1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004, 2005 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 * Registers:: Naming and describing the hardware registers.
35 * Register Classes:: Defining the classes of hardware registers.
36 * Stack and Calling:: Defining which way the stack grows and by how much.
37 * Varargs:: Defining the varargs macros.
38 * Trampolines:: Code set up at run time to enter a nested function.
39 * Library Calls:: Controlling how library routines are implicitly called.
40 * Addressing Modes:: Defining addressing modes valid for memory operands.
41 * Condition Code:: Defining how insns update the condition code.
42 * Costs:: Defining relative costs of different operations.
43 * Scheduling:: Adjusting the behavior of the instruction scheduler.
44 * Sections:: Dividing storage into text, data, and other sections.
45 * PIC:: Macros for position independent code.
46 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
47 * Debugging Info:: Defining the format of debugging output.
48 * Floating Point:: Handling floating point for cross-compilers.
49 * Mode Switching:: Insertion of mode-switching instructions.
50 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
51 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
52 * PCH Target:: Validity checking for precompiled headers.
53 * C++ ABI:: Controlling C++ ABI changes.
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 @option{-malt-abi},
157 @option{-EB}, and @option{-mspoo} cause different multilibs to be chosen:
160 #define TARGET_OPTION_TRANSLATE_TABLE \
161 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
162 @{ "-compat", "-EB -malign=4 -mspoo" @}
166 @defmac DRIVER_SELF_SPECS
167 A list of specs for the driver itself. It should be a suitable
168 initializer for an array of strings, with no surrounding braces.
170 The driver applies these specs to its own command line between loading
171 default @file{specs} files (but not command-line specified ones) and
172 choosing the multilib directory or running any subcommands. It
173 applies them in the order given, so each spec can depend on the
174 options added by earlier ones. It is also possible to remove options
175 using @samp{%<@var{option}} in the usual way.
177 This macro can be useful when a port has several interdependent target
178 options. It provides a way of standardizing the command line so
179 that the other specs are easier to write.
181 Do not define this macro if it does not need to do anything.
184 @defmac OPTION_DEFAULT_SPECS
185 A list of specs used to support configure-time default options (i.e.@:
186 @option{--with} options) in the driver. It should be a suitable initializer
187 for an array of structures, each containing two strings, without the
188 outermost pair of surrounding braces.
190 The first item in the pair is the name of the default. This must match
191 the code in @file{config.gcc} for the target. The second item is a spec
192 to apply if a default with this name was specified. The string
193 @samp{%(VALUE)} in the spec will be replaced by the value of the default
194 everywhere it occurs.
196 The driver will apply these specs to its own command line between loading
197 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
198 the same mechanism as @code{DRIVER_SELF_SPECS}.
200 Do not define this macro if it does not need to do anything.
204 A C string constant that tells the GCC driver program options to
205 pass to CPP@. It can also specify how to translate options you
206 give to GCC into options for GCC to pass to the CPP@.
208 Do not define this macro if it does not need to do anything.
211 @defmac CPLUSPLUS_CPP_SPEC
212 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
213 than C@. If you do not define this macro, then the value of
214 @code{CPP_SPEC} (if any) will be used instead.
218 A C string constant that tells the GCC driver program options to
219 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
221 It can also specify how to translate options you give to GCC into options
222 for GCC to pass to front ends.
224 Do not define this macro if it does not need to do anything.
228 A C string constant that tells the GCC driver program options to
229 pass to @code{cc1plus}. It can also specify how to translate options you
230 give to GCC into options for GCC to pass to the @code{cc1plus}.
232 Do not define this macro if it does not need to do anything.
233 Note that everything defined in CC1_SPEC is already passed to
234 @code{cc1plus} so there is no need to duplicate the contents of
235 CC1_SPEC in CC1PLUS_SPEC@.
239 A C string constant that tells the GCC driver program options to
240 pass to the assembler. It can also specify how to translate options
241 you give to GCC into options for GCC to pass to the assembler.
242 See the file @file{sun3.h} for an example of this.
244 Do not define this macro if it does not need to do anything.
247 @defmac ASM_FINAL_SPEC
248 A C string constant that tells the GCC driver program how to
249 run any programs which cleanup after the normal assembler.
250 Normally, this is not needed. See the file @file{mips.h} for
253 Do not define this macro if it does not need to do anything.
256 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
257 Define this macro, with no value, if the driver should give the assembler
258 an argument consisting of a single dash, @option{-}, to instruct it to
259 read from its standard input (which will be a pipe connected to the
260 output of the compiler proper). This argument is given after any
261 @option{-o} option specifying the name of the output file.
263 If you do not define this macro, the assembler is assumed to read its
264 standard input if given no non-option arguments. If your assembler
265 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
266 see @file{mips.h} for instance.
270 A C string constant that tells the GCC driver program options to
271 pass to the linker. It can also specify how to translate options you
272 give to GCC into options for GCC to pass to the linker.
274 Do not define this macro if it does not need to do anything.
278 Another C string constant used much like @code{LINK_SPEC}. The difference
279 between the two is that @code{LIB_SPEC} is used at the end of the
280 command given to the linker.
282 If this macro is not defined, a default is provided that
283 loads the standard C library from the usual place. See @file{gcc.c}.
287 Another C string constant that tells the GCC driver program
288 how and when to place a reference to @file{libgcc.a} into the
289 linker command line. This constant is placed both before and after
290 the value of @code{LIB_SPEC}.
292 If this macro is not defined, the GCC driver provides a default that
293 passes the string @option{-lgcc} to the linker.
296 @defmac REAL_LIBGCC_SPEC
297 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
298 @code{LIBGCC_SPEC} is not directly used by the driver program but is
299 instead modified to refer to different versions of @file{libgcc.a}
300 depending on the values of the command line flags @option{-static},
301 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
302 targets where these modifications are inappropriate, define
303 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
304 driver how to place a reference to @file{libgcc} on the link command
305 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
308 @defmac USE_LD_AS_NEEDED
309 A macro that controls the modifications to @code{LIBGCC_SPEC}
310 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
311 generated that uses --as-needed and the shared libgcc in place of the
312 static exception handler library, when linking without any of
313 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
317 If defined, this C string constant is added to @code{LINK_SPEC}.
318 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
319 the modifications to @code{LIBGCC_SPEC} mentioned in
320 @code{REAL_LIBGCC_SPEC}.
323 @defmac STARTFILE_SPEC
324 Another C string constant used much like @code{LINK_SPEC}. The
325 difference between the two is that @code{STARTFILE_SPEC} is used at
326 the very beginning of the command given to the linker.
328 If this macro is not defined, a default is provided that loads the
329 standard C startup file from the usual place. See @file{gcc.c}.
333 Another C string constant used much like @code{LINK_SPEC}. The
334 difference between the two is that @code{ENDFILE_SPEC} is used at
335 the very end of the command given to the linker.
337 Do not define this macro if it does not need to do anything.
340 @defmac THREAD_MODEL_SPEC
341 GCC @code{-v} will print the thread model GCC was configured to use.
342 However, this doesn't work on platforms that are multilibbed on thread
343 models, such as AIX 4.3. On such platforms, define
344 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
345 blanks that names one of the recognized thread models. @code{%*}, the
346 default value of this macro, will expand to the value of
347 @code{thread_file} set in @file{config.gcc}.
350 @defmac SYSROOT_SUFFIX_SPEC
351 Define this macro to add a suffix to the target sysroot when GCC is
352 configured with a sysroot. This will cause GCC to search for usr/lib,
353 et al, within sysroot+suffix.
356 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
357 Define this macro to add a headers_suffix to the target sysroot when
358 GCC is configured with a sysroot. This will cause GCC to pass the
359 updated sysroot+headers_suffix to CPP, causing it to search for
360 usr/include, et al, within sysroot+headers_suffix.
364 Define this macro to provide additional specifications to put in the
365 @file{specs} file that can be used in various specifications like
368 The definition should be an initializer for an array of structures,
369 containing a string constant, that defines the specification name, and a
370 string constant that provides the specification.
372 Do not define this macro if it does not need to do anything.
374 @code{EXTRA_SPECS} is useful when an architecture contains several
375 related targets, which have various @code{@dots{}_SPECS} which are similar
376 to each other, and the maintainer would like one central place to keep
379 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
380 define either @code{_CALL_SYSV} when the System V calling sequence is
381 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
384 The @file{config/rs6000/rs6000.h} target file defines:
387 #define EXTRA_SPECS \
388 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
390 #define CPP_SYS_DEFAULT ""
393 The @file{config/rs6000/sysv.h} target file defines:
397 "%@{posix: -D_POSIX_SOURCE @} \
398 %@{mcall-sysv: -D_CALL_SYSV @} \
399 %@{!mcall-sysv: %(cpp_sysv_default) @} \
400 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
402 #undef CPP_SYSV_DEFAULT
403 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
406 while the @file{config/rs6000/eabiaix.h} target file defines
407 @code{CPP_SYSV_DEFAULT} as:
410 #undef CPP_SYSV_DEFAULT
411 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
415 @defmac LINK_LIBGCC_SPECIAL_1
416 Define this macro if the driver program should find the library
417 @file{libgcc.a}. If you do not define this macro, the driver program will pass
418 the argument @option{-lgcc} to tell the linker to do the search.
421 @defmac LINK_GCC_C_SEQUENCE_SPEC
422 The sequence in which libgcc and libc are specified to the linker.
423 By default this is @code{%G %L %G}.
426 @defmac LINK_COMMAND_SPEC
427 A C string constant giving the complete command line need to execute the
428 linker. When you do this, you will need to update your port each time a
429 change is made to the link command line within @file{gcc.c}. Therefore,
430 define this macro only if you need to completely redefine the command
431 line for invoking the linker and there is no other way to accomplish
432 the effect you need. Overriding this macro may be avoidable by overriding
433 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
436 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
437 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
438 directories from linking commands. Do not give it a nonzero value if
439 removing duplicate search directories changes the linker's semantics.
442 @defmac MULTILIB_DEFAULTS
443 Define this macro as a C expression for the initializer of an array of
444 string to tell the driver program which options are defaults for this
445 target and thus do not need to be handled specially when using
446 @code{MULTILIB_OPTIONS}.
448 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
449 the target makefile fragment or if none of the options listed in
450 @code{MULTILIB_OPTIONS} are set by default.
451 @xref{Target Fragment}.
454 @defmac RELATIVE_PREFIX_NOT_LINKDIR
455 Define this macro to tell @command{gcc} that it should only translate
456 a @option{-B} prefix into a @option{-L} linker option if the prefix
457 indicates an absolute file name.
460 @defmac MD_EXEC_PREFIX
461 If defined, this macro is an additional prefix to try after
462 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
463 when the @option{-b} option is used, or the compiler is built as a cross
464 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
465 to the list of directories used to find the assembler in @file{configure.in}.
468 @defmac STANDARD_STARTFILE_PREFIX
469 Define this macro as a C string constant if you wish to override the
470 standard choice of @code{libdir} as the default prefix to
471 try when searching for startup files such as @file{crt0.o}.
472 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
473 is built as a cross compiler.
476 @defmac STANDARD_STARTFILE_PREFIX_1
477 Define this macro as a C string constant if you wish to override the
478 standard choice of @code{/lib} as a prefix to try after the default prefix
479 when searching for startup files such as @file{crt0.o}.
480 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
481 is built as a cross compiler.
484 @defmac STANDARD_STARTFILE_PREFIX_2
485 Define this macro as a C string constant if you wish to override the
486 standard choice of @code{/lib} as yet another prefix to try after the
487 default prefix when searching for startup files such as @file{crt0.o}.
488 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
489 is built as a cross compiler.
492 @defmac MD_STARTFILE_PREFIX
493 If defined, this macro supplies an additional prefix to try after the
494 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
495 @option{-b} option is used, or when the compiler is built as a cross
499 @defmac MD_STARTFILE_PREFIX_1
500 If defined, this macro supplies yet another prefix to try after the
501 standard prefixes. It is not searched when the @option{-b} option is
502 used, or when the compiler is built as a cross compiler.
505 @defmac INIT_ENVIRONMENT
506 Define this macro as a C string constant if you wish to set environment
507 variables for programs called by the driver, such as the assembler and
508 loader. The driver passes the value of this macro to @code{putenv} to
509 initialize the necessary environment variables.
512 @defmac LOCAL_INCLUDE_DIR
513 Define this macro as a C string constant if you wish to override the
514 standard choice of @file{/usr/local/include} as the default prefix to
515 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
516 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
518 Cross compilers do not search either @file{/usr/local/include} or its
522 @defmac MODIFY_TARGET_NAME
523 Define this macro if you wish to define command-line switches that
524 modify the default target name.
526 For each switch, you can include a string to be appended to the first
527 part of the configuration name or a string to be deleted from the
528 configuration name, if present. The definition should be an initializer
529 for an array of structures. Each array element should have three
530 elements: the switch name (a string constant, including the initial
531 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
532 indicate whether the string should be inserted or deleted, and the string
533 to be inserted or deleted (a string constant).
535 For example, on a machine where @samp{64} at the end of the
536 configuration name denotes a 64-bit target and you want the @option{-32}
537 and @option{-64} switches to select between 32- and 64-bit targets, you would
541 #define MODIFY_TARGET_NAME \
542 @{ @{ "-32", DELETE, "64"@}, \
543 @{"-64", ADD, "64"@}@}
547 @defmac SYSTEM_INCLUDE_DIR
548 Define this macro as a C string constant if you wish to specify a
549 system-specific directory to search for header files before the standard
550 directory. @code{SYSTEM_INCLUDE_DIR} comes before
551 @code{STANDARD_INCLUDE_DIR} in the search order.
553 Cross compilers do not use this macro and do not search the directory
557 @defmac STANDARD_INCLUDE_DIR
558 Define this macro as a C string constant if you wish to override the
559 standard choice of @file{/usr/include} as the default prefix to
560 try when searching for header files.
562 Cross compilers ignore this macro and do not search either
563 @file{/usr/include} or its replacement.
566 @defmac STANDARD_INCLUDE_COMPONENT
567 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
568 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
569 If you do not define this macro, no component is used.
572 @defmac INCLUDE_DEFAULTS
573 Define this macro if you wish to override the entire default search path
574 for include files. For a native compiler, the default search path
575 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
576 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
577 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
578 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
579 and specify private search areas for GCC@. The directory
580 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
582 The definition should be an initializer for an array of structures.
583 Each array element should have four elements: the directory name (a
584 string constant), the component name (also a string constant), a flag
585 for C++-only directories,
586 and a flag showing that the includes in the directory don't need to be
587 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
588 the array with a null element.
590 The component name denotes what GNU package the include file is part of,
591 if any, in all uppercase letters. For example, it might be @samp{GCC}
592 or @samp{BINUTILS}. If the package is part of a vendor-supplied
593 operating system, code the component name as @samp{0}.
595 For example, here is the definition used for VAX/VMS:
598 #define INCLUDE_DEFAULTS \
600 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
601 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
602 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
609 Here is the order of prefixes tried for exec files:
613 Any prefixes specified by the user with @option{-B}.
616 The environment variable @code{GCC_EXEC_PREFIX}, if any.
619 The directories specified by the environment variable @code{COMPILER_PATH}.
622 The macro @code{STANDARD_EXEC_PREFIX}.
625 @file{/usr/lib/gcc/}.
628 The macro @code{MD_EXEC_PREFIX}, if any.
631 Here is the order of prefixes tried for startfiles:
635 Any prefixes specified by the user with @option{-B}.
638 The environment variable @code{GCC_EXEC_PREFIX}, if any.
641 The directories specified by the environment variable @code{LIBRARY_PATH}
642 (or port-specific name; native only, cross compilers do not use this).
645 The macro @code{STANDARD_EXEC_PREFIX}.
648 @file{/usr/lib/gcc/}.
651 The macro @code{MD_EXEC_PREFIX}, if any.
654 The macro @code{MD_STARTFILE_PREFIX}, if any.
657 The macro @code{STANDARD_STARTFILE_PREFIX}.
666 @node Run-time Target
667 @section Run-time Target Specification
668 @cindex run-time target specification
669 @cindex predefined macros
670 @cindex target specifications
672 @c prevent bad page break with this line
673 Here are run-time target specifications.
675 @defmac TARGET_CPU_CPP_BUILTINS ()
676 This function-like macro expands to a block of code that defines
677 built-in preprocessor macros and assertions for the target cpu, using
678 the functions @code{builtin_define}, @code{builtin_define_std} and
679 @code{builtin_assert}. When the front end
680 calls this macro it provides a trailing semicolon, and since it has
681 finished command line option processing your code can use those
684 @code{builtin_assert} takes a string in the form you pass to the
685 command-line option @option{-A}, such as @code{cpu=mips}, and creates
686 the assertion. @code{builtin_define} takes a string in the form
687 accepted by option @option{-D} and unconditionally defines the macro.
689 @code{builtin_define_std} takes a string representing the name of an
690 object-like macro. If it doesn't lie in the user's namespace,
691 @code{builtin_define_std} defines it unconditionally. Otherwise, it
692 defines a version with two leading underscores, and another version
693 with two leading and trailing underscores, and defines the original
694 only if an ISO standard was not requested on the command line. For
695 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
696 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
697 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
698 defines only @code{_ABI64}.
700 You can also test for the C dialect being compiled. The variable
701 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
702 or @code{clk_objective_c}. Note that if we are preprocessing
703 assembler, this variable will be @code{clk_c} but the function-like
704 macro @code{preprocessing_asm_p()} will return true, so you might want
705 to check for that first. If you need to check for strict ANSI, the
706 variable @code{flag_iso} can be used. The function-like macro
707 @code{preprocessing_trad_p()} can be used to check for traditional
711 @defmac TARGET_OS_CPP_BUILTINS ()
712 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
713 and is used for the target operating system instead.
716 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
717 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
718 and is used for the target object format. @file{elfos.h} uses this
719 macro to define @code{__ELF__}, so you probably do not need to define
723 @deftypevar {extern int} target_flags
724 This variable is declared in @file{options.h}, which is included before
725 any target-specific headers.
728 @deftypevar {Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
729 This variable specifies the initial value of @code{target_flags}.
730 Its default setting is 0.
732 If the target defines @code{TARGET_SWITCHES}, the null
733 @code{TARGET_SWITCHES} entry will override this value.
736 @deftypefn {Target Hook} bool TARGET_HANDLE_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
737 This hook is called whenever the user specifies one of the
738 target-specific options described by the @file{.opt} definition files
739 (@pxref{Options}). It has the opportunity to do some option-specific
740 processing and should return true if the option is valid. The default
741 definition does nothing but return true.
743 @var{code} specifies the @code{OPT_@var{name}} enumeration value
744 associated with the selected option; @var{name} is just a rendering of
745 the option name in which non-alphanumeric characters are replaced by
746 underscores. @var{arg} specifies the string argument and is null if
747 no argument was given. If the option is flagged as a @code{UInteger}
748 (@pxref{Option properties}), @var{value} is the numeric value of the
749 argument. Otherwise @var{value} is 1 if the positive form of the
750 option was used and 0 if the ``no-'' form was.
753 @cindex optional hardware or system features
754 @cindex features, optional, in system conventions
756 @defmac TARGET_@var{featurename}
757 This series of macros is to allow compiler command arguments to
758 enable or disable the use of optional features of the target machine.
759 For example, one machine description serves both the 68000 and
760 the 68020; a command argument tells the compiler whether it should
761 use 68020-only instructions or not. This command argument works
762 by means of a macro @code{TARGET_68020} that tests a bit in
765 Define a macro @code{TARGET_@var{featurename}} for each such option.
766 Its definition should test a bit in @code{target_flags}. It is
767 recommended that a helper macro @code{MASK_@var{featurename}}
768 is defined for each bit-value to test, and used in
769 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
773 #define TARGET_MASK_68020 1
774 #define TARGET_68020 (target_flags & MASK_68020)
777 One place where these macros are used is in the condition-expressions
778 of instruction patterns. Note how @code{TARGET_68020} appears
779 frequently in the 68000 machine description file, @file{m68k.md}.
780 Another place they are used is in the definitions of the other
781 macros in the @file{@var{machine}.h} file.
784 @defmac TARGET_SWITCHES
785 This macro defines names of command options to set and clear
786 bits in @code{target_flags}. Its definition is an initializer
787 with a subgrouping for each command option.
789 Each subgrouping contains a string constant, that defines the option
790 name, a number, which contains the bits to set in
791 @code{target_flags}, and a second string which is the description
792 displayed by @option{--help}. If the number is negative then the bits specified
793 by the number are cleared instead of being set. If the description
794 string is present but empty, then no help information will be displayed
795 for that option, but it will not count as an undocumented option. The
796 actual option name is made by appending @samp{-m} to the specified name.
797 Non-empty description strings should be marked with @code{N_(@dots{})} for
798 @command{xgettext}. Please do not mark empty strings because the empty
799 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
800 of the message catalog with meta information, not the empty string.
802 In addition to the description for @option{--help},
803 more detailed documentation for each option should be added to
806 One of the subgroupings should have a null string. The number in
807 this grouping is the default value for @code{target_flags}. Any
808 target options act starting with that value.
810 Here is an example which defines @option{-m68000} and @option{-m68020}
811 with opposite meanings, and picks the latter as the default:
814 #define TARGET_SWITCHES \
815 @{ @{ "68020", MASK_68020, "" @}, \
816 @{ "68000", -MASK_68020, \
817 N_("Compile for the 68000") @}, \
818 @{ "", MASK_68020, "" @}, \
822 This macro is being kept for compatibility with older backends.
823 New targets should use option definition files instead.
827 @defmac TARGET_OPTIONS
828 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
829 options that have values. Its definition is an initializer with a
830 subgrouping for each command option.
832 Each subgrouping contains a string constant, that defines the option
833 name, the address of a variable, a description string, and a value.
834 Non-empty description strings should be marked with @code{N_(@dots{})}
835 for @command{xgettext}. Please do not mark empty strings because the
836 empty string is reserved by GNU gettext. @code{gettext("")} returns the
837 header entry of the message catalog with meta information, not the empty
840 If the value listed in the table is @code{NULL}, then the variable, type
841 @code{char *}, is set to the variable part of the given option if the
842 fixed part matches. In other words, if the first part of the option
843 matches what's in the table, the variable will be set to point to the
844 rest of the option. This allows the user to specify a value for that
845 option. The actual option name is made by appending @samp{-m} to the
846 specified name. Again, each option should also be documented in
849 If the value listed in the table is non-@code{NULL}, then the option
850 must match the option in the table exactly (with @samp{-m}), and the
851 variable is set to point to the value listed in the table.
853 Here is an example which defines @option{-mshort-data-@var{number}}. If the
854 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
855 will be set to the string @code{"512"}.
858 extern char *m88k_short_data;
859 #define TARGET_OPTIONS \
860 @{ @{ "short-data-", &m88k_short_data, \
861 N_("Specify the size of the short data section"), 0 @} @}
864 Here is a variant of the above that allows the user to also specify
865 just @option{-mshort-data} where a default of @code{"64"} is used.
868 extern char *m88k_short_data;
869 #define TARGET_OPTIONS \
870 @{ @{ "short-data-", &m88k_short_data, \
871 N_("Specify the size of the short data section"), 0 @} \
872 @{ "short-data", &m88k_short_data, "", "64" @},
876 Here is an example which defines @option{-mno-alu}, @option{-malu1}, and
877 @option{-malu2} as a three-state switch, along with suitable macros for
878 checking the state of the option (documentation is elided for brevity).
882 char *chip_alu = ""; /* @r{Specify default here.} */
885 extern char *chip_alu;
886 #define TARGET_OPTIONS \
887 @{ @{ "no-alu", &chip_alu, "", "" @}, \
888 @{ "alu1", &chip_alu, "", "1" @}, \
889 @{ "alu2", &chip_alu, "", "2" @}, @}
890 #define TARGET_ALU (chip_alu[0] != '\0')
891 #define TARGET_ALU1 (chip_alu[0] == '1')
892 #define TARGET_ALU2 (chip_alu[0] == '2')
895 This macro is being kept for compatibility with older backends.
896 New targets should use option definition files instead.
900 @defmac TARGET_VERSION
901 This macro is a C statement to print on @code{stderr} a string
902 describing the particular machine description choice. Every machine
903 description should define @code{TARGET_VERSION}. For example:
907 #define TARGET_VERSION \
908 fprintf (stderr, " (68k, Motorola syntax)");
910 #define TARGET_VERSION \
911 fprintf (stderr, " (68k, MIT syntax)");
916 @defmac OVERRIDE_OPTIONS
917 Sometimes certain combinations of command options do not make sense on
918 a particular target machine. You can define a macro
919 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
920 defined, is executed once just after all the command options have been
923 Don't use this macro to turn on various extra optimizations for
924 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
927 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
928 Some machines may desire to change what optimizations are performed for
929 various optimization levels. This macro, if defined, is executed once
930 just after the optimization level is determined and before the remainder
931 of the command options have been parsed. Values set in this macro are
932 used as the default values for the other command line options.
