than aligned accesses, for example if they are emulated in a trap
5223 When this macro is nonzero, the compiler will act as if
5224 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5225 moves. This can cause significantly more instructions to be produced.
5226 Therefore, do not set this macro nonzero if unaligned accesses only add a
5227 cycle or two to the time for a memory access.
5229 If the value of this macro is always zero, it need not be defined. If
5230 this macro is defined, it should produce a nonzero value when
5231 @code{STRICT_ALIGNMENT} is nonzero.
5235 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5236 which a sequence of insns should be generated instead of a
5237 string move insn or a library call. Increasing the value will always
5238 make code faster, but eventually incurs high cost in increased code size.
5240 Note that on machines where the corresponding move insn is a
5241 @code{define_expand} that emits a sequence of insns, this macro counts
5242 the number of such sequences.
5244 If you don't define this, a reasonable default is used.
5247 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5248 A C expression used to determine whether @code{move_by_pieces} will be used to
5249 copy a chunk of memory, or whether some other block move mechanism
5250 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5251 than @code{MOVE_RATIO}.
5254 @defmac MOVE_MAX_PIECES
5255 A C expression used by @code{move_by_pieces} to determine the largest unit
5256 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5260 The threshold of number of scalar move insns, @emph{below} which a sequence
5261 of insns should be generated to clear memory instead of a string clear insn
5262 or a library call. Increasing the value will always make code faster, but
5263 eventually incurs high cost in increased code size.
5265 If you don't define this, a reasonable default is used.
5268 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5269 A C expression used to determine whether @code{clear_by_pieces} will be used
5270 to clear a chunk of memory, or whether some other block clear mechanism
5271 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5272 than @code{CLEAR_RATIO}.
5275 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5276 A C expression used to determine whether @code{store_by_pieces} will be
5277 used to set a chunk of memory to a constant value, or whether some other
5278 mechanism will be used. Used by @code{__builtin_memset} when storing
5279 values other than constant zero and by @code{__builtin_strcpy} when
5280 when called with a constant source string.
5281 Defaults to @code{MOVE_BY_PIECES_P}.
5284 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5285 A C expression used to determine whether a load postincrement is a good
5286 thing to use for a given mode. Defaults to the value of
5287 @code{HAVE_POST_INCREMENT}.
5290 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5291 A C expression used to determine whether a load postdecrement is a good
5292 thing to use for a given mode. Defaults to the value of
5293 @code{HAVE_POST_DECREMENT}.
5296 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5297 A C expression used to determine whether a load preincrement is a good
5298 thing to use for a given mode. Defaults to the value of
5299 @code{HAVE_PRE_INCREMENT}.
5302 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5303 A C expression used to determine whether a load predecrement is a good
5304 thing to use for a given mode. Defaults to the value of
5305 @code{HAVE_PRE_DECREMENT}.
5308 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5309 A C expression used to determine whether a store postincrement is a good
5310 thing to use for a given mode. Defaults to the value of
5311 @code{HAVE_POST_INCREMENT}.
5314 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5315 A C expression used to determine whether a store postdecrement is a good
5316 thing to use for a given mode. Defaults to the value of
5317 @code{HAVE_POST_DECREMENT}.
5320 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5321 This macro is used to determine whether a store preincrement is a good
5322 thing to use for a given mode. Defaults to the value of
5323 @code{HAVE_PRE_INCREMENT}.
5326 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5327 This macro is used to determine whether a store predecrement is a good
5328 thing to use for a given mode. Defaults to the value of
5329 @code{HAVE_PRE_DECREMENT}.
5332 @defmac NO_FUNCTION_CSE
5333 Define this macro if it is as good or better to call a constant
5334 function address than to call an address kept in a register.
5337 @defmac NO_RECURSIVE_FUNCTION_CSE
5338 Define this macro if it is as good or better for a function to call
5339 itself with an explicit address than to call an address kept in a
5343 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5344 Define this macro if a non-short-circuit operation produced by
5345 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5346 @code{BRANCH_COST} is greater than or equal to the value 2.
5349 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5350 This target hook describes the relative costs of RTL expressions.
5352 The cost may depend on the precise form of the expression, which is
5353 available for examination in @var{x}, and the rtx code of the expression
5354 in which it is contained, found in @var{outer_code}. @var{code} is the
5355 expression code---redundant, since it can be obtained with
5356 @code{GET_CODE (@var{x})}.
5358 In implementing this hook, you can use the construct
5359 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5362 On entry to the hook, @code{*@var{total}} contains a default estimate
5363 for the cost of the expression. The hook should modify this value as
5366 The hook returns true when all subexpressions of @var{x} have been
5367 processed, and false when @code{rtx_cost} should recurse.
5370 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5371 This hook computes the cost of an addressing mode that contains
5372 @var{address}. If not defined, the cost is computed from
5373 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5375 For most CISC machines, the default cost is a good approximation of the
5376 true cost of the addressing mode. However, on RISC machines, all
5377 instructions normally have the same length and execution time. Hence
5378 all addresses will have equal costs.
5380 In cases where more than one form of an address is known, the form with
5381 the lowest cost will be used. If multiple forms have the same, lowest,
5382 cost, the one that is the most complex will be used.
5384 For example, suppose an address that is equal to the sum of a register
5385 and a constant is used twice in the same basic block. When this macro
5386 is not defined, the address will be computed in a register and memory
5387 references will be indirect through that register. On machines where
5388 the cost of the addressing mode containing the sum is no higher than
5389 that of a simple indirect reference, this will produce an additional
5390 instruction and possibly require an additional register. Proper
5391 specification of this macro eliminates this overhead for such machines.
5393 This hook is never called with an invalid address.
5395 On machines where an address involving more than one register is as
5396 cheap as an address computation involving only one register, defining
5397 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5398 be live over a region of code where only one would have been if
5399 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5400 should be considered in the definition of this macro. Equivalent costs
5401 should probably only be given to addresses with different numbers of
5402 registers on machines with lots of registers.
5406 @section Adjusting the Instruction Scheduler
5408 The instruction scheduler may need a fair amount of machine-specific
5409 adjustment in order to produce good code. GCC provides several target
5410 hooks for this purpose. It is usually enough to define just a few of
5411 them: try the first ones in this list first.
5413 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5414 This hook returns the maximum number of instructions that can ever
5415 issue at the same time on the target machine. The default is one.
5416 Although the insn scheduler can define itself the possibility of issue
5417 an insn on the same cycle, the value can serve as an additional
5418 constraint to issue insns on the same simulated processor cycle (see
5419 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5420 This value must be constant over the entire compilation. If you need
5421 it to vary depending on what the instructions are, you must use
5422 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5424 For the automaton based pipeline interface, you could define this hook
5425 to return the value of the macro @code{MAX_DFA_ISSUE_RATE}.
5428 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5429 This hook is executed by the scheduler after it has scheduled an insn
5430 from the ready list. It should return the number of insns which can
5431 still be issued in the current cycle. The default is
5432 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5433 @code{USE}, which normally are not counted against the issue rate.
5434 You should define this hook if some insns take more machine resources
5435 than others, so that fewer insns can follow them in the same cycle.
5436 @var{file} is either a null pointer, or a stdio stream to write any
5437 debug output to. @var{verbose} is the verbose level provided by
5438 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5442 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5443 This function corrects the value of @var{cost} based on the
5444 relationship between @var{insn} and @var{dep_insn} through the
5445 dependence @var{link}. It should return the new value. The default
5446 is to make no adjustment to @var{cost}. This can be used for example
5447 to specify to the scheduler using the traditional pipeline description
5448 that an output- or anti-dependence does not incur the same cost as a
5449 data-dependence. If the scheduler using the automaton based pipeline
5450 description, the cost of anti-dependence is zero and the cost of
5451 output-dependence is maximum of one and the difference of latency
5452 times of the first and the second insns. If these values are not
5453 acceptable, you could use the hook to modify them too. See also
5454 @pxref{Automaton pipeline description}.
5457 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5458 This hook adjusts the integer scheduling priority @var{priority} of
5459 @var{insn}. It should return the new priority. Reduce the priority to
5460 execute @var{insn} earlier, increase the priority to execute @var{insn}
5461 later. Do not define this hook if you do not need to adjust the
5462 scheduling priorities of insns.
5465 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5466 This hook is executed by the scheduler after it has scheduled the ready
5467 list, to allow the machine description to reorder it (for example to
5468 combine two small instructions together on @samp{VLIW} machines).
5469 @var{file} is either a null pointer, or a stdio stream to write any
5470 debug output to. @var{verbose} is the verbose level provided by
5471 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5472 list of instructions that are ready to be scheduled. @var{n_readyp} is
5473 a pointer to the number of elements in the ready list. The scheduler
5474 reads the ready list in reverse order, starting with
5475 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5476 is the timer tick of the scheduler. You may modify the ready list and
5477 the number of ready insns. The return value is the number of insns that
5478 can issue this cycle; normally this is just @code{issue_rate}. See also
5479 @samp{TARGET_SCHED_REORDER2}.
5482 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5483 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5484 function is called whenever the scheduler starts a new cycle. This one
5485 is called once per iteration over a cycle, immediately after
5486 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5487 return the number of insns to be scheduled in the same cycle. Defining
5488 this hook can be useful if there are frequent situations where
5489 scheduling one insn causes other insns to become ready in the same
5490 cycle. These other insns can then be taken into account properly.
5493 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5494 This hook is called after evaluation forward dependencies of insns in
5495 chain given by two parameter values (@var{head} and @var{tail}
5496 correspondingly) but before insns scheduling of the insn chain. For
5497 example, it can be used for better insn classification if it requires
5498 analysis of dependencies. This hook can use backward and forward
5499 dependencies of the insn scheduler because they are already
5503 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5504 This hook is executed by the scheduler at the beginning of each block of
5505 instructions that are to be scheduled. @var{file} is either a null
5506 pointer, or a stdio stream to write any debug output to. @var{verbose}
5507 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5508 @var{max_ready} is the maximum number of insns in the current scheduling
5509 region that can be live at the same time. This can be used to allocate
5510 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5513 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5514 This hook is executed by the scheduler at the end of each block of
5515 instructions that are to be scheduled. It can be used to perform
5516 cleanup of any actions done by the other scheduling hooks. @var{file}
5517 is either a null pointer, or a stdio stream to write any debug output
5518 to. @var{verbose} is the verbose level provided by
5519 @option{-fsched-verbose-@var{n}}.
5522 @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void)
5523 This hook is called many times during insn scheduling. If the hook
5524 returns nonzero, the automaton based pipeline description is used for
5525 insn scheduling. Otherwise the traditional pipeline description is
5526 used. The default is usage of the traditional pipeline description.
5528 You should also remember that to simplify the insn scheduler sources
5529 an empty traditional pipeline description interface is generated even
5530 if there is no a traditional pipeline description in the @file{.md}
5531 file. The same is true for the automaton based pipeline description.
5532 That means that you should be accurate in defining the hook.
5535 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5536 The hook returns an RTL insn. The automaton state used in the
5537 pipeline hazard recognizer is changed as if the insn were scheduled
5538 when the new simulated processor cycle starts. Usage of the hook may
5539 simplify the automaton pipeline description for some @acronym{VLIW}
5540 processors. If the hook is defined, it is used only for the automaton
5541 based pipeline description. The default is not to change the state
5542 when the new simulated processor cycle starts.
5545 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5546 The hook can be used to initialize data used by the previous hook.
5549 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5550 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5551 to changed the state as if the insn were scheduled when the new
5552 simulated processor cycle finishes.
5555 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5556 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5557 used to initialize data used by the previous hook.
5560 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5561 This hook controls better choosing an insn from the ready insn queue
5562 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5563 chooses the first insn from the queue. If the hook returns a positive
5564 value, an additional scheduler code tries all permutations of
5565 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5566 subsequent ready insns to choose an insn whose issue will result in
5567 maximal number of issued insns on the same cycle. For the
5568 @acronym{VLIW} processor, the code could actually solve the problem of
5569 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5570 rules of @acronym{VLIW} packing are described in the automaton.
5572 This code also could be used for superscalar @acronym{RISC}
5573 processors. Let us consider a superscalar @acronym{RISC} processor
5574 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5575 @var{B}, some insns can be executed only in pipelines @var{B} or
5576 @var{C}, and one insn can be executed in pipeline @var{B}. The
5577 processor may issue the 1st insn into @var{A} and the 2nd one into
5578 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5579 until the next cycle. If the scheduler issues the 3rd insn the first,
5580 the processor could issue all 3 insns per cycle.
5582 Actually this code demonstrates advantages of the automaton based
5583 pipeline hazard recognizer. We try quickly and easy many insn
5584 schedules to choose the best one.
5586 The default is no multipass scheduling.
5589 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5591 This hook controls what insns from the ready insn queue will be
5592 considered for the multipass insn scheduling. If the hook returns
5593 zero for insn passed as the parameter, the insn will be not chosen to
5596 The default is that any ready insns can be chosen to be issued.
5599 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5601 This hook is called by the insn scheduler before issuing insn passed
5602 as the third parameter on given cycle. If the hook returns nonzero,
5603 the insn is not issued on given processors cycle. Instead of that,
5604 the processor cycle is advanced. If the value passed through the last
5605 parameter is zero, the insn ready queue is not sorted on the new cycle
5606 start as usually. The first parameter passes file for debugging
5607 output. The second one passes the scheduler verbose level of the
5608 debugging output. The forth and the fifth parameter values are
5609 correspondingly processor cycle on which the previous insn has been
5610 issued and the current processor cycle.
5613 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void)
5614 The @acronym{DFA}-based scheduler could take the insertion of nop
5615 operations for better insn scheduling into account. It can be done
5616 only if the multi-pass insn scheduling works (see hook
5617 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}).
5619 Let us consider a @acronym{VLIW} processor insn with 3 slots. Each
5620 insn can be placed only in one of the three slots. We have 3 ready
5621 insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be
5622 placed only in the 1st slot, @var{B} can be placed only in the 3rd
5623 slot. We described the automaton which does not permit empty slot
5624 gaps between insns (usually such description is simpler). Without
5625 this code the scheduler would place each insn in 3 separate
5626 @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd
5627 slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is
5628 the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook
5629 @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or
5630 create the nop insns.
5632 You should remember that the scheduler does not insert the nop insns.
5633 It is not wise because of the following optimizations. The scheduler
5634 only considers such possibility to improve the result schedule. The
5635 nop insns should be inserted lately, e.g. on the final phase.
5638 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index})
5639 This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert
5640 nop operations for better insn scheduling when @acronym{DFA}-based
5641 scheduler makes multipass insn scheduling (see also description of
5642 hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook
5643 returns a nop insn with given @var{index}. The indexes start with
5644 zero. The hook should return @code{NULL} if there are no more nop
5645 insns with indexes greater than given index.
5648 @deftypefn {Target Hook} bool IS_COSTLY_DEPENDENCE (rtx @var{insn1}, rtx @var{insn2}, rtx @var{dep_link}, int @var{dep_cost}, int @var{distance})
5649 This hook is used to define which dependences are considered costly by
5650 the target, so costly that it is not advisable to schedule the insns that
5651 are involved in the dependence too close to one another. The parameters
5652 to this hook are as follows: The second parameter @var{insn2} is dependent
5653 upon the first parameter @var{insn1}. The dependence between @var{insn1}
5654 and @var{insn2} is represented by the third parameter @var{dep_link}. The
5655 fourth parameter @var{cost} is the cost of the dependence, and the fifth
5656 parameter @var{distance} is the distance in cycles between the two insns.
5657 The hook returns @code{true} if considering the distance between the two
5658 insns the dependence between them is considered costly by the target,
5659 and @code{false} otherwise.