934 @var{level} is the optimization level specified; 2 if @option{-O2} is
935 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
937 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
939 You should not use this macro to change options that are not
940 machine-specific. These should uniformly selected by the same
941 optimization level on all supported machines. Use this macro to enable
942 machine-specific optimizations.
944 @strong{Do not examine @code{write_symbols} in
945 this macro!} The debugging options are not supposed to alter the
949 @defmac CAN_DEBUG_WITHOUT_FP
950 Define this macro if debugging can be performed even without a frame
951 pointer. If this macro is defined, GCC will turn on the
952 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
955 @node Per-Function Data
956 @section Defining data structures for per-function information.
957 @cindex per-function data
958 @cindex data structures
960 If the target needs to store information on a per-function basis, GCC
961 provides a macro and a couple of variables to allow this. Note, just
962 using statics to store the information is a bad idea, since GCC supports
963 nested functions, so you can be halfway through encoding one function
964 when another one comes along.
966 GCC defines a data structure called @code{struct function} which
967 contains all of the data specific to an individual function. This
968 structure contains a field called @code{machine} whose type is
969 @code{struct machine_function *}, which can be used by targets to point
970 to their own specific data.
972 If a target needs per-function specific data it should define the type
973 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
974 This macro should be used to initialize the function pointer
975 @code{init_machine_status}. This pointer is explained below.
977 One typical use of per-function, target specific data is to create an
978 RTX to hold the register containing the function's return address. This
979 RTX can then be used to implement the @code{__builtin_return_address}
980 function, for level 0.
982 Note---earlier implementations of GCC used a single data area to hold
983 all of the per-function information. Thus when processing of a nested
984 function began the old per-function data had to be pushed onto a
985 stack, and when the processing was finished, it had to be popped off the
986 stack. GCC used to provide function pointers called
987 @code{save_machine_status} and @code{restore_machine_status} to handle
988 the saving and restoring of the target specific information. Since the
989 single data area approach is no longer used, these pointers are no
992 @defmac INIT_EXPANDERS
993 Macro called to initialize any target specific information. This macro
994 is called once per function, before generation of any RTL has begun.
995 The intention of this macro is to allow the initialization of the
996 function pointer @code{init_machine_status}.
999 @deftypevar {void (*)(struct function *)} init_machine_status
1000 If this function pointer is non-@code{NULL} it will be called once per
1001 function, before function compilation starts, in order to allow the
1002 target to perform any target specific initialization of the
1003 @code{struct function} structure. It is intended that this would be
1004 used to initialize the @code{machine} of that structure.
1006 @code{struct machine_function} structures are expected to be freed by GC@.
1007 Generally, any memory that they reference must be allocated by using
1008 @code{ggc_alloc}, including the structure itself.
1011 @node Storage Layout
1012 @section Storage Layout
1013 @cindex storage layout
1015 Note that the definitions of the macros in this table which are sizes or
1016 alignments measured in bits do not need to be constant. They can be C
1017 expressions that refer to static variables, such as the @code{target_flags}.
1018 @xref{Run-time Target}.
1020 @defmac BITS_BIG_ENDIAN
1021 Define this macro to have the value 1 if the most significant bit in a
1022 byte has the lowest number; otherwise define it to have the value zero.
1023 This means that bit-field instructions count from the most significant
1024 bit. If the machine has no bit-field instructions, then this must still
1025 be defined, but it doesn't matter which value it is defined to. This
1026 macro need not be a constant.
1028 This macro does not affect the way structure fields are packed into
1029 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
1032 @defmac BYTES_BIG_ENDIAN
1033 Define this macro to have the value 1 if the most significant byte in a
1034 word has the lowest number. This macro need not be a constant.
1037 @defmac WORDS_BIG_ENDIAN
1038 Define this macro to have the value 1 if, in a multiword object, the
1039 most significant word has the lowest number. This applies to both
1040 memory locations and registers; GCC fundamentally assumes that the
1041 order of words in memory is the same as the order in registers. This
1042 macro need not be a constant.
1045 @defmac LIBGCC2_WORDS_BIG_ENDIAN
1046 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
1047 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
1048 used only when compiling @file{libgcc2.c}. Typically the value will be set
1049 based on preprocessor defines.
1052 @defmac FLOAT_WORDS_BIG_ENDIAN
1053 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
1054 @code{TFmode} floating point numbers are stored in memory with the word
1055 containing the sign bit at the lowest address; otherwise define it to
1056 have the value 0. This macro need not be a constant.
1058 You need not define this macro if the ordering is the same as for
1059 multi-word integers.
1062 @defmac BITS_PER_UNIT
1063 Define this macro to be the number of bits in an addressable storage
1064 unit (byte). If you do not define this macro the default is 8.
1067 @defmac BITS_PER_WORD
1068 Number of bits in a word. If you do not define this macro, the default
1069 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
1072 @defmac MAX_BITS_PER_WORD
1073 Maximum number of bits in a word. If this is undefined, the default is
1074 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
1075 largest value that @code{BITS_PER_WORD} can have at run-time.
1078 @defmac UNITS_PER_WORD
1079 Number of storage units in a word; normally the size of a general-purpose
1080 register, a power of two from 1 or 8.
1083 @defmac MIN_UNITS_PER_WORD
1084 Minimum number of units in a word. If this is undefined, the default is
1085 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
1086 smallest value that @code{UNITS_PER_WORD} can have at run-time.
1089 @defmac UNITS_PER_SIMD_WORD
1090 Number of units in the vectors that the vectorizer can produce.
1091 The default is equal to @code{UNITS_PER_WORD}, because the vectorizer
1092 can do some transformations even in absence of specialized @acronym{SIMD}
1096 @defmac POINTER_SIZE
1097 Width of a pointer, in bits. You must specify a value no wider than the
1098 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1099 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1100 a value the default is @code{BITS_PER_WORD}.
1103 @defmac POINTERS_EXTEND_UNSIGNED
1104 A C expression whose value is greater than zero if pointers that need to be
1105 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
1106 be zero-extended and zero if they are to be sign-extended. If the value
1107 is less then zero then there must be an "ptr_extend" instruction that
1108 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
1110 You need not define this macro if the @code{POINTER_SIZE} is equal
1111 to the width of @code{Pmode}.
1114 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1115 A macro to update @var{m} and @var{unsignedp} when an object whose type
1116 is @var{type} and which has the specified mode and signedness is to be
1117 stored in a register. This macro is only called when @var{type} is a
1120 On most RISC machines, which only have operations that operate on a full
1121 register, define this macro to set @var{m} to @code{word_mode} if
1122 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1123 cases, only integer modes should be widened because wider-precision
1124 floating-point operations are usually more expensive than their narrower
1127 For most machines, the macro definition does not change @var{unsignedp}.
1128 However, some machines, have instructions that preferentially handle
1129 either signed or unsigned quantities of certain modes. For example, on
1130 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1131 sign-extend the result to 64 bits. On such machines, set
1132 @var{unsignedp} according to which kind of extension is more efficient.
1134 Do not define this macro if it would never modify @var{m}.
1137 @defmac PROMOTE_FUNCTION_MODE
1138 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
1139 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1140 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1142 The default is @code{PROMOTE_MODE}.
1145 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1146 This target hook should return @code{true} if the promotion described by
1147 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1151 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1152 This target hook should return @code{true} if the promotion described by
1153 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1156 If this target hook returns @code{true}, @code{FUNCTION_VALUE} must
1157 perform the same promotions done by @code{PROMOTE_FUNCTION_MODE}.
1160 @defmac PARM_BOUNDARY
1161 Normal alignment required for function parameters on the stack, in
1162 bits. All stack parameters receive at least this much alignment
1163 regardless of data type. On most machines, this is the same as the
1167 @defmac STACK_BOUNDARY
1168 Define this macro to the minimum alignment enforced by hardware for the
1169 stack pointer on this machine. The definition is a C expression for the
1170 desired alignment (measured in bits). This value is used as a default
1171 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1172 this should be the same as @code{PARM_BOUNDARY}.
1175 @defmac PREFERRED_STACK_BOUNDARY
1176 Define this macro if you wish to preserve a certain alignment for the
1177 stack pointer, greater than what the hardware enforces. The definition
1178 is a C expression for the desired alignment (measured in bits). This
1179 macro must evaluate to a value equal to or larger than
1180 @code{STACK_BOUNDARY}.
1183 @defmac FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1184 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1185 not guaranteed by the runtime and we should emit code to align the stack
1186 at the beginning of @code{main}.
1188 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1189 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1190 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1191 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1192 be momentarily unaligned while pushing arguments.
1195 @defmac FUNCTION_BOUNDARY
1196 Alignment required for a function entry point, in bits.
1199 @defmac BIGGEST_ALIGNMENT
1200 Biggest alignment that any data type can require on this machine, in bits.
1203 @defmac MINIMUM_ATOMIC_ALIGNMENT
1204 If defined, the smallest alignment, in bits, that can be given to an
1205 object that can be referenced in one operation, without disturbing any
1206 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1207 on machines that don't have byte or half-word store operations.
1210 @defmac BIGGEST_FIELD_ALIGNMENT
1211 Biggest alignment that any structure or union field can require on this
1212 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1213 structure and union fields only, unless the field alignment has been set
1214 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1217 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1218 An expression for the alignment of a structure field @var{field} if the
1219 alignment computed in the usual way (including applying of
1220 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1221 alignment) is @var{computed}. It overrides alignment only if the
1222 field alignment has not been set by the
1223 @code{__attribute__ ((aligned (@var{n})))} construct.
1226 @defmac MAX_OFILE_ALIGNMENT
1227 Biggest alignment supported by the object file format of this machine.
1228 Use this macro to limit the alignment which can be specified using the
1229 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1230 the default value is @code{BIGGEST_ALIGNMENT}.
1233 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1234 If defined, a C expression to compute the alignment for a variable in
1235 the static store. @var{type} is the data type, and @var{basic-align} is
1236 the alignment that the object would ordinarily have. The value of this
1237 macro is used instead of that alignment to align the object.
1239 If this macro is not defined, then @var{basic-align} is used.
1242 One use of this macro is to increase alignment of medium-size data to
1243 make it all fit in fewer cache lines. Another is to cause character
1244 arrays to be word-aligned so that @code{strcpy} calls that copy
1245 constants to character arrays can be done inline.
1248 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1249 If defined, a C expression to compute the alignment given to a constant
1250 that is being placed in memory. @var{constant} is the constant and
1251 @var{basic-align} is the alignment that the object would ordinarily
1252 have. The value of this macro is used instead of that alignment to
1255 If this macro is not defined, then @var{basic-align} is used.
1257 The typical use of this macro is to increase alignment for string
1258 constants to be word aligned so that @code{strcpy} calls that copy
1259 constants can be done inline.
1262 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1263 If defined, a C expression to compute the alignment for a variable in
1264 the local store. @var{type} is the data type, and @var{basic-align} is
1265 the alignment that the object would ordinarily have. The value of this
1266 macro is used instead of that alignment to align the object.
1268 If this macro is not defined, then @var{basic-align} is used.
1270 One use of this macro is to increase alignment of medium-size data to
1271 make it all fit in fewer cache lines.
1274 @defmac EMPTY_FIELD_BOUNDARY
1275 Alignment in bits to be given to a structure bit-field that follows an
1276 empty field such as @code{int : 0;}.
1278 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1281 @defmac STRUCTURE_SIZE_BOUNDARY
1282 Number of bits which any structure or union's size must be a multiple of.
1283 Each structure or union's size is rounded up to a multiple of this.
1285 If you do not define this macro, the default is the same as
1286 @code{BITS_PER_UNIT}.
1289 @defmac STRICT_ALIGNMENT
1290 Define this macro to be the value 1 if instructions will fail to work
1291 if given data not on the nominal alignment. If instructions will merely
1292 go slower in that case, define this macro as 0.
1295 @defmac PCC_BITFIELD_TYPE_MATTERS
1296 Define this if you wish to imitate the way many other C compilers handle
1297 alignment of bit-fields and the structures that contain them.
1299 The behavior is that the type written for a named bit-field (@code{int},
1300 @code{short}, or other integer type) imposes an alignment for the entire
1301 structure, as if the structure really did contain an ordinary field of
1302 that type. In addition, the bit-field is placed within the structure so
1303 that it would fit within such a field, not crossing a boundary for it.
1305 Thus, on most machines, a named bit-field whose type is written as
1306 @code{int} would not cross a four-byte boundary, and would force
1307 four-byte alignment for the whole structure. (The alignment used may
1308 not be four bytes; it is controlled by the other alignment parameters.)
1310 An unnamed bit-field will not affect the alignment of the containing
1313 If the macro is defined, its definition should be a C expression;
1314 a nonzero value for the expression enables this behavior.
1316 Note that if this macro is not defined, or its value is zero, some
1317 bit-fields may cross more than one alignment boundary. The compiler can
1318 support such references if there are @samp{insv}, @samp{extv}, and
1319 @samp{extzv} insns that can directly reference memory.
1321 The other known way of making bit-fields work is to define
1322 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1323 Then every structure can be accessed with fullwords.
1325 Unless the machine has bit-field instructions or you define
1326 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1327 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1329 If your aim is to make GCC use the same conventions for laying out
1330 bit-fields as are used by another compiler, here is how to investigate
1331 what the other compiler does. Compile and run this program:
1350 printf ("Size of foo1 is %d\n",
1351 sizeof (struct foo1));
1352 printf ("Size of foo2 is %d\n",
1353 sizeof (struct foo2));
1358 If this prints 2 and 5, then the compiler's behavior is what you would
1359 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1362 @defmac BITFIELD_NBYTES_LIMITED
1363 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1364 to aligning a bit-field within the structure.
1367 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1368 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1369 whether unnamed bitfields affect the alignment of the containing
1370 structure. The hook should return true if the structure should inherit
1371 the alignment requirements of an unnamed bitfield's type.
1374 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1375 Return 1 if a structure or array containing @var{field} should be accessed using
1378 If @var{field} is the only field in the structure, @var{mode} is its
1379 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1380 case where structures of one field would require the structure's mode to
1381 retain the field's mode.
1383 Normally, this is not needed. See the file @file{c4x.h} for an example
1384 of how to use this macro to prevent a structure having a floating point
1385 field from being accessed in an integer mode.
1388 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1389 Define this macro as an expression for the alignment of a type (given
1390 by @var{type} as a tree node) if the alignment computed in the usual
1391 way is @var{computed} and the alignment explicitly specified was
1394 The default is to use @var{specified} if it is larger; otherwise, use
1395 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1398 @defmac MAX_FIXED_MODE_SIZE
1399 An integer expression for the size in bits of the largest integer
1400 machine mode that should actually be used. All integer machine modes of
1401 this size or smaller can be used for structures and unions with the
1402 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1403 (DImode)} is assumed.
1406 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1407 If defined, an expression of type @code{enum machine_mode} that
1408 specifies the mode of the save area operand of a
1409 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1410 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1411 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1412 having its mode specified.
1414 You need not define this macro if it always returns @code{Pmode}. You
1415 would most commonly define this macro if the
1416 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1420 @defmac STACK_SIZE_MODE
1421 If defined, an expression of type @code{enum machine_mode} that
1422 specifies the mode of the size increment operand of an
1423 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1425 You need not define this macro if it always returns @code{word_mode}.
1426 You would most commonly define this macro if the @code{allocate_stack}
1427 pattern needs to support both a 32- and a 64-bit mode.
1430 @defmac TARGET_FLOAT_FORMAT
1431 A code distinguishing the floating point format of the target machine.
1432 There are four defined values:
1435 @item IEEE_FLOAT_FORMAT
1436 This code indicates IEEE floating point. It is the default; there is no
1437 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1439 @item VAX_FLOAT_FORMAT
1440 This code indicates the ``F float'' (for @code{float}) and ``D float''
1441 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1443 @item IBM_FLOAT_FORMAT
1444 This code indicates the format used on the IBM System/370.
1446 @item C4X_FLOAT_FORMAT
1447 This code indicates the format used on the TMS320C3x/C4x.
1450 If your target uses a floating point format other than these, you must
1451 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1452 it to @file{real.c}.
1454 The ordering of the component words of floating point values stored in
1455 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1458 @defmac MODE_HAS_NANS (@var{mode})
1459 When defined, this macro should be true if @var{mode} has a NaN
1460 representation. The compiler assumes that NaNs are not equal to
1461 anything (including themselves) and that addition, subtraction,
1462 multiplication and division all return NaNs when one operand is
1465 By default, this macro is true if @var{mode} is a floating-point
1466 mode and the target floating-point format is IEEE@.
1469 @defmac MODE_HAS_INFINITIES (@var{mode})
1470 This macro should be true if @var{mode} can represent infinity. At
1471 present, the compiler uses this macro to decide whether @samp{x - x}
1472 is always defined. By default, the macro is true when @var{mode}
1473 is a floating-point mode and the target format is IEEE@.
1476 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1477 True if @var{mode} distinguishes between positive and negative zero.
1478 The rules are expected to follow the IEEE standard:
1482 @samp{x + x} has the same sign as @samp{x}.
1485 If the sum of two values with opposite sign is zero, the result is
1486 positive for all rounding modes expect towards @minus{}infinity, for
1487 which it is negative.
1490 The sign of a product or quotient is negative when exactly one
1491 of the operands is negative.
1494 The default definition is true if @var{mode} is a floating-point
1495 mode and the target format is IEEE@.
1498 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1499 If defined, this macro should be true for @var{mode} if it has at
1500 least one rounding mode in which @samp{x} and @samp{-x} can be
1501 rounded to numbers of different magnitude. Two such modes are
1502 towards @minus{}infinity and towards +infinity.
1504 The default definition of this macro is true if @var{mode} is
1505 a floating-point mode and the target format is IEEE@.
1508 @defmac ROUND_TOWARDS_ZERO
1509 If defined, this macro should be true if the prevailing rounding
1510 mode is towards zero. A true value has the following effects:
1514 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1517 @file{libgcc.a}'s floating-point emulator will round towards zero
1518 rather than towards nearest.
1521 The compiler's floating-point emulator will round towards zero after
1522 doing arithmetic, and when converting from the internal float format to
1526 The macro does not affect the parsing of string literals. When the
1527 primary rounding mode is towards zero, library functions like
1528 @code{strtod} might still round towards nearest, and the compiler's
1529 parser should behave like the target's @code{strtod} where possible.
1531 Not defining this macro is equivalent to returning zero.
1534 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1535 This macro should return true if floats with @var{size}
1536 bits do not have a NaN or infinity representation, but use the largest
1537 exponent for normal numbers instead.
1539 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1540 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1541 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1542 floating-point arithmetic.
1544 The default definition of this macro returns false for all sizes.
1547 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1548 This target hook should return @code{true} a vector is opaque. That
1549 is, if no cast is needed when copying a vector value of type
1550 @var{type} into another vector lvalue of the same size. Vector opaque
1551 types cannot be initialized. The default is that there are no such
1555 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1556 This target hook returns @code{true} if bit-fields in the given
1557 @var{record_type} are to be laid out following the rules of Microsoft
1558 Visual C/C++, namely: (i) a bit-field won't share the same storage
1559 unit with the previous bit-field if their underlying types have
1560 different sizes, and the bit-field will be aligned to the highest
1561 alignment of the underlying types of itself and of the previous
1562 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1563 the whole enclosing structure, even if it is unnamed; except that
1564 (iii) a zero-sized bit-field will be disregarded unless it follows
1565 another bit-field of nonzero size. If this hook returns @code{true},
1566 other macros that control bit-field layout are ignored.
1568 When a bit-field is inserted into a packed record, the whole size
1569 of the underlying type is used by one or more same-size adjacent
1570 bit-fields (that is, if its long:3, 32 bits is used in the record,
1571 and any additional adjacent long bit-fields are packed into the same
1572 chunk of 32 bits. However, if the size changes, a new field of that
1573 size is allocated). In an unpacked record, this is the same as using
1574 alignment, but not equivalent when packing.
1576 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1577 the latter will take precedence. If @samp{__attribute__((packed))} is
1578 used on a single field when MS bit-fields are in use, it will take
1579 precedence for that field, but the alignment of the rest of the structure
1580 may affect its placement.
1583 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_FUNDAMENTAL_TYPE (tree @var{type})
1584 If your target defines any fundamental types, define this hook to
1585 return the appropriate encoding for these types as part of a C++
1586 mangled name. The @var{type} argument is the tree structure
1587 representing the type to be mangled. The hook may be applied to trees
1588 which are not target-specific fundamental types; it should return
1589 @code{NULL} for all such types, as well as arguments it does not
1590 recognize. If the return value is not @code{NULL}, it must point to
1591 a statically-allocated string constant.
1593 Target-specific fundamental types might be new fundamental types or
1594 qualified versions of ordinary fundamental types. Encode new
1595 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1596 is the name used for the type in source code, and @var{n} is the
1597 length of @var{name} in decimal. Encode qualified versions of
1598 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1599 @var{name} is the name used for the type qualifier in source code,
1600 @var{n} is the length of @var{name} as above, and @var{code} is the
1601 code used to represent the unqualified version of this type. (See
1602 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1603 codes.) In both cases the spaces are for clarity; do not include any
1604 spaces in your string.
1606 The default version of this hook always returns @code{NULL}, which is
1607 appropriate for a target that does not define any new fundamental
1612 @section Layout of Source Language Data Types
1614 These macros define the sizes and other characteristics of the standard
1615 basic data types used in programs being compiled. Unlike the macros in
1616 the previous section, these apply to specific features of C and related
1617 languages, rather than to fundamental aspects of storage layout.
1619 @defmac INT_TYPE_SIZE
1620 A C expression for the size in bits of the type @code{int} on the
1621 target machine. If you don't define this, the default is one word.
1624 @defmac SHORT_TYPE_SIZE
1625 A C expression for the size in bits of the type @code{short} on the
1626 target machine. If you don't define this, the default is half a word.
1627 (If this would be less than one storage unit, it is rounded up to one
1631 @defmac LONG_TYPE_SIZE
1632 A C expression for the size in bits of the type @code{long} on the
1633 target machine. If you don't define this, the default is one word.
1636 @defmac ADA_LONG_TYPE_SIZE
1637 On some machines, the size used for the Ada equivalent of the type
1638 @code{long} by a native Ada compiler differs from that used by C@. In
1639 that situation, define this macro to be a C expression to be used for
1640 the size of that type. If you don't define this, the default is the
1641 value of @code{LONG_TYPE_SIZE}.
1644 @defmac LONG_LONG_TYPE_SIZE
1645 A C expression for the size in bits of the type @code{long long} on the
1646 target machine. If you don't define this, the default is two
1647 words. If you want to support GNU Ada on your machine, the value of this
1648 macro must be at least 64.
1651 @defmac CHAR_TYPE_SIZE
1652 A C expression for the size in bits of the type @code{char} on the
1653 target machine. If you don't define this, the default is
1654 @code{BITS_PER_UNIT}.
1657 @defmac BOOL_TYPE_SIZE
1658 A C expression for the size in bits of the C++ type @code{bool} and
1659 C99 type @code{_Bool} on the target machine. If you don't define
1660 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1663 @defmac FLOAT_TYPE_SIZE
1664 A C expression for the size in bits of the type @code{float} on the
1665 target machine. If you don't define this, the default is one word.
1668 @defmac DOUBLE_TYPE_SIZE
1669 A C expression for the size in bits of the type @code{double} on the
1670 target machine. If you don't define this, the default is two
1674 @defmac LONG_DOUBLE_TYPE_SIZE
1675 A C expression for the size in bits of the type @code{long double} on
1676 the target machine. If you don't define this, the default is two
1680 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1681 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1682 if you want routines in @file{libgcc2.a} for a size other than
1683 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1684 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1687 @defmac LIBGCC2_HAS_DF_MODE
1688 Define this macro if neither @code{LIBGCC2_DOUBLE_TYPE_SIZE} nor
1689 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1690 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1691 anyway. If you don't define this and either @code{LIBGCC2_DOUBLE_TYPE_SIZE}
1692 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1696 @defmac LIBGCC2_HAS_XF_MODE
1697 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1698 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1699 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1700 is 80 then the default is 1, otherwise it is 0.
1703 @defmac LIBGCC2_HAS_TF_MODE
1704 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1705 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1706 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1707 is 128 then the default is 1, otherwise it is 0.
1710 @defmac TARGET_FLT_EVAL_METHOD
1711 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1712 assuming, if applicable, that the floating-point control word is in its
1713 default state. If you do not define this macro the value of
1714 @code{FLT_EVAL_METHOD} will be zero.
1717 @defmac WIDEST_HARDWARE_FP_SIZE
1718 A C expression for the size in bits of the widest floating-point format
1719 supported by the hardware. If you define this macro, you must specify a
1720 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1721 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1725 @defmac DEFAULT_SIGNED_CHAR
1726 An expression whose value is 1 or 0, according to whether the type
1727 @code{char} should be signed or unsigned by default. The user can
1728 always override this default with the options @option{-fsigned-char}
1729 and @option{-funsigned-char}.
1732 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1733 This target hook should return true if the compiler should give an
1734 @code{enum} type only as many bytes as it takes to represent the range
1735 of possible values of that type. It should return false if all
1736 @code{enum} types should be allocated like @code{int}.
1738 The default is to return false.
1742 A C expression for a string describing the name of the data type to use
1743 for size values. The typedef name @code{size_t} is defined using the
1744 contents of the string.
1746 The string can contain more than one keyword. If so, separate them with
1747 spaces, and write first any length keyword, then @code{unsigned} if
1748 appropriate, and finally @code{int}. The string must exactly match one
1749 of the data type names defined in the function
1750 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1751 omit @code{int} or change the order---that would cause the compiler to
1754 If you don't define this macro, the default is @code{"long unsigned
1758 @defmac PTRDIFF_TYPE
1759 A C expression for a string describing the name of the data type to use
1760 for the result of subtracting two pointers. The typedef name
1761 @code{ptrdiff_t} is defined using the contents of the string. See
1762 @code{SIZE_TYPE} above for more information.
1764 If you don't define this macro, the default is @code{"long int"}.
1768 A C expression for a string describing the name of the data type to use
1769 for wide characters. The typedef name @code{wchar_t} is defined using
1770 the contents of the string. See @code{SIZE_TYPE} above for more
1773 If you don't define this macro, the default is @code{"int"}.
1776 @defmac WCHAR_TYPE_SIZE
1777 A C expression for the size in bits of the data type for wide
1778 characters. This is used in @code{cpp}, which cannot make use of
1783 A C expression for a string describing the name of the data type to
1784 use for wide characters passed to @code{printf} and returned from
1785 @code{getwc}. The typedef name @code{wint_t} is defined using the
1786 contents of the string. See @code{SIZE_TYPE} above for more
1789 If you don't define this macro, the default is @code{"unsigned int"}.
1793 A C expression for a string describing the name of the data type that
1794 can represent any value of any standard or extended signed integer type.
1795 The typedef name @code{intmax_t} is defined using the contents of the
1796 string. See @code{SIZE_TYPE} above for more information.
1798 If you don't define this macro, the default is the first of
1799 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1800 much precision as @code{long long int}.
1803 @defmac UINTMAX_TYPE
1804 A C expression for a string describing the name of the data type that
1805 can represent any value of any standard or extended unsigned integer
1806 type. The typedef name @code{uintmax_t} is defined using the contents
1807 of the string. See @code{SIZE_TYPE} above for more information.
1809 If you don't define this macro, the default is the first of
1810 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1811 unsigned int"} that has as much precision as @code{long long unsigned
1815 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1816 The C++ compiler represents a pointer-to-member-function with a struct
1823 ptrdiff_t vtable_index;
1830 The C++ compiler must use one bit to indicate whether the function that
1831 will be called through a pointer-to-member-function is virtual.
1832 Normally, we assume that the low-order bit of a function pointer must
1833 always be zero. Then, by ensuring that the vtable_index is odd, we can
1834 distinguish which variant of the union is in use. But, on some
1835 platforms function pointers can be odd, and so this doesn't work. In
1836 that case, we use the low-order bit of the @code{delta} field, and shift
1837 the remainder of the @code{delta} field to the left.
1839 GCC will automatically make the right selection about where to store
1840 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1841 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1842 set such that functions always start at even addresses, but the lowest
1843 bit of pointers to functions indicate whether the function at that
1844 address is in ARM or Thumb mode. If this is the case of your
1845 architecture, you should define this macro to
1846 @code{ptrmemfunc_vbit_in_delta}.
1848 In general, you should not have to define this macro. On architectures
1849 in which function addresses are always even, according to
1850 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1851 @code{ptrmemfunc_vbit_in_pfn}.
1854 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1855 Normally, the C++ compiler uses function pointers in vtables. This
1856 macro allows the target to change to use ``function descriptors''
1857 instead. Function descriptors are found on targets for whom a
1858 function pointer is actually a small data structure. Normally the
1859 data structure consists of the actual code address plus a data
1860 pointer to which the function's data is relative.
1862 If vtables are used, the value of this macro should be the number
1863 of words that the function descriptor occupies.
1866 @defmac TARGET_VTABLE_ENTRY_ALIGN
1867 By default, the vtable entries are void pointers, the so the alignment
1868 is the same as pointer alignment. The value of this macro specifies
1869 the alignment of the vtable entry in bits. It should be defined only
1870 when special alignment is necessary. */
1873 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1874 There are a few non-descriptor entries in the vtable at offsets below
1875 zero. If these entries must be padded (say, to preserve the alignment
1876 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1877 of words in each data entry.
1881 @section Register Usage
1882 @cindex register usage
1884 This section explains how to describe what registers the target machine
1885 has, and how (in general) they can be used.
1887 The description of which registers a specific instruction can use is
1888 done with register classes; see @ref{Register Classes}. For information
1889 on using registers to access a stack frame, see @ref{Frame Registers}.
1890 For passing values in registers, see @ref{Register Arguments}.
1891 For returning values in registers, see @ref{Scalar Return}.
1894 * Register Basics:: Number and kinds of registers.
1895 * Allocation Order:: Order in which registers are allocated.
1896 * Values in Registers:: What kinds of values each reg can hold.
1897 * Leaf Functions:: Renumbering registers for leaf functions.
1898 * Stack Registers:: Handling a register stack such as 80387.
1901 @node Register Basics
1902 @subsection Basic Characteristics of Registers
1904 @c prevent bad page break with this line
1905 Registers have various characteristics.
1907 @defmac FIRST_PSEUDO_REGISTER
1908 Number of hardware registers known to the compiler. They receive
1909 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1910 pseudo register's number really is assigned the number
1911 @code{FIRST_PSEUDO_REGISTER}.
1914 @defmac FIXED_REGISTERS
1915 @cindex fixed register
1916 An initializer that says which registers are used for fixed purposes
1917 all throughout the compiled code and are therefore not available for
1918 general allocation. These would include the stack pointer, the frame
1919 pointer (except on machines where that can be used as a general
1920 register when no frame pointer is needed), the program counter on
1921 machines where that is considered one of the addressable registers,
1922 and any other numbered register with a standard use.
1924 This information is expressed as a sequence of numbers, separated by
1925 commas and surrounded by braces. The @var{n}th number is 1 if
1926 register @var{n} is fixed, 0 otherwise.
1928 The table initialized from this macro, and the table initialized by
1929 the following one, may be overridden at run time either automatically,
1930 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1931 the user with the command options @option{-ffixed-@var{reg}},
1932 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1935 @defmac CALL_USED_REGISTERS
1936 @cindex call-used register
1937 @cindex call-clobbered register
1938 @cindex call-saved register
1939 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1940 clobbered (in general) by function calls as well as for fixed
1941 registers. This macro therefore identifies the registers that are not
1942 available for general allocation of values that must live across
1945 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1946 automatically saves it on function entry and restores it on function
1947 exit, if the register is used within the function.
1950 @defmac CALL_REALLY_USED_REGISTERS
1951 @cindex call-used register
1952 @cindex call-clobbered register
1953 @cindex call-saved register
1954 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1955 that the entire set of @code{FIXED_REGISTERS} be included.
1956 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1957 This macro is optional. If not specified, it defaults to the value
1958 of @code{CALL_USED_REGISTERS}.
1961 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1962 @cindex call-used register
1963 @cindex call-clobbered register
1964 @cindex call-saved register
1965 A C expression that is nonzero if it is not permissible to store a
1966 value of mode @var{mode} in hard register number @var{regno} across a
1967 call without some part of it being clobbered. For most machines this
1968 macro need not be defined. It is only required for machines that do not
1969 preserve the entire contents of a register across a call.
1973 @findex call_used_regs
1976 @findex reg_class_contents
1977 @defmac CONDITIONAL_REGISTER_USAGE
1978 Zero or more C statements that may conditionally modify five variables
1979 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1980 @code{reg_names}, and @code{reg_class_contents}, to take into account
1981 any dependence of these register sets on target flags. The first three
1982 of these are of type @code{char []} (interpreted as Boolean vectors).
1983 @code{global_regs} is a @code{const char *[]}, and
1984 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1985 called, @code{fixed_regs}, @code{call_used_regs},
1986 @code{reg_class_contents}, and @code{reg_names} have been initialized
1987 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1988 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1989 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1990 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1991 command options have been applied.
1993 You need not define this macro if it has no work to do.
1995 @cindex disabling certain registers
1996 @cindex controlling register usage
1997 If the usage of an entire class of registers depends on the target
1998 flags, you may indicate this to GCC by using this macro to modify
1999 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
2000 registers in the classes which should not be used by GCC@. Also define
2001 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
2002 to return @code{NO_REGS} if it
2003 is called with a letter for a class that shouldn't be used.
2005 (However, if this class is not included in @code{GENERAL_REGS} and all
2006 of the insn patterns whose constraints permit this class are
2007 controlled by target switches, then GCC will automatically avoid using
2008 these registers when the target switches are opposed to them.)
2011 @defmac INCOMING_REGNO (@var{out})
2012 Define this macro if the target machine has register windows. This C
2013 expression returns the register number as seen by the called function
2014 corresponding to the register number @var{out} as seen by the calling
2015 function. Return @var{out} if register number @var{out} is not an
2019 @defmac OUTGOING_REGNO (@var{in})
2020 Define this macro if the target machine has register windows. This C
2021 expression returns the register number as seen by the calling function
2022 corresponding to the register number @var{in} as seen by the called
2023 function. Return @var{in} if register number @var{in} is not an inbound
2027 @defmac LOCAL_REGNO (@var{regno})
2028 Define this macro if the target machine has register windows. This C
2029 expression returns true if the register is call-saved but is in the
2030 register window. Unlike most call-saved registers, such registers
2031 need not be explicitly restored on function exit or during non-local
2036 If the program counter has a register number, define this as that
2037 register number. Otherwise, do not define it.
2040 @node Allocation Order
2041 @subsection Order of Allocation of Registers
2042 @cindex order of register allocation
2043 @cindex register allocation order
2045 @c prevent bad page break with this line
2046 Registers are allocated in order.
2048 @defmac REG_ALLOC_ORDER
2049 If defined, an initializer for a vector of integers, containing the
2050 numbers of hard registers in the order in which GCC should prefer
2051 to use them (from most preferred to least).
2053 If this macro is not defined, registers are used lowest numbered first
2054 (all else being equal).
2056 One use of this macro is on machines where the highest numbered
2057 registers must always be saved and the save-multiple-registers
2058 instruction supports only sequences of consecutive registers. On such
2059 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2060 the highest numbered allocable register first.
2063 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
2064 A C statement (sans semicolon) to choose the order in which to allocate
2065 hard registers for pseudo-registers local to a basic block.
2067 Store the desired register order in the array @code{reg_alloc_order}.
2068 Element 0 should be the register to allocate first; element 1, the next
2069 register; and so on.
2071 The macro body should not assume anything about the contents of
2072 @code{reg_alloc_order} before execution of the macro.
2074 On most machines, it is not necessary to define this macro.
2077 @node Values in Registers
2078 @subsection How Values Fit in Registers
2080 This section discusses the macros that describe which kinds of values
2081 (specifically, which machine modes) each register can hold, and how many
2082 consecutive registers are needed for a given mode.
2084 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2085 A C expression for the number of consecutive hard registers, starting
2086 at register number @var{regno}, required to hold a value of mode
2089 On a machine where all registers are exactly one word, a suitable
2090 definition of this macro is
2093 #define HARD_REGNO_NREGS(REGNO, MODE) \
2094 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2099 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2100 Define this macro if the natural size of registers that hold values
2101 of mode @var{mode} is not the word size. It is a C expression that
2102 should give the natural size in bytes for the specified mode. It is
2103 used by the register allocator to try to optimize its results. This
2104 happens for example on SPARC 64-bit where the natural size of
2105 floating-point registers is still 32-bit.
2108 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2109 A C expression that is nonzero if it is permissible to store a value
2110 of mode @var{mode} in hard register number @var{regno} (or in several
2111 registers starting with that one). For a machine where all registers
2112 are equivalent, a suitable definition is
2115 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2118 You need not include code to check for the numbers of fixed registers,
2119 because the allocation mechanism considers them to be always occupied.
2121 @cindex register pairs
2122 On some machines, double-precision values must be kept in even/odd
2123 register pairs. You can implement that by defining this macro to reject
2124 odd register numbers for such modes.
2126 The minimum requirement for a mode to be OK in a register is that the
2127 @samp{mov@var{mode}} instruction pattern support moves between the
2128 register and other hard register in the same class and that moving a
2129 value into the register and back out not alter it.
2131 Since the same instruction used to move @code{word_mode} will work for
2132 all narrower integer modes, it is not necessary on any machine for
2133 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2134 you define patterns @samp{movhi}, etc., to take advantage of this. This
2135 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2136 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2139 Many machines have special registers for floating point arithmetic.
2140 Often people assume that floating point machine modes are allowed only
2141 in floating point registers. This is not true. Any registers that
2142 can hold integers can safely @emph{hold} a floating point machine
2143 mode, whether or not floating arithmetic can be done on it in those
2144 registers. Integer move instructions can be used to move the values.
2146 On some machines, though, the converse is true: fixed-point machine
2147 modes may not go in floating registers. This is true if the floating
2148 registers normalize any value stored in them, because storing a
2149 non-floating value there would garble it. In this case,
2150 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2151 floating registers. But if the floating registers do not automatically
2152 normalize, if you can store any bit pattern in one and retrieve it
2153 unchanged without a trap, then any machine mode may go in a floating
2154 register, so you can define this macro to say so.
2156 The primary significance of special floating registers is rather that
2157 they are the registers acceptable in floating point arithmetic
2158 instructions. However, this is of no concern to
2159 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2160 constraints for those instructions.
2162 On some machines, the floating registers are especially slow to access,
2163 so that it is better to store a value in a stack frame than in such a
2164 register if floating point arithmetic is not being done. As long as the
2165 floating registers are not in class @code{GENERAL_REGS}, they will not
2166 be used unless some pattern's constraint asks for one.
2169 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2170 A C expression that is nonzero if it is OK to rename a hard register
2171 @var{from} to another hard register @var{to}.
2173 One common use of this macro is to prevent renaming of a register to
2174 another register that is not saved by a prologue in an interrupt
2177 The default is always nonzero.
2180 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2181 A C expression that is nonzero if a value of mode
2182 @var{mode1} is accessible in mode @var{mode2} without copying.
2184 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2185 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2186 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2187 should be nonzero. If they differ for any @var{r}, you should define
2188 this macro to return zero unless some other mechanism ensures the
2189 accessibility of the value in a narrower mode.
2191 You should define this macro to return nonzero in as many cases as
2192 possible since doing so will allow GCC to perform better register
2196 @defmac AVOID_CCMODE_COPIES
2197 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2198 registers. You should only define this macro if support for copying to/from
2199 @code{CCmode} is incomplete.
2202 @node Leaf Functions
2203 @subsection Handling Leaf Functions
2205 @cindex leaf functions
2206 @cindex functions, leaf
2207 On some machines, a leaf function (i.e., one which makes no calls) can run
2208 more efficiently if it does not make its own register window. Often this
2209 means it is required to receive its arguments in the registers where they
2210 are passed by the caller, instead of the registers where they would
2213 The special treatment for leaf functions generally applies only when
2214 other conditions are met; for example, often they may use only those
2215 registers for its own variables and temporaries. We use the term ``leaf
2216 function'' to mean a function that is suitable for this special
2217 handling, so that functions with no calls are not necessarily ``leaf
2220 GCC assigns register numbers before it knows whether the function is
2221 suitable for leaf function treatment. So it needs to renumber the
2222 registers in order to output a leaf function. The following macros
2225 @defmac LEAF_REGISTERS
2226 Name of a char vector, indexed by hard register number, which
2227 contains 1 for a register that is allowable in a candidate for leaf
2230 If leaf function treatment involves renumbering the registers, then the
2231 registers marked here should be the ones before renumbering---those that
2232 GCC would ordinarily allocate. The registers which will actually be
2233 used in the assembler code, after renumbering, should not be marked with 1
2236 Define this macro only if the target machine offers a way to optimize
2237 the treatment of leaf functions.
2240 @defmac LEAF_REG_REMAP (@var{regno})
2241 A C expression whose value is the register number to which @var{regno}
2242 should be renumbered, when a function is treated as a leaf function.
2244 If @var{regno} is a register number which should not appear in a leaf
2245 function before renumbering, then the expression should yield @minus{}1, which
2246 will cause the compiler to abort.
2248 Define this macro only if the target machine offers a way to optimize the
2249 treatment of leaf functions, and registers need to be renumbered to do
2253 @findex current_function_is_leaf
2254 @findex current_function_uses_only_leaf_regs
2255 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2256 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2257 specially. They can test the C variable @code{current_function_is_leaf}
2258 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2259 set prior to local register allocation and is valid for the remaining
2260 compiler passes. They can also test the C variable
2261 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2262 functions which only use leaf registers.
2263 @code{current_function_uses_only_leaf_regs} is valid after all passes
2264 that modify the instructions have been run and is only useful if
2265 @code{LEAF_REGISTERS} is defined.
2266 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2267 @c of the next paragraph?! --mew 2feb93
2269 @node Stack Registers
2270 @subsection Registers That Form a Stack
2272 There are special features to handle computers where some of the
2273 ``registers'' form a stack. Stack registers are normally written by
2274 pushing onto the stack, and are numbered relative to the top of the
2277 Currently, GCC can only handle one group of stack-like registers, and
2278 they must be consecutively numbered. Furthermore, the existing
2279 support for stack-like registers is specific to the 80387 floating
2280 point coprocessor. If you have a new architecture that uses
2281 stack-like registers, you will need to do substantial work on
2282 @file{reg-stack.c} and write your machine description to cooperate
2283 with it, as well as defining these macros.
2286 Define this if the machine has any stack-like registers.
2289 @defmac FIRST_STACK_REG
2290 The number of the first stack-like register. This one is the top
2294 @defmac LAST_STACK_REG
2295 The number of the last stack-like register. This one is the bottom of
2299 @node Register Classes
2300 @section Register Classes
2301 @cindex register class definitions
2302 @cindex class definitions, register
2304 On many machines, the numbered registers are not all equivalent.
2305 For example, certain registers may not be allowed for indexed addressing;
2306 certain registers may not be allowed in some instructions. These machine
2307 restrictions are described to the compiler using @dfn{register classes}.
2309 You define a number of register classes, giving each one a name and saying
2310 which of the registers belong to it. Then you can specify register classes
2311 that are allowed as operands to particular instruction patterns.
2315 In general, each register will belong to several classes. In fact, one
2316 class must be named @code{ALL_REGS} and contain all the registers. Another
2317 class must be named @code{NO_REGS} and contain no registers. Often the
2318 union of two classes will be another class; however, this is not required.
2320 @findex GENERAL_REGS
2321 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2322 terribly special about the name, but the operand constraint letters
2323 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2324 the same as @code{ALL_REGS}, just define it as a macro which expands
2327 Order the classes so that if class @var{x} is contained in class @var{y}
2328 then @var{x} has a lower class number than @var{y}.
2330 The way classes other than @code{GENERAL_REGS} are specified in operand
2331 constraints is through machine-dependent operand constraint letters.
2332 You can define such letters to correspond to various classes, then use
2333 them in operand constraints.
2335 You should define a class for the union of two classes whenever some
2336 instruction allows both classes. For example, if an instruction allows
2337 either a floating point (coprocessor) register or a general register for a
2338 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2339 which includes both of them. Otherwise you will get suboptimal code.
2341 You must also specify certain redundant information about the register
2342 classes: for each class, which classes contain it and which ones are
2343 contained in it; for each pair of classes, the largest class contained
2346 When a value occupying several consecutive registers is expected in a
2347 certain class, all the registers used must belong to that class.
2348 Therefore, register classes cannot be used to enforce a requirement for
2349 a register pair to start with an even-numbered register. The way to
2350 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2352 Register classes used for input-operands of bitwise-and or shift
2353 instructions have a special requirement: each such class must have, for
2354 each fixed-point machine mode, a subclass whose registers can transfer that
2355 mode to or from memory. For example, on some machines, the operations for
2356 single-byte values (@code{QImode}) are limited to certain registers. When
2357 this is so, each register class that is used in a bitwise-and or shift
2358 instruction must have a subclass consisting of registers from which
2359 single-byte values can be loaded or stored. This is so that
2360 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2362 @deftp {Data type} {enum reg_class}
2363 An enumerated type that must be defined with all the register class names
2364 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2365 must be the last register class, followed by one more enumerated value,
2366 @code{LIM_REG_CLASSES}, which is not a register class but rather
2367 tells how many classes there are.
2369 Each register class has a number, which is the value of casting
2370 the class name to type @code{int}. The number serves as an index
2371 in many of the tables described below.
2374 @defmac N_REG_CLASSES
2375 The number of distinct register classes, defined as follows:
2378 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2382 @defmac REG_CLASS_NAMES
2383 An initializer containing the names of the register classes as C string
2384 constants. These names are used in writing some of the debugging dumps.
2387 @defmac REG_CLASS_CONTENTS
2388 An initializer containing the contents of the register classes, as integers
2389 which are bit masks. The @var{n}th integer specifies the contents of class
2390 @var{n}. The way the integer @var{mask} is interpreted is that
2391 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2393 When the machine has more than 32 registers, an integer does not suffice.
2394 Then the integers are replaced by sub-initializers, braced groupings containing
2395 several integers. Each sub-initializer must be suitable as an initializer
2396 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2397 In this situation, the first integer in each sub-initializer corresponds to
2398 registers 0 through 31, the second integer to registers 32 through 63, and
2402 @defmac REGNO_REG_CLASS (@var{regno})
2403 A C expression whose value is a register class containing hard register
2404 @var{regno}. In general there is more than one such class; choose a class
2405 which is @dfn{minimal}, meaning that no smaller class also contains the
2409 @defmac BASE_REG_CLASS
2410 A macro whose definition is the name of the class to which a valid
2411 base register must belong. A base register is one used in an address
2412 which is the register value plus a displacement.
2415 @defmac MODE_BASE_REG_CLASS (@var{mode})
2416 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2417 the selection of a base register in a mode dependent manner. If
2418 @var{mode} is VOIDmode then it should return the same value as
2419 @code{BASE_REG_CLASS}.
2422 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2423 A C expression whose value is the register class to which a valid
2424 base register must belong in order to be used in a base plus index
2425 register address. You should define this macro if base plus index
2426 addresses have different requirements than other base register uses.
2429 @defmac INDEX_REG_CLASS
2430 A macro whose definition is the name of the class to which a valid
2431 index register must belong. An index register is one used in an
2432 address where its value is either multiplied by a scale factor or
2433 added to another register (as well as added to a displacement).