5661 Defining this hook can be useful in multiple-issue out-of-order machines,
5662 where (a) it's practically hopeless to predict the actual data/resource
5663 delays, however: (b) there's a better chance to predict the actual grouping
5664 that will be formed, and (c) correctly emulating the grouping can be very
5665 important. In such targets one may want to allow issuing dependent insns
5666 closer to one another - i.e, closer than the dependence distance; however,
5667 not in cases of "costly dependences", which this hooks allows to define.
5670 Macros in the following table are generated by the program
5671 @file{genattr} and can be useful for writing the hooks.
5673 @defmac TRADITIONAL_PIPELINE_INTERFACE
5674 The macro definition is generated if there is a traditional pipeline
5675 description in @file{.md} file. You should also remember that to
5676 simplify the insn scheduler sources an empty traditional pipeline
5677 description interface is generated even if there is no a traditional
5678 pipeline description in the @file{.md} file. The macro can be used to
5679 distinguish the two types of the traditional interface.
5682 @defmac DFA_PIPELINE_INTERFACE
5683 The macro definition is generated if there is an automaton pipeline
5684 description in @file{.md} file. You should also remember that to
5685 simplify the insn scheduler sources an empty automaton pipeline
5686 description interface is generated even if there is no an automaton
5687 pipeline description in the @file{.md} file. The macro can be used to
5688 distinguish the two types of the automaton interface.
5691 @defmac MAX_DFA_ISSUE_RATE
5692 The macro definition is generated in the automaton based pipeline
5693 description interface. Its value is calculated from the automaton
5694 based pipeline description and is equal to maximal number of all insns
5695 described in constructions @samp{define_insn_reservation} which can be
5696 issued on the same processor cycle.
5700 @section Dividing the Output into Sections (Texts, Data, @dots{})
5701 @c the above section title is WAY too long. maybe cut the part between
5702 @c the (...)? --mew 10feb93
5704 An object file is divided into sections containing different types of
5705 data. In the most common case, there are three sections: the @dfn{text
5706 section}, which holds instructions and read-only data; the @dfn{data
5707 section}, which holds initialized writable data; and the @dfn{bss
5708 section}, which holds uninitialized data. Some systems have other kinds
5711 The compiler must tell the assembler when to switch sections. These
5712 macros control what commands to output to tell the assembler this. You
5713 can also define additional sections.
5715 @defmac TEXT_SECTION_ASM_OP
5716 A C expression whose value is a string, including spacing, containing the
5717 assembler operation that should precede instructions and read-only data.
5718 Normally @code{"\t.text"} is right.
5721 @defmac TEXT_SECTION
5722 A C statement that switches to the default section containing instructions.
5723 Normally this is not needed, as simply defining @code{TEXT_SECTION_ASM_OP}
5724 is enough. The MIPS port uses this to sort all functions after all data
5728 @defmac HOT_TEXT_SECTION_NAME
5729 If defined, a C string constant for the name of the section containing most
5730 frequently executed functions of the program. If not defined, GCC will provide
5731 a default definition if the target supports named sections.
5734 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5735 If defined, a C string constant for the name of the section containing unlikely
5736 executed functions in the program.
5739 @defmac DATA_SECTION_ASM_OP
5740 A C expression whose value is a string, including spacing, containing the
5741 assembler operation to identify the following data as writable initialized
5742 data. Normally @code{"\t.data"} is right.
5745 @defmac READONLY_DATA_SECTION_ASM_OP
5746 A C expression whose value is a string, including spacing, containing the
5747 assembler operation to identify the following data as read-only initialized
5751 @defmac READONLY_DATA_SECTION
5752 A macro naming a function to call to switch to the proper section for
5753 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5754 if defined, else fall back to @code{text_section}.
5756 The most common definition will be @code{data_section}, if the target
5757 does not have a special read-only data section, and does not put data
5758 in the text section.
5761 @defmac SHARED_SECTION_ASM_OP
5762 If defined, a C expression whose value is a string, including spacing,
5763 containing the assembler operation to identify the following data as
5764 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5767 @defmac BSS_SECTION_ASM_OP
5768 If defined, a C expression whose value is a string, including spacing,
5769 containing the assembler operation to identify the following data as
5770 uninitialized global data. If not defined, and neither
5771 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5772 uninitialized global data will be output in the data section if
5773 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5777 @defmac SHARED_BSS_SECTION_ASM_OP
5778 If defined, a C expression whose value is a string, including spacing,
5779 containing the assembler operation to identify the following data as
5780 uninitialized global shared data. If not defined, and
5781 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5784 @defmac INIT_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 initialization code. If not defined, GCC will assume such a section does
5791 @defmac FINI_SECTION_ASM_OP
5792 If defined, a C expression whose value is a string, including spacing,
5793 containing the assembler operation to identify the following data as
5794 finalization code. If not defined, GCC will assume such a section does
5798 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5799 If defined, an ASM statement that switches to a different section
5800 via @var{section_op}, calls @var{function}, and switches back to
5801 the text section. This is used in @file{crtstuff.c} if
5802 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5803 to initialization and finalization functions from the init and fini
5804 sections. By default, this macro uses a simple function call. Some
5805 ports need hand-crafted assembly code to avoid dependencies on
5806 registers initialized in the function prologue or to ensure that
5807 constant pools don't end up too far way in the text section.
5810 @defmac FORCE_CODE_SECTION_ALIGN
5811 If defined, an ASM statement that aligns a code section to some
5812 arbitrary boundary. This is used to force all fragments of the
5813 @code{.init} and @code{.fini} sections to have to same alignment
5814 and thus prevent the linker from having to add any padding.
5819 @defmac EXTRA_SECTIONS
5820 A list of names for sections other than the standard two, which are
5821 @code{in_text} and @code{in_data}. You need not define this macro
5822 on a system with no other sections (that GCC needs to use).
5825 @findex text_section
5826 @findex data_section
5827 @defmac EXTRA_SECTION_FUNCTIONS
5828 One or more functions to be defined in @file{varasm.c}. These
5829 functions should do jobs analogous to those of @code{text_section} and
5830 @code{data_section}, for your additional sections. Do not define this
5831 macro if you do not define @code{EXTRA_SECTIONS}.
5834 @defmac JUMP_TABLES_IN_TEXT_SECTION
5835 Define this macro to be an expression with a nonzero value if jump
5836 tables (for @code{tablejump} insns) should be output in the text
5837 section, along with the assembler instructions. Otherwise, the
5838 readonly data section is used.
5840 This macro is irrelevant if there is no separate readonly data section.
5843 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5844 Switches to the appropriate section for output of @var{exp}. You can
5845 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5846 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5847 requires link-time relocations. Bit 0 is set when variable contains
5848 local relocations only, while bit 1 is set for global relocations.
5849 Select the section by calling @code{data_section} or one of the
5850 alternatives for other sections. @var{align} is the constant alignment
5853 The default version of this function takes care of putting read-only
5854 variables in @code{readonly_data_section}.
5857 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5858 Build up a unique section name, expressed as a @code{STRING_CST} node,
5859 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5860 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5861 the initial value of @var{exp} requires link-time relocations.
5863 The default version of this function appends the symbol name to the
5864 ELF section name that would normally be used for the symbol. For
5865 example, the function @code{foo} would be placed in @code{.text.foo}.
5866 Whatever the actual target object format, this is often good enough.
5869 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
5870 Switches to the appropriate section for output of constant pool entry
5871 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
5872 constant in RTL@. The argument @var{mode} is redundant except in the
5873 case of a @code{const_int} rtx. Select the section by calling
5874 @code{readonly_data_section} or one of the alternatives for other
5875 sections. @var{align} is the constant alignment in bits.
5877 The default version of this function takes care of putting symbolic
5878 constants in @code{flag_pic} mode in @code{data_section} and everything
5879 else in @code{readonly_data_section}.
5882 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
5883 Define this hook if references to a symbol or a constant must be
5884 treated differently depending on something about the variable or
5885 function named by the symbol (such as what section it is in).
5887 The hook is executed immediately after rtl has been created for
5888 @var{decl}, which may be a variable or function declaration or
5889 an entry in the constant pool. In either case, @var{rtl} is the
5890 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
5891 in this hook; that field may not have been initialized yet.
5893 In the case of a constant, it is safe to assume that the rtl is
5894 a @code{mem} whose address is a @code{symbol_ref}. Most decls
5895 will also have this form, but that is not guaranteed. Global
5896 register variables, for instance, will have a @code{reg} for their
5897 rtl. (Normally the right thing to do with such unusual rtl is
5900 The @var{new_decl_p} argument will be true if this is the first time
5901 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
5902 be false for subsequent invocations, which will happen for duplicate
5903 declarations. Whether or not anything must be done for the duplicate
5904 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
5905 @var{new_decl_p} is always true when the hook is called for a constant.
5907 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
5908 The usual thing for this hook to do is to record flags in the
5909 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
5910 Historically, the name string was modified if it was necessary to
5911 encode more than one bit of information, but this practice is now
5912 discouraged; use @code{SYMBOL_REF_FLAGS}.
5914 The default definition of this hook, @code{default_encode_section_info}
5915 in @file{varasm.c}, sets a number of commonly-useful bits in
5916 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
5917 before overriding it.
5920 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
5921 Decode @var{name} and return the real name part, sans
5922 the characters that @code{TARGET_ENCODE_SECTION_INFO}
5926 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
5927 Returns true if @var{exp} should be placed into a ``small data'' section.
5928 The default version of this hook always returns false.
5931 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
5932 Contains the value true if the target places read-only
5933 ``small data'' into a separate section. The default value is false.
5936 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
5937 Returns true if @var{exp} names an object for which name resolution
5938 rules must resolve to the current ``module'' (dynamic shared library
5939 or executable image).
5941 The default version of this hook implements the name resolution rules
5942 for ELF, which has a looser model of global name binding than other
5943 currently supported object file formats.
5946 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
5947 Contains the value true if the target supports thread-local storage.
5948 The default value is false.
5953 @section Position Independent Code
5954 @cindex position independent code
5957 This section describes macros that help implement generation of position
5958 independent code. Simply defining these macros is not enough to
5959 generate valid PIC; you must also add support to the macros
5960 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5961 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5962 @samp{movsi} to do something appropriate when the source operand
5963 contains a symbolic address. You may also need to alter the handling of
5964 switch statements so that they use relative addresses.
5965 @c i rearranged the order of the macros above to try to force one of
5966 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5968 @defmac PIC_OFFSET_TABLE_REGNUM
5969 The register number of the register used to address a table of static
5970 data addresses in memory. In some cases this register is defined by a
5971 processor's ``application binary interface'' (ABI)@. When this macro
5972 is defined, RTL is generated for this register once, as with the stack
5973 pointer and frame pointer registers. If this macro is not defined, it
5974 is up to the machine-dependent files to allocate such a register (if
5975 necessary). Note that this register must be fixed when in use (e.g.@:
5976 when @code{flag_pic} is true).
5979 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5980 Define this macro if the register defined by
5981 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
5982 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5985 @defmac FINALIZE_PIC
5986 By generating position-independent code, when two different programs (A
5987 and B) share a common library (libC.a), the text of the library can be
5988 shared whether or not the library is linked at the same address for both
5989 programs. In some of these environments, position-independent code
5990 requires not only the use of different addressing modes, but also
5991 special code to enable the use of these addressing modes.
5993 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
5994 codes once the function is being compiled into assembly code, but not
5995 before. (It is not done before, because in the case of compiling an
5996 inline function, it would lead to multiple PIC prologues being
5997 included in functions which used inline functions and were compiled to
6001 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6002 A C expression that is nonzero if @var{x} is a legitimate immediate
6003 operand on the target machine when generating position independent code.
6004 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6005 check this. You can also assume @var{flag_pic} is true, so you need not
6006 check it either. You need not define this macro if all constants
6007 (including @code{SYMBOL_REF}) can be immediate operands when generating
6008 position independent code.
6011 @node Assembler Format
6012 @section Defining the Output Assembler Language
6014 This section describes macros whose principal purpose is to describe how
6015 to write instructions in assembler language---rather than what the
6019 * File Framework:: Structural information for the assembler file.
6020 * Data Output:: Output of constants (numbers, strings, addresses).
6021 * Uninitialized Data:: Output of uninitialized variables.
6022 * Label Output:: Output and generation of labels.
6023 * Initialization:: General principles of initialization
6024 and termination routines.
6025 * Macros for Initialization::
6026 Specific macros that control the handling of
6027 initialization and termination routines.
6028 * Instruction Output:: Output of actual instructions.
6029 * Dispatch Tables:: Output of jump tables.
6030 * Exception Region Output:: Output of exception region code.
6031 * Alignment Output:: Pseudo ops for alignment and skipping data.
6034 @node File Framework
6035 @subsection The Overall Framework of an Assembler File
6036 @cindex assembler format
6037 @cindex output of assembler code
6039 @c prevent bad page break with this line
6040 This describes the overall framework of an assembly file.
6042 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6043 @findex default_file_start
6044 Output to @code{asm_out_file} any text which the assembler expects to
6045 find at the beginning of a file. The default behavior is controlled
6046 by two flags, documented below. Unless your target's assembler is
6047 quite unusual, if you override the default, you should call
6048 @code{default_file_start} at some point in your target hook. This
6049 lets other target files rely on these variables.
6052 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6053 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6054 printed as the very first line in the assembly file, unless
6055 @option{-fverbose-asm} is in effect. (If that macro has been defined
6056 to the empty string, this variable has no effect.) With the normal
6057 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6058 assembler that it need not bother stripping comments or extra
6059 whitespace from its input. This allows it to work a bit faster.
6061 The default is false. You should not set it to true unless you have
6062 verified that your port does not generate any extra whitespace or
6063 comments that will cause GAS to issue errors in NO_APP mode.
6066 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6067 If this flag is true, @code{output_file_directive} will be called
6068 for the primary source file, immediately after printing
6069 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6070 this to be done. The default is false.
6073 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6074 Output to @code{asm_out_file} any text which the assembler expects
6075 to find at the end of a file. The default is to output nothing.
6078 @deftypefun void file_end_indicate_exec_stack ()
6079 Some systems use a common convention, the @samp{.note.GNU-stack}
6080 special section, to indicate whether or not an object file relies on
6081 the stack being executable. If your system uses this convention, you
6082 should define @code{TARGET_ASM_FILE_END} to this function. If you
6083 need to do other things in that hook, have your hook function call
6087 @defmac ASM_COMMENT_START
6088 A C string constant describing how to begin a comment in the target
6089 assembler language. The compiler assumes that the comment will end at
6090 the end of the line.
6094 A C string constant for text to be output before each @code{asm}
6095 statement or group of consecutive ones. Normally this is
6096 @code{"#APP"}, which is a comment that has no effect on most
6097 assemblers but tells the GNU assembler that it must check the lines
6098 that follow for all valid assembler constructs.
6102 A C string constant for text to be output after each @code{asm}
6103 statement or group of consecutive ones. Normally this is
6104 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6105 time-saving assumptions that are valid for ordinary compiler output.
6108 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6109 A C statement to output COFF information or DWARF debugging information
6110 which indicates that filename @var{name} is the current source file to
6111 the stdio stream @var{stream}.
6113 This macro need not be defined if the standard form of output
6114 for the file format in use is appropriate.
6117 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6118 A C statement to output the string @var{string} to the stdio stream
6119 @var{stream}. If you do not call the function @code{output_quoted_string}
6120 in your config files, GCC will only call it to output filenames to
6121 the assembler source. So you can use it to canonicalize the format
6122 of the filename using this macro.