2436 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2437 For the constraint at the start of @var{str}, which starts with the letter
2438 @var{c}, return the length. This allows you to have register class /
2439 constant / extra constraints that are longer than a single letter;
2440 you don't need to define this macro if you can do with single-letter
2441 constraints only. The definition of this macro should use
2442 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2443 to handle specially.
2444 There are some sanity checks in genoutput.c that check the constraint lengths
2445 for the md file, so you can also use this macro to help you while you are
2446 transitioning from a byzantine single-letter-constraint scheme: when you
2447 return a negative length for a constraint you want to re-use, genoutput
2448 will complain about every instance where it is used in the md file.
2451 @defmac REG_CLASS_FROM_LETTER (@var{char})
2452 A C expression which defines the machine-dependent operand constraint
2453 letters for register classes. If @var{char} is such a letter, the
2454 value should be the register class corresponding to it. Otherwise,
2455 the value should be @code{NO_REGS}. The register letter @samp{r},
2456 corresponding to class @code{GENERAL_REGS}, will not be passed
2457 to this macro; you do not need to handle it.
2460 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2461 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2462 passed in @var{str}, so that you can use suffixes to distinguish between
2466 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2467 A C expression which is nonzero if register number @var{num} is
2468 suitable for use as a base register in operand addresses. It may be
2469 either a suitable hard register or a pseudo register that has been
2470 allocated such a hard register.
2473 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2474 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2475 that expression may examine the mode of the memory reference in
2476 @var{mode}. You should define this macro if the mode of the memory
2477 reference affects whether a register may be used as a base register. If
2478 you define this macro, the compiler will use it instead of
2479 @code{REGNO_OK_FOR_BASE_P}.
2482 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2483 A C expression which is nonzero if register number @var{num} is suitable for
2484 use as a base register in base plus index operand addresses, accessing
2485 memory in mode @var{mode}. It may be either a suitable hard register or a
2486 pseudo register that has been allocated such a hard register. You should
2487 define this macro if base plus index addresses have different requirements
2488 than other base register uses.
2491 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2492 A C expression which is nonzero if register number @var{num} is
2493 suitable for use as an index register in operand addresses. It may be
2494 either a suitable hard register or a pseudo register that has been
2495 allocated such a hard register.
2497 The difference between an index register and a base register is that
2498 the index register may be scaled. If an address involves the sum of
2499 two registers, neither one of them scaled, then either one may be
2500 labeled the ``base'' and the other the ``index''; but whichever
2501 labeling is used must fit the machine's constraints of which registers
2502 may serve in each capacity. The compiler will try both labelings,
2503 looking for one that is valid, and will reload one or both registers
2504 only if neither labeling works.
2507 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2508 A C expression that places additional restrictions on the register class
2509 to use when it is necessary to copy value @var{x} into a register in class
2510 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2511 another, smaller class. On many machines, the following definition is
2515 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2518 Sometimes returning a more restrictive class makes better code. For
2519 example, on the 68000, when @var{x} is an integer constant that is in range
2520 for a @samp{moveq} instruction, the value of this macro is always
2521 @code{DATA_REGS} as long as @var{class} includes the data registers.
2522 Requiring a data register guarantees that a @samp{moveq} will be used.
2524 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2525 @var{class} is if @var{x} is a legitimate constant which cannot be
2526 loaded into some register class. By returning @code{NO_REGS} you can
2527 force @var{x} into a memory location. For example, rs6000 can load
2528 immediate values into general-purpose registers, but does not have an
2529 instruction for loading an immediate value into a floating-point
2530 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2531 @var{x} is a floating-point constant. If the constant can't be loaded
2532 into any kind of register, code generation will be better if
2533 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2534 of using @code{PREFERRED_RELOAD_CLASS}.
2537 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2538 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2539 input reloads. If you don't define this macro, the default is to use
2540 @var{class}, unchanged.
2543 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2544 A C expression that places additional restrictions on the register class
2545 to use when it is necessary to be able to hold a value of mode
2546 @var{mode} in a reload register for which class @var{class} would
2549 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2550 there are certain modes that simply can't go in certain reload classes.
2552 The value is a register class; perhaps @var{class}, or perhaps another,
2555 Don't define this macro unless the target machine has limitations which
2556 require the macro to do something nontrivial.
2559 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2560 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2561 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2562 Many machines have some registers that cannot be copied directly to or
2563 from memory or even from other types of registers. An example is the
2564 @samp{MQ} register, which on most machines, can only be copied to or
2565 from general registers, but not memory. Some machines allow copying all
2566 registers to and from memory, but require a scratch register for stores
2567 to some memory locations (e.g., those with symbolic address on the RT,
2568 and those with certain symbolic address on the SPARC when compiling
2569 PIC)@. In some cases, both an intermediate and a scratch register are
2572 You should define these macros to indicate to the reload phase that it may
2573 need to allocate at least one register for a reload in addition to the
2574 register to contain the data. Specifically, if copying @var{x} to a
2575 register @var{class} in @var{mode} requires an intermediate register,
2576 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2577 largest register class all of whose registers can be used as
2578 intermediate registers or scratch registers.
2580 If copying a register @var{class} in @var{mode} to @var{x} requires an
2581 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2582 should be defined to return the largest register class required. If the
2583 requirements for input and output reloads are the same, the macro
2584 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2587 The values returned by these macros are often @code{GENERAL_REGS}.
2588 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2589 can be directly copied to or from a register of @var{class} in
2590 @var{mode} without requiring a scratch register. Do not define this
2591 macro if it would always return @code{NO_REGS}.
2593 If a scratch register is required (either with or without an
2594 intermediate register), you should define patterns for
2595 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2596 (@pxref{Standard Names}. These patterns, which will normally be
2597 implemented with a @code{define_expand}, should be similar to the
2598 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2601 Define constraints for the reload register and scratch register that
2602 contain a single register class. If the original reload register (whose
2603 class is @var{class}) can meet the constraint given in the pattern, the
2604 value returned by these macros is used for the class of the scratch
2605 register. Otherwise, two additional reload registers are required.
2606 Their classes are obtained from the constraints in the insn pattern.
2608 @var{x} might be a pseudo-register or a @code{subreg} of a
2609 pseudo-register, which could either be in a hard register or in memory.
2610 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2611 in memory and the hard register number if it is in a register.
2613 These macros should not be used in the case where a particular class of
2614 registers can only be copied to memory and not to another class of
2615 registers. In that case, secondary reload registers are not needed and
2616 would not be helpful. Instead, a stack location must be used to perform
2617 the copy and the @code{mov@var{m}} pattern should use memory as an
2618 intermediate storage. This case often occurs between floating-point and
2622 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2623 Certain machines have the property that some registers cannot be copied
2624 to some other registers without using memory. Define this macro on
2625 those machines to be a C expression that is nonzero if objects of mode
2626 @var{m} in registers of @var{class1} can only be copied to registers of
2627 class @var{class2} by storing a register of @var{class1} into memory
2628 and loading that memory location into a register of @var{class2}.
2630 Do not define this macro if its value would always be zero.
2633 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2634 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2635 allocates a stack slot for a memory location needed for register copies.
2636 If this macro is defined, the compiler instead uses the memory location
2637 defined by this macro.
2639 Do not define this macro if you do not define
2640 @code{SECONDARY_MEMORY_NEEDED}.
2643 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2644 When the compiler needs a secondary memory location to copy between two
2645 registers of mode @var{mode}, it normally allocates sufficient memory to
2646 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2647 load operations in a mode that many bits wide and whose class is the
2648 same as that of @var{mode}.
2650 This is right thing to do on most machines because it ensures that all
2651 bits of the register are copied and prevents accesses to the registers
2652 in a narrower mode, which some machines prohibit for floating-point
2655 However, this default behavior is not correct on some machines, such as
2656 the DEC Alpha, that store short integers in floating-point registers
2657 differently than in integer registers. On those machines, the default
2658 widening will not work correctly and you must define this macro to
2659 suppress that widening in some cases. See the file @file{alpha.h} for
2662 Do not define this macro if you do not define
2663 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2664 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2667 @defmac SMALL_REGISTER_CLASSES
2668 On some machines, it is risky to let hard registers live across arbitrary
2669 insns. Typically, these machines have instructions that require values
2670 to be in specific registers (like an accumulator), and reload will fail
2671 if the required hard register is used for another purpose across such an
2674 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2675 value on these machines. When this macro has a nonzero value, the
2676 compiler will try to minimize the lifetime of hard registers.
2678 It is always safe to define this macro with a nonzero value, but if you
2679 unnecessarily define it, you will reduce the amount of optimizations
2680 that can be performed in some cases. If you do not define this macro
2681 with a nonzero value when it is required, the compiler will run out of
2682 spill registers and print a fatal error message. For most machines, you
2683 should not define this macro at all.
2686 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2687 A C expression whose value is nonzero if pseudos that have been assigned
2688 to registers of class @var{class} would likely be spilled because
2689 registers of @var{class} are needed for spill registers.
2691 The default value of this macro returns 1 if @var{class} has exactly one
2692 register and zero otherwise. On most machines, this default should be
2693 used. Only define this macro to some other expression if pseudos
2694 allocated by @file{local-alloc.c} end up in memory because their hard
2695 registers were needed for spill registers. If this macro returns nonzero
2696 for those classes, those pseudos will only be allocated by
2697 @file{global.c}, which knows how to reallocate the pseudo to another
2698 register. If there would not be another register available for
2699 reallocation, you should not change the definition of this macro since
2700 the only effect of such a definition would be to slow down register
2704 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2705 A C expression for the maximum number of consecutive registers
2706 of class @var{class} needed to hold a value of mode @var{mode}.
2708 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2709 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2710 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2711 @var{mode})} for all @var{regno} values in the class @var{class}.
2713 This macro helps control the handling of multiple-word values
2717 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2718 If defined, a C expression that returns nonzero for a @var{class} for which
2719 a change from mode @var{from} to mode @var{to} is invalid.
2721 For the example, loading 32-bit integer or floating-point objects into
2722 floating-point registers on the Alpha extends them to 64 bits.
2723 Therefore loading a 64-bit object and then storing it as a 32-bit object
2724 does not store the low-order 32 bits, as would be the case for a normal
2725 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2729 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2730 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2731 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2735 Three other special macros describe which operands fit which constraint
2738 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2739 A C expression that defines the machine-dependent operand constraint
2740 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2741 particular ranges of integer values. If @var{c} is one of those
2742 letters, the expression should check that @var{value}, an integer, is in
2743 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2744 not one of those letters, the value should be 0 regardless of
2748 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2749 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2750 string passed in @var{str}, so that you can use suffixes to distinguish
2751 between different variants.
2754 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2755 A C expression that defines the machine-dependent operand constraint
2756 letters that specify particular ranges of @code{const_double} values
2757 (@samp{G} or @samp{H}).
2759 If @var{c} is one of those letters, the expression should check that
2760 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2761 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2762 letters, the value should be 0 regardless of @var{value}.
2764 @code{const_double} is used for all floating-point constants and for
2765 @code{DImode} fixed-point constants. A given letter can accept either
2766 or both kinds of values. It can use @code{GET_MODE} to distinguish
2767 between these kinds.
2770 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2771 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2772 string passed in @var{str}, so that you can use suffixes to distinguish
2773 between different variants.
2776 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2777 A C expression that defines the optional machine-dependent constraint
2778 letters that can be used to segregate specific types of operands, usually
2779 memory references, for the target machine. Any letter that is not
2780 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2781 @code{REG_CLASS_FROM_CONSTRAINT}
2782 may be used. Normally this macro will not be defined.
2784 If it is required for a particular target machine, it should return 1
2785 if @var{value} corresponds to the operand type represented by the
2786 constraint letter @var{c}. If @var{c} is not defined as an extra
2787 constraint, the value returned should be 0 regardless of @var{value}.
2789 For example, on the ROMP, load instructions cannot have their output
2790 in r0 if the memory reference contains a symbolic address. Constraint
2791 letter @samp{Q} is defined as representing a memory address that does
2792 @emph{not} contain a symbolic address. An alternative is specified with
2793 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2794 alternative specifies @samp{m} on the input and a register class that
2795 does not include r0 on the output.
2798 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2799 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2800 in @var{str}, so that you can use suffixes to distinguish between different
2804 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2805 A C expression that defines the optional machine-dependent constraint
2806 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2807 be treated like memory constraints by the reload pass.
2809 It should return 1 if the operand type represented by the constraint
2810 at the start of @var{str}, the first letter of which is the letter @var{c},
2811 comprises a subset of all memory references including
2812 all those whose address is simply a base register. This allows the reload
2813 pass to reload an operand, if it does not directly correspond to the operand
2814 type of @var{c}, by copying its address into a base register.
2816 For example, on the S/390, some instructions do not accept arbitrary
2817 memory references, but only those that do not make use of an index
2818 register. The constraint letter @samp{Q} is defined via
2819 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2820 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2821 a @samp{Q} constraint can handle any memory operand, because the
2822 reload pass knows it can be reloaded by copying the memory address
2823 into a base register if required. This is analogous to the way
2824 a @samp{o} constraint can handle any memory operand.
2827 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2828 A C expression that defines the optional machine-dependent constraint
2829 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2830 @code{EXTRA_CONSTRAINT_STR}, that should
2831 be treated like address constraints by the reload pass.
2833 It should return 1 if the operand type represented by the constraint
2834 at the start of @var{str}, which starts with the letter @var{c}, comprises
2835 a subset of all memory addresses including
2836 all those that consist of just a base register. This allows the reload
2837 pass to reload an operand, if it does not directly correspond to the operand
2838 type of @var{str}, by copying it into a base register.
2840 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2841 be used with the @code{address_operand} predicate. It is treated
2842 analogously to the @samp{p} constraint.
2845 @node Stack and Calling
2846 @section Stack Layout and Calling Conventions
2847 @cindex calling conventions
2849 @c prevent bad page break with this line
2850 This describes the stack layout and calling conventions.
2854 * Exception Handling::
2859 * Register Arguments::
2861 * Aggregate Return::
2869 @subsection Basic Stack Layout
2870 @cindex stack frame layout
2871 @cindex frame layout
2873 @c prevent bad page break with this line
2874 Here is the basic stack layout.
2876 @defmac STACK_GROWS_DOWNWARD
2877 Define this macro if pushing a word onto the stack moves the stack
2878 pointer to a smaller address.
2880 When we say, ``define this macro if @dots{}'', it means that the
2881 compiler checks this macro only with @code{#ifdef} so the precise
2882 definition used does not matter.
2885 @defmac STACK_PUSH_CODE
2886 This macro defines the operation used when something is pushed
2887 on the stack. In RTL, a push operation will be
2888 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2890 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2891 and @code{POST_INC}. Which of these is correct depends on
2892 the stack direction and on whether the stack pointer points
2893 to the last item on the stack or whether it points to the
2894 space for the next item on the stack.
2896 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2897 defined, which is almost always right, and @code{PRE_INC} otherwise,
2898 which is often wrong.
2901 @defmac FRAME_GROWS_DOWNWARD
2902 Define this macro if the addresses of local variable slots are at negative
2903 offsets from the frame pointer.
2906 @defmac ARGS_GROW_DOWNWARD
2907 Define this macro if successive arguments to a function occupy decreasing
2908 addresses on the stack.
2911 @defmac STARTING_FRAME_OFFSET
2912 Offset from the frame pointer to the first local variable slot to be allocated.
2914 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2915 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2916 Otherwise, it is found by adding the length of the first slot to the
2917 value @code{STARTING_FRAME_OFFSET}.
2918 @c i'm not sure if the above is still correct.. had to change it to get
2919 @c rid of an overfull. --mew 2feb93
2922 @defmac STACK_ALIGNMENT_NEEDED
2923 Define to zero to disable final alignment of the stack during reload.
2924 The nonzero default for this macro is suitable for most ports.
2926 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2927 is a register save block following the local block that doesn't require
2928 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2929 stack alignment and do it in the backend.
2932 @defmac STACK_POINTER_OFFSET
2933 Offset from the stack pointer register to the first location at which
2934 outgoing arguments are placed. If not specified, the default value of
2935 zero is used. This is the proper value for most machines.
2937 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2938 the first location at which outgoing arguments are placed.
2941 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2942 Offset from the argument pointer register to the first argument's
2943 address. On some machines it may depend on the data type of the
2946 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2947 the first argument's address.
2950 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2951 Offset from the stack pointer register to an item dynamically allocated
2952 on the stack, e.g., by @code{alloca}.
2954 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2955 length of the outgoing arguments. The default is correct for most
2956 machines. See @file{function.c} for details.
2959 @defmac INITIAL_FRAME_ADDRESS_RTX
2960 A C expression whose value is RTL representing the address of the initial
2961 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
2962 @code{DYNAMIC_CHAIN_ADDRESS}.
2963 If you don't define this macro, the default is to return
2964 @code{hard_frame_pointer_rtx}.
2965 This default is usually correct unless @code{-fomit-frame-pointer} is in
2967 Define this macro in order to make @code{__builtin_frame_address (0)} and
2968 @code{__builtin_return_address (0)} work even in absence of a hard frame pointer.
2971 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2972 A C expression whose value is RTL representing the address in a stack
2973 frame where the pointer to the caller's frame is stored. Assume that
2974 @var{frameaddr} is an RTL expression for the address of the stack frame
2977 If you don't define this macro, the default is to return the value
2978 of @var{frameaddr}---that is, the stack frame address is also the
2979 address of the stack word that points to the previous frame.
2982 @defmac SETUP_FRAME_ADDRESSES
2983 If defined, a C expression that produces the machine-specific code to
2984 setup the stack so that arbitrary frames can be accessed. For example,
2985 on the SPARC, we must flush all of the register windows to the stack
2986 before we can access arbitrary stack frames. You will seldom need to
2990 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
2991 This target hook should return an rtx that is used to store
2992 the address of the current frame into the built in @code{setjmp} buffer.
2993 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2994 machines. One reason you may need to define this target hook is if
2995 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2998 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2999 A C expression whose value is RTL representing the value of the return
3000 address for the frame @var{count} steps up from the current frame, after
3001 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3002 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3003 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3005 The value of the expression must always be the correct address when
3006 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
3007 determine the return address of other frames.
3010 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3011 Define this if the return address of a particular stack frame is accessed
3012 from the frame pointer of the previous stack frame.
3015 @defmac INCOMING_RETURN_ADDR_RTX
3016 A C expression whose value is RTL representing the location of the
3017 incoming return address at the beginning of any function, before the
3018 prologue. This RTL is either a @code{REG}, indicating that the return
3019 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3022 You only need to define this macro if you want to support call frame
3023 debugging information like that provided by DWARF 2.
3025 If this RTL is a @code{REG}, you should also define
3026 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3029 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3030 A C expression whose value is an integer giving a DWARF 2 column
3031 number that may be used as an alternate return column. This should
3032 be defined only if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3033 general register, but an alternate column needs to be used for
3037 @defmac DWARF_ZERO_REG
3038 A C expression whose value is an integer giving a DWARF 2 register
3039 number that is considered to always have the value zero. This should
3040 only be defined if the target has an architected zero register, and
3041 someone decided it was a good idea to use that register number to
3042 terminate the stack backtrace. New ports should avoid this.
3045 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3046 This target hook allows the backend to emit frame-related insns that
3047 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3048 info engine will invoke it on insns of the form
3050 (set (reg) (unspec [...] UNSPEC_INDEX))
3054 (set (reg) (unspec_volatile [...] UNSPECV_INDEX)).
3056 to let the backend emit the call frame instructions. @var{label} is
3057 the CFI label attached to the insn, @var{pattern} is the pattern of
3058 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3061 @defmac INCOMING_FRAME_SP_OFFSET
3062 A C expression whose value is an integer giving the offset, in bytes,
3063 from the value of the stack pointer register to the top of the stack
3064 frame at the beginning of any function, before the prologue. The top of
3065 the frame is defined to be the value of the stack pointer in the
3066 previous frame, just before the call instruction.
3068 You only need to define this macro if you want to support call frame
3069 debugging information like that provided by DWARF 2.
3072 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3073 A C expression whose value is an integer giving the offset, in bytes,
3074 from the argument pointer to the canonical frame address (cfa). The
3075 final value should coincide with that calculated by
3076 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3077 during virtual register instantiation.
3079 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
3080 which is correct for most machines; in general, the arguments are found
3081 immediately before the stack frame. Note that this is not the case on
3082 some targets that save registers into the caller's frame, such as SPARC
3083 and rs6000, and so such targets need to define this macro.
3085 You only need to define this macro if the default is incorrect, and you
3086 want to support call frame debugging information like that provided by
3090 @node Exception Handling
3091 @subsection Exception Handling Support
3092 @cindex exception handling
3094 @defmac EH_RETURN_DATA_REGNO (@var{N})
3095 A C expression whose value is the @var{N}th register number used for
3096 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3097 @var{N} registers are usable.
3099 The exception handling library routines communicate with the exception
3100 handlers via a set of agreed upon registers. Ideally these registers
3101 should be call-clobbered; it is possible to use call-saved registers,
3102 but may negatively impact code size. The target must support at least
3103 2 data registers, but should define 4 if there are enough free registers.
3105 You must define this macro if you want to support call frame exception
3106 handling like that provided by DWARF 2.
3109 @defmac EH_RETURN_STACKADJ_RTX
3110 A C expression whose value is RTL representing a location in which
3111 to store a stack adjustment to be applied before function return.
3112 This is used to unwind the stack to an exception handler's call frame.
3113 It will be assigned zero on code paths that return normally.
3115 Typically this is a call-clobbered hard register that is otherwise
3116 untouched by the epilogue, but could also be a stack slot.
3118 Do not define this macro if the stack pointer is saved and restored
3119 by the regular prolog and epilog code in the call frame itself; in
3120 this case, the exception handling library routines will update the
3121 stack location to be restored in place. Otherwise, you must define
3122 this macro if you want to support call frame exception handling like
3123 that provided by DWARF 2.
3126 @defmac EH_RETURN_HANDLER_RTX
3127 A C expression whose value is RTL representing a location in which
3128 to store the address of an exception handler to which we should
3129 return. It will not be assigned on code paths that return normally.
3131 Typically this is the location in the call frame at which the normal
3132 return address is stored. For targets that return by popping an
3133 address off the stack, this might be a memory address just below
3134 the @emph{target} call frame rather than inside the current call
3135 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3136 been assigned, so it may be used to calculate the location of the
3139 Some targets have more complex requirements than storing to an
3140 address calculable during initial code generation. In that case
3141 the @code{eh_return} instruction pattern should be used instead.
3143 If you want to support call frame exception handling, you must
3144 define either this macro or the @code{eh_return} instruction pattern.
3147 @defmac RETURN_ADDR_OFFSET
3148 If defined, an integer-valued C expression for which rtl will be generated
3149 to add it to the exception handler address before it is searched in the
3150 exception handling tables, and to subtract it again from the address before
3151 using it to return to the exception handler.
3154 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3155 This macro chooses the encoding of pointers embedded in the exception
3156 handling sections. If at all possible, this should be defined such
3157 that the exception handling section will not require dynamic relocations,
3158 and so may be read-only.
3160 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3161 @var{global} is true if the symbol may be affected by dynamic relocations.
3162 The macro should return a combination of the @code{DW_EH_PE_*} defines
3163 as found in @file{dwarf2.h}.
3165 If this macro is not defined, pointers will not be encoded but
3166 represented directly.
3169 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3170 This macro allows the target to emit whatever special magic is required
3171 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3172 Generic code takes care of pc-relative and indirect encodings; this must
3173 be defined if the target uses text-relative or data-relative encodings.
3175 This is a C statement that branches to @var{done} if the format was
3176 handled. @var{encoding} is the format chosen, @var{size} is the number
3177 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3181 @defmac MD_UNWIND_SUPPORT
3182 A string specifying a file to be #include'd in unwind-dw2.c. The file
3183 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3186 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3187 This macro allows the target to add cpu and operating system specific
3188 code to the call-frame unwinder for use when there is no unwind data
3189 available. The most common reason to implement this macro is to unwind
3190 through signal frames.
3192 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3193 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3194 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3195 for the address of the code being executed and @code{context->cfa} for
3196 the stack pointer value. If the frame can be decoded, the register save
3197 addresses should be updated in @var{fs} and the macro should evaluate to
3198 @code{_URC_NO_REASON}. If the frame cannot be decoded, the macro should
3199 evaluate to @code{_URC_END_OF_STACK}.