6125 @defmac ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
6126 A C statement to output DBX or SDB debugging information before code
6127 for line number @var{line} of the current source file to the
6128 stdio stream @var{stream}. @var{counter} is the number of time the
6129 macro was invoked, including the current invocation; it is intended
6130 to generate unique labels in the assembly output.
6132 This macro need not be defined if the standard form of debugging
6133 information for the debugger in use is appropriate.
6136 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6137 A C statement to output something to the assembler file to handle a
6138 @samp{#ident} directive containing the text @var{string}. If this
6139 macro is not defined, nothing is output for a @samp{#ident} directive.
6142 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6143 Output assembly directives to switch to section @var{name}. The section
6144 should have attributes as specified by @var{flags}, which is a bit mask
6145 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6146 is nonzero, it contains an alignment in bytes to be used for the section,
6147 otherwise some target default should be used. Only targets that must
6148 specify an alignment within the section directive need pay attention to
6149 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6152 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6153 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6156 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6157 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6158 based on a variable or function decl, a section name, and whether or not the
6159 declaration's initializer may contain runtime relocations. @var{decl} may be
6160 null, in which case read-write data should be assumed.
6162 The default version if this function handles choosing code vs data,
6163 read-only vs read-write data, and @code{flag_pic}. You should only
6164 need to override this if your target has special flags that might be
6165 set via @code{__attribute__}.
6170 @subsection Output of Data
6173 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6174 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6175 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6176 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6177 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6178 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6179 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6180 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6181 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6182 These hooks specify assembly directives for creating certain kinds
6183 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6184 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6185 aligned two-byte object, and so on. Any of the hooks may be
6186 @code{NULL}, indicating that no suitable directive is available.
6188 The compiler will print these strings at the start of a new line,
6189 followed immediately by the object's initial value. In most cases,
6190 the string should contain a tab, a pseudo-op, and then another tab.
6193 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6194 The @code{assemble_integer} function uses this hook to output an
6195 integer object. @var{x} is the object's value, @var{size} is its size
6196 in bytes and @var{aligned_p} indicates whether it is aligned. The
6197 function should return @code{true} if it was able to output the
6198 object. If it returns false, @code{assemble_integer} will try to
6199 split the object into smaller parts.
6201 The default implementation of this hook will use the
6202 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6203 when the relevant string is @code{NULL}.
6206 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6207 A C statement to recognize @var{rtx} patterns that
6208 @code{output_addr_const} can't deal with, and output assembly code to
6209 @var{stream} corresponding to the pattern @var{x}. This may be used to
6210 allow machine-dependent @code{UNSPEC}s to appear within constants.
6212 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6213 @code{goto fail}, so that a standard error message is printed. If it
6214 prints an error message itself, by calling, for example,
6215 @code{output_operand_lossage}, it may just complete normally.
6218 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6219 A C statement to output to the stdio stream @var{stream} an assembler
6220 instruction to assemble a string constant containing the @var{len}
6221 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6222 @code{char *} and @var{len} a C expression of type @code{int}.
6224 If the assembler has a @code{.ascii} pseudo-op as found in the
6225 Berkeley Unix assembler, do not define the macro
6226 @code{ASM_OUTPUT_ASCII}.
6229 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6230 A C statement to output word @var{n} of a function descriptor for
6231 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6232 is defined, and is otherwise unused.
6235 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6236 You may define this macro as a C expression. You should define the
6237 expression to have a nonzero value if GCC should output the constant
6238 pool for a function before the code for the function, or a zero value if
6239 GCC should output the constant pool after the function. If you do
6240 not define this macro, the usual case, GCC will output the constant
6241 pool before the function.
6244 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6245 A C statement to output assembler commands to define the start of the
6246 constant pool for a function. @var{funname} is a string giving
6247 the name of the function. Should the return type of the function
6248 be required, it can be obtained via @var{fundecl}. @var{size}
6249 is the size, in bytes, of the constant pool that will be written
6250 immediately after this call.
6252 If no constant-pool prefix is required, the usual case, this macro need
6256 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6257 A C statement (with or without semicolon) to output a constant in the
6258 constant pool, if it needs special treatment. (This macro need not do
6259 anything for RTL expressions that can be output normally.)
6261 The argument @var{file} is the standard I/O stream to output the
6262 assembler code on. @var{x} is the RTL expression for the constant to
6263 output, and @var{mode} is the machine mode (in case @var{x} is a
6264 @samp{const_int}). @var{align} is the required alignment for the value
6265 @var{x}; you should output an assembler directive to force this much
6268 The argument @var{labelno} is a number to use in an internal label for
6269 the address of this pool entry. The definition of this macro is
6270 responsible for outputting the label definition at the proper place.
6271 Here is how to do this:
6274 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6277 When you output a pool entry specially, you should end with a
6278 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6279 entry from being output a second time in the usual manner.
6281 You need not define this macro if it would do nothing.
6284 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6285 A C statement to output assembler commands to at the end of the constant
6286 pool for a function. @var{funname} is a string giving the name of the
6287 function. Should the return type of the function be required, you can
6288 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6289 constant pool that GCC wrote immediately before this call.
6291 If no constant-pool epilogue is required, the usual case, you need not
6295 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6296 Define this macro as a C expression which is nonzero if @var{C} is
6297 used as a logical line separator by the assembler.
6299 If you do not define this macro, the default is that only
6300 the character @samp{;} is treated as a logical line separator.
6303 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6304 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6305 These target hooks are C string constants, describing the syntax in the
6306 assembler for grouping arithmetic expressions. If not overridden, they
6307 default to normal parentheses, which is correct for most assemblers.
6310 These macros are provided by @file{real.h} for writing the definitions
6311 of @code{ASM_OUTPUT_DOUBLE} and the like:
6313 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6314 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6315 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6316 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6317 floating point representation, and store its bit pattern in the variable
6318 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6319 be a simple @code{long int}. For the others, it should be an array of
6320 @code{long int}. The number of elements in this array is determined by
6321 the size of the desired target floating point data type: 32 bits of it
6322 go in each @code{long int} array element. Each array element holds 32
6323 bits of the result, even if @code{long int} is wider than 32 bits on the
6326 The array element values are designed so that you can print them out
6327 using @code{fprintf} in the order they should appear in the target
6331 @node Uninitialized Data
6332 @subsection Output of Uninitialized Variables
6334 Each of the macros in this section is used to do the whole job of
6335 outputting a single uninitialized variable.
6337 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6338 A C statement (sans semicolon) to output to the stdio stream
6339 @var{stream} the assembler definition of a common-label named
6340 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6341 is the size rounded up to whatever alignment the caller wants.
6343 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6344 output the name itself; before and after that, output the additional
6345 assembler syntax for defining the name, and a newline.
6347 This macro controls how the assembler definitions of uninitialized
6348 common global variables are output.
6351 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6352 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6353 separate, explicit argument. If you define this macro, it is used in
6354 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6355 handling the required alignment of the variable. The alignment is specified
6356 as the number of bits.
6359 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6360 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6361 variable to be output, if there is one, or @code{NULL_TREE} if there
6362 is no corresponding variable. If you define this macro, GCC will use it
6363 in place of both @code{ASM_OUTPUT_COMMON} and
6364 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6365 the variable's decl in order to chose what to output.
6368 @defmac ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6369 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6370 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6374 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6375 A C statement (sans semicolon) to output to the stdio stream
6376 @var{stream} the assembler definition of uninitialized global @var{decl} named
6377 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6378 is the size rounded up to whatever alignment the caller wants.
6380 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6381 defining this macro. If unable, use the expression
6382 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6383 before and after that, output the additional assembler syntax for defining
6384 the name, and a newline.
6386 This macro controls how the assembler definitions of uninitialized global
6387 variables are output. This macro exists to properly support languages like
6388 C++ which do not have @code{common} data. However, this macro currently
6389 is not defined for all targets. If this macro and
6390 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6391 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6392 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6395 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6396 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6397 separate, explicit argument. If you define this macro, it is used in
6398 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6399 handling the required alignment of the variable. The alignment is specified
6400 as the number of bits.
6402 Try to use function @code{asm_output_aligned_bss} defined in file
6403 @file{varasm.c} when defining this macro.
6406 @defmac ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6407 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6408 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6412 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6413 A C statement (sans semicolon) to output to the stdio stream
6414 @var{stream} the assembler definition of a local-common-label named
6415 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6416 is the size rounded up to whatever alignment the caller wants.
6418 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6419 output the name itself; before and after that, output the additional
6420 assembler syntax for defining the name, and a newline.
6422 This macro controls how the assembler definitions of uninitialized
6423 static variables are output.
6426 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6427 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6428 separate, explicit argument. If you define this macro, it is used in
6429 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6430 handling the required alignment of the variable. The alignment is specified
6431 as the number of bits.
6434 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6435 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6436 variable to be output, if there is one, or @code{NULL_TREE} if there
6437 is no corresponding variable. If you define this macro, GCC will use it
6438 in place of both @code{ASM_OUTPUT_DECL} and
6439 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6440 the variable's decl in order to chose what to output.
6443 @defmac ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6444 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6445 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6450 @subsection Output and Generation of Labels
6452 @c prevent bad page break with this line
6453 This is about outputting labels.
6455 @findex assemble_name
6456 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6457 A C statement (sans semicolon) to output to the stdio stream
6458 @var{stream} the assembler definition of a label named @var{name}.
6459 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6460 output the name itself; before and after that, output the additional
6461 assembler syntax for defining the name, and a newline. A default
6462 definition of this macro is provided which is correct for most systems.
6466 A C string containing the appropriate assembler directive to specify the
6467 size of a symbol, without any arguments. On systems that use ELF, the
6468 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6469 systems, the default is not to define this macro.
6471 Define this macro only if it is correct to use the default definitions
6472 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6473 for your system. If you need your own custom definitions of those
6474 macros, or if you do not need explicit symbol sizes at all, do not
6478 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6479 A C statement (sans semicolon) to output to the stdio stream
6480 @var{stream} a directive telling the assembler that the size of the
6481 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6482 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6486 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6487 A C statement (sans semicolon) to output to the stdio stream
6488 @var{stream} a directive telling the assembler to calculate the size of
6489 the symbol @var{name} by subtracting its address from the current
6492 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6493 provided. The default assumes that the assembler recognizes a special
6494 @samp{.} symbol as referring to the current address, and can calculate
6495 the difference between this and another symbol. If your assembler does
6496 not recognize @samp{.} or cannot do calculations with it, you will need
6497 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6501 A C string containing the appropriate assembler directive to specify the
6502 type of a symbol, without any arguments. On systems that use ELF, the
6503 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6504 systems, the default is not to define this macro.
6506 Define this macro only if it is correct to use the default definition of
6507 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6508 custom definition of this macro, or if you do not need explicit symbol
6509 types at all, do not define this macro.
6512 @defmac TYPE_OPERAND_FMT
6513 A C string which specifies (using @code{printf} syntax) the format of
6514 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6515 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6516 the default is not to define this macro.
6518 Define this macro only if it is correct to use the default definition of
6519 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6520 custom definition of this macro, or if you do not need explicit symbol
6521 types at all, do not define this macro.
6524 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6525 A C statement (sans semicolon) to output to the stdio stream
6526 @var{stream} a directive telling the assembler that the type of the
6527 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6528 that string is always either @samp{"function"} or @samp{"object"}, but
6529 you should not count on this.
6531 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6532 definition of this macro is provided.
6535 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6536 A C statement (sans semicolon) to output to the stdio stream
6537 @var{stream} any text necessary for declaring the name @var{name} of a
6538 function which is being defined. This macro is responsible for
6539 outputting the label definition (perhaps using
6540 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6541 @code{FUNCTION_DECL} tree node representing the function.
6543 If this macro is not defined, then the function name is defined in the
6544 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6546 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6550 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6551 A C statement (sans semicolon) to output to the stdio stream
6552 @var{stream} any text necessary for declaring the size of a function
6553 which is being defined. The argument @var{name} is the name of the
6554 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6555 representing the function.
6557 If this macro is not defined, then the function size is not defined.
6559 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6563 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6564 A C statement (sans semicolon) to output to the stdio stream
6565 @var{stream} any text necessary for declaring the name @var{name} of an
6566 initialized variable which is being defined. This macro must output the
6567 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6568 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6570 If this macro is not defined, then the variable name is defined in the
6571 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6573 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6574 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6577 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
6578 A C statement (sans semicolon) to output to the stdio stream
6579 @var{stream} any text necessary for declaring the name @var{name} of a
6580 constant which is being defined. This macro is responsible for
6581 outputting the label definition (perhaps using
6582 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
6583 value of the constant, and @var{size} is the size of the constant
6584 in bytes. @var{name} will be an internal label.
6586 If this macro is not defined, then the @var{name} is defined in the
6587 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6589 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6593 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6594 A C statement (sans semicolon) to output to the stdio stream
6595 @var{stream} any text necessary for claiming a register @var{regno}
6596 for a global variable @var{decl} with name @var{name}.
6598 If you don't define this macro, that is equivalent to defining it to do
6602 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6603 A C statement (sans semicolon) to finish up declaring a variable name
6604 once the compiler has processed its initializer fully and thus has had a
6605 chance to determine the size of an array when controlled by an
6606 initializer. This is used on systems where it's necessary to declare
6607 something about the size of the object.
6609 If you don't define this macro, that is equivalent to defining it to do
6612 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6613 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6616 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6617 This target hook is a function to output to the stdio stream
6618 @var{stream} some commands that will make the label @var{name} global;
6619 that is, available for reference from other files.
6621 The default implementation relies on a proper definition of
6622 @code{GLOBAL_ASM_OP}.
6625 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6626 A C statement (sans semicolon) to output to the stdio stream
6627 @var{stream} some commands that will make the label @var{name} weak;
6628 that is, available for reference from other files but only used if
6629 no other definition is available. Use the expression
6630 @code{assemble_name (@var{stream}, @var{name})} to output the name
6631 itself; before and after that, output the additional assembler syntax
6632 for making that name weak, and a newline.
6634 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6635 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6639 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6640 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6641 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6642 or variable decl. If @var{value} is not @code{NULL}, this C statement
6643 should output to the stdio stream @var{stream} assembler code which
6644 defines (equates) the weak symbol @var{name} to have the value
6645 @var{value}. If @var{value} is @code{NULL}, it should output commands
6646 to make @var{name} weak.
6649 @defmac SUPPORTS_WEAK
6650 A C expression which evaluates to true if the target supports weak symbols.
6652 If you don't define this macro, @file{defaults.h} provides a default
6653 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6654 is defined, the default definition is @samp{1}; otherwise, it is
6655 @samp{0}. Define this macro if you want to control weak symbol support
6656 with a compiler flag such as @option{-melf}.
6659 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
6660 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6661 public symbol such that extra copies in multiple translation units will
6662 be discarded by the linker. Define this macro if your object file
6663 format provides support for this concept, such as the @samp{COMDAT}
6664 section flags in the Microsoft Windows PE/COFF format, and this support
6665 requires changes to @var{decl}, such as putting it in a separate section.
6668 @defmac SUPPORTS_ONE_ONLY
6669 A C expression which evaluates to true if the target supports one-only
6672 If you don't define this macro, @file{varasm.c} provides a default
6673 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6674 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6675 you want to control one-only symbol support with a compiler flag, or if
6676 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6677 be emitted as one-only.
6680 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6681 This target hook is a function to output to @var{asm_out_file} some
6682 commands that will make the symbol(s) associated with @var{decl} have
6683 hidden, protected or internal visibility as specified by @var{visibility}.
6686 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6687 A C statement (sans semicolon) to output to the stdio stream
6688 @var{stream} any text necessary for declaring the name of an external
6689 symbol named @var{name} which is referenced in this compilation but
6690 not defined. The value of @var{decl} is the tree node for the
6693 This macro need not be defined if it does not need to output anything.
6694 The GNU assembler and most Unix assemblers don't require anything.
6697 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
6698 This target hook is a function to output to @var{asm_out_file} an assembler
6699 pseudo-op to declare a library function name external. The name of the
6700 library function is given by @var{symref}, which is a @code{symbol_ref}.
6703 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6704 A C statement (sans semicolon) to output to the stdio stream
6705 @var{stream} a reference in assembler syntax to a label named
6706 @var{name}. This should add @samp{_} to the front of the name, if that
6707 is customary on your operating system, as it is in most Berkeley Unix
6708 systems. This macro is used in @code{assemble_name}.
6711 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6712 A C statement (sans semicolon) to output a reference to
6713 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6714 will be used to output the name of the symbol. This macro may be used
6715 to modify the way a symbol is referenced depending on information
6716 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6719 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6720 A C statement (sans semicolon) to output a reference to @var{buf}, the
6721 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6722 @code{assemble_name} will be used to output the name of the symbol.
6723 This macro is not used by @code{output_asm_label}, or the @code{%l}
6724 specifier that calls it; the intention is that this macro should be set
6725 when it is necessary to output a label differently when its address is
6729 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6730 A function to output to the stdio stream @var{stream} a label whose
6731 name is made from the string @var{prefix} and the number @var{labelno}.
6733 It is absolutely essential that these labels be distinct from the labels
6734 used for user-level functions and variables. Otherwise, certain programs
6735 will have name conflicts with internal labels.
6737 It is desirable to exclude internal labels from the symbol table of the
6738 object file. Most assemblers have a naming convention for labels that
6739 should be excluded; on many systems, the letter @samp{L} at the
6740 beginning of a label has this effect. You should find out what
6741 convention your system uses, and follow it.
6743 The default version of this function utilizes ASM_GENERATE_INTERNAL_LABEL.
6746 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6747 A C statement to output to the stdio stream @var{stream} a debug info
6748 label whose name is made from the string @var{prefix} and the number
6749 @var{num}. This is useful for VLIW targets, where debug info labels
6750 may need to be treated differently than branch target labels. On some
6751 systems, branch target labels must be at the beginning of instruction
6752 bundles, but debug info labels can occur in the middle of instruction
6755 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
6759 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6760 A C statement to store into the string @var{string} a label whose name
6761 is made from the string @var{prefix} and the number @var{num}.
6763 This string, when output subsequently by @code{assemble_name}, should
6764 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
6765 with the same @var{prefix} and @var{num}.
6767 If the string begins with @samp{*}, then @code{assemble_name} will
6768 output the rest of the string unchanged. It is often convenient for
6769 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6770 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6771 to output the string, and may change it. (Of course,
6772 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6773 you should know what it does on your machine.)
6776 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6777 A C expression to assign to @var{outvar} (which is a variable of type
6778 @code{char *}) a newly allocated string made from the string
6779 @var{name} and the number @var{number}, with some suitable punctuation
6780 added. Use @code{alloca} to get space for the string.
6782 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6783 produce an assembler label for an internal static variable whose name is
6784 @var{name}. Therefore, the string must be such as to result in valid
6785 assembler code. The argument @var{number} is different each time this
6786 macro is executed; it prevents conflicts between similarly-named
6787 internal static variables in different scopes.
6789 Ideally this string should not be a valid C identifier, to prevent any
6790 conflict with the user's own symbols. Most assemblers allow periods
6791 or percent signs in assembler symbols; putting at least one of these
6792 between the name and the number will suffice.
6794 If this macro is not defined, a default definition will be provided
6795 which is correct for most systems.
6798 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6799 A C statement to output to the stdio stream @var{stream} assembler code
6800 which defines (equates) the symbol @var{name} to have the value @var{value}.
6803 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6804 correct for most systems.
6807 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6808 A C statement to output to the stdio stream @var{stream} assembler code
6809 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6810 to have the value of the tree node @var{decl_of_value}. This macro will
6811 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6812 the tree nodes are available.
6815 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6816 correct for most systems.
6819 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6820 A C statement to output to the stdio stream @var{stream} assembler code
6821 which defines (equates) the weak symbol @var{name} to have the value
6822 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6823 an undefined weak symbol.
6825 Define this macro if the target only supports weak aliases; define
6826 @code{ASM_OUTPUT_DEF} instead if possible.
6829 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6830 Define this macro to override the default assembler names used for
6831 Objective-C methods.
6833 The default name is a unique method number followed by the name of the
6834 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6835 the category is also included in the assembler name (e.g.@:
6838 These names are safe on most systems, but make debugging difficult since
6839 the method's selector is not present in the name. Therefore, particular
6840 systems define other ways of computing names.
6842 @var{buf} is an expression of type @code{char *} which gives you a
6843 buffer in which to store the name; its length is as long as
6844 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6845 50 characters extra.
6847 The argument @var{is_inst} specifies whether the method is an instance
6848 method or a class method; @var{class_name} is the name of the class;
6849 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6850 in a category); and @var{sel_name} is the name of the selector.
6852 On systems where the assembler can handle quoted names, you can use this
6853 macro to provide more human-readable names.
6856 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
6857 A C statement (sans semicolon) to output to the stdio stream
6858 @var{stream} commands to declare that the label @var{name} is an
6859 Objective-C class reference. This is only needed for targets whose
6860 linkers have special support for NeXT-style runtimes.
6863 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6864 A C statement (sans semicolon) to output to the stdio stream
6865 @var{stream} commands to declare that the label @var{name} is an
6866 unresolved Objective-C class reference. This is only needed for targets
6867 whose linkers have special support for NeXT-style runtimes.
6870 @node Initialization
6871 @subsection How Initialization Functions Are Handled
6872 @cindex initialization routines
6873 @cindex termination routines
6874 @cindex constructors, output of
6875 @cindex destructors, output of
6877 The compiled code for certain languages includes @dfn{constructors}
6878 (also called @dfn{initialization routines})---functions to initialize
6879 data in the program when the program is started. These functions need
6880 to be called before the program is ``started''---that is to say, before
6881 @code{main} is called.
6883 Compiling some languages generates @dfn{destructors} (also called
6884 @dfn{termination routines}) that should be called when the program
6887 To make the initialization and termination functions work, the compiler
6888 must output something in the assembler code to cause those functions to
6889 be called at the appropriate time. When you port the compiler to a new
6890 system, you need to specify how to do this.
6892 There are two major ways that GCC currently supports the execution of
6893 initialization and termination functions. Each way has two variants.
6894 Much of the structure is common to all four variations.
6896 @findex __CTOR_LIST__
6897 @findex __DTOR_LIST__
6898 The linker must build two lists of these functions---a list of
6899 initialization functions, called @code{__CTOR_LIST__}, and a list of
6900 termination functions, called @code{__DTOR_LIST__}.
6902 Each list always begins with an ignored function pointer (which may hold
6903 0, @minus{}1, or a count of the function pointers after it, depending on
6904 the environment). This is followed by a series of zero or more function
6905 pointers to constructors (or destructors), followed by a function
6906 pointer containing zero.
6908 Depending on the operating system and its executable file format, either
6909 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6910 time and exit time. Constructors are called in reverse order of the
6911 list; destructors in forward order.
6913 The best way to handle static constructors works only for object file
6914 formats which provide arbitrarily-named sections. A section is set
6915 aside for a list of constructors, and another for a list of destructors.
6916 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6917 object file that defines an initialization function also puts a word in
6918 the constructor section to point to that function. The linker
6919 accumulates all these words into one contiguous @samp{.ctors} section.
6920 Termination functions are handled similarly.
6922 This method will be chosen as the default by @file{target-def.h} if
6923 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
6924 support arbitrary sections, but does support special designated
6925 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
6926 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
6928 When arbitrary sections are available, there are two variants, depending
6929 upon how the code in @file{crtstuff.c} is called. On systems that
6930 support a @dfn{.init} section which is executed at program startup,
6931 parts of @file{crtstuff.c} are compiled into that section. The
6932 program is linked by the @command{gcc} driver like this:
6935 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
6938 The prologue of a function (@code{__init}) appears in the @code{.init}
6939 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
6940 for the function @code{__fini} in the @dfn{.fini} section. Normally these
6941 files are provided by the operating system or by the GNU C library, but
6942 are provided by GCC for a few targets.
6944 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
6945 compiled from @file{crtstuff.c}. They contain, among other things, code
6946 fragments within the @code{.init} and @code{.fini} sections that branch
6947 to routines in the @code{.text} section. The linker will pull all parts
6948 of a section together, which results in a complete @code{__init} function
6949 that invokes the routines we need at startup.
6951 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
6954 If no init section is available, when GCC compiles any function called
6955 @code{main} (or more accurately, any function designated as a program
6956 entry point by the language front end calling @code{expand_main_function}),
6957 it inserts a procedure call to @code{__main} as the first executable code
6958 after the function prologue. The @code{__main} function is defined
6959 in @file{libgcc2.c} and runs the global constructors.
6961 In file formats that don't support arbitrary sections, there are again
6962 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
6963 and an `a.out' format must be used. In this case,
6964 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
6965 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
6966 and with the address of the void function containing the initialization
6967 code as its value. The GNU linker recognizes this as a request to add
6968 the value to a @dfn{set}; the values are accumulated, and are eventually
6969 placed in the executable as a vector in the format described above, with
6970 a leading (ignored) count and a trailing zero element.
6971 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
6972 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
6973 the compilation of @code{main} to call @code{__main} as above, starting
6974 the initialization process.
6976 The last variant uses neither arbitrary sections nor the GNU linker.
6977 This is preferable when you want to do dynamic linking and when using
6978 file formats which the GNU linker does not support, such as `ECOFF'@. In
6979 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
6980 termination functions are recognized simply by their names. This requires
6981 an extra program in the linkage step, called @command{collect2}. This program
6982 pretends to be the linker, for use with GCC; it does its job by running
6983 the ordinary linker, but also arranges to include the vectors of
6984 initialization and termination functions. These functions are called
6985 via @code{__main} as described above. In order to use this method,
6986 @code{use_collect2} must be defined in the target in @file{config.gcc}.
6989 The following section describes the specific macros that control and
6990 customize the handling of initialization and termination functions.
6993 @node Macros for Initialization
6994 @subsection Macros Controlling Initialization Routines
6996 Here are the macros that control how the compiler handles initialization
6997 and termination functions:
6999 @defmac INIT_SECTION_ASM_OP
7000 If defined, a C string constant, including spacing, for the assembler
7001 operation to identify the following data as initialization code. If not
7002 defined, GCC will assume such a section does not exist. When you are
7003 using special sections for initialization and termination functions, this
7004 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7005 run the initialization functions.
7008 @defmac HAS_INIT_SECTION
7009 If defined, @code{main} will not call @code{__main} as described above.
7010 This macro should be defined for systems that control start-up code
7011 on a symbol-by-symbol basis, such as OSF/1, and should not
7012 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7015 @defmac LD_INIT_SWITCH
7016 If defined, a C string constant for a switch that tells the linker that
7017 the following symbol is an initialization routine.
7020 @defmac LD_FINI_SWITCH
7021 If defined, a C string constant for a switch that tells the linker that
7022 the following symbol is a finalization routine.
7025 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7026 If defined, a C statement that will write a function that can be
7027 automatically called when a shared library is loaded. The function
7028 should call @var{func}, which takes no arguments. If not defined, and
7029 the object format requires an explicit initialization function, then a
7030 function called @code{_GLOBAL__DI} will be generated.
7032 This function and the following one are used by collect2 when linking a
7033 shared library that needs constructors or destructors, or has DWARF2
7034 exception tables embedded in the code.
7037 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7038 If defined, a C statement that will write a function that can be
7039 automatically called when a shared library is unloaded. The function
7040 should call @var{func}, which takes no arguments. If not defined, and
7041 the object format requires an explicit finalization function, then a
7042 function called @code{_GLOBAL__DD} will be generated.
7045 @defmac INVOKE__main
7046 If defined, @code{main} will call @code{__main} despite the presence of
7047 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7048 where the init section is not actually run automatically, but is still
7049 useful for collecting the lists of constructors and destructors.
7052 @defmac SUPPORTS_INIT_PRIORITY
7053 If nonzero, the C++ @code{init_priority} attribute is supported and the
7054 compiler should emit instructions to control the order of initialization
7055 of objects. If zero, the compiler will issue an error message upon
7056 encountering an @code{init_priority} attribute.
7059 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7060 This value is true if the target supports some ``native'' method of
7061 collecting constructors and destructors to be run at startup and exit.
7062 It is false if we must use @command{collect2}.
7065 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7066 If defined, a function that outputs assembler code to arrange to call
7067 the function referenced by @var{symbol} at initialization time.
7069 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7070 no arguments and with no return value. If the target supports initialization
7071 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7072 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7074 If this macro is not defined by the target, a suitable default will
7075 be chosen if (1) the target supports arbitrary section names, (2) the
7076 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7080 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7081 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7082 functions rather than initialization functions.
7085 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7086 generated for the generated object file will have static linkage.
7088 If your system uses @command{collect2} as the means of processing
7089 constructors, then that program normally uses @command{nm} to scan
7090 an object file for constructor functions to be called.
7092 On certain kinds of systems, you can define this macro to make
7093 @command{collect2} work faster (and, in some cases, make it work at all):
7095 @defmac OBJECT_FORMAT_COFF
7096 Define this macro if the system uses COFF (Common Object File Format)
7097 object files, so that @command{collect2} can assume this format and scan
7098 object files directly for dynamic constructor/destructor functions.
7100 This macro is effective only in a native compiler; @command{collect2} as
7101 part of a cross compiler always uses @command{nm} for the target machine.
7104 @defmac COLLECT_PARSE_FLAG (@var{flag})
7105 Define this macro to be C code that examines @command{collect2} command
7106 line option @var{flag} and performs special actions if
7107 @command{collect2} needs to behave differently depending on @var{flag}.
7110 @defmac REAL_NM_FILE_NAME
7111 Define this macro as a C string constant containing the file name to use
7112 to execute @command{nm}. The default is to search the path normally for
7115 If your system supports shared libraries and has a program to list the
7116 dynamic dependencies of a given library or executable, you can define
7117 these macros to enable support for running initialization and
7118 termination functions in shared libraries:
7122 Define this macro to a C string constant containing the name of the program
7123 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7126 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7127 Define this macro to be C code that extracts filenames from the output
7128 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7129 of type @code{char *} that points to the beginning of a line of output
7130 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7131 code must advance @var{ptr} to the beginning of the filename on that
7132 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7135 @node Instruction Output
7136 @subsection Output of Assembler Instructions
7138 @c prevent bad page break with this line
7139 This describes assembler instruction output.
7141 @defmac REGISTER_NAMES
7142 A C initializer containing the assembler's names for the machine
7143 registers, each one as a C string constant. This is what translates
7144 register numbers in the compiler into assembler language.
7147 @defmac ADDITIONAL_REGISTER_NAMES
7148 If defined, a C initializer for an array of structures containing a name
7149 and a register number. This macro defines additional names for hard
7150 registers, thus allowing the @code{asm} option in declarations to refer
7151 to registers using alternate names.
7154 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7155 Define this macro if you are using an unusual assembler that
7156 requires different names for the machine instructions.
7158 The definition is a C statement or statements which output an
7159 assembler instruction opcode to the stdio stream @var{stream}. The
7160 macro-operand @var{ptr} is a variable of type @code{char *} which
7161 points to the opcode name in its ``internal'' form---the form that is
7162 written in the machine description. The definition should output the
7163 opcode name to @var{stream}, performing any translation you desire, and
7164 increment the variable @var{ptr} to point at the end of the opcode
7165 so that it will not be output twice.