3201 For proper signal handling in Java this macro is accompanied by
3202 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3205 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3206 This macro allows the target to add operating system specific code to the
3207 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3208 usually used for signal or interrupt frames.
3210 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3211 @var{context} is an @code{_Unwind_Context};
3212 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3213 for the abi and context in the @code{.unwabi} directive. If the
3214 @code{.unwabi} directive can be handled, the register save addresses should
3215 be updated in @var{fs}.
3218 @defmac TARGET_USES_WEAK_UNWIND_INFO
3219 A C expression that evaluates to true if the target requires unwind
3220 info to be given comdat linkage. Define it to be @code{1} if comdat
3221 linkage is necessary. The default is @code{0}.
3224 @node Stack Checking
3225 @subsection Specifying How Stack Checking is Done
3227 GCC will check that stack references are within the boundaries of
3228 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3232 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3233 will assume that you have arranged for stack checking to be done at
3234 appropriate places in the configuration files, e.g., in
3235 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3239 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3240 called @code{check_stack} in your @file{md} file, GCC will call that
3241 pattern with one argument which is the address to compare the stack
3242 value against. You must arrange for this pattern to report an error if
3243 the stack pointer is out of range.
3246 If neither of the above are true, GCC will generate code to periodically
3247 ``probe'' the stack pointer using the values of the macros defined below.
3250 Normally, you will use the default values of these macros, so GCC
3251 will use the third approach.
3253 @defmac STACK_CHECK_BUILTIN
3254 A nonzero value if stack checking is done by the configuration files in a
3255 machine-dependent manner. You should define this macro if stack checking
3256 is require by the ABI of your machine or if you would like to have to stack
3257 checking in some more efficient way than GCC's portable approach.
3258 The default value of this macro is zero.
3261 @defmac STACK_CHECK_PROBE_INTERVAL
3262 An integer representing the interval at which GCC must generate stack
3263 probe instructions. You will normally define this macro to be no larger
3264 than the size of the ``guard pages'' at the end of a stack area. The
3265 default value of 4096 is suitable for most systems.
3268 @defmac STACK_CHECK_PROBE_LOAD
3269 A integer which is nonzero if GCC should perform the stack probe
3270 as a load instruction and zero if GCC should use a store instruction.
3271 The default is zero, which is the most efficient choice on most systems.
3274 @defmac STACK_CHECK_PROTECT
3275 The number of bytes of stack needed to recover from a stack overflow,
3276 for languages where such a recovery is supported. The default value of
3277 75 words should be adequate for most machines.
3280 @defmac STACK_CHECK_MAX_FRAME_SIZE
3281 The maximum size of a stack frame, in bytes. GCC will generate probe
3282 instructions in non-leaf functions to ensure at least this many bytes of
3283 stack are available. If a stack frame is larger than this size, stack
3284 checking will not be reliable and GCC will issue a warning. The
3285 default is chosen so that GCC only generates one instruction on most
3286 systems. You should normally not change the default value of this macro.
3289 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3290 GCC uses this value to generate the above warning message. It
3291 represents the amount of fixed frame used by a function, not including
3292 space for any callee-saved registers, temporaries and user variables.
3293 You need only specify an upper bound for this amount and will normally
3294 use the default of four words.
3297 @defmac STACK_CHECK_MAX_VAR_SIZE
3298 The maximum size, in bytes, of an object that GCC will place in the
3299 fixed area of the stack frame when the user specifies
3300 @option{-fstack-check}.
3301 GCC computed the default from the values of the above macros and you will
3302 normally not need to override that default.
3306 @node Frame Registers
3307 @subsection Registers That Address the Stack Frame
3309 @c prevent bad page break with this line
3310 This discusses registers that address the stack frame.
3312 @defmac STACK_POINTER_REGNUM
3313 The register number of the stack pointer register, which must also be a
3314 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3315 the hardware determines which register this is.
3318 @defmac FRAME_POINTER_REGNUM
3319 The register number of the frame pointer register, which is used to
3320 access automatic variables in the stack frame. On some machines, the
3321 hardware determines which register this is. On other machines, you can
3322 choose any register you wish for this purpose.
3325 @defmac HARD_FRAME_POINTER_REGNUM
3326 On some machines the offset between the frame pointer and starting
3327 offset of the automatic variables is not known until after register
3328 allocation has been done (for example, because the saved registers are
3329 between these two locations). On those machines, define
3330 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3331 be used internally until the offset is known, and define
3332 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3333 used for the frame pointer.
3335 You should define this macro only in the very rare circumstances when it
3336 is not possible to calculate the offset between the frame pointer and
3337 the automatic variables until after register allocation has been
3338 completed. When this macro is defined, you must also indicate in your
3339 definition of @code{ELIMINABLE_REGS} how to eliminate
3340 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3341 or @code{STACK_POINTER_REGNUM}.
3343 Do not define this macro if it would be the same as
3344 @code{FRAME_POINTER_REGNUM}.
3347 @defmac ARG_POINTER_REGNUM
3348 The register number of the arg pointer register, which is used to access
3349 the function's argument list. On some machines, this is the same as the
3350 frame pointer register. On some machines, the hardware determines which
3351 register this is. On other machines, you can choose any register you
3352 wish for this purpose. If this is not the same register as the frame
3353 pointer register, then you must mark it as a fixed register according to
3354 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3355 (@pxref{Elimination}).
3358 @defmac RETURN_ADDRESS_POINTER_REGNUM
3359 The register number of the return address pointer register, which is used to
3360 access the current function's return address from the stack. On some
3361 machines, the return address is not at a fixed offset from the frame
3362 pointer or stack pointer or argument pointer. This register can be defined
3363 to point to the return address on the stack, and then be converted by
3364 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3366 Do not define this macro unless there is no other way to get the return
3367 address from the stack.
3370 @defmac STATIC_CHAIN_REGNUM
3371 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3372 Register numbers used for passing a function's static chain pointer. If
3373 register windows are used, the register number as seen by the called
3374 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3375 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3376 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3379 The static chain register need not be a fixed register.
3381 If the static chain is passed in memory, these macros should not be
3382 defined; instead, the next two macros should be defined.
3385 @defmac STATIC_CHAIN
3386 @defmacx STATIC_CHAIN_INCOMING
3387 If the static chain is passed in memory, these macros provide rtx giving
3388 @code{mem} expressions that denote where they are stored.
3389 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3390 as seen by the calling and called functions, respectively. Often the former
3391 will be at an offset from the stack pointer and the latter at an offset from
3394 @findex stack_pointer_rtx
3395 @findex frame_pointer_rtx
3396 @findex arg_pointer_rtx
3397 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3398 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3399 macros and should be used to refer to those items.
3401 If the static chain is passed in a register, the two previous macros should
3405 @defmac DWARF_FRAME_REGISTERS
3406 This macro specifies the maximum number of hard registers that can be
3407 saved in a call frame. This is used to size data structures used in
3408 DWARF2 exception handling.
3410 Prior to GCC 3.0, this macro was needed in order to establish a stable
3411 exception handling ABI in the face of adding new hard registers for ISA
3412 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3413 in the number of hard registers. Nevertheless, this macro can still be
3414 used to reduce the runtime memory requirements of the exception handling
3415 routines, which can be substantial if the ISA contains a lot of
3416 registers that are not call-saved.
3418 If this macro is not defined, it defaults to
3419 @code{FIRST_PSEUDO_REGISTER}.
3422 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3424 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3425 for backward compatibility in pre GCC 3.0 compiled code.
3427 If this macro is not defined, it defaults to
3428 @code{DWARF_FRAME_REGISTERS}.
3431 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3433 Define this macro if the target's representation for dwarf registers
3434 is different than the internal representation for unwind column.
3435 Given a dwarf register, this macro should return the internal unwind
3436 column number to use instead.
3438 See the PowerPC's SPE target for an example.
3441 @defmac DWARF_FRAME_REGNUM (@var{regno})
3443 Define this macro if the target's representation for dwarf registers
3444 used in .eh_frame or .debug_frame is different from that used in other
3445 debug info sections. Given a GCC hard register number, this macro
3446 should return the .eh_frame register number. The default is
3447 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3451 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3453 Define this macro to map register numbers held in the call frame info
3454 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3455 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3456 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3457 return @code{@var{regno}}.
3462 @subsection Eliminating Frame Pointer and Arg Pointer
3464 @c prevent bad page break with this line
3465 This is about eliminating the frame pointer and arg pointer.
3467 @defmac FRAME_POINTER_REQUIRED
3468 A C expression which is nonzero if a function must have and use a frame
3469 pointer. This expression is evaluated in the reload pass. If its value is
3470 nonzero the function will have a frame pointer.
3472 The expression can in principle examine the current function and decide
3473 according to the facts, but on most machines the constant 0 or the
3474 constant 1 suffices. Use 0 when the machine allows code to be generated
3475 with no frame pointer, and doing so saves some time or space. Use 1
3476 when there is no possible advantage to avoiding a frame pointer.
3478 In certain cases, the compiler does not know how to produce valid code
3479 without a frame pointer. The compiler recognizes those cases and
3480 automatically gives the function a frame pointer regardless of what
3481 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3484 In a function that does not require a frame pointer, the frame pointer
3485 register can be allocated for ordinary usage, unless you mark it as a
3486 fixed register. See @code{FIXED_REGISTERS} for more information.
3489 @findex get_frame_size
3490 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3491 A C statement to store in the variable @var{depth-var} the difference
3492 between the frame pointer and the stack pointer values immediately after
3493 the function prologue. The value would be computed from information
3494 such as the result of @code{get_frame_size ()} and the tables of
3495 registers @code{regs_ever_live} and @code{call_used_regs}.
3497 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3498 need not be defined. Otherwise, it must be defined even if
3499 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3500 case, you may set @var{depth-var} to anything.
3503 @defmac ELIMINABLE_REGS
3504 If defined, this macro specifies a table of register pairs used to
3505 eliminate unneeded registers that point into the stack frame. If it is not
3506 defined, the only elimination attempted by the compiler is to replace
3507 references to the frame pointer with references to the stack pointer.
3509 The definition of this macro is a list of structure initializations, each
3510 of which specifies an original and replacement register.
3512 On some machines, the position of the argument pointer is not known until
3513 the compilation is completed. In such a case, a separate hard register
3514 must be used for the argument pointer. This register can be eliminated by
3515 replacing it with either the frame pointer or the argument pointer,
3516 depending on whether or not the frame pointer has been eliminated.
3518 In this case, you might specify:
3520 #define ELIMINABLE_REGS \
3521 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3522 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3523 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3526 Note that the elimination of the argument pointer with the stack pointer is
3527 specified first since that is the preferred elimination.
3530 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3531 A C expression that returns nonzero if the compiler is allowed to try
3532 to replace register number @var{from-reg} with register number
3533 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3534 is defined, and will usually be the constant 1, since most of the cases
3535 preventing register elimination are things that the compiler already
3539 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3540 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3541 specifies the initial difference between the specified pair of
3542 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3546 @node Stack Arguments
3547 @subsection Passing Function Arguments on the Stack
3548 @cindex arguments on stack
3549 @cindex stack arguments
3551 The macros in this section control how arguments are passed
3552 on the stack. See the following section for other macros that
3553 control passing certain arguments in registers.
3555 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3556 This target hook returns @code{true} if an argument declared in a
3557 prototype as an integral type smaller than @code{int} should actually be
3558 passed as an @code{int}. In addition to avoiding errors in certain
3559 cases of mismatch, it also makes for better code on certain machines.
3560 The default is to not promote prototypes.
3564 A C expression. If nonzero, push insns will be used to pass
3566 If the target machine does not have a push instruction, set it to zero.
3567 That directs GCC to use an alternate strategy: to
3568 allocate the entire argument block and then store the arguments into
3569 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3572 @defmac PUSH_ARGS_REVERSED
3573 A C expression. If nonzero, function arguments will be evaluated from
3574 last to first, rather than from first to last. If this macro is not
3575 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3576 and args grow in opposite directions, and 0 otherwise.
3579 @defmac PUSH_ROUNDING (@var{npushed})
3580 A C expression that is the number of bytes actually pushed onto the
3581 stack when an instruction attempts to push @var{npushed} bytes.
3583 On some machines, the definition
3586 #define PUSH_ROUNDING(BYTES) (BYTES)
3590 will suffice. But on other machines, instructions that appear
3591 to push one byte actually push two bytes in an attempt to maintain
3592 alignment. Then the definition should be
3595 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3599 @findex current_function_outgoing_args_size
3600 @defmac ACCUMULATE_OUTGOING_ARGS
3601 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3602 will be computed and placed into the variable
3603 @code{current_function_outgoing_args_size}. No space will be pushed
3604 onto the stack for each call; instead, the function prologue should
3605 increase the stack frame size by this amount.
3607 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3611 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3612 Define this macro if functions should assume that stack space has been
3613 allocated for arguments even when their values are passed in
3616 The value of this macro is the size, in bytes, of the area reserved for
3617 arguments passed in registers for the function represented by @var{fndecl},
3618 which can be zero if GCC is calling a library function.
3620 This space can be allocated by the caller, or be a part of the
3621 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3624 @c above is overfull. not sure what to do. --mew 5feb93 did
3625 @c something, not sure if it looks good. --mew 10feb93
3627 @defmac OUTGOING_REG_PARM_STACK_SPACE
3628 Define this if it is the responsibility of the caller to allocate the area
3629 reserved for arguments passed in registers.
3631 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3632 whether the space for these arguments counts in the value of
3633 @code{current_function_outgoing_args_size}.
3636 @defmac STACK_PARMS_IN_REG_PARM_AREA
3637 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3638 stack parameters don't skip the area specified by it.
3639 @c i changed this, makes more sens and it should have taken care of the
3640 @c overfull.. not as specific, tho. --mew 5feb93
3642 Normally, when a parameter is not passed in registers, it is placed on the
3643 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3644 suppresses this behavior and causes the parameter to be passed on the
3645 stack in its natural location.
3648 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3649 A C expression that should indicate the number of bytes of its own
3650 arguments that a function pops on returning, or 0 if the
3651 function pops no arguments and the caller must therefore pop them all
3652 after the function returns.
3654 @var{fundecl} is a C variable whose value is a tree node that describes
3655 the function in question. Normally it is a node of type
3656 @code{FUNCTION_DECL} that describes the declaration of the function.
3657 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3659 @var{funtype} is a C variable whose value is a tree node that
3660 describes the function in question. Normally it is a node of type
3661 @code{FUNCTION_TYPE} that describes the data type of the function.
3662 From this it is possible to obtain the data types of the value and
3663 arguments (if known).
3665 When a call to a library function is being considered, @var{fundecl}
3666 will contain an identifier node for the library function. Thus, if
3667 you need to distinguish among various library functions, you can do so
3668 by their names. Note that ``library function'' in this context means
3669 a function used to perform arithmetic, whose name is known specially
3670 in the compiler and was not mentioned in the C code being compiled.
3672 @var{stack-size} is the number of bytes of arguments passed on the
3673 stack. If a variable number of bytes is passed, it is zero, and
3674 argument popping will always be the responsibility of the calling function.
3676 On the VAX, all functions always pop their arguments, so the definition
3677 of this macro is @var{stack-size}. On the 68000, using the standard
3678 calling convention, no functions pop their arguments, so the value of
3679 the macro is always 0 in this case. But an alternative calling
3680 convention is available in which functions that take a fixed number of
3681 arguments pop them but other functions (such as @code{printf}) pop
3682 nothing (the caller pops all). When this convention is in use,
3683 @var{funtype} is examined to determine whether a function takes a fixed
3684 number of arguments.
3687 @defmac CALL_POPS_ARGS (@var{cum})
3688 A C expression that should indicate the number of bytes a call sequence
3689 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3690 when compiling a function call.
3692 @var{cum} is the variable in which all arguments to the called function
3693 have been accumulated.
3695 On certain architectures, such as the SH5, a call trampoline is used
3696 that pops certain registers off the stack, depending on the arguments
3697 that have been passed to the function. Since this is a property of the
3698 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3702 @node Register Arguments
3703 @subsection Passing Arguments in Registers
3704 @cindex arguments in registers
3705 @cindex registers arguments
3707 This section describes the macros which let you control how various
3708 types of arguments are passed in registers or how they are arranged in
3711 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3712 A C expression that controls whether a function argument is passed
3713 in a register, and which register.
3715 The arguments are @var{cum}, which summarizes all the previous
3716 arguments; @var{mode}, the machine mode of the argument; @var{type},
3717 the data type of the argument as a tree node or 0 if that is not known
3718 (which happens for C support library functions); and @var{named},
3719 which is 1 for an ordinary argument and 0 for nameless arguments that
3720 correspond to @samp{@dots{}} in the called function's prototype.
3721 @var{type} can be an incomplete type if a syntax error has previously
3724 The value of the expression is usually either a @code{reg} RTX for the
3725 hard register in which to pass the argument, or zero to pass the
3726 argument on the stack.
3728 For machines like the VAX and 68000, where normally all arguments are
3729 pushed, zero suffices as a definition.
3731 The value of the expression can also be a @code{parallel} RTX@. This is
3732 used when an argument is passed in multiple locations. The mode of the
3733 @code{parallel} should be the mode of the entire argument. The
3734 @code{parallel} holds any number of @code{expr_list} pairs; each one
3735 describes where part of the argument is passed. In each
3736 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3737 register in which to pass this part of the argument, and the mode of the
3738 register RTX indicates how large this part of the argument is. The
3739 second operand of the @code{expr_list} is a @code{const_int} which gives
3740 the offset in bytes into the entire argument of where this part starts.
3741 As a special exception the first @code{expr_list} in the @code{parallel}
3742 RTX may have a first operand of zero. This indicates that the entire
3743 argument is also stored on the stack.
3745 The last time this macro is called, it is called with @code{MODE ==
3746 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3747 pattern as operands 2 and 3 respectively.
3749 @cindex @file{stdarg.h} and register arguments
3750 The usual way to make the ISO library @file{stdarg.h} work on a machine
3751 where some arguments are usually passed in registers, is to cause
3752 nameless arguments to be passed on the stack instead. This is done
3753 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3755 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3756 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3757 You may use the hook @code{targetm.calls.must_pass_in_stack}
3758 in the definition of this macro to determine if this argument is of a
3759 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3760 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3761 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3762 defined, the argument will be computed in the stack and then loaded into
3766 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type})
3767 This target hook should return @code{true} if we should not pass @var{type}
3768 solely in registers. The file @file{expr.h} defines a
3769 definition that is usually appropriate, refer to @file{expr.h} for additional
3773 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3774 Define this macro if the target machine has ``register windows'', so
3775 that the register in which a function sees an arguments is not
3776 necessarily the same as the one in which the caller passed the
3779 For such machines, @code{FUNCTION_ARG} computes the register in which
3780 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3781 be defined in a similar fashion to tell the function being called
3782 where the arguments will arrive.
3784 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3785 serves both purposes.
3788 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3789 This target hook returns the number of bytes at the beginning of an
3790 argument that must be put in registers. The value must be zero for
3791 arguments that are passed entirely in registers or that are entirely
3792 pushed on the stack.
3794 On some machines, certain arguments must be passed partially in
3795 registers and partially in memory. On these machines, typically the
3796 first few words of arguments are passed in registers, and the rest
3797 on the stack. If a multi-word argument (a @code{double} or a
3798 structure) crosses that boundary, its first few words must be passed
3799 in registers and the rest must be pushed. This macro tells the
3800 compiler when this occurs, and how many bytes should go in registers.
3802 @code{FUNCTION_ARG} for these arguments should return the first
3803 register to be used by the caller for this argument; likewise
3804 @code{FUNCTION_INCOMING_ARG}, for the called function.
3807 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3808 This target hook should return @code{true} if an argument at the
3809 position indicated by @var{cum} should be passed by reference. This
3810 predicate is queried after target independent reasons for being
3811 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
3813 If the hook returns true, a copy of that argument is made in memory and a
3814 pointer to the argument is passed instead of the argument itself.
3815 The pointer is passed in whatever way is appropriate for passing a pointer
3819 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3820 The function argument described by the parameters to this hook is
3821 known to be passed by reference. The hook should return true if the
3822 function argument should be copied by the callee instead of copied
3825 For any argument for which the hook returns true, if it can be
3826 determined that the argument is not modified, then a copy need
3829 The default version of this hook always returns false.
3832 @defmac CUMULATIVE_ARGS
3833 A C type for declaring a variable that is used as the first argument of
3834 @code{FUNCTION_ARG} and other related values. For some target machines,
3835 the type @code{int} suffices and can hold the number of bytes of
3838 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3839 arguments that have been passed on the stack. The compiler has other
3840 variables to keep track of that. For target machines on which all
3841 arguments are passed on the stack, there is no need to store anything in
3842 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3843 should not be empty, so use @code{int}.
3846 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3847 A C statement (sans semicolon) for initializing the variable
3848 @var{cum} for the state at the beginning of the argument list. The
3849 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3850 is the tree node for the data type of the function which will receive
3851 the args, or 0 if the args are to a compiler support library function.
3852 For direct calls that are not libcalls, @var{fndecl} contain the
3853 declaration node of the function. @var{fndecl} is also set when
3854 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3855 being compiled. @var{n_named_args} is set to the number of named
3856 arguments, including a structure return address if it is passed as a
3857 parameter, when making a call. When processing incoming arguments,
3858 @var{n_named_args} is set to @minus{}1.
3860 When processing a call to a compiler support library function,
3861 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3862 contains the name of the function, as a string. @var{libname} is 0 when
3863 an ordinary C function call is being processed. Thus, each time this
3864 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3865 never both of them at once.
3868 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3869 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3870 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3871 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3872 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3873 0)} is used instead.
3876 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3877 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3878 finding the arguments for the function being compiled. If this macro is
3879 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3881 The value passed for @var{libname} is always 0, since library routines
3882 with special calling conventions are never compiled with GCC@. The
3883 argument @var{libname} exists for symmetry with
3884 @code{INIT_CUMULATIVE_ARGS}.
3885 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3886 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3889 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3890 A C statement (sans semicolon) to update the summarizer variable
3891 @var{cum} to advance past an argument in the argument list. The
3892 values @var{mode}, @var{type} and @var{named} describe that argument.
3893 Once this is done, the variable @var{cum} is suitable for analyzing
3894 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3896 This macro need not do anything if the argument in question was passed
3897 on the stack. The compiler knows how to track the amount of stack space
3898 used for arguments without any special help.
3901 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3902 If defined, a C expression which determines whether, and in which direction,
3903 to pad out an argument with extra space. The value should be of type
3904 @code{enum direction}: either @code{upward} to pad above the argument,
3905 @code{downward} to pad below, or @code{none} to inhibit padding.
3907 The @emph{amount} of padding is always just enough to reach the next
3908 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3911 This macro has a default definition which is right for most systems.
3912 For little-endian machines, the default is to pad upward. For
3913 big-endian machines, the default is to pad downward for an argument of
3914 constant size shorter than an @code{int}, and upward otherwise.
3917 @defmac PAD_VARARGS_DOWN
3918 If defined, a C expression which determines whether the default
3919 implementation of va_arg will attempt to pad down before reading the
3920 next argument, if that argument is smaller than its aligned space as
3921 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3922 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3925 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3926 Specify padding for the last element of a block move between registers and
3927 memory. @var{first} is nonzero if this is the only element. Defining this
3928 macro allows better control of register function parameters on big-endian
3929 machines, without using @code{PARALLEL} rtl. In particular,
3930 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3931 registers, as there is no longer a "wrong" part of a register; For example,
3932 a three byte aggregate may be passed in the high part of a register if so
3936 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3937 If defined, a C expression that gives the alignment boundary, in bits,
3938 of an argument with the specified mode and type. If it is not defined,
3939 @code{PARM_BOUNDARY} is used for all arguments.