7167 In fact, your macro definition may process less than the entire opcode
7168 name, or more than the opcode name; but if you want to process text
7169 that includes @samp{%}-sequences to substitute operands, you must take
7170 care of the substitution yourself. Just be sure to increment
7171 @var{ptr} over whatever text should not be output normally.
7173 @findex recog_data.operand
7174 If you need to look at the operand values, they can be found as the
7175 elements of @code{recog_data.operand}.
7177 If the macro definition does nothing, the instruction is output
7181 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7182 If defined, a C statement to be executed just prior to the output of
7183 assembler code for @var{insn}, to modify the extracted operands so
7184 they will be output differently.
7186 Here the argument @var{opvec} is the vector containing the operands
7187 extracted from @var{insn}, and @var{noperands} is the number of
7188 elements of the vector which contain meaningful data for this insn.
7189 The contents of this vector are what will be used to convert the insn
7190 template into assembler code, so you can change the assembler output
7191 by changing the contents of the vector.
7193 This macro is useful when various assembler syntaxes share a single
7194 file of instruction patterns; by defining this macro differently, you
7195 can cause a large class of instructions to be output differently (such
7196 as with rearranged operands). Naturally, variations in assembler
7197 syntax affecting individual insn patterns ought to be handled by
7198 writing conditional output routines in those patterns.
7200 If this macro is not defined, it is equivalent to a null statement.
7203 @defmac FINAL_PRESCAN_LABEL
7204 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
7205 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
7206 @var{noperands} will be zero.
7209 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7210 A C compound statement to output to stdio stream @var{stream} the
7211 assembler syntax for an instruction operand @var{x}. @var{x} is an
7214 @var{code} is a value that can be used to specify one of several ways
7215 of printing the operand. It is used when identical operands must be
7216 printed differently depending on the context. @var{code} comes from
7217 the @samp{%} specification that was used to request printing of the
7218 operand. If the specification was just @samp{%@var{digit}} then
7219 @var{code} is 0; if the specification was @samp{%@var{ltr}
7220 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7223 If @var{x} is a register, this macro should print the register's name.
7224 The names can be found in an array @code{reg_names} whose type is
7225 @code{char *[]}. @code{reg_names} is initialized from
7226 @code{REGISTER_NAMES}.
7228 When the machine description has a specification @samp{%@var{punct}}
7229 (a @samp{%} followed by a punctuation character), this macro is called
7230 with a null pointer for @var{x} and the punctuation character for
7234 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7235 A C expression which evaluates to true if @var{code} is a valid
7236 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7237 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7238 punctuation characters (except for the standard one, @samp{%}) are used
7242 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7243 A C compound statement to output to stdio stream @var{stream} the
7244 assembler syntax for an instruction operand that is a memory reference
7245 whose address is @var{x}. @var{x} is an RTL expression.
7247 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7248 On some machines, the syntax for a symbolic address depends on the
7249 section that the address refers to. On these machines, define the hook
7250 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7251 @code{symbol_ref}, and then check for it here. @xref{Assembler
7255 @findex dbr_sequence_length
7256 @defmac DBR_OUTPUT_SEQEND (@var{file})
7257 A C statement, to be executed after all slot-filler instructions have
7258 been output. If necessary, call @code{dbr_sequence_length} to
7259 determine the number of slots filled in a sequence (zero if not
7260 currently outputting a sequence), to decide how many no-ops to output,
7263 Don't define this macro if it has nothing to do, but it is helpful in
7264 reading assembly output if the extent of the delay sequence is made
7265 explicit (e.g.@: with white space).
7268 @findex final_sequence
7269 Note that output routines for instructions with delay slots must be
7270 prepared to deal with not being output as part of a sequence
7271 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7272 found.) The variable @code{final_sequence} is null when not
7273 processing a sequence, otherwise it contains the @code{sequence} rtx
7277 @defmac REGISTER_PREFIX
7278 @defmacx LOCAL_LABEL_PREFIX
7279 @defmacx USER_LABEL_PREFIX
7280 @defmacx IMMEDIATE_PREFIX
7281 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7282 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7283 @file{final.c}). These are useful when a single @file{md} file must
7284 support multiple assembler formats. In that case, the various @file{tm.h}
7285 files can define these macros differently.
7288 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
7289 If defined this macro should expand to a series of @code{case}
7290 statements which will be parsed inside the @code{switch} statement of
7291 the @code{asm_fprintf} function. This allows targets to define extra
7292 printf formats which may useful when generating their assembler
7293 statements. Note that uppercase letters are reserved for future
7294 generic extensions to asm_fprintf, and so are not available to target
7295 specific code. The output file is given by the parameter @var{file}.
7296 The varargs input pointer is @var{argptr} and the rest of the format
7297 string, starting the character after the one that is being switched
7298 upon, is pointed to by @var{format}.
7301 @defmac ASSEMBLER_DIALECT
7302 If your target supports multiple dialects of assembler language (such as
7303 different opcodes), define this macro as a C expression that gives the
7304 numeric index of the assembler language dialect to use, with zero as the
7307 If this macro is defined, you may use constructs of the form
7309 @samp{@{option0|option1|option2@dots{}@}}
7312 in the output templates of patterns (@pxref{Output Template}) or in the
7313 first argument of @code{asm_fprintf}. This construct outputs
7314 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7315 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7316 within these strings retain their usual meaning. If there are fewer
7317 alternatives within the braces than the value of
7318 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7320 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7321 @samp{@}} do not have any special meaning when used in templates or
7322 operands to @code{asm_fprintf}.
7324 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7325 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7326 the variations in assembler language syntax with that mechanism. Define
7327 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7328 if the syntax variant are larger and involve such things as different
7329 opcodes or operand order.
7332 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7333 A C expression to output to @var{stream} some assembler code
7334 which will push hard register number @var{regno} onto the stack.
7335 The code need not be optimal, since this macro is used only when
7339 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7340 A C expression to output to @var{stream} some assembler code
7341 which will pop hard register number @var{regno} off of the stack.
7342 The code need not be optimal, since this macro is used only when
7346 @node Dispatch Tables
7347 @subsection Output of Dispatch Tables
7349 @c prevent bad page break with this line
7350 This concerns dispatch tables.
7352 @cindex dispatch table
7353 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7354 A C statement to output to the stdio stream @var{stream} an assembler
7355 pseudo-instruction to generate a difference between two labels.
7356 @var{value} and @var{rel} are the numbers of two internal labels. The
7357 definitions of these labels are output using
7358 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7359 way here. For example,
7362 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7363 @var{value}, @var{rel})
7366 You must provide this macro on machines where the addresses in a
7367 dispatch table are relative to the table's own address. If defined, GCC
7368 will also use this macro on all machines when producing PIC@.
7369 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7370 mode and flags can be read.
7373 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7374 This macro should be provided on machines where the addresses
7375 in a dispatch table are absolute.
7377 The definition should be a C statement to output to the stdio stream
7378 @var{stream} an assembler pseudo-instruction to generate a reference to
7379 a label. @var{value} is the number of an internal label whose
7380 definition is output using @code{(*targetm.asm_out.internal_label)}.
7384 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7388 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7389 Define this if the label before a jump-table needs to be output
7390 specially. The first three arguments are the same as for
7391 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7392 jump-table which follows (a @code{jump_insn} containing an
7393 @code{addr_vec} or @code{addr_diff_vec}).
7395 This feature is used on system V to output a @code{swbeg} statement
7398 If this macro is not defined, these labels are output with
7399 @code{(*targetm.asm_out.internal_label)}.
7402 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7403 Define this if something special must be output at the end of a
7404 jump-table. The definition should be a C statement to be executed
7405 after the assembler code for the table is written. It should write
7406 the appropriate code to stdio stream @var{stream}. The argument
7407 @var{table} is the jump-table insn, and @var{num} is the label-number
7408 of the preceding label.
7410 If this macro is not defined, nothing special is output at the end of
7414 @node Exception Region Output
7415 @subsection Assembler Commands for Exception Regions
7417 @c prevent bad page break with this line
7419 This describes commands marking the start and the end of an exception
7422 @defmac EH_FRAME_SECTION_NAME
7423 If defined, a C string constant for the name of the section containing
7424 exception handling frame unwind information. If not defined, GCC will
7425 provide a default definition if the target supports named sections.
7426 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7428 You should define this symbol if your target supports DWARF 2 frame
7429 unwind information and the default definition does not work.
7432 @defmac EH_FRAME_IN_DATA_SECTION
7433 If defined, DWARF 2 frame unwind information will be placed in the
7434 data section even though the target supports named sections. This
7435 might be necessary, for instance, if the system linker does garbage
7436 collection and sections cannot be marked as not to be collected.
7438 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7442 @defmac MASK_RETURN_ADDR
7443 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7444 that it does not contain any extraneous set bits in it.
7447 @defmac DWARF2_UNWIND_INFO
7448 Define this macro to 0 if your target supports DWARF 2 frame unwind
7449 information, but it does not yet work with exception handling.
7450 Otherwise, if your target supports this information (if it defines
7451 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7452 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7455 If this macro is defined to 1, the DWARF 2 unwinder will be the default
7456 exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by
7459 If this macro is defined to anything, the DWARF 2 unwinder will be used
7460 instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case.
7463 @defmac MUST_USE_SJLJ_EXCEPTIONS
7464 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
7465 runtime-variable. In that case, @file{except.h} cannot correctly
7466 determine the corresponding definition of
7467 @code{MUST_USE_SJLJ_EXCEPTIONS}, so the target must provide it directly.
7470 @defmac DWARF_CIE_DATA_ALIGNMENT
7471 This macro need only be defined if the target might save registers in the
7472 function prologue at an offset to the stack pointer that is not aligned to
7473 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7474 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7475 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7476 the target supports DWARF 2 frame unwind information.
7479 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7480 If defined, a function that switches to the section in which the main
7481 exception table is to be placed (@pxref{Sections}). The default is a
7482 function that switches to a section named @code{.gcc_except_table} on
7483 machines that support named sections via
7484 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7485 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7486 @code{readonly_data_section}.
7489 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7490 If defined, a function that switches to the section in which the DWARF 2
7491 frame unwind information to be placed (@pxref{Sections}). The default
7492 is a function that outputs a standard GAS section directive, if
7493 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7494 directive followed by a synthetic label.
7497 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7498 Contains the value true if the target should add a zero word onto the
7499 end of a Dwarf-2 frame info section when used for exception handling.
7500 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7504 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
7505 Given a register, this hook should return a parallel of registers to
7506 represent where to find the register pieces. Define this hook if the
7507 register and its mode are represented in Dwarf in non-contiguous
7508 locations, or if the register should be represented in more than one
7509 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
7510 If not defined, the default is to return @code{NULL_RTX}.
7513 @node Alignment Output
7514 @subsection Assembler Commands for Alignment
7516 @c prevent bad page break with this line
7517 This describes commands for alignment.
7519 @defmac JUMP_ALIGN (@var{label})
7520 The alignment (log base 2) to put in front of @var{label}, which is
7521 a common destination of jumps and has no fallthru incoming edge.
7523 This macro need not be defined if you don't want any special alignment
7524 to be done at such a time. Most machine descriptions do not currently
7527 Unless it's necessary to inspect the @var{label} parameter, it is better
7528 to set the variable @var{align_jumps} in the target's
7529 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7530 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7533 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
7534 The alignment (log base 2) to put in front of @var{label}, which follows
7537 This macro need not be defined if you don't want any special alignment
7538 to be done at such a time. Most machine descriptions do not currently
7542 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7543 The maximum number of bytes to skip when applying
7544 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7545 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7548 @defmac LOOP_ALIGN (@var{label})
7549 The alignment (log base 2) to put in front of @var{label}, which follows
7550 a @code{NOTE_INSN_LOOP_BEG} note.
7552 This macro need not be defined if you don't want any special alignment
7553 to be done at such a time. Most machine descriptions do not currently
7556 Unless it's necessary to inspect the @var{label} parameter, it is better
7557 to set the variable @code{align_loops} in the target's
7558 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7559 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7562 @defmac LOOP_ALIGN_MAX_SKIP
7563 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7564 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7567 @defmac LABEL_ALIGN (@var{label})
7568 The alignment (log base 2) to put in front of @var{label}.
7569 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7570 the maximum of the specified values is used.
7572 Unless it's necessary to inspect the @var{label} parameter, it is better
7573 to set the variable @code{align_labels} in the target's
7574 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7575 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7578 @defmac LABEL_ALIGN_MAX_SKIP
7579 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7580 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7583 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7584 A C statement to output to the stdio stream @var{stream} an assembler
7585 instruction to advance the location counter by @var{nbytes} bytes.
7586 Those bytes should be zero when loaded. @var{nbytes} will be a C
7587 expression of type @code{int}.
7590 @defmac ASM_NO_SKIP_IN_TEXT
7591 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7592 text section because it fails to put zeros in the bytes that are skipped.
7593 This is true on many Unix systems, where the pseudo--op to skip bytes
7594 produces no-op instructions rather than zeros when used in the text
7598 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7599 A C statement to output to the stdio stream @var{stream} an assembler
7600 command to advance the location counter to a multiple of 2 to the
7601 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7604 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7605 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7606 for padding, if necessary.
7609 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7610 A C statement to output to the stdio stream @var{stream} an assembler
7611 command to advance the location counter to a multiple of 2 to the
7612 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7613 satisfy the alignment request. @var{power} and @var{max_skip} will be
7614 a C expression of type @code{int}.
7618 @node Debugging Info
7619 @section Controlling Debugging Information Format
7621 @c prevent bad page break with this line
7622 This describes how to specify debugging information.
7625 * All Debuggers:: Macros that affect all debugging formats uniformly.
7626 * DBX Options:: Macros enabling specific options in DBX format.
7627 * DBX Hooks:: Hook macros for varying DBX format.
7628 * File Names and DBX:: Macros controlling output of file names in DBX format.
7629 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7630 * VMS Debug:: Macros for VMS debug format.
7634 @subsection Macros Affecting All Debugging Formats
7636 @c prevent bad page break with this line
7637 These macros affect all debugging formats.
7639 @defmac DBX_REGISTER_NUMBER (@var{regno})
7640 A C expression that returns the DBX register number for the compiler
7641 register number @var{regno}. In the default macro provided, the value
7642 of this expression will be @var{regno} itself. But sometimes there are
7643 some registers that the compiler knows about and DBX does not, or vice
7644 versa. In such cases, some register may need to have one number in the
7645 compiler and another for DBX@.
7647 If two registers have consecutive numbers inside GCC, and they can be
7648 used as a pair to hold a multiword value, then they @emph{must} have
7649 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7650 Otherwise, debuggers will be unable to access such a pair, because they
7651 expect register pairs to be consecutive in their own numbering scheme.
7653 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7654 does not preserve register pairs, then what you must do instead is
7655 redefine the actual register numbering scheme.
7658 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
7659 A C expression that returns the integer offset value for an automatic
7660 variable having address @var{x} (an RTL expression). The default
7661 computation assumes that @var{x} is based on the frame-pointer and
7662 gives the offset from the frame-pointer. This is required for targets
7663 that produce debugging output for DBX or COFF-style debugging output
7664 for SDB and allow the frame-pointer to be eliminated when the
7665 @option{-g} options is used.
7668 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7669 A C expression that returns the integer offset value for an argument
7670 having address @var{x} (an RTL expression). The nominal offset is
7674 @defmac PREFERRED_DEBUGGING_TYPE
7675 A C expression that returns the type of debugging output GCC should
7676 produce when the user specifies just @option{-g}. Define
7677 this if you have arranged for GCC to support more than one format of
7678 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7679 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7680 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7682 When the user specifies @option{-ggdb}, GCC normally also uses the
7683 value of this macro to select the debugging output format, but with two
7684 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7685 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7686 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7687 defined, GCC uses @code{DBX_DEBUG}.
7689 The value of this macro only affects the default debugging output; the
7690 user can always get a specific type of output by using @option{-gstabs},
7691 @option{-gcoff}, @option{-gdwarf-1}, @option{-gdwarf-2}, @option{-gxcoff},
7696 @subsection Specific Options for DBX Output
7698 @c prevent bad page break with this line
7699 These are specific options for DBX output.
7701 @defmac DBX_DEBUGGING_INFO
7702 Define this macro if GCC should produce debugging output for DBX
7703 in response to the @option{-g} option.
7706 @defmac XCOFF_DEBUGGING_INFO
7707 Define this macro if GCC should produce XCOFF format debugging output
7708 in response to the @option{-g} option. This is a variant of DBX format.
7711 @defmac DEFAULT_GDB_EXTENSIONS
7712 Define this macro to control whether GCC should by default generate
7713 GDB's extended version of DBX debugging information (assuming DBX-format
7714 debugging information is enabled at all). If you don't define the
7715 macro, the default is 1: always generate the extended information
7716 if there is any occasion to.
7719 @defmac DEBUG_SYMS_TEXT
7720 Define this macro if all @code{.stabs} commands should be output while
7721 in the text section.
7724 @defmac ASM_STABS_OP
7725 A C string constant, including spacing, naming the assembler pseudo op to
7726 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7727 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7728 applies only to DBX debugging information format.
7731 @defmac ASM_STABD_OP
7732 A C string constant, including spacing, naming the assembler pseudo op to
7733 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7734 value is the current location. If you don't define this macro,
7735 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7739 @defmac ASM_STABN_OP
7740 A C string constant, including spacing, naming the assembler pseudo op to
7741 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7742 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7743 macro applies only to DBX debugging information format.
7746 @defmac DBX_NO_XREFS
7747 Define this macro if DBX on your system does not support the construct
7748 @samp{xs@var{tagname}}. On some systems, this construct is used to
7749 describe a forward reference to a structure named @var{tagname}.
7750 On other systems, this construct is not supported at all.
7753 @defmac DBX_CONTIN_LENGTH
7754 A symbol name in DBX-format debugging information is normally
7755 continued (split into two separate @code{.stabs} directives) when it
7756 exceeds a certain length (by default, 80 characters). On some
7757 operating systems, DBX requires this splitting; on others, splitting
7758 must not be done. You can inhibit splitting by defining this macro
7759 with the value zero. You can override the default splitting-length by
7760 defining this macro as an expression for the length you desire.
7763 @defmac DBX_CONTIN_CHAR
7764 Normally continuation is indicated by adding a @samp{\} character to
7765 the end of a @code{.stabs} string when a continuation follows. To use
7766 a different character instead, define this macro as a character
7767 constant for the character you want to use. Do not define this macro
7768 if backslash is correct for your system.
7771 @defmac DBX_STATIC_STAB_DATA_SECTION
7772 Define this macro if it is necessary to go to the data section before
7773 outputting the @samp{.stabs} pseudo-op for a non-global static
7777 @defmac DBX_TYPE_DECL_STABS_CODE
7778 The value to use in the ``code'' field of the @code{.stabs} directive
7779 for a typedef. The default is @code{N_LSYM}.
7782 @defmac DBX_STATIC_CONST_VAR_CODE
7783 The value to use in the ``code'' field of the @code{.stabs} directive
7784 for a static variable located in the text section. DBX format does not
7785 provide any ``right'' way to do this. The default is @code{N_FUN}.
7788 @defmac DBX_REGPARM_STABS_CODE
7789 The value to use in the ``code'' field of the @code{.stabs} directive
7790 for a parameter passed in registers. DBX format does not provide any
7791 ``right'' way to do this. The default is @code{N_RSYM}.
7794 @defmac DBX_REGPARM_STABS_LETTER
7795 The letter to use in DBX symbol data to identify a symbol as a parameter
7796 passed in registers. DBX format does not customarily provide any way to
7797 do this. The default is @code{'P'}.
7800 @defmac DBX_MEMPARM_STABS_LETTER
7801 The letter to use in DBX symbol data to identify a symbol as a stack
7802 parameter. The default is @code{'p'}.
7805 @defmac DBX_FUNCTION_FIRST
7806 Define this macro if the DBX information for a function and its
7807 arguments should precede the assembler code for the function. Normally,
7808 in DBX format, the debugging information entirely follows the assembler
7812 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
7813 Define this macro if the value of a symbol describing the scope of a
7814 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7815 of the enclosing function. Normally, GCC uses an absolute address.
7818 @defmac DBX_USE_BINCL
7819 Define this macro if GCC should generate @code{N_BINCL} and
7820 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7821 macro also directs GCC to output a type number as a pair of a file
7822 number and a type number within the file. Normally, GCC does not
7823 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7824 number for a type number.
7828 @subsection Open-Ended Hooks for DBX Format
7830 @c prevent bad page break with this line
7831 These are hooks for DBX format.
7833 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7834 Define this macro to say how to output to @var{stream} the debugging
7835 information for the start of a scope level for variable names. The
7836 argument @var{name} is the name of an assembler symbol (for use with
7837 @code{assemble_name}) whose value is the address where the scope begins.
7840 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7841 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7844 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
7845 Define this macro if the target machine requires special handling to
7846 output an @code{N_FUN} entry for the function @var{decl}.
7849 @defmac DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7850 Define this macro if the target machine requires special output at the
7851 end of the debugging information for a function. The definition should
7852 be a C statement (sans semicolon) to output the appropriate information
7853 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7857 @defmac DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7858 Define this macro if you need to control the order of output of the
7859 standard data types at the beginning of compilation. The argument
7860 @var{syms} is a @code{tree} which is a chain of all the predefined
7861 global symbols, including names of data types.
7863 Normally, DBX output starts with definitions of the types for integers
7864 and characters, followed by all the other predefined types of the
7865 particular language in no particular order.
7867 On some machines, it is necessary to output different particular types
7868 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
7869 those symbols in the necessary order. Any predefined types that you
7870 don't explicitly output will be output afterward in no particular order.
7872 Be careful not to define this macro so that it works only for C@. There
7873 are no global variables to access most of the built-in types, because
7874 another language may have another set of types. The way to output a
7875 particular type is to look through @var{syms} to see if you can find it.
7881 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7882 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
7884 dbxout_symbol (decl);
7890 This does nothing if the expected type does not exist.
7892 See the function @code{init_decl_processing} in @file{c-decl.c} to find
7893 the names to use for all the built-in C types.
7895 Here is another way of finding a particular type:
7897 @c this is still overfull. --mew 10feb93
7901 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7902 if (TREE_CODE (decl) == TYPE_DECL
7903 && (TREE_CODE (TREE_TYPE (decl))
7905 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
7906 && TYPE_UNSIGNED (TREE_TYPE (decl)))
7908 /* @r{This must be @code{unsigned short}.} */
7909 dbxout_symbol (decl);
7916 @defmac NO_DBX_FUNCTION_END
7917 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7918 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
7919 On those machines, define this macro to turn this feature off without
7920 disturbing the rest of the gdb extensions.
7923 @node File Names and DBX
7924 @subsection File Names in DBX Format
7926 @c prevent bad page break with this line
7927 This describes file names in DBX format.
7929 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7930 A C statement to output DBX debugging information to the stdio stream
7931 @var{stream} which indicates that file @var{name} is the main source
7932 file---the file specified as the input file for compilation.
7933 This macro is called only once, at the beginning of compilation.
7935 This macro need not be defined if the standard form of output
7936 for DBX debugging information is appropriate.
7939 @defmac DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
7940 A C statement to output DBX debugging information to the stdio stream
7941 @var{stream} which indicates that the current directory during
7942 compilation is named @var{name}.
7944 This macro need not be defined if the standard form of output
7945 for DBX debugging information is appropriate.
7948 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7949 A C statement to output DBX debugging information at the end of
7950 compilation of the main source file @var{name}.
7952 If you don't define this macro, nothing special is output at the end
7953 of compilation, which is correct for most machines.
7958 @subsection Macros for SDB and DWARF Output
7960 @c prevent bad page break with this line
7961 Here are macros for SDB and DWARF output.
7963 @defmac SDB_DEBUGGING_INFO
7964 Define this macro if GCC should produce COFF-style debugging output
7965 for SDB in response to the @option{-g} option.
7968 @defmac DWARF_DEBUGGING_INFO
7969 Define this macro if GCC should produce dwarf format debugging output
7970 in response to the @option{-g} option.
7973 @defmac DWARF2_DEBUGGING_INFO
7974 Define this macro if GCC should produce dwarf version 2 format
7975 debugging output in response to the @option{-g} option.
7977 To support optional call frame debugging information, you must also
7978 define @code{INCOMING_RETURN_ADDR_RTX} and either set
7979 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
7980 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
7981 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
7984 @defmac DWARF2_FRAME_INFO
7985 Define this macro to a nonzero value if GCC should always output
7986 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
7987 (@pxref{Exception Region Output} is nonzero, GCC will output this
7988 information not matter how you define @code{DWARF2_FRAME_INFO}.
7991 @defmac LINKER_DOES_NOT_WORK_WITH_DWARF2
7992 Define this macro if the linker does not work with Dwarf version 2.
7993 Normally, if the user specifies only @option{-ggdb} GCC will use Dwarf
7994 version 2 if available; this macro disables this. See the description
7995 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
7998 @defmac DWARF2_GENERATE_TEXT_SECTION_LABEL
7999 By default, the Dwarf 2 debugging information generator will generate a
8000 label to mark the beginning of the text section. If it is better simply
8001 to use the name of the text section itself, rather than an explicit label,
8002 to indicate the beginning of the text section, define this macro to zero.
8005 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8006 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8007 line debug info sections. This will result in much more compact line number
8008 tables, and hence is desirable if it works.
8011 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8012 A C statement to issue assembly directives that create a difference
8013 between the two given labels, using an integer of the given size.
8016 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8017 A C statement to issue assembly directives that create a
8018 section-relative reference to the given label, using an integer of the
8022 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8023 A C statement to issue assembly directives that create a self-relative
8024 reference to the given label, using an integer of the given size.
8027 @defmac PUT_SDB_@dots{}
8028 Define these macros to override the assembler syntax for the special
8029 SDB assembler directives. See @file{sdbout.c} for a list of these
8030 macros and their arguments. If the standard syntax is used, you need
8031 not define them yourself.
8035 Some assemblers do not support a semicolon as a delimiter, even between
8036 SDB assembler directives. In that case, define this macro to be the
8037 delimiter to use (usually @samp{\n}). It is not necessary to define
8038 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8042 @defmac SDB_GENERATE_FAKE
8043 Define this macro to override the usual method of constructing a dummy
8044 name for anonymous structure and union types. See @file{sdbout.c} for
8048 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8049 Define this macro to allow references to unknown structure,
8050 union, or enumeration tags to be emitted. Standard COFF does not
8051 allow handling of unknown references, MIPS ECOFF has support for
8055 @defmac SDB_ALLOW_FORWARD_REFERENCES
8056 Define this macro to allow references to structure, union, or
8057 enumeration tags that have not yet been seen to be handled. Some
8058 assemblers choke if forward tags are used, while some require it.
8063 @subsection Macros for VMS Debug Format
8065 @c prevent bad page break with this line
8066 Here are macros for VMS debug format.
8068 @defmac VMS_DEBUGGING_INFO
8069 Define this macro if GCC should produce debugging output for VMS
8070 in response to the @option{-g} option. The default behavior for VMS
8071 is to generate minimal debug info for a traceback in the absence of
8072 @option{-g} unless explicitly overridden with @option{-g0}. This
8073 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8074 @code{OVERRIDE_OPTIONS}.
8077 @node Floating Point
8078 @section Cross Compilation and Floating Point
8079 @cindex cross compilation and floating point
8080 @cindex floating point and cross compilation
8082 While all modern machines use twos-complement representation for integers,
8083 there are a variety of representations for floating point numbers. This
8084 means that in a cross-compiler the representation of floating point numbers
8085 in the compiled program may be different from that used in the machine
8086 doing the compilation.
8088 Because different representation systems may offer different amounts of
8089 range and precision, all floating point constants must be represented in
8090 the target machine's format. Therefore, the cross compiler cannot
8091 safely use the host machine's floating point arithmetic; it must emulate
8092 the target's arithmetic. To ensure consistency, GCC always uses
8093 emulation to work with floating point values, even when the host and
8094 target floating point formats are identical.
8096 The following macros are provided by @file{real.h} for the compiler to
8097 use. All parts of the compiler which generate or optimize
8098 floating-point calculations must use these macros. They may evaluate
8099 their operands more than once, so operands must not have side effects.
8101 @defmac REAL_VALUE_TYPE
8102 The C data type to be used to hold a floating point value in the target
8103 machine's format. Typically this is a @code{struct} containing an
8104 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8108 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8109 Compares for equality the two values, @var{x} and @var{y}. If the target
8110 floating point format supports negative zeroes and/or NaNs,
8111 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8112 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8115 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8116 Tests whether @var{x} is less than @var{y}.
8119 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8120 Truncates @var{x} to a signed integer, rounding toward zero.
8123 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8124 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8125 @var{x} is negative, returns zero.
8128 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8129 Converts @var{string} into a floating point number in the target machine's
8130 representation for mode @var{mode}. This routine can handle both
8131 decimal and hexadecimal floating point constants, using the syntax
8132 defined by the C language for both.
8135 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8136 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8139 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8140 Determines whether @var{x} represents infinity (positive or negative).
8143 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8144 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8147 @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})
8148 Calculates an arithmetic operation on the two floating point values
8149 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8152 The operation to be performed is specified by @var{code}. Only the
8153 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8154 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8156 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8157 target's floating point format cannot represent infinity, it will call
8158 @code{abort}. Callers should check for this situation first, using
8159 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8162 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8163 Returns the negative of the floating point value @var{x}.
8166 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8167 Returns the absolute value of @var{x}.
8170 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8171 Truncates the floating point value @var{x} to fit in @var{mode}. The
8172 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8173 appropriate bit pattern to be output asa floating constant whose
8174 precision accords with mode @var{mode}.
8177 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8178 Converts a floating point value @var{x} into a double-precision integer
8179 which is then stored into @var{low} and @var{high}. If the value is not
8180 integral, it is truncated.
8183 @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})
8184 Converts a double-precision integer found in @var{low} and @var{high},
8185 into a floating point value which is then stored into @var{x}. The
8186 value is truncated to fit in mode @var{mode}.
8189 @node Mode Switching
8190 @section Mode Switching Instructions
8191 @cindex mode switching
8192 The following macros control mode switching optimizations:
8194 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8195 Define this macro if the port needs extra instructions inserted for mode
8196 switching in an optimizing compilation.
8198 For an example, the SH4 can perform both single and double precision
8199 floating point operations, but to perform a single precision operation,
8200 the FPSCR PR bit has to be cleared, while for a double precision
8201 operation, this bit has to be set. Changing the PR bit requires a general
8202 purpose register as a scratch register, hence these FPSCR sets have to
8203 be inserted before reload, i.e.@: you can't put this into instruction emitting
8204 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
8206 You can have multiple entities that are mode-switched, and select at run time
8207 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8208 return nonzero for any @var{entity} that needs mode-switching.
8209 If you define this macro, you also have to define
8210 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8211 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8212 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
8216 @defmac NUM_MODES_FOR_MODE_SWITCHING
8217 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8218 initializer for an array of integers. Each initializer element
8219 N refers to an entity that needs mode switching, and specifies the number
8220 of different modes that might need to be set for this entity.