3942 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3943 A C expression that is nonzero if @var{regno} is the number of a hard
3944 register in which function arguments are sometimes passed. This does
3945 @emph{not} include implicit arguments such as the static chain and
3946 the structure-value address. On many machines, no registers can be
3947 used for this purpose since all function arguments are pushed on the
3951 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
3952 This hook should return true if parameter of type @var{type} are passed
3953 as two scalar parameters. By default, GCC will attempt to pack complex
3954 arguments into the target's word size. Some ABIs require complex arguments
3955 to be split and treated as their individual components. For example, on
3956 AIX64, complex floats should be passed in a pair of floating point
3957 registers, even though a complex float would fit in one 64-bit floating
3960 The default value of this hook is @code{NULL}, which is treated as always
3964 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
3965 This hook returns a type node for @code{va_list} for the target.
3966 The default version of the hook returns @code{void*}.
3969 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
3970 This hook performs target-specific gimplification of
3971 @code{VA_ARG_EXPR}. The first two parameters correspond to the
3972 arguments to @code{va_arg}; the latter two are as in
3973 @code{gimplify.c:gimplify_expr}.
3976 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
3977 Define this to return nonzero if the port can handle pointers
3978 with machine mode @var{mode}. The default version of this
3979 hook returns true for both @code{ptr_mode} and @code{Pmode}.
3982 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
3983 Define this to return nonzero if the port is prepared to handle
3984 insns involving scalar mode @var{mode}. For a scalar mode to be
3985 considered supported, all the basic arithmetic and comparisons
3988 The default version of this hook returns true for any mode
3989 required to handle the basic C types (as defined by the port).
3990 Included here are the double-word arithmetic supported by the
3991 code in @file{optabs.c}.
3994 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
3995 Define this to return nonzero if the port is prepared to handle
3996 insns involving vector mode @var{mode}. At the very least, it
3997 must have move patterns for this mode.
4001 @subsection How Scalar Function Values Are Returned
4002 @cindex return values in registers
4003 @cindex values, returned by functions
4004 @cindex scalars, returned as values
4006 This section discusses the macros that control returning scalars as
4007 values---values that can fit in registers.
4009 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4010 A C expression to create an RTX representing the place where a
4011 function returns a value of data type @var{valtype}. @var{valtype} is
4012 a tree node representing a data type. Write @code{TYPE_MODE
4013 (@var{valtype})} to get the machine mode used to represent that type.
4014 On many machines, only the mode is relevant. (Actually, on most
4015 machines, scalar values are returned in the same place regardless of
4018 The value of the expression is usually a @code{reg} RTX for the hard
4019 register where the return value is stored. The value can also be a
4020 @code{parallel} RTX, if the return value is in multiple places. See
4021 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
4023 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply the same
4024 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
4027 If the precise function being called is known, @var{func} is a tree
4028 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4029 pointer. This makes it possible to use a different value-returning
4030 convention for specific functions when all their calls are
4033 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
4034 types, because these are returned in another way. See
4035 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4038 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
4039 Define this macro if the target machine has ``register windows''
4040 so that the register in which a function returns its value is not
4041 the same as the one in which the caller sees the value.
4043 For such machines, @code{FUNCTION_VALUE} computes the register in which
4044 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
4045 defined in a similar fashion to tell the function where to put the
4048 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
4049 @code{FUNCTION_VALUE} serves both purposes.
4051 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
4052 aggregate data types, because these are returned in another way. See
4053 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4056 @defmac LIBCALL_VALUE (@var{mode})
4057 A C expression to create an RTX representing the place where a library
4058 function returns a value of mode @var{mode}. If the precise function
4059 being called is known, @var{func} is a tree node
4060 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4061 pointer. This makes it possible to use a different value-returning
4062 convention for specific functions when all their calls are
4065 Note that ``library function'' in this context means a compiler
4066 support routine, used to perform arithmetic, whose name is known
4067 specially by the compiler and was not mentioned in the C code being
4070 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
4071 data types, because none of the library functions returns such types.
4074 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4075 A C expression that is nonzero if @var{regno} is the number of a hard
4076 register in which the values of called function may come back.
4078 A register whose use for returning values is limited to serving as the
4079 second of a pair (for a value of type @code{double}, say) need not be
4080 recognized by this macro. So for most machines, this definition
4084 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4087 If the machine has register windows, so that the caller and the called
4088 function use different registers for the return value, this macro
4089 should recognize only the caller's register numbers.
4092 @defmac APPLY_RESULT_SIZE
4093 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4094 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4095 saving and restoring an arbitrary return value.
4098 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
4099 This hook should return true if values of type @var{type} are returned
4100 at the most significant end of a register (in other words, if they are
4101 padded at the least significant end). You can assume that @var{type}
4102 is returned in a register; the caller is required to check this.
4104 Note that the register provided by @code{FUNCTION_VALUE} must be able
4105 to hold the complete return value. For example, if a 1-, 2- or 3-byte
4106 structure is returned at the most significant end of a 4-byte register,
4107 @code{FUNCTION_VALUE} should provide an @code{SImode} rtx.
4110 @node Aggregate Return
4111 @subsection How Large Values Are Returned
4112 @cindex aggregates as return values
4113 @cindex large return values
4114 @cindex returning aggregate values
4115 @cindex structure value address
4117 When a function value's mode is @code{BLKmode} (and in some other
4118 cases), the value is not returned according to @code{FUNCTION_VALUE}
4119 (@pxref{Scalar Return}). Instead, the caller passes the address of a
4120 block of memory in which the value should be stored. This address
4121 is called the @dfn{structure value address}.
4123 This section describes how to control returning structure values in
4126 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
4127 This target hook should return a nonzero value to say to return the
4128 function value in memory, just as large structures are always returned.
4129 Here @var{type} will be the data type of the value, and @var{fntype}
4130 will be the type of the function doing the returning, or @code{NULL} for
4133 Note that values of mode @code{BLKmode} must be explicitly handled
4134 by this function. Also, the option @option{-fpcc-struct-return}
4135 takes effect regardless of this macro. On most systems, it is
4136 possible to leave the hook undefined; this causes a default
4137 definition to be used, whose value is the constant 1 for @code{BLKmode}
4138 values, and 0 otherwise.
4140 Do not use this hook to indicate that structures and unions should always
4141 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4145 @defmac DEFAULT_PCC_STRUCT_RETURN
4146 Define this macro to be 1 if all structure and union return values must be
4147 in memory. Since this results in slower code, this should be defined
4148 only if needed for compatibility with other compilers or with an ABI@.
4149 If you define this macro to be 0, then the conventions used for structure
4150 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4153 If not defined, this defaults to the value 1.
4156 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4157 This target hook should return the location of the structure value
4158 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4159 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4160 be @code{NULL}, for libcalls. You do not need to define this target
4161 hook if the address is always passed as an ``invisible'' first
4164 On some architectures the place where the structure value address
4165 is found by the called function is not the same place that the
4166 caller put it. This can be due to register windows, or it could
4167 be because the function prologue moves it to a different place.
4168 @var{incoming} is @code{true} when the location is needed in
4169 the context of the called function, and @code{false} in the context of
4172 If @var{incoming} is @code{true} and the address is to be found on the
4173 stack, return a @code{mem} which refers to the frame pointer.
4176 @defmac PCC_STATIC_STRUCT_RETURN
4177 Define this macro if the usual system convention on the target machine
4178 for returning structures and unions is for the called function to return
4179 the address of a static variable containing the value.
4181 Do not define this if the usual system convention is for the caller to
4182 pass an address to the subroutine.
4184 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4185 nothing when you use @option{-freg-struct-return} mode.
4189 @subsection Caller-Saves Register Allocation
4191 If you enable it, GCC can save registers around function calls. This
4192 makes it possible to use call-clobbered registers to hold variables that
4193 must live across calls.
4195 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4196 A C expression to determine whether it is worthwhile to consider placing
4197 a pseudo-register in a call-clobbered hard register and saving and
4198 restoring it around each function call. The expression should be 1 when
4199 this is worth doing, and 0 otherwise.
4201 If you don't define this macro, a default is used which is good on most
4202 machines: @code{4 * @var{calls} < @var{refs}}.
4205 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4206 A C expression specifying which mode is required for saving @var{nregs}
4207 of a pseudo-register in call-clobbered hard register @var{regno}. If
4208 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4209 returned. For most machines this macro need not be defined since GCC
4210 will select the smallest suitable mode.
4213 @node Function Entry
4214 @subsection Function Entry and Exit
4215 @cindex function entry and exit
4219 This section describes the macros that output function entry
4220 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4222 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4223 If defined, a function that outputs the assembler code for entry to a
4224 function. The prologue is responsible for setting up the stack frame,
4225 initializing the frame pointer register, saving registers that must be
4226 saved, and allocating @var{size} additional bytes of storage for the
4227 local variables. @var{size} is an integer. @var{file} is a stdio
4228 stream to which the assembler code should be output.
4230 The label for the beginning of the function need not be output by this
4231 macro. That has already been done when the macro is run.
4233 @findex regs_ever_live
4234 To determine which registers to save, the macro can refer to the array
4235 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4236 @var{r} is used anywhere within the function. This implies the function
4237 prologue should save register @var{r}, provided it is not one of the
4238 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4239 @code{regs_ever_live}.)
4241 On machines that have ``register windows'', the function entry code does
4242 not save on the stack the registers that are in the windows, even if
4243 they are supposed to be preserved by function calls; instead it takes
4244 appropriate steps to ``push'' the register stack, if any non-call-used
4245 registers are used in the function.
4247 @findex frame_pointer_needed
4248 On machines where functions may or may not have frame-pointers, the
4249 function entry code must vary accordingly; it must set up the frame
4250 pointer if one is wanted, and not otherwise. To determine whether a
4251 frame pointer is in wanted, the macro can refer to the variable
4252 @code{frame_pointer_needed}. The variable's value will be 1 at run
4253 time in a function that needs a frame pointer. @xref{Elimination}.
4255 The function entry code is responsible for allocating any stack space
4256 required for the function. This stack space consists of the regions
4257 listed below. In most cases, these regions are allocated in the
4258 order listed, with the last listed region closest to the top of the
4259 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4260 the highest address if it is not defined). You can use a different order
4261 for a machine if doing so is more convenient or required for
4262 compatibility reasons. Except in cases where required by standard
4263 or by a debugger, there is no reason why the stack layout used by GCC
4264 need agree with that used by other compilers for a machine.
4267 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4268 If defined, a function that outputs assembler code at the end of a
4269 prologue. This should be used when the function prologue is being
4270 emitted as RTL, and you have some extra assembler that needs to be
4271 emitted. @xref{prologue instruction pattern}.
4274 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4275 If defined, a function that outputs assembler code at the start of an
4276 epilogue. This should be used when the function epilogue is being
4277 emitted as RTL, and you have some extra assembler that needs to be
4278 emitted. @xref{epilogue instruction pattern}.
4281 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4282 If defined, a function that outputs the assembler code for exit from a
4283 function. The epilogue is responsible for restoring the saved
4284 registers and stack pointer to their values when the function was
4285 called, and returning control to the caller. This macro takes the
4286 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4287 registers to restore are determined from @code{regs_ever_live} and
4288 @code{CALL_USED_REGISTERS} in the same way.
4290 On some machines, there is a single instruction that does all the work
4291 of returning from the function. On these machines, give that
4292 instruction the name @samp{return} and do not define the macro
4293 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4295 Do not define a pattern named @samp{return} if you want the
4296 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4297 switches to control whether return instructions or epilogues are used,
4298 define a @samp{return} pattern with a validity condition that tests the
4299 target switches appropriately. If the @samp{return} pattern's validity
4300 condition is false, epilogues will be used.
4302 On machines where functions may or may not have frame-pointers, the
4303 function exit code must vary accordingly. Sometimes the code for these
4304 two cases is completely different. To determine whether a frame pointer
4305 is wanted, the macro can refer to the variable
4306 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4307 a function that needs a frame pointer.
4309 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4310 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4311 The C variable @code{current_function_is_leaf} is nonzero for such a
4312 function. @xref{Leaf Functions}.
4314 On some machines, some functions pop their arguments on exit while
4315 others leave that for the caller to do. For example, the 68020 when
4316 given @option{-mrtd} pops arguments in functions that take a fixed
4317 number of arguments.
4319 @findex current_function_pops_args
4320 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4321 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4322 needs to know what was decided. The variable that is called
4323 @code{current_function_pops_args} is the number of bytes of its
4324 arguments that a function should pop. @xref{Scalar Return}.
4325 @c what is the "its arguments" in the above sentence referring to, pray
4326 @c tell? --mew 5feb93
4331 @findex current_function_pretend_args_size
4332 A region of @code{current_function_pretend_args_size} bytes of
4333 uninitialized space just underneath the first argument arriving on the
4334 stack. (This may not be at the very start of the allocated stack region
4335 if the calling sequence has pushed anything else since pushing the stack
4336 arguments. But usually, on such machines, nothing else has been pushed
4337 yet, because the function prologue itself does all the pushing.) This
4338 region is used on machines where an argument may be passed partly in
4339 registers and partly in memory, and, in some cases to support the
4340 features in @code{<stdarg.h>}.
4343 An area of memory used to save certain registers used by the function.
4344 The size of this area, which may also include space for such things as
4345 the return address and pointers to previous stack frames, is
4346 machine-specific and usually depends on which registers have been used
4347 in the function. Machines with register windows often do not require
4351 A region of at least @var{size} bytes, possibly rounded up to an allocation
4352 boundary, to contain the local variables of the function. On some machines,
4353 this region and the save area may occur in the opposite order, with the
4354 save area closer to the top of the stack.
4357 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4358 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4359 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4360 argument lists of the function. @xref{Stack Arguments}.
4363 @defmac EXIT_IGNORE_STACK
4364 Define this macro as a C expression that is nonzero if the return
4365 instruction or the function epilogue ignores the value of the stack
4366 pointer; in other words, if it is safe to delete an instruction to
4367 adjust the stack pointer before a return from the function. The
4370 Note that this macro's value is relevant only for functions for which
4371 frame pointers are maintained. It is never safe to delete a final
4372 stack adjustment in a function that has no frame pointer, and the
4373 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4376 @defmac EPILOGUE_USES (@var{regno})
4377 Define this macro as a C expression that is nonzero for registers that are
4378 used by the epilogue or the @samp{return} pattern. The stack and frame
4379 pointer registers are already be assumed to be used as needed.
4382 @defmac EH_USES (@var{regno})
4383 Define this macro as a C expression that is nonzero for registers that are
4384 used by the exception handling mechanism, and so should be considered live
4385 on entry to an exception edge.
4388 @defmac DELAY_SLOTS_FOR_EPILOGUE
4389 Define this macro if the function epilogue contains delay slots to which
4390 instructions from the rest of the function can be ``moved''. The
4391 definition should be a C expression whose value is an integer
4392 representing the number of delay slots there.
4395 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4396 A C expression that returns 1 if @var{insn} can be placed in delay
4397 slot number @var{n} of the epilogue.
4399 The argument @var{n} is an integer which identifies the delay slot now
4400 being considered (since different slots may have different rules of
4401 eligibility). It is never negative and is always less than the number
4402 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4403 If you reject a particular insn for a given delay slot, in principle, it
4404 may be reconsidered for a subsequent delay slot. Also, other insns may
4405 (at least in principle) be considered for the so far unfilled delay
4408 @findex current_function_epilogue_delay_list
4409 @findex final_scan_insn
4410 The insns accepted to fill the epilogue delay slots are put in an RTL
4411 list made with @code{insn_list} objects, stored in the variable
4412 @code{current_function_epilogue_delay_list}. The insn for the first
4413 delay slot comes first in the list. Your definition of the macro
4414 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4415 outputting the insns in this list, usually by calling
4416 @code{final_scan_insn}.
4418 You need not define this macro if you did not define
4419 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4422 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, tree @var{function})
4423 A function that outputs the assembler code for a thunk
4424 function, used to implement C++ virtual function calls with multiple
4425 inheritance. The thunk acts as a wrapper around a virtual function,
4426 adjusting the implicit object parameter before handing control off to
4429 First, emit code to add the integer @var{delta} to the location that
4430 contains the incoming first argument. Assume that this argument
4431 contains a pointer, and is the one used to pass the @code{this} pointer
4432 in C++. This is the incoming argument @emph{before} the function prologue,
4433 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4434 all other incoming arguments.
4436 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4437 made after adding @code{delta}. In particular, if @var{p} is the
4438 adjusted pointer, the following adjustment should be made:
4441 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4444 After the additions, emit code to jump to @var{function}, which is a
4445 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4446 not touch the return address. Hence returning from @var{FUNCTION} will
4447 return to whoever called the current @samp{thunk}.
4449 The effect must be as if @var{function} had been called directly with
4450 the adjusted first argument. This macro is responsible for emitting all
4451 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4452 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4454 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4455 have already been extracted from it.) It might possibly be useful on
4456 some targets, but probably not.
4458 If you do not define this macro, the target-independent code in the C++
4459 front end will generate a less efficient heavyweight thunk that calls
4460 @var{function} instead of jumping to it. The generic approach does
4461 not support varargs.
4464 @deftypefn {Target Hook} bool TARGET_ASM_CAN_OUTPUT_MI_THUNK (tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, tree @var{function})
4465 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4466 to output the assembler code for the thunk function specified by the
4467 arguments it is passed, and false otherwise. In the latter case, the
4468 generic approach will be used by the C++ front end, with the limitations
4473 @subsection Generating Code for Profiling
4474 @cindex profiling, code generation
4476 These macros will help you generate code for profiling.
4478 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4479 A C statement or compound statement to output to @var{file} some
4480 assembler code to call the profiling subroutine @code{mcount}.
4483 The details of how @code{mcount} expects to be called are determined by
4484 your operating system environment, not by GCC@. To figure them out,
4485 compile a small program for profiling using the system's installed C
4486 compiler and look at the assembler code that results.
4488 Older implementations of @code{mcount} expect the address of a counter
4489 variable to be loaded into some register. The name of this variable is
4490 @samp{LP} followed by the number @var{labelno}, so you would generate
4491 the name using @samp{LP%d} in a @code{fprintf}.
4494 @defmac PROFILE_HOOK
4495 A C statement or compound statement to output to @var{file} some assembly
4496 code to call the profiling subroutine @code{mcount} even the target does
4497 not support profiling.
4500 @defmac NO_PROFILE_COUNTERS
4501 Define this macro if the @code{mcount} subroutine on your system does
4502 not need a counter variable allocated for each function. This is true
4503 for almost all modern implementations. If you define this macro, you
4504 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4507 @defmac PROFILE_BEFORE_PROLOGUE
4508 Define this macro if the code for function profiling should come before
4509 the function prologue. Normally, the profiling code comes after.
4513 @subsection Permitting tail calls
4516 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4517 True if it is ok to do sibling call optimization for the specified
4518 call expression @var{exp}. @var{decl} will be the called function,
4519 or @code{NULL} if this is an indirect call.
4521 It is not uncommon for limitations of calling conventions to prevent
4522 tail calls to functions outside the current unit of translation, or
4523 during PIC compilation. The hook is used to enforce these restrictions,
4524 as the @code{sibcall} md pattern can not fail, or fall over to a
4525 ``normal'' call. The criteria for successful sibling call optimization
4526 may vary greatly between different architectures.
4530 @section Implementing the Varargs Macros
4531 @cindex varargs implementation
4533 GCC comes with an implementation of @code{<varargs.h>} and
4534 @code{<stdarg.h>} that work without change on machines that pass arguments
4535 on the stack. Other machines require their own implementations of
4536 varargs, and the two machine independent header files must have
4537 conditionals to include it.
4539 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4540 the calling convention for @code{va_start}. The traditional
4541 implementation takes just one argument, which is the variable in which
4542 to store the argument pointer. The ISO implementation of
4543 @code{va_start} takes an additional second argument. The user is
4544 supposed to write the last named argument of the function here.
4546 However, @code{va_start} should not use this argument. The way to find
4547 the end of the named arguments is with the built-in functions described
4550 @defmac __builtin_saveregs ()
4551 Use this built-in function to save the argument registers in memory so
4552 that the varargs mechanism can access them. Both ISO and traditional
4553 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4554 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4556 On some machines, @code{__builtin_saveregs} is open-coded under the
4557 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4558 other machines, it calls a routine written in assembler language,
4559 found in @file{libgcc2.c}.
4561 Code generated for the call to @code{__builtin_saveregs} appears at the
4562 beginning of the function, as opposed to where the call to
4563 @code{__builtin_saveregs} is written, regardless of what the code is.
4564 This is because the registers must be saved before the function starts
4565 to use them for its own purposes.
4566 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4570 @defmac __builtin_args_info (@var{category})
4571 Use this built-in function to find the first anonymous arguments in
4574 In general, a machine may have several categories of registers used for
4575 arguments, each for a particular category of data types. (For example,
4576 on some machines, floating-point registers are used for floating-point
4577 arguments while other arguments are passed in the general registers.)
4578 To make non-varargs functions use the proper calling convention, you
4579 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4580 registers in each category have been used so far
4582 @code{__builtin_args_info} accesses the same data structure of type
4583 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4584 with it, with @var{category} specifying which word to access. Thus, the
4585 value indicates the first unused register in a given category.
4587 Normally, you would use @code{__builtin_args_info} in the implementation
4588 of @code{va_start}, accessing each category just once and storing the
4589 value in the @code{va_list} object. This is because @code{va_list} will
4590 have to update the values, and there is no way to alter the
4591 values accessed by @code{__builtin_args_info}.
4594 @defmac __builtin_next_arg (@var{lastarg})
4595 This is the equivalent of @code{__builtin_args_info}, for stack
4596 arguments. It returns the address of the first anonymous stack
4597 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4598 returns the address of the location above the first anonymous stack
4599 argument. Use it in @code{va_start} to initialize the pointer for
4600 fetching arguments from the stack. Also use it in @code{va_start} to
4601 verify that the second parameter @var{lastarg} is the last named argument
4602 of the current function.
4605 @defmac __builtin_classify_type (@var{object})
4606 Since each machine has its own conventions for which data types are
4607 passed in which kind of register, your implementation of @code{va_arg}
4608 has to embody these conventions. The easiest way to categorize the
4609 specified data type is to use @code{__builtin_classify_type} together
4610 with @code{sizeof} and @code{__alignof__}.
4612 @code{__builtin_classify_type} ignores the value of @var{object},
4613 considering only its data type. It returns an integer describing what
4614 kind of type that is---integer, floating, pointer, structure, and so on.
4616 The file @file{typeclass.h} defines an enumeration that you can use to
4617 interpret the values of @code{__builtin_classify_type}.
4620 These machine description macros help implement varargs:
4622 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4623 If defined, this hook produces the machine-specific code for a call to
4624 @code{__builtin_saveregs}. This code will be moved to the very
4625 beginning of the function, before any parameter access are made. The
4626 return value of this function should be an RTX that contains the value
4627 to use as the return of @code{__builtin_saveregs}.
4630 @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})
4631 This target hook offers an alternative to using
4632 @code{__builtin_saveregs} and defining the hook
4633 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4634 register arguments into the stack so that all the arguments appear to
4635 have been passed consecutively on the stack. Once this is done, you can
4636 use the standard implementation of varargs that works for machines that
4637 pass all their arguments on the stack.
4639 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4640 structure, containing the values that are obtained after processing the
4641 named arguments. The arguments @var{mode} and @var{type} describe the
4642 last named argument---its machine mode and its data type as a tree node.
4644 The target hook should do two things: first, push onto the stack all the
4645 argument registers @emph{not} used for the named arguments, and second,
4646 store the size of the data thus pushed into the @code{int}-valued
4647 variable pointed to by @var{pretend_args_size}. The value that you
4648 store here will serve as additional offset for setting up the stack
4651 Because you must generate code to push the anonymous arguments at
4652 compile time without knowing their data types,
4653 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4654 have just a single category of argument register and use it uniformly
4657 If the argument @var{second_time} is nonzero, it means that the
4658 arguments of the function are being analyzed for the second time. This
4659 happens for an inline function, which is not actually compiled until the
4660 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4661 not generate any instructions in this case.
4664 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4665 Define this hook to return @code{true} if the location where a function
4666 argument is passed depends on whether or not it is a named argument.
4668 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4669 is set for varargs and stdarg functions. If this hook returns
4670 @code{true}, the @var{named} argument is always true for named
4671 arguments, and false for unnamed arguments. If it returns @code{false},
4672 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
4673 then all arguments are treated as named. Otherwise, all named arguments
4674 except the last are treated as named.
4676 You need not define this hook if it always returns zero.
4679 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4680 If you need to conditionally change ABIs so that one works with
4681 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4682 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4683 defined, then define this hook to return @code{true} if
4684 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4685 Otherwise, you should not define this hook.