8221 The position of the initializer in the initializer - starting counting at
8222 zero - determines the integer that is used to refer to the mode-switched
8224 In macros that take mode arguments / yield a mode result, modes are
8225 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8226 switch is needed / supplied.
8229 @defmac MODE_NEEDED (@var{entity}, @var{insn})
8230 @var{entity} is an integer specifying a mode-switched entity. If
8231 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8232 return an integer value not larger than the corresponding element in
8233 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8234 be switched into prior to the execution of @var{insn}.
8237 @defmac MODE_AFTER (@var{mode}, @var{insn})
8238 If this macro is defined, it is evaluated for every @var{insn} during
8239 mode switching. It determines the mode that an insn results in (if
8240 different from the incoming mode).
8243 @defmac MODE_ENTRY (@var{entity})
8244 If this macro is defined, it is evaluated for every @var{entity} that needs
8245 mode switching. It should evaluate to an integer, which is a mode that
8246 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
8247 is defined then @code{MODE_EXIT} must be defined.
8250 @defmac MODE_EXIT (@var{entity})
8251 If this macro is defined, it is evaluated for every @var{entity} that needs
8252 mode switching. It should evaluate to an integer, which is a mode that
8253 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
8254 is defined then @code{MODE_ENTRY} must be defined.
8257 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8258 This macro specifies the order in which modes for @var{entity} are processed.
8259 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8260 lowest. The value of the macro should be an integer designating a mode
8261 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8262 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8263 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8266 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8267 Generate one or more insns to set @var{entity} to @var{mode}.
8268 @var{hard_reg_live} is the set of hard registers live at the point where
8269 the insn(s) are to be inserted.
8272 @node Target Attributes
8273 @section Defining target-specific uses of @code{__attribute__}
8274 @cindex target attributes
8275 @cindex machine attributes
8276 @cindex attributes, target-specific
8278 Target-specific attributes may be defined for functions, data and types.
8279 These are described using the following target hooks; they also need to
8280 be documented in @file{extend.texi}.
8282 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8283 If defined, this target hook points to an array of @samp{struct
8284 attribute_spec} (defined in @file{tree.h}) specifying the machine
8285 specific attributes for this target and some of the restrictions on the
8286 entities to which these attributes are applied and the arguments they
8290 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8291 If defined, this target hook is a function which returns zero if the attributes on
8292 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8293 and two if they are nearly compatible (which causes a warning to be
8294 generated). If this is not defined, machine-specific attributes are
8295 supposed always to be compatible.
8298 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8299 If defined, this target hook is a function which assigns default attributes to
8300 newly defined @var{type}.
8303 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8304 Define this target hook if the merging of type attributes needs special
8305 handling. If defined, the result is a list of the combined
8306 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8307 that @code{comptypes} has already been called and returned 1. This
8308 function may call @code{merge_attributes} to handle machine-independent
8312 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8313 Define this target hook if the merging of decl attributes needs special
8314 handling. If defined, the result is a list of the combined
8315 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8316 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8317 when this is needed are when one attribute overrides another, or when an
8318 attribute is nullified by a subsequent definition. This function may
8319 call @code{merge_attributes} to handle machine-independent merging.
8321 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8322 If the only target-specific handling you require is @samp{dllimport} for
8323 Windows targets, you should define the macro
8324 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function
8325 called @code{merge_dllimport_decl_attributes} which can then be defined
8326 as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done
8327 in @file{i386/cygwin.h} and @file{i386/i386.c}, for example.
8330 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8331 Define this target hook if you want to be able to add attributes to a decl
8332 when it is being created. This is normally useful for back ends which
8333 wish to implement a pragma by using the attributes which correspond to
8334 the pragma's effect. The @var{node} argument is the decl which is being
8335 created. The @var{attr_ptr} argument is a pointer to the attribute list
8336 for this decl. The list itself should not be modified, since it may be
8337 shared with other decls, but attributes may be chained on the head of
8338 the list and @code{*@var{attr_ptr}} modified to point to the new
8339 attributes, or a copy of the list may be made if further changes are
8343 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8345 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8346 into the current function, despite its having target-specific
8347 attributes, @code{false} otherwise. By default, if a function has a
8348 target specific attribute attached to it, it will not be inlined.
8351 @node MIPS Coprocessors
8352 @section Defining coprocessor specifics for MIPS targets.
8353 @cindex MIPS coprocessor-definition macros
8355 The MIPS specification allows MIPS implementations to have as many as 4
8356 coprocessors, each with as many as 32 private registers. gcc supports
8357 accessing these registers and transferring values between the registers
8358 and memory using asm-ized variables. For example:
8361 register unsigned int cp0count asm ("c0r1");
8367 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8368 names may be added as described below, or the default names may be
8369 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8371 Coprocessor registers are assumed to be epilogue-used; sets to them will
8372 be preserved even if it does not appear that the register is used again
8373 later in the function.
8375 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8376 the FPU. One accesses COP1 registers through standard mips
8377 floating-point support; they are not included in this mechanism.
8379 There is one macro used in defining the MIPS coprocessor interface which
8380 you may want to override in subtargets; it is described below.
8382 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
8383 A comma-separated list (with leading comma) of pairs describing the
8384 alternate names of coprocessor registers. The format of each entry should be
8386 @{ @var{alternatename}, @var{register_number}@}
8392 @section Parameters for Precompiled Header Validity Checking
8393 @cindex parameters, precompiled headers
8395 @deftypefn {Target Hook} void * TARGET_GET_PCH_VALIDITY (size_t * @var{sz})
8396 Define this hook if your target needs to check a different collection
8397 of flags than the default, which is every flag defined by
8398 @code{TARGET_SWITCHES} and @code{TARGET_OPTIONS}. It should return
8399 some data which will be saved in the PCH file and presented to
8400 @code{TARGET_PCH_VALID_P} later; it should set @code{SZ} to the size
8404 @deftypefn {Target Hook} const char * TARGET_PCH_VALID_P (const void * @var{data}, size_t @var{sz})
8405 Define this hook if your target needs to check a different collection of
8406 flags than the default, which is every flag defined by @code{TARGET_SWITCHES}
8407 and @code{TARGET_OPTIONS}. It is given data which came from
8408 @code{TARGET_GET_PCH_VALIDITY} (in this version of this compiler, so there
8409 is no need for extensive validity checking). It returns @code{NULL} if
8410 it is safe to load a PCH file with this data, or a suitable error message
8411 if not. The error message will be presented to the user, so it should
8416 @section Miscellaneous Parameters
8417 @cindex parameters, miscellaneous
8419 @c prevent bad page break with this line
8420 Here are several miscellaneous parameters.
8422 @defmac PREDICATE_CODES
8423 Define this if you have defined special-purpose predicates in the file
8424 @file{@var{machine}.c}. This macro is called within an initializer of an
8425 array of structures. The first field in the structure is the name of a
8426 predicate and the second field is an array of rtl codes. For each
8427 predicate, list all rtl codes that can be in expressions matched by the
8428 predicate. The list should have a trailing comma. Here is an example
8429 of two entries in the list for a typical RISC machine:
8432 #define PREDICATE_CODES \
8433 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8434 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8437 Defining this macro does not affect the generated code (however,
8438 incorrect definitions that omit an rtl code that may be matched by the
8439 predicate can cause the compiler to malfunction). Instead, it allows
8440 the table built by @file{genrecog} to be more compact and efficient,
8441 thus speeding up the compiler. The most important predicates to include
8442 in the list specified by this macro are those used in the most insn
8445 For each predicate function named in @code{PREDICATE_CODES}, a
8446 declaration will be generated in @file{insn-codes.h}.
8449 @defmac SPECIAL_MODE_PREDICATES
8450 Define this if you have special predicates that know special things
8451 about modes. Genrecog will warn about certain forms of
8452 @code{match_operand} without a mode; if the operand predicate is
8453 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8456 Here is an example from the IA-32 port (@code{ext_register_operand}
8457 specially checks for @code{HImode} or @code{SImode} in preparation
8458 for a byte extraction from @code{%ah} etc.).
8461 #define SPECIAL_MODE_PREDICATES \
8462 "ext_register_operand",
8466 @defmac CASE_VECTOR_MODE
8467 An alias for a machine mode name. This is the machine mode that
8468 elements of a jump-table should have.
8471 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8472 Optional: return the preferred mode for an @code{addr_diff_vec}
8473 when the minimum and maximum offset are known. If you define this,
8474 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8475 To make this work, you also have to define @code{INSN_ALIGN} and
8476 make the alignment for @code{addr_diff_vec} explicit.
8477 The @var{body} argument is provided so that the offset_unsigned and scale
8478 flags can be updated.
8481 @defmac CASE_VECTOR_PC_RELATIVE
8482 Define this macro to be a C expression to indicate when jump-tables
8483 should contain relative addresses. If jump-tables never contain
8484 relative addresses, then you need not define this macro.
8487 @defmac CASE_DROPS_THROUGH
8488 Define this if control falls through a @code{case} insn when the index
8489 value is out of range. This means the specified default-label is
8490 actually ignored by the @code{case} insn proper.
8493 @defmac CASE_VALUES_THRESHOLD
8494 Define this to be the smallest number of different values for which it
8495 is best to use a jump-table instead of a tree of conditional branches.
8496 The default is four for machines with a @code{casesi} instruction and
8497 five otherwise. This is best for most machines.
8500 @defmac CASE_USE_BIT_TESTS
8501 Define this macro to be a C expression to indicate whether C switch
8502 statements may be implemented by a sequence of bit tests. This is
8503 advantageous on processors that can efficiently implement left shift
8504 of 1 by the number of bits held in a register, but inappropriate on
8505 targets that would require a loop. By default, this macro returns
8506 @code{true} if the target defines an @code{ashlsi3} pattern, and
8507 @code{false} otherwise.
8510 @defmac WORD_REGISTER_OPERATIONS
8511 Define this macro if operations between registers with integral mode
8512 smaller than a word are always performed on the entire register.
8513 Most RISC machines have this property and most CISC machines do not.
8516 @defmac LOAD_EXTEND_OP (@var{mode})
8517 Define this macro to be a C expression indicating when insns that read
8518 memory in @var{mode}, an integral mode narrower than a word, set the
8519 bits outside of @var{mode} to be either the sign-extension or the
8520 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8521 of @var{mode} for which the
8522 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8523 @code{NIL} for other modes.
8525 This macro is not called with @var{mode} non-integral or with a width
8526 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8527 value in this case. Do not define this macro if it would always return
8528 @code{NIL}. On machines where this macro is defined, you will normally
8529 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8532 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
8533 Define this macro if loading short immediate values into registers sign
8537 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
8538 Define this macro if the same instructions that convert a floating
8539 point number to a signed fixed point number also convert validly to an
8544 The maximum number of bytes that a single instruction can move quickly
8545 between memory and registers or between two memory locations.
8548 @defmac MAX_MOVE_MAX
8549 The maximum number of bytes that a single instruction can move quickly
8550 between memory and registers or between two memory locations. If this
8551 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8552 constant value that is the largest value that @code{MOVE_MAX} can have
8556 @defmac SHIFT_COUNT_TRUNCATED
8557 A C expression that is nonzero if on this machine the number of bits
8558 actually used for the count of a shift operation is equal to the number
8559 of bits needed to represent the size of the object being shifted. When
8560 this macro is nonzero, the compiler will assume that it is safe to omit
8561 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8562 truncates the count of a shift operation. On machines that have
8563 instructions that act on bit-fields at variable positions, which may
8564 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8565 also enables deletion of truncations of the values that serve as
8566 arguments to bit-field instructions.
8568 If both types of instructions truncate the count (for shifts) and
8569 position (for bit-field operations), or if no variable-position bit-field
8570 instructions exist, you should define this macro.
8572 However, on some machines, such as the 80386 and the 680x0, truncation
8573 only applies to shift operations and not the (real or pretended)
8574 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8575 such machines. Instead, add patterns to the @file{md} file that include
8576 the implied truncation of the shift instructions.
8578 You need not define this macro if it would always have the value of zero.
8581 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8582 A C expression which is nonzero if on this machine it is safe to
8583 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8584 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8585 operating on it as if it had only @var{outprec} bits.
8587 On many machines, this expression can be 1.
8589 @c rearranged this, removed the phrase "it is reported that". this was
8590 @c to fix an overfull hbox. --mew 10feb93
8591 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8592 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8593 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8594 such cases may improve things.
8597 @defmac STORE_FLAG_VALUE
8598 A C expression describing the value returned by a comparison operator
8599 with an integral mode and stored by a store-flag instruction
8600 (@samp{s@var{cond}}) when the condition is true. This description must
8601 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8602 comparison operators whose results have a @code{MODE_INT} mode.
8604 A value of 1 or @minus{}1 means that the instruction implementing the
8605 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8606 and 0 when the comparison is false. Otherwise, the value indicates
8607 which bits of the result are guaranteed to be 1 when the comparison is
8608 true. This value is interpreted in the mode of the comparison
8609 operation, which is given by the mode of the first operand in the
8610 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8611 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8614 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8615 generate code that depends only on the specified bits. It can also
8616 replace comparison operators with equivalent operations if they cause
8617 the required bits to be set, even if the remaining bits are undefined.
8618 For example, on a machine whose comparison operators return an
8619 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8620 @samp{0x80000000}, saying that just the sign bit is relevant, the
8624 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8631 (ashift:SI @var{x} (const_int @var{n}))
8635 where @var{n} is the appropriate shift count to move the bit being
8636 tested into the sign bit.
8638 There is no way to describe a machine that always sets the low-order bit
8639 for a true value, but does not guarantee the value of any other bits,
8640 but we do not know of any machine that has such an instruction. If you
8641 are trying to port GCC to such a machine, include an instruction to
8642 perform a logical-and of the result with 1 in the pattern for the
8643 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8645 Often, a machine will have multiple instructions that obtain a value
8646 from a comparison (or the condition codes). Here are rules to guide the
8647 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8652 Use the shortest sequence that yields a valid definition for
8653 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8654 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8655 comparison operators to do so because there may be opportunities to
8656 combine the normalization with other operations.
8659 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8660 slightly preferred on machines with expensive jumps and 1 preferred on
8664 As a second choice, choose a value of @samp{0x80000001} if instructions
8665 exist that set both the sign and low-order bits but do not define the
8669 Otherwise, use a value of @samp{0x80000000}.
8672 Many machines can produce both the value chosen for
8673 @code{STORE_FLAG_VALUE} and its negation in the same number of
8674 instructions. On those machines, you should also define a pattern for
8675 those cases, e.g., one matching
8678 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8681 Some machines can also perform @code{and} or @code{plus} operations on
8682 condition code values with less instructions than the corresponding
8683 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8684 machines, define the appropriate patterns. Use the names @code{incscc}
8685 and @code{decscc}, respectively, for the patterns which perform
8686 @code{plus} or @code{minus} operations on condition code values. See
8687 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8688 find such instruction sequences on other machines.
8690 If this macro is not defined, the default value, 1, is used. You need
8691 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8692 instructions, or if the value generated by these instructions is 1.
8695 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
8696 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8697 returned when comparison operators with floating-point results are true.
8698 Define this macro on machine that have comparison operations that return
8699 floating-point values. If there are no such operations, do not define
8703 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8704 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8705 A C expression that evaluates to true if the architecture defines a value
8706 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
8707 should be set to this value. If this macro is not defined, the value of
8708 @code{clz} or @code{ctz} is assumed to be undefined.
8710 This macro must be defined if the target's expansion for @code{ffs}
8711 relies on a particular value to get correct results. Otherwise it
8712 is not necessary, though it may be used to optimize some corner cases.
8714 Note that regardless of this macro the ``definedness'' of @code{clz}
8715 and @code{ctz} at zero do @emph{not} extend to the builtin functions
8716 visible to the user. Thus one may be free to adjust the value at will
8717 to match the target expansion of these operations without fear of
8722 An alias for the machine mode for pointers. On most machines, define
8723 this to be the integer mode corresponding to the width of a hardware
8724 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8725 On some machines you must define this to be one of the partial integer
8726 modes, such as @code{PSImode}.
8728 The width of @code{Pmode} must be at least as large as the value of
8729 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8730 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8734 @defmac FUNCTION_MODE
8735 An alias for the machine mode used for memory references to functions
8736 being called, in @code{call} RTL expressions. On most machines this
8737 should be @code{QImode}.
8740 @defmac INTEGRATE_THRESHOLD (@var{decl})
8741 A C expression for the maximum number of instructions above which the
8742 function @var{decl} should not be inlined. @var{decl} is a
8743 @code{FUNCTION_DECL} node.
8745 The default definition of this macro is 64 plus 8 times the number of
8746 arguments that the function accepts. Some people think a larger
8747 threshold should be used on RISC machines.
8750 @defmac STDC_0_IN_SYSTEM_HEADERS
8751 In normal operation, the preprocessor expands @code{__STDC__} to the
8752 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8753 hosts, like Solaris, the system compiler uses a different convention,
8754 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8755 strict conformance to the C Standard.
8757 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8758 convention when processing system header files, but when processing user
8759 files @code{__STDC__} will always expand to 1.
8762 @defmac NO_IMPLICIT_EXTERN_C
8763 Define this macro if the system header files support C++ as well as C@.
8764 This macro inhibits the usual method of using system header files in
8765 C++, which is to pretend that the file's contents are enclosed in
8766 @samp{extern "C" @{@dots{}@}}.
8771 @defmac REGISTER_TARGET_PRAGMAS ()
8772 Define this macro if you want to implement any target-specific pragmas.
8773 If defined, it is a C expression which makes a series of calls to
8774 @code{c_register_pragma} for each pragma. The macro may also do any
8775 setup required for the pragmas.
8777 The primary reason to define this macro is to provide compatibility with
8778 other compilers for the same target. In general, we discourage
8779 definition of target-specific pragmas for GCC@.
8781 If the pragma can be implemented by attributes then you should consider
8782 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8784 Preprocessor macros that appear on pragma lines are not expanded. All
8785 @samp{#pragma} directives that do not match any registered pragma are
8786 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8789 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
8791 Each call to @code{c_register_pragma} establishes one pragma. The
8792 @var{callback} routine will be called when the preprocessor encounters a
8796 #pragma [@var{space}] @var{name} @dots{}
8799 @var{space} is the case-sensitive namespace of the pragma, or
8800 @code{NULL} to put the pragma in the global namespace. The callback
8801 routine receives @var{pfile} as its first argument, which can be passed
8802 on to cpplib's functions if necessary. You can lex tokens after the
8803 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8804 callback will be silently ignored. The end of the line is indicated by
8805 a token of type @code{CPP_EOF}
8807 For an example use of this routine, see @file{c4x.h} and the callback
8808 routines defined in @file{c4x-c.c}.
8810 Note that the use of @code{c_lex} is specific to the C and C++
8811 compilers. It will not work in the Java or Fortran compilers, or any
8812 other language compilers for that matter. Thus if @code{c_lex} is going
8813 to be called from target-specific code, it must only be done so when
8814 building the C and C++ compilers. This can be done by defining the
8815 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8816 target entry in the @file{config.gcc} file. These variables should name
8817 the target-specific, language-specific object file which contains the
8818 code that uses @code{c_lex}. Note it will also be necessary to add a
8819 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8820 how to build this object file.
8825 @defmac HANDLE_SYSV_PRAGMA
8826 Define this macro (to a value of 1) if you want the System V style
8827 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8828 [=<value>]} to be supported by gcc.
8830 The pack pragma specifies the maximum alignment (in bytes) of fields
8831 within a structure, in much the same way as the @samp{__aligned__} and
8832 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8833 the behavior to the default.
8835 A subtlety for Microsoft Visual C/C++ style bit-field packing
8836 (e.g. -mms-bitfields) for targets that support it:
8837 When a bit-field is inserted into a packed record, the whole size
8838 of the underlying type is used by one or more same-size adjacent
8839 bit-fields (that is, if its long:3, 32 bits is used in the record,
8840 and any additional adjacent long bit-fields are packed into the same
8841 chunk of 32 bits. However, if the size changes, a new field of that
8844 If both MS bit-fields and @samp{__attribute__((packed))} are used,
8845 the latter will take precedence. If @samp{__attribute__((packed))} is
8846 used on a single field when MS bit-fields are in use, it will take
8847 precedence for that field, but the alignment of the rest of the structure
8848 may affect its placement.
8850 The weak pragma only works if @code{SUPPORTS_WEAK} and
8851 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8852 of specifically named weak labels, optionally with a value.
8857 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
8858 Define this macro (to a value of 1) if you want to support the Win32
8859 style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma
8860 pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment
8861 (in bytes) of fields within a structure, in much the same way as the
8862 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8863 pack value of zero resets the behavior to the default. Successive
8864 invocations of this pragma cause the previous values to be stacked, so
8865 that invocations of @samp{#pragma pack(pop)} will return to the previous
8869 @defmac DOLLARS_IN_IDENTIFIERS
8870 Define this macro to control use of the character @samp{$} in
8871 identifier names for the C family of languages. 0 means @samp{$} is
8872 not allowed by default; 1 means it is allowed. 1 is the default;
8873 there is no need to define this macro in that case.
8876 @defmac NO_DOLLAR_IN_LABEL
8877 Define this macro if the assembler does not accept the character
8878 @samp{$} in label names. By default constructors and destructors in
8879 G++ have @samp{$} in the identifiers. If this macro is defined,
8880 @samp{.} is used instead.
8883 @defmac NO_DOT_IN_LABEL
8884 Define this macro if the assembler does not accept the character
8885 @samp{.} in label names. By default constructors and destructors in G++
8886 have names that use @samp{.}. If this macro is defined, these names
8887 are rewritten to avoid @samp{.}.
8890 @defmac DEFAULT_MAIN_RETURN
8891 Define this macro if the target system expects every program's @code{main}
8892 function to return a standard ``success'' value by default (if no other
8893 value is explicitly returned).
8895 The definition should be a C statement (sans semicolon) to generate the
8896 appropriate rtl instructions. It is used only when compiling the end of
8900 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
8901 Define this macro as a C expression that is nonzero if it is safe for the
8902 delay slot scheduler to place instructions in the delay slot of @var{insn},
8903 even if they appear to use a resource set or clobbered in @var{insn}.
8904 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8905 every @code{call_insn} has this behavior. On machines where some @code{insn}
8906 or @code{jump_insn} is really a function call and hence has this behavior,
8907 you should define this macro.
8909 You need not define this macro if it would always return zero.
8912 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
8913 Define this macro as a C expression that is nonzero if it is safe for the
8914 delay slot scheduler to place instructions in the delay slot of @var{insn},
8915 even if they appear to set or clobber a resource referenced in @var{insn}.
8916 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8917 some @code{insn} or @code{jump_insn} is really a function call and its operands
8918 are registers whose use is actually in the subroutine it calls, you should
8919 define this macro. Doing so allows the delay slot scheduler to move
8920 instructions which copy arguments into the argument registers into the delay
8923 You need not define this macro if it would always return zero.
8926 @defmac MULTIPLE_SYMBOL_SPACES
8927 Define this macro if in some cases global symbols from one translation
8928 unit may not be bound to undefined symbols in another translation unit
8929 without user intervention. For instance, under Microsoft Windows
8930 symbols must be explicitly imported from shared libraries (DLLs).
8933 @defmac MD_ASM_CLOBBERS (@var{clobbers})
8934 A C statement that adds to @var{clobbers} @code{STRING_CST} trees for
8935 any hard regs the port wishes to automatically clobber for all asms.
8938 @defmac MAX_INTEGER_COMPUTATION_MODE
8939 Define this to the largest integer machine mode which can be used for
8940 operations other than load, store and copy operations.
8942 You need only define this macro if the target holds values larger than
8943 @code{word_mode} in general purpose registers. Most targets should not define
8947 @defmac MATH_LIBRARY
8948 Define this macro as a C string constant for the linker argument to link
8949 in the system math library, or @samp{""} if the target does not have a
8950 separate math library.
8952 You need only define this macro if the default of @samp{"-lm"} is wrong.
8955 @defmac LIBRARY_PATH_ENV
8956 Define this macro as a C string constant for the environment variable that
8957 specifies where the linker should look for libraries.
8959 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
8963 @defmac TARGET_HAS_F_SETLKW
8964 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
8965 Note that this functionality is part of POSIX@.
8966 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
8967 to use file locking when exiting a program, which avoids race conditions
8968 if the program has forked.
8971 @defmac MAX_CONDITIONAL_EXECUTE
8973 A C expression for the maximum number of instructions to execute via
8974 conditional execution instructions instead of a branch. A value of
8975 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8976 1 if it does use cc0.
8979 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
8980 Used if the target needs to perform machine-dependent modifications on the
8981 conditionals used for turning basic blocks into conditionally executed code.
8982 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
8983 contains information about the currently processed blocks. @var{true_expr}
8984 and @var{false_expr} are the tests that are used for converting the
8985 then-block and the else-block, respectively. Set either @var{true_expr} or
8986 @var{false_expr} to a null pointer if the tests cannot be converted.
8989 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
8990 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
8991 if-statements into conditions combined by @code{and} and @code{or} operations.
8992 @var{bb} contains the basic block that contains the test that is currently
8993 being processed and about to be turned into a condition.
8996 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
8997 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
8998 be converted to conditional execution format. @var{ce_info} points to
8999 a data structure, @code{struct ce_if_block}, which contains information
9000 about the currently processed blocks.
9003 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
9004 A C expression to perform any final machine dependent modifications in
9005 converting code to conditional execution. The involved basic blocks
9006 can be found in the @code{struct ce_if_block} structure that is pointed
9007 to by @var{ce_info}.
9010 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
9011 A C expression to cancel any machine dependent modifications in
9012 converting code to conditional execution. The involved basic blocks
9013 can be found in the @code{struct ce_if_block} structure that is pointed
9014 to by @var{ce_info}.
9017 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
9018 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9019 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9022 @defmac IFCVT_EXTRA_FIELDS
9023 If defined, it should expand to a set of field declarations that will be
9024 added to the @code{struct ce_if_block} structure. These should be initialized
9025 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9028 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
9029 If non-null, this hook performs a target-specific pass over the
9030 instruction stream. The compiler will run it at all optimization levels,
9031 just before the point at which it normally does delayed-branch scheduling.
9033 The exact purpose of the hook varies from target to target. Some use
9034 it to do transformations that are necessary for correctness, such as
9035 laying out in-function constant pools or avoiding hardware hazards.
9036 Others use it as an opportunity to do some machine-dependent optimizations.
9038 You need not implement the hook if it has nothing to do. The default
9042 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9043 Define this hook if you have any machine-specific built-in functions
9044 that need to be defined. It should be a function that performs the
9047 Machine specific built-in functions can be useful to expand special machine
9048 instructions that would otherwise not normally be generated because
9049 they have no equivalent in the source language (for example, SIMD vector
9050 instructions or prefetch instructions).
9052 To create a built-in function, call the function @code{builtin_function}
9053 which is defined by the language front end. You can use any type nodes set
9054 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9055 only language front ends that use those two functions will call
9056 @samp{TARGET_INIT_BUILTINS}.
9059 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9061 Expand a call to a machine specific built-in function that was set up by
9062 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9063 function call; the result should go to @var{target} if that is
9064 convenient, and have mode @var{mode} if that is convenient.
9065 @var{subtarget} may be used as the target for computing one of
9066 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9067 ignored. This function should return the result of the call to the
9071 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
9073 Take a branch insn in @var{branch1} and another in @var{branch2}.
9074 Return true if redirecting @var{branch1} to the destination of
9075 @var{branch2} is possible.
9077 On some targets, branches may have a limited range. Optimizing the
9078 filling of delay slots can result in branches being redirected, and this
9079 may in turn cause a branch offset to overflow.
9082 @defmac ALLOCATE_INITIAL_VALUE (@var{hard_reg})
9084 When the initial value of a hard register has been copied in a pseudo
9085 register, it is often not necessary to actually allocate another register
9086 to this pseudo register, because the original hard register or a stack slot
9087 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9088 defined, is called at the start of register allocation once for each
9089 hard register that had its initial value copied by using
9090 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9091 Possible values are @code{NULL_RTX}, if you don't want
9092 to do any special allocation, a @code{REG} rtx---that would typically be
9093 the hard register itself, if it is known not to be clobbered---or a
9095 If you are returning a @code{MEM}, this is only a hint for the allocator;
9096 it might decide to use another register anyways.
9097 You may use @code{current_function_leaf_function} in the definition of the
9098 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9099 register in question will not be clobbered.
9102 @defmac TARGET_OBJECT_SUFFIX
9103 Define this macro to be a C string representing the suffix for object
9104 files on your target machine. If you do not define this macro, GCC will
9105 use @samp{.o} as the suffix for object files.
9108 @defmac TARGET_EXECUTABLE_SUFFIX
9109 Define this macro to be a C string representing the suffix to be
9110 automatically added to executable files on your target machine. If you
9111 do not define this macro, GCC will use the null string as the suffix for
9115 @defmac COLLECT_EXPORT_LIST
9116 If defined, @code{collect2} will scan the individual object files
9117 specified on its command line and create an export list for the linker.
9118 Define this macro for systems like AIX, where the linker discards
9119 object files that are not referenced from @code{main} and uses export
9123 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
9124 Define this macro to a C expression representing a variant of the
9125 method call @var{mdecl}, if Java Native Interface (JNI) methods
9126 must be invoked differently from other methods on your target.
9127 For example, on 32-bit Windows, JNI methods must be invoked using
9128 the @code{stdcall} calling convention and this macro is then
9129 defined as this expression:
9132 build_type_attribute_variant (@var{mdecl},
9134 (get_identifier ("stdcall"),
9139 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9140 This target hook returns @code{true} past the point in which new jump
9141 instructions could be created. On machines that require a register for
9142 every jump such as the SHmedia ISA of SH5, this point would typically be
9143 reload, so this target hook should be defined to a function such as:
9147 cannot_modify_jumps_past_reload_p ()
9149 return (reload_completed || reload_in_progress);
9154 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
9155 This target hook returns a register class for which branch target register
9156 optimizations should be applied. All registers in this class should be
9157 usable interchangeably. After reload, registers in this class will be
9158 re-allocated and loads will be hoisted out of loops and be subjected
9159 to inter-block scheduling.
9162 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
9163 Branch target register optimization will by default exclude callee-saved
9165 that are not already live during the current function; if this target hook
9166 returns true, they will be included. The target code must than make sure
9167 that all target registers in the class returned by
9168 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
9169 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
9170 epilogues have already been generated. Note, even if you only return
9171 true when @var{after_prologue_epilogue_gen} is false, you still are likely
9172 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
9173 to reserve space for caller-saved target registers.
9176 @defmac POWI_MAX_MULTS
9177 If defined, this macro is interpreted as a signed integer C expression
9178 that specifies the maximum number of floating point multiplications
9179 that should be emitted when expanding exponentiation by an integer
9180 constant inline. When this value is defined, exponentiation requiring
9181 more than this number of multiplications is implemented by calling the
9182 system library's @code{pow}, @code{powf} or @code{powl} routines.
9183 The default value places no upper bound on the multiplication count.