4689 @section Trampolines for Nested Functions
4690 @cindex trampolines for nested functions
4691 @cindex nested functions, trampolines for
4693 A @dfn{trampoline} is a small piece of code that is created at run time
4694 when the address of a nested function is taken. It normally resides on
4695 the stack, in the stack frame of the containing function. These macros
4696 tell GCC how to generate code to allocate and initialize a
4699 The instructions in the trampoline must do two things: load a constant
4700 address into the static chain register, and jump to the real address of
4701 the nested function. On CISC machines such as the m68k, this requires
4702 two instructions, a move immediate and a jump. Then the two addresses
4703 exist in the trampoline as word-long immediate operands. On RISC
4704 machines, it is often necessary to load each address into a register in
4705 two parts. Then pieces of each address form separate immediate
4708 The code generated to initialize the trampoline must store the variable
4709 parts---the static chain value and the function address---into the
4710 immediate operands of the instructions. On a CISC machine, this is
4711 simply a matter of copying each address to a memory reference at the
4712 proper offset from the start of the trampoline. On a RISC machine, it
4713 may be necessary to take out pieces of the address and store them
4716 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4717 A C statement to output, on the stream @var{file}, assembler code for a
4718 block of data that contains the constant parts of a trampoline. This
4719 code should not include a label---the label is taken care of
4722 If you do not define this macro, it means no template is needed
4723 for the target. Do not define this macro on systems where the block move
4724 code to copy the trampoline into place would be larger than the code
4725 to generate it on the spot.
4728 @defmac TRAMPOLINE_SECTION
4729 The name of a subroutine to switch to the section in which the
4730 trampoline template is to be placed (@pxref{Sections}). The default is
4731 a value of @samp{readonly_data_section}, which places the trampoline in
4732 the section containing read-only data.
4735 @defmac TRAMPOLINE_SIZE
4736 A C expression for the size in bytes of the trampoline, as an integer.
4739 @defmac TRAMPOLINE_ALIGNMENT
4740 Alignment required for trampolines, in bits.
4742 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4743 is used for aligning trampolines.
4746 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4747 A C statement to initialize the variable parts of a trampoline.
4748 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4749 an RTX for the address of the nested function; @var{static_chain} is an
4750 RTX for the static chain value that should be passed to the function
4754 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4755 A C statement that should perform any machine-specific adjustment in
4756 the address of the trampoline. Its argument contains the address that
4757 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4758 used for a function call should be different from the address in which
4759 the template was stored, the different address should be assigned to
4760 @var{addr}. If this macro is not defined, @var{addr} will be used for
4763 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4764 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4765 If this macro is not defined, by default the trampoline is allocated as
4766 a stack slot. This default is right for most machines. The exceptions
4767 are machines where it is impossible to execute instructions in the stack
4768 area. On such machines, you may have to implement a separate stack,
4769 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4770 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4772 @var{fp} points to a data structure, a @code{struct function}, which
4773 describes the compilation status of the immediate containing function of
4774 the function which the trampoline is for. The stack slot for the
4775 trampoline is in the stack frame of this containing function. Other
4776 allocation strategies probably must do something analogous with this
4780 Implementing trampolines is difficult on many machines because they have
4781 separate instruction and data caches. Writing into a stack location
4782 fails to clear the memory in the instruction cache, so when the program
4783 jumps to that location, it executes the old contents.
4785 Here are two possible solutions. One is to clear the relevant parts of
4786 the instruction cache whenever a trampoline is set up. The other is to
4787 make all trampolines identical, by having them jump to a standard
4788 subroutine. The former technique makes trampoline execution faster; the
4789 latter makes initialization faster.
4791 To clear the instruction cache when a trampoline is initialized, define
4792 the following macro.
4794 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4795 If defined, expands to a C expression clearing the @emph{instruction
4796 cache} in the specified interval. The definition of this macro would
4797 typically be a series of @code{asm} statements. Both @var{beg} and
4798 @var{end} are both pointer expressions.
4801 The operating system may also require the stack to be made executable
4802 before calling the trampoline. To implement this requirement, define
4803 the following macro.
4805 @defmac ENABLE_EXECUTE_STACK
4806 Define this macro if certain operations must be performed before executing
4807 code located on the stack. The macro should expand to a series of C
4808 file-scope constructs (e.g.@: functions) and provide a unique entry point
4809 named @code{__enable_execute_stack}. The target is responsible for
4810 emitting calls to the entry point in the code, for example from the
4811 @code{INITIALIZE_TRAMPOLINE} macro.
4814 To use a standard subroutine, define the following macro. In addition,
4815 you must make sure that the instructions in a trampoline fill an entire
4816 cache line with identical instructions, or else ensure that the
4817 beginning of the trampoline code is always aligned at the same point in
4818 its cache line. Look in @file{m68k.h} as a guide.
4820 @defmac TRANSFER_FROM_TRAMPOLINE
4821 Define this macro if trampolines need a special subroutine to do their
4822 work. The macro should expand to a series of @code{asm} statements
4823 which will be compiled with GCC@. They go in a library function named
4824 @code{__transfer_from_trampoline}.
4826 If you need to avoid executing the ordinary prologue code of a compiled
4827 C function when you jump to the subroutine, you can do so by placing a
4828 special label of your own in the assembler code. Use one @code{asm}
4829 statement to generate an assembler label, and another to make the label
4830 global. Then trampolines can use that label to jump directly to your
4831 special assembler code.
4835 @section Implicit Calls to Library Routines
4836 @cindex library subroutine names
4837 @cindex @file{libgcc.a}
4839 @c prevent bad page break with this line
4840 Here is an explanation of implicit calls to library routines.
4842 @defmac DECLARE_LIBRARY_RENAMES
4843 This macro, if defined, should expand to a piece of C code that will get
4844 expanded when compiling functions for libgcc.a. It can be used to
4845 provide alternate names for GCC's internal library functions if there
4846 are ABI-mandated names that the compiler should provide.
4849 @findex init_one_libfunc
4850 @findex set_optab_libfunc
4851 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
4852 This hook should declare additional library routines or rename
4853 existing ones, using the functions @code{set_optab_libfunc} and
4854 @code{init_one_libfunc} defined in @file{optabs.c}.
4855 @code{init_optabs} calls this macro after initializing all the normal
4858 The default is to do nothing. Most ports don't need to define this hook.
4861 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4862 This macro should return @code{true} if the library routine that
4863 implements the floating point comparison operator @var{comparison} in
4864 mode @var{mode} will return a boolean, and @var{false} if it will
4867 GCC's own floating point libraries return tristates from the
4868 comparison operators, so the default returns false always. Most ports
4869 don't need to define this macro.
4872 @defmac TARGET_LIB_INT_CMP_BIASED
4873 This macro should evaluate to @code{true} if the integer comparison
4874 functions (like @code{__cmpdi2}) return 0 to indicate that the first
4875 operand is smaller than the second, 1 to indicate that they are equal,
4876 and 2 to indicate that the first operand is greater than the second.
4877 If this macro evaluates to @code{false} the comparison functions return
4878 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
4879 in @file{libgcc.a}, you do not need to define this macro.
4882 @cindex US Software GOFAST, floating point emulation library
4883 @cindex floating point emulation library, US Software GOFAST
4884 @cindex GOFAST, floating point emulation library
4885 @findex gofast_maybe_init_libfuncs
4886 @defmac US_SOFTWARE_GOFAST
4887 Define this macro if your system C library uses the US Software GOFAST
4888 library to provide floating point emulation.
4890 In addition to defining this macro, your architecture must set
4891 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
4892 else call that function from its version of that hook. It is defined
4893 in @file{config/gofast.h}, which must be included by your
4894 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
4897 If this macro is defined, the
4898 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
4899 false for @code{SFmode} and @code{DFmode} comparisons.
4902 @cindex @code{EDOM}, implicit usage
4905 The value of @code{EDOM} on the target machine, as a C integer constant
4906 expression. If you don't define this macro, GCC does not attempt to
4907 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4908 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4911 If you do not define @code{TARGET_EDOM}, then compiled code reports
4912 domain errors by calling the library function and letting it report the
4913 error. If mathematical functions on your system use @code{matherr} when
4914 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4915 that @code{matherr} is used normally.
4918 @cindex @code{errno}, implicit usage
4919 @defmac GEN_ERRNO_RTX
4920 Define this macro as a C expression to create an rtl expression that
4921 refers to the global ``variable'' @code{errno}. (On certain systems,
4922 @code{errno} may not actually be a variable.) If you don't define this
4923 macro, a reasonable default is used.
4926 @cindex C99 math functions, implicit usage
4927 @defmac TARGET_C99_FUNCTIONS
4928 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4929 @code{sinf} and similarly for other functions defined by C99 standard. The
4930 default is nonzero that should be proper value for most modern systems, however
4931 number of existing systems lacks support for these functions in the runtime so
4932 they needs this macro to be redefined to 0.
4935 @defmac NEXT_OBJC_RUNTIME
4936 Define this macro to generate code for Objective-C message sending using
4937 the calling convention of the NeXT system. This calling convention
4938 involves passing the object, the selector and the method arguments all
4939 at once to the method-lookup library function.
4941 The default calling convention passes just the object and the selector
4942 to the lookup function, which returns a pointer to the method.
4945 @node Addressing Modes
4946 @section Addressing Modes
4947 @cindex addressing modes
4949 @c prevent bad page break with this line
4950 This is about addressing modes.
4952 @defmac HAVE_PRE_INCREMENT
4953 @defmacx HAVE_PRE_DECREMENT
4954 @defmacx HAVE_POST_INCREMENT
4955 @defmacx HAVE_POST_DECREMENT
4956 A C expression that is nonzero if the machine supports pre-increment,
4957 pre-decrement, post-increment, or post-decrement addressing respectively.
4960 @defmac HAVE_PRE_MODIFY_DISP
4961 @defmacx HAVE_POST_MODIFY_DISP
4962 A C expression that is nonzero if the machine supports pre- or
4963 post-address side-effect generation involving constants other than
4964 the size of the memory operand.
4967 @defmac HAVE_PRE_MODIFY_REG
4968 @defmacx HAVE_POST_MODIFY_REG
4969 A C expression that is nonzero if the machine supports pre- or
4970 post-address side-effect generation involving a register displacement.
4973 @defmac CONSTANT_ADDRESS_P (@var{x})
4974 A C expression that is 1 if the RTX @var{x} is a constant which
4975 is a valid address. On most machines, this can be defined as
4976 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4977 in which constant addresses are supported.
4980 @defmac CONSTANT_P (@var{x})
4981 @code{CONSTANT_P}, which is defined by target-independent code,
4982 accepts integer-values expressions whose values are not explicitly
4983 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4984 expressions and @code{const} arithmetic expressions, in addition to
4985 @code{const_int} and @code{const_double} expressions.
4988 @defmac MAX_REGS_PER_ADDRESS
4989 A number, the maximum number of registers that can appear in a valid
4990 memory address. Note that it is up to you to specify a value equal to
4991 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4995 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4996 A C compound statement with a conditional @code{goto @var{label};}
4997 executed if @var{x} (an RTX) is a legitimate memory address on the
4998 target machine for a memory operand of mode @var{mode}.
5000 It usually pays to define several simpler macros to serve as
5001 subroutines for this one. Otherwise it may be too complicated to
5004 This macro must exist in two variants: a strict variant and a
5005 non-strict one. The strict variant is used in the reload pass. It
5006 must be defined so that any pseudo-register that has not been
5007 allocated a hard register is considered a memory reference. In
5008 contexts where some kind of register is required, a pseudo-register
5009 with no hard register must be rejected.
5011 The non-strict variant is used in other passes. It must be defined to
5012 accept all pseudo-registers in every context where some kind of
5013 register is required.
5015 @findex REG_OK_STRICT
5016 Compiler source files that want to use the strict variant of this
5017 macro define the macro @code{REG_OK_STRICT}. You should use an
5018 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
5019 in that case and the non-strict variant otherwise.
5021 Subroutines to check for acceptable registers for various purposes (one
5022 for base registers, one for index registers, and so on) are typically
5023 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
5024 Then only these subroutine macros need have two variants; the higher
5025 levels of macros may be the same whether strict or not.
5027 Normally, constant addresses which are the sum of a @code{symbol_ref}
5028 and an integer are stored inside a @code{const} RTX to mark them as
5029 constant. Therefore, there is no need to recognize such sums
5030 specifically as legitimate addresses. Normally you would simply
5031 recognize any @code{const} as legitimate.
5033 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5034 sums that are not marked with @code{const}. It assumes that a naked
5035 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5036 naked constant sums as illegitimate addresses, so that none of them will
5037 be given to @code{PRINT_OPERAND_ADDRESS}.
5039 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5040 On some machines, whether a symbolic address is legitimate depends on
5041 the section that the address refers to. On these machines, define the
5042 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5043 into the @code{symbol_ref}, and then check for it here. When you see a
5044 @code{const}, you will have to look inside it to find the
5045 @code{symbol_ref} in order to determine the section. @xref{Assembler
5049 @defmac REG_OK_FOR_BASE_P (@var{x})
5050 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
5051 RTX) is valid for use as a base register. For hard registers, it
5052 should always accept those which the hardware permits and reject the
5053 others. Whether the macro accepts or rejects pseudo registers must be
5054 controlled by @code{REG_OK_STRICT} as described above. This usually
5055 requires two variant definitions, of which @code{REG_OK_STRICT}
5056 controls the one actually used.
5059 @defmac REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
5060 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
5061 that expression may examine the mode of the memory reference in
5062 @var{mode}. You should define this macro if the mode of the memory
5063 reference affects whether a register may be used as a base register. If
5064 you define this macro, the compiler will use it instead of
5065 @code{REG_OK_FOR_BASE_P}.
5068 @defmac REG_MODE_OK_FOR_REG_BASE_P (@var{x}, @var{mode})
5069 A C expression which is nonzero if @var{x} (assumed to be a @code{reg} RTX)
5070 is suitable for use as a base register in base plus index operand addresses,
5071 accessing memory in mode @var{mode}. It may be either a suitable hard
5072 register or a pseudo register that has been allocated such a hard register.
5073 You should define this macro if base plus index addresses have different
5074 requirements than other base register uses.
5077 @defmac REG_OK_FOR_INDEX_P (@var{x})
5078 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
5079 RTX) is valid for use as an index register.
5081 The difference between an index register and a base register is that
5082 the index register may be scaled. If an address involves the sum of
5083 two registers, neither one of them scaled, then either one may be
5084 labeled the ``base'' and the other the ``index''; but whichever
5085 labeling is used must fit the machine's constraints of which registers
5086 may serve in each capacity. The compiler will try both labelings,
5087 looking for one that is valid, and will reload one or both registers
5088 only if neither labeling works.
5091 @defmac FIND_BASE_TERM (@var{x})
5092 A C expression to determine the base term of address @var{x}.
5093 This macro is used in only one place: `find_base_term' in alias.c.
5095 It is always safe for this macro to not be defined. It exists so
5096 that alias analysis can understand machine-dependent addresses.
5098 The typical use of this macro is to handle addresses containing
5099 a label_ref or symbol_ref within an UNSPEC@.
5102 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
5103 A C compound statement that attempts to replace @var{x} with a valid
5104 memory address for an operand of mode @var{mode}. @var{win} will be a
5105 C statement label elsewhere in the code; the macro definition may use
5108 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
5112 to avoid further processing if the address has become legitimate.
5114 @findex break_out_memory_refs
5115 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5116 and @var{oldx} will be the operand that was given to that function to produce
5119 The code generated by this macro should not alter the substructure of
5120 @var{x}. If it transforms @var{x} into a more legitimate form, it
5121 should assign @var{x} (which will always be a C variable) a new value.
5123 It is not necessary for this macro to come up with a legitimate
5124 address. The compiler has standard ways of doing so in all cases. In
5125 fact, it is safe to omit this macro. But often a
5126 machine-dependent strategy can generate better code.
5129 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5130 A C compound statement that attempts to replace @var{x}, which is an address
5131 that needs reloading, with a valid memory address for an operand of mode
5132 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5133 It is not necessary to define this macro, but it might be useful for
5134 performance reasons.
5136 For example, on the i386, it is sometimes possible to use a single
5137 reload register instead of two by reloading a sum of two pseudo
5138 registers into a register. On the other hand, for number of RISC
5139 processors offsets are limited so that often an intermediate address
5140 needs to be generated in order to address a stack slot. By defining
5141 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5142 generated for adjacent some stack slots can be made identical, and thus
5145 @emph{Note}: This macro should be used with caution. It is necessary
5146 to know something of how reload works in order to effectively use this,
5147 and it is quite easy to produce macros that build in too much knowledge
5148 of reload internals.
5150 @emph{Note}: This macro must be able to reload an address created by a
5151 previous invocation of this macro. If it fails to handle such addresses
5152 then the compiler may generate incorrect code or abort.
5155 The macro definition should use @code{push_reload} to indicate parts that
5156 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5157 suitable to be passed unaltered to @code{push_reload}.
5159 The code generated by this macro must not alter the substructure of
5160 @var{x}. If it transforms @var{x} into a more legitimate form, it
5161 should assign @var{x} (which will always be a C variable) a new value.
5162 This also applies to parts that you change indirectly by calling
5165 @findex strict_memory_address_p
5166 The macro definition may use @code{strict_memory_address_p} to test if
5167 the address has become legitimate.
5170 If you want to change only a part of @var{x}, one standard way of doing
5171 this is to use @code{copy_rtx}. Note, however, that is unshares only a
5172 single level of rtl. Thus, if the part to be changed is not at the
5173 top level, you'll need to replace first the top level.
5174 It is not necessary for this macro to come up with a legitimate
5175 address; but often a machine-dependent strategy can generate better code.
5178 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5179 A C statement or compound statement with a conditional @code{goto
5180 @var{label};} executed if memory address @var{x} (an RTX) can have
5181 different meanings depending on the machine mode of the memory
5182 reference it is used for or if the address is valid for some modes
5185 Autoincrement and autodecrement addresses typically have mode-dependent
5186 effects because the amount of the increment or decrement is the size
5187 of the operand being addressed. Some machines have other mode-dependent
5188 addresses. Many RISC machines have no mode-dependent addresses.
5190 You may assume that @var{addr} is a valid address for the machine.
5193 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5194 A C expression that is nonzero if @var{x} is a legitimate constant for
5195 an immediate operand on the target machine. You can assume that
5196 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5197 @samp{1} is a suitable definition for this macro on machines where
5198 anything @code{CONSTANT_P} is valid.
5201 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5202 This hook is used to undo the possibly obfuscating effects of the
5203 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5204 macros. Some backend implementations of these macros wrap symbol
5205 references inside an @code{UNSPEC} rtx to represent PIC or similar
5206 addressing modes. This target hook allows GCC's optimizers to understand
5207 the semantics of these opaque @code{UNSPEC}s by converting them back
5208 into their original form.
5211 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x})
5212 This hook should return true if @var{x} is of a form that cannot (or
5213 should not) be spilled to the constant pool. The default version of
5214 this hook returns false.
5216 The primary reason to define this hook is to prevent reload from
5217 deciding that a non-legitimate constant would be better reloaded
5218 from the constant pool instead of spilling and reloading a register
5219 holding the constant. This restriction is often true of addresses
5220 of TLS symbols for various targets.
5223 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5224 This hook should return the DECL of a function @var{f} that given an
5225 address @var{addr} as an argument returns a mask @var{m} that can be
5226 used to extract from two vectors the relevant data that resides in
5227 @var{addr} in case @var{addr} is not properly aligned.
5229 The autovectrizer, when vectorizing a load operation from an address
5230 @var{addr} that may be unaligned, will generate two vector loads from
5231 the two aligned addresses around @var{addr}. It then generates a
5232 @code{REALIGN_LOAD} operation to extract the relevant data from the
5233 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5234 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5235 the third argument, @var{OFF}, defines how the data will be extracted
5236 from these two vectors: if @var{OFF} is 0, then the returned vector is
5237 @var{v2}; otherwise, the returned vector is composed from the last
5238 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5239 @var{OFF} elements of @var{v2}.
5241 If this hook is defined, the autovectorizer will generate a call
5242 to @var{f} (using the DECL tree that this hook returns) and will
5243 use the return value of @var{f} as the argument @var{OFF} to
5244 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5245 should comply with the semantics expected by @code{REALIGN_LOAD}
5247 If this hook is not defined, then @var{addr} will be used as
5248 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5249 log2(@var{VS})-1 bits of @var{addr} will be considered.
5252 @node Condition Code
5253 @section Condition Code Status
5254 @cindex condition code status
5256 @c prevent bad page break with this line
5257 This describes the condition code status.
5260 The file @file{conditions.h} defines a variable @code{cc_status} to
5261 describe how the condition code was computed (in case the interpretation of
5262 the condition code depends on the instruction that it was set by). This
5263 variable contains the RTL expressions on which the condition code is
5264 currently based, and several standard flags.
5266 Sometimes additional machine-specific flags must be defined in the machine
5267 description header file. It can also add additional machine-specific
5268 information by defining @code{CC_STATUS_MDEP}.
5270 @defmac CC_STATUS_MDEP
5271 C code for a data type which is used for declaring the @code{mdep}
5272 component of @code{cc_status}. It defaults to @code{int}.
5274 This macro is not used on machines that do not use @code{cc0}.
5277 @defmac CC_STATUS_MDEP_INIT
5278 A C expression to initialize the @code{mdep} field to ``empty''.
5279 The default definition does nothing, since most machines don't use
5280 the field anyway. If you want to use the field, you should probably
5281 define this macro to initialize it.
5283 This macro is not used on machines that do not use @code{cc0}.
5286 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5287 A C compound statement to set the components of @code{cc_status}
5288 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5289 this macro's responsibility to recognize insns that set the condition
5290 code as a byproduct of other activity as well as those that explicitly
5293 This macro is not used on machines that do not use @code{cc0}.
5295 If there are insns that do not set the condition code but do alter
5296 other machine registers, this macro must check to see whether they
5297 invalidate the expressions that the condition code is recorded as
5298 reflecting. For example, on the 68000, insns that store in address
5299 registers do not set the condition code, which means that usually
5300 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5301 insns. But suppose that the previous insn set the condition code
5302 based on location @samp{a4@@(102)} and the current insn stores a new
5303 value in @samp{a4}. Although the condition code is not changed by
5304 this, it will no longer be true that it reflects the contents of
5305 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5306 @code{cc_status} in this case to say that nothing is known about the
5307 condition code value.
5309 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5310 with the results of peephole optimization: insns whose patterns are
5311 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5312 constants which are just the operands. The RTL structure of these
5313 insns is not sufficient to indicate what the insns actually do. What
5314 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5315 @code{CC_STATUS_INIT}.
5317 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5318 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5319 @samp{cc}. This avoids having detailed information about patterns in
5320 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5323 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5324 Returns a mode from class @code{MODE_CC} to be used when comparison
5325 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5326 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5327 @pxref{Jump Patterns} for a description of the reason for this
5331 #define SELECT_CC_MODE(OP,X,Y) \
5332 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5333 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5334 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5335 || GET_CODE (X) == NEG) \
5336 ? CC_NOOVmode : CCmode))
5339 You should define this macro if and only if you define extra CC modes
5340 in @file{@var{machine}-modes.def}.
5343 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5344 On some machines not all possible comparisons are defined, but you can
5345 convert an invalid comparison into a valid one. For example, the Alpha
5346 does not have a @code{GT} comparison, but you can use an @code{LT}
5347 comparison instead and swap the order of the operands.
5349 On such machines, define this macro to be a C statement to do any
5350 required conversions. @var{code} is the initial comparison code
5351 and @var{op0} and @var{op1} are the left and right operands of the
5352 comparison, respectively. You should modify @var{code}, @var{op0}, and
5353 @var{op1} as required.
5355 GCC will not assume that the comparison resulting from this macro is
5356 valid but will see if the resulting insn matches a pattern in the
5359 You need not define this macro if it would never change the comparison
5363 @defmac REVERSIBLE_CC_MODE (@var{mode})
5364 A C expression whose value is one if it is always safe to reverse a
5365 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5366 can ever return @var{mode} for a floating-point inequality comparison,
5367 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5369 You need not define this macro if it would always returns zero or if the
5370 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5371 For example, here is the definition used on the SPARC, where floating-point
5372 inequality comparisons are always given @code{CCFPEmode}:
5375 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5379 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5380 A C expression whose value is reversed condition code of the @var{code} for
5381 comparison done in CC_MODE @var{mode}. The macro is used only in case
5382 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5383 machine has some non-standard way how to reverse certain conditionals. For
5384 instance in case all floating point conditions are non-trapping, compiler may
5385 freely convert unordered compares to ordered one. Then definition may look
5389 #define REVERSE_CONDITION(CODE, MODE) \
5390 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5391 : reverse_condition_maybe_unordered (CODE))
5395 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
5396 A C expression that returns true if the conditional execution predicate
5397 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
5398 versa. Define this to return 0 if the target has conditional execution
5399 predicates that cannot be reversed safely. There is no need to validate
5400 that the arguments of op1 and op2 are the same, this is done separately.
5401 If no expansion is specified, this macro is defined as follows:
5404 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5405 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
5409 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5410 On targets which do not use @code{(cc0)}, and which use a hard
5411 register rather than a pseudo-register to hold condition codes, the
5412 regular CSE passes are often not able to identify cases in which the
5413 hard register is set to a common value. Use this hook to enable a
5414 small pass which optimizes such cases. This hook should return true
5415 to enable this pass, and it should set the integers to which its
5416 arguments point to the hard register numbers used for condition codes.
5417 When there is only one such register, as is true on most systems, the
5418 integer pointed to by the second argument should be set to
5419 @code{INVALID_REGNUM}.
5421 The default version of this hook returns false.
5424 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5425 On targets which use multiple condition code modes in class
5426 @code{MODE_CC}, it is sometimes the case that a comparison can be
5427 validly done in more than one mode. On such a system, define this
5428 target hook to take two mode arguments and to return a mode in which
5429 both comparisons may be validly done. If there is no such mode,
5430 return @code{VOIDmode}.
5432 The default version of this hook checks whether the modes are the
5433 same. If they are, it returns that mode. If they are different, it
5434 returns @code{VOIDmode}.
5438 @section Describing Relative Costs of Operations
5439 @cindex costs of instructions
5440 @cindex relative costs
5441 @cindex speed of instructions
5443 These macros let you describe the relative speed of various operations
5444 on the target machine.
5446 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5447 A C expression for the cost of moving data of mode @var{mode} from a
5448 register in class @var{from} to one in class @var{to}. The classes are
5449 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5450 value of 2 is the default; other values are interpreted relative to
5453 It is not required that the cost always equal 2 when @var{from} is the
5454 same as @var{to}; on some machines it is expensive to move between
5455 registers if they are not general registers.
5457 If reload sees an insn consisting of a single @code{set} between two
5458 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5459 classes returns a value of 2, reload does not check to ensure that the
5460 constraints of the insn are met. Setting a cost of other than 2 will
5461 allow reload to verify that the constraints are met. You should do this
5462 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5465 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5466 A C expression for the cost of moving data of mode @var{mode} between a
5467 register of class @var{class} and memory; @var{in} is zero if the value
5468 is to be written to memory, nonzero if it is to be read in. This cost
5469 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5470 registers and memory is more expensive than between two registers, you
5471 should define this macro to express the relative cost.
5473 If you do not define this macro, GCC uses a default cost of 4 plus
5474 the cost of copying via a secondary reload register, if one is
5475 needed. If your machine requires a secondary reload register to copy
5476 between memory and a register of @var{class} but the reload mechanism is
5477 more complex than copying via an intermediate, define this macro to
5478 reflect the actual cost of the move.
5480 GCC defines the function @code{memory_move_secondary_cost} if
5481 secondary reloads are needed. It computes the costs due to copying via
5482 a secondary register. If your machine copies from memory using a
5483 secondary register in the conventional way but the default base value of
5484 4 is not correct for your machine, define this macro to add some other
5485 value to the result of that function. The arguments to that function
5486 are the same as to this macro.
5490 A C expression for the cost of a branch instruction. A value of 1 is
5491 the default; other values are interpreted relative to that.
5494 Here are additional macros which do not specify precise relative costs,
5495 but only that certain actions are more expensive than GCC would
5498 @defmac SLOW_BYTE_ACCESS
5499 Define this macro as a C expression which is nonzero if accessing less
5500 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5501 faster than accessing a word of memory, i.e., if such access
5502 require more than one instruction or if there is no difference in cost
5503 between byte and (aligned) word loads.
5505 When this macro is not defined, the compiler will access a field by
5506 finding the smallest containing object; when it is defined, a fullword
5507 load will be used if alignment permits. Unless bytes accesses are
5508 faster than word accesses, using word accesses is preferable since it
5509 may eliminate subsequent memory access if subsequent accesses occur to
5510 other fields in the same word of the structure, but to different bytes.
5513 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5514 Define this macro to be the value 1 if memory accesses described by the
5515 @var{mode} and @var{alignment} parameters have a cost many times greater
5516 than aligned accesses, for example if they are emulated in a trap
5519 When this macro is nonzero, the compiler will act as if
5520 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5521 moves. This can cause significantly more instructions to be produced.
5522 Therefore, do not set this macro nonzero if unaligned accesses only add a
5523 cycle or two to the time for a memory access.
5525 If the value of this macro is always zero, it need not be defined. If
5526 this macro is defined, it should produce a nonzero value when
5527 @code{STRICT_ALIGNMENT} is nonzero.
5531 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5532 which a sequence of insns should be generated instead of a
5533 string move insn or a library call. Increasing the value will always
5534 make code faster, but eventually incurs high cost in increased code size.
5536 Note that on machines where the corresponding move insn is a
5537 @code{define_expand} that emits a sequence of insns, this macro counts
5538 the number of such sequences.
5540 If you don't define this, a reasonable default is used.
5543 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5544 A C expression used to determine whether @code{move_by_pieces} will be used to
5545 copy a chunk of memory, or whether some other block move mechanism
5546 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5547 than @code{MOVE_RATIO}.
5550 @defmac MOVE_MAX_PIECES
5551 A C expression used by @code{move_by_pieces} to determine the largest unit
5552 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5556 The threshold of number of scalar move insns, @emph{below} which a sequence
5557 of insns should be generated to clear memory instead of a string clear insn
5558 or a library call. Increasing the value will always make code faster, but
5559 eventually incurs high cost in increased code size.
5561 If you don't define this, a reasonable default is used.
5564 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5565 A C expression used to determine whether @code{clear_by_pieces} will be used
5566 to clear a chunk of memory, or whether some other block clear mechanism
5567 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5568 than @code{CLEAR_RATIO}.
5571 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5572 A C expression used to determine whether @code{store_by_pieces} will be
5573 used to set a chunk of memory to a constant value, or whether some other
5574 mechanism will be used. Used by @code{__builtin_memset} when storing
5575 values other than constant zero and by @code{__builtin_strcpy} when
5576 when called with a constant source string.
5577 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5578 than @code{MOVE_RATIO}.
5581 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5582 A C expression used to determine whether a load postincrement is a good
5583 thing to use for a given mode. Defaults to the value of
5584 @code{HAVE_POST_INCREMENT}.
5587 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5588 A C expression used to determine whether a load postdecrement is a good
5589 thing to use for a given mode. Defaults to the value of
5590 @code{HAVE_POST_DECREMENT}.
5593 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5594 A C expression used to determine whether a load preincrement is a good
5595 thing to use for a given mode. Defaults to the value of
5596 @code{HAVE_PRE_INCREMENT}.
5599 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5600 A C expression used to determine whether a load predecrement is a good
5601 thing to use for a given mode. Defaults to the value of
5602 @code{HAVE_PRE_DECREMENT}.
5605 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5606 A C expression used to determine whether a store postincrement is a good
5607 thing to use for a given mode. Defaults to the value of
5608 @code{HAVE_POST_INCREMENT}.
5611 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5612 A C expression used to determine whether a store postdecrement is a good
5613 thing to use for a given mode. Defaults to the value of
5614 @code{HAVE_POST_DECREMENT}.
5617 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5618 This macro is used to determine whether a store preincrement is a good
5619 thing to use for a given mode. Defaults to the value of
5620 @code{HAVE_PRE_INCREMENT}.
5623 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5624 This macro is used to determine whether a store predecrement is a good
5625 thing to use for a given mode. Defaults to the value of
5626 @code{HAVE_PRE_DECREMENT}.
5629 @defmac NO_FUNCTION_CSE
5630 Define this macro if it is as good or better to call a constant
5631 function address than to call an address kept in a register.
5634 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5635 Define this macro if a non-short-circuit operation produced by
5636 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5637 @code{BRANCH_COST} is greater than or equal to the value 2.
5640 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5641 This target hook describes the relative costs of RTL expressions.
5643 The cost may depend on the precise form of the expression, which is
5644 available for examination in @var{x}, and the rtx code of the expression
5645 in which it is contained, found in @var{outer_code}. @var{code} is the
5646 expression code---redundant, since it can be obtained with
5647 @code{GET_CODE (@var{x})}.
5649 In implementing this hook, you can use the construct
5650 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5653 On entry to the hook, @code{*@var{total}} contains a default estimate
5654 for the cost of the expression. The hook should modify this value as
5655 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
5656 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
5657 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
5659 When optimizing for code size, i.e.@: when @code{optimize_size} is
5660 nonzero, this target hook should be used to estimate the relative
5661 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
5663 The hook returns true when all subexpressions of @var{x} have been
5664 processed, and false when @code{rtx_cost} should recurse.
5667 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5668 This hook computes the cost of an addressing mode that contains
5669 @var{address}. If not defined, the cost is computed from
5670 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5672 For most CISC machines, the default cost is a good approximation of the
5673 true cost of the addressing mode. However, on RISC machines, all
5674 instructions normally have the same length and execution time. Hence
5675 all addresses will have equal costs.
5677 In cases where more than one form of an address is known, the form with
5678 the lowest cost will be used. If multiple forms have the same, lowest,
5679 cost, the one that is the most complex will be used.
5681 For example, suppose an address that is equal to the sum of a register
5682 and a constant is used twice in the same basic block. When this macro
5683 is not defined, the address will be computed in a register and memory
5684 references will be indirect through that register. On machines where
5685 the cost of the addressing mode containing the sum is no higher than
5686 that of a simple indirect reference, this will produce an additional
5687 instruction and possibly require an additional register. Proper
5688 specification of this macro eliminates this overhead for such machines.
5690 This hook is never called with an invalid address.
5692 On machines where an address involving more than one register is as
5693 cheap as an address computation involving only one register, defining
5694 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5695 be live over a region of code where only one would have been if
5696 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5697 should be considered in the definition of this macro. Equivalent costs
5698 should probably only be given to addresses with different numbers of
5699 registers on machines with lots of registers.
5703 @section Adjusting the Instruction Scheduler
5705 The instruction scheduler may need a fair amount of machine-specific
5706 adjustment in order to produce good code. GCC provides several target
5707 hooks for this purpose. It is usually enough to define just a few of
5708 them: try the first ones in this list first.
5710 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5711 This hook returns the maximum number of instructions that can ever
5712 issue at the same time on the target machine. The default is one.
5713 Although the insn scheduler can define itself the possibility of issue
5714 an insn on the same cycle, the value can serve as an additional
5715 constraint to issue insns on the same simulated processor cycle (see
5716 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5717 This value must be constant over the entire compilation. If you need
5718 it to vary depending on what the instructions are, you must use
5719 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5721 You could define this hook to return the value of the macro
5722 @code{MAX_DFA_ISSUE_RATE}.
5725 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5726 This hook is executed by the scheduler after it has scheduled an insn
5727 from the ready list. It should return the number of insns which can
5728 still be issued in the current cycle. The default is
5729 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5730 @code{USE}, which normally are not counted against the issue rate.
5731 You should define this hook if some insns take more machine resources
5732 than others, so that fewer insns can follow them in the same cycle.
5733 @var{file} is either a null pointer, or a stdio stream to write any
5734 debug output to. @var{verbose} is the verbose level provided by
5735 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5739 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5740 This function corrects the value of @var{cost} based on the
5741 relationship between @var{insn} and @var{dep_insn} through the
5742 dependence @var{link}. It should return the new value. The default
5743 is to make no adjustment to @var{cost}. This can be used for example
5744 to specify to the scheduler using the traditional pipeline description
5745 that an output- or anti-dependence does not incur the same cost as a
5746 data-dependence. If the scheduler using the automaton based pipeline
5747 description, the cost of anti-dependence is zero and the cost of
5748 output-dependence is maximum of one and the difference of latency
5749 times of the first and the second insns. If these values are not
5750 acceptable, you could use the hook to modify them too. See also
5751 @pxref{Processor pipeline description}.
5754 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5755 This hook adjusts the integer scheduling priority @var{priority} of
5756 @var{insn}. It should return the new priority. Reduce the priority to
5757 execute @var{insn} earlier, increase the priority to execute @var{insn}
5758 later. Do not define this hook if you do not need to adjust the
5759 scheduling priorities of insns.
5762 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5763 This hook is executed by the scheduler after it has scheduled the ready
5764 list, to allow the machine description to reorder it (for example to
5765 combine two small instructions together on @samp{VLIW} machines).
5766 @var{file} is either a null pointer, or a stdio stream to write any
5767 debug output to. @var{verbose} is the verbose level provided by
5768 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5769 list of instructions that are ready to be scheduled. @var{n_readyp} is
5770 a pointer to the number of elements in the ready list. The scheduler
5771 reads the ready list in reverse order, starting with
5772 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5773 is the timer tick of the scheduler. You may modify the ready list and
5774 the number of ready insns. The return value is the number of insns that
5775 can issue this cycle; normally this is just @code{issue_rate}. See also
5776 @samp{TARGET_SCHED_REORDER2}.
5779 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5780 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5781 function is called whenever the scheduler starts a new cycle. This one
5782 is called once per iteration over a cycle, immediately after
5783 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5784 return the number of insns to be scheduled in the same cycle. Defining
5785 this hook can be useful if there are frequent situations where
5786 scheduling one insn causes other insns to become ready in the same
5787 cycle. These other insns can then be taken into account properly.
5790 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5791 This hook is called after evaluation forward dependencies of insns in
5792 chain given by two parameter values (@var{head} and @var{tail}
5793 correspondingly) but before insns scheduling of the insn chain. For
5794 example, it can be used for better insn classification if it requires
5795 analysis of dependencies. This hook can use backward and forward
5796 dependencies of the insn scheduler because they are already
5800 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5801 This hook is executed by the scheduler at the beginning of each block of
5802 instructions that are to be scheduled. @var{file} is either a null
5803 pointer, or a stdio stream to write any debug output to. @var{verbose}
5804 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5805 @var{max_ready} is the maximum number of insns in the current scheduling
5806 region that can be live at the same time. This can be used to allocate
5807 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
5810 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5811 This hook is executed by the scheduler at the end of each block of
5812 instructions that are to be scheduled. It can be used to perform
5813 cleanup of any actions done by the other scheduling hooks. @var{file}
5814 is either a null pointer, or a stdio stream to write any debug output
5815 to. @var{verbose} is the verbose level provided by
5816 @option{-fsched-verbose-@var{n}}.
5819 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
5820 This hook is executed by the scheduler after function level initializations.
5821 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5822 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5823 @var{old_max_uid} is the maximum insn uid when scheduling begins.
5826 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
5827 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
5828 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5829 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5832 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5833 The hook returns an RTL insn. The automaton state used in the
5834 pipeline hazard recognizer is changed as if the insn were scheduled
5835 when the new simulated processor cycle starts. Usage of the hook may
5836 simplify the automaton pipeline description for some @acronym{VLIW}
5837 processors. If the hook is defined, it is used only for the automaton
5838 based pipeline description. The default is not to change the state
5839 when the new simulated processor cycle starts.
5842 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5843 The hook can be used to initialize data used by the previous hook.
5846 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5847 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5848 to changed the state as if the insn were scheduled when the new
5849 simulated processor cycle finishes.
5852 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5853 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5854 used to initialize data used by the previous hook.
5857 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5858 This hook controls better choosing an insn from the ready insn queue
5859 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5860 chooses the first insn from the queue. If the hook returns a positive
5861 value, an additional scheduler code tries all permutations of
5862 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5863 subsequent ready insns to choose an insn whose issue will result in
5864 maximal number of issued insns on the same cycle. For the
5865 @acronym{VLIW} processor, the code could actually solve the problem of
5866 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5867 rules of @acronym{VLIW} packing are described in the automaton.
5869 This code also could be used for superscalar @acronym{RISC}
5870 processors. Let us consider a superscalar @acronym{RISC} processor
5871 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5872 @var{B}, some insns can be executed only in pipelines @var{B} or
5873 @var{C}, and one insn can be executed in pipeline @var{B}. The
5874 processor may issue the 1st insn into @var{A} and the 2nd one into
5875 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5876 until the next cycle. If the scheduler issues the 3rd insn the first,
5877 the processor could issue all 3 insns per cycle.
5879 Actually this code demonstrates advantages of the automaton based
5880 pipeline hazard recognizer. We try quickly and easy many insn
5881 schedules to choose the best one.
5883 The default is no multipass scheduling.
5886 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5888 This hook controls what insns from the ready insn queue will be
5889 considered for the multipass insn scheduling. If the hook returns
5890 zero for insn passed as the parameter, the insn will be not chosen to
5893 The default is that any ready insns can be chosen to be issued.
5896 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5898 This hook is called by the insn scheduler before issuing insn passed
5899 as the third parameter on given cycle. If the hook returns nonzero,
5900 the insn is not issued on given processors cycle. Instead of that,
5901 the processor cycle is advanced. If the value passed through the last
5902 parameter is zero, the insn ready queue is not sorted on the new cycle
5903 start as usually. The first parameter passes file for debugging
5904 output. The second one passes the scheduler verbose level of the
5905 debugging output. The forth and the fifth parameter values are
5906 correspondingly processor cycle on which the previous insn has been
5907 issued and the current processor cycle.
5910 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (rtx @var{insn1}, rtx @var{insn2}, rtx @var{dep_link}, int @var{dep_cost}, int @var{distance})
5911 This hook is used to define which dependences are considered costly by
5912 the target, so costly that it is not advisable to schedule the insns that
5913 are involved in the dependence too close to one another. The parameters
5914 to this hook are as follows: The second parameter @var{insn2} is dependent
5915 upon the first parameter @var{insn1}. The dependence between @var{insn1}
5916 and @var{insn2} is represented by the third parameter @var{dep_link}. The
5917 fourth parameter @var{cost} is the cost of the dependence, and the fifth
5918 parameter @var{distance} is the distance in cycles between the two insns.
5919 The hook returns @code{true} if considering the distance between the two
5920 insns the dependence between them is considered costly by the target,
5921 and @code{false} otherwise.
5923 Defining this hook can be useful in multiple-issue out-of-order machines,
5924 where (a) it's practically hopeless to predict the actual data/resource
5925 delays, however: (b) there's a better chance to predict the actual grouping
5926 that will be formed, and (c) correctly emulating the grouping can be very
5927 important. In such targets one may want to allow issuing dependent insns
5928 closer to one another---i.e., closer than the dependence distance; however,
5929 not in cases of "costly dependences", which this hooks allows to define.
5932 Macros in the following table are generated by the program
5933 @file{genattr} and can be useful for writing the hooks.
5935 @defmac MAX_DFA_ISSUE_RATE
5936 The macro definition is generated in the automaton based pipeline
5937 description interface. Its value is calculated from the automaton
5938 based pipeline description and is equal to maximal number of all insns
5939 described in constructions @samp{define_insn_reservation} which can be
5940 issued on the same processor cycle.
5944 @section Dividing the Output into Sections (Texts, Data, @dots{})
5945 @c the above section title is WAY too long. maybe cut the part between
5946 @c the (...)? --mew 10feb93
5948 An object file is divided into sections containing different types of
5949 data. In the most common case, there are three sections: the @dfn{text
5950 section}, which holds instructions and read-only data; the @dfn{data
5951 section}, which holds initialized writable data; and the @dfn{bss
5952 section}, which holds uninitialized data. Some systems have other kinds
5955 The compiler must tell the assembler when to switch sections. These
5956 macros control what commands to output to tell the assembler this. You
5957 can also define additional sections.
5959 @defmac TEXT_SECTION_ASM_OP
5960 A C expression whose value is a string, including spacing, containing the
5961 assembler operation that should precede instructions and read-only data.
5962 Normally @code{"\t.text"} is right.
5965 @defmac HOT_TEXT_SECTION_NAME
5966 If defined, a C string constant for the name of the section containing most
5967 frequently executed functions of the program. If not defined, GCC will provide
5968 a default definition if the target supports named sections.
5971 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5972 If defined, a C string constant for the name of the section containing unlikely
5973 executed functions in the program.
5976 @defmac DATA_SECTION_ASM_OP
5977 A C expression whose value is a string, including spacing, containing the
5978 assembler operation to identify the following data as writable initialized
5979 data. Normally @code{"\t.data"} is right.
5982 @defmac READONLY_DATA_SECTION_ASM_OP
5983 A C expression whose value is a string, including spacing, containing the
5984 assembler operation to identify the following data as read-only initialized
5988 @defmac READONLY_DATA_SECTION
5989 A macro naming a function to call to switch to the proper section for
5990 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5991 if defined, else fall back to @code{text_section}.
5993 The most common definition will be @code{data_section}, if the target
5994 does not have a special read-only data section, and does not put data
5995 in the text section.
5998 @defmac BSS_SECTION_ASM_OP
5999 If defined, a C expression whose value is a string, including spacing,
6000 containing the assembler operation to identify the following data as
6001 uninitialized global data. If not defined, and neither
6002 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
6003 uninitialized global data will be output in the data section if
6004 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6008 @defmac INIT_SECTION_ASM_OP
6009 If defined, a C expression whose value is a string, including spacing,
6010 containing the assembler operation to identify the following data as
6011 initialization code. If not defined, GCC will assume such a section does
6015 @defmac FINI_SECTION_ASM_OP
6016 If defined, a C expression whose value is a string, including spacing,
6017 containing the assembler operation to identify the following data as
6018 finalization code. If not defined, GCC will assume such a section does
6022 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6023 If defined, an ASM statement that switches to a different section
6024 via @var{section_op}, calls @var{function}, and switches back to
6025 the text section. This is used in @file{crtstuff.c} if
6026 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6027 to initialization and finalization functions from the init and fini
6028 sections. By default, this macro uses a simple function call. Some
6029 ports need hand-crafted assembly code to avoid dependencies on
6030 registers initialized in the function prologue or to ensure that
6031 constant pools don't end up too far way in the text section.
6034 @defmac FORCE_CODE_SECTION_ALIGN
6035 If defined, an ASM statement that aligns a code section to some
6036 arbitrary boundary. This is used to force all fragments of the
6037 @code{.init} and @code{.fini} sections to have to same alignment
6038 and thus prevent the linker from having to add any padding.
6043 @defmac EXTRA_SECTIONS
6044 A list of names for sections other than the standard two, which are
6045 @code{in_text} and @code{in_data}. You need not define this macro
6046 on a system with no other sections (that GCC needs to use).
6049 @findex text_section
6050 @findex data_section
6051 @defmac EXTRA_SECTION_FUNCTIONS
6052 One or more functions to be defined in @file{varasm.c}. These
6053 functions should do jobs analogous to those of @code{text_section} and
6054 @code{data_section}, for your additional sections. Do not define this
6055 macro if you do not define @code{EXTRA_SECTIONS}.
6058 @defmac JUMP_TABLES_IN_TEXT_SECTION
6059 Define this macro to be an expression with a nonzero value if jump
6060 tables (for @code{tablejump} insns) should be output in the text
6061 section, along with the assembler instructions. Otherwise, the
6062 readonly data section is used.
6064 This macro is irrelevant if there is no separate readonly data section.
6067 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6068 Switches to the appropriate section for output of @var{exp}. You can
6069 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6070 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6071 requires link-time relocations. Bit 0 is set when variable contains
6072 local relocations only, while bit 1 is set for global relocations.
6073 Select the section by calling @code{data_section} or one of the
6074 alternatives for other sections. @var{align} is the constant alignment
6077 The default version of this function takes care of putting read-only
6078 variables in @code{readonly_data_section}.
6080 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6083 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6084 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6085 for @code{FUNCTION_DECL}s as well as for variables and constants.
6087 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6088 function has been determined to be likely to be called, and nonzero if
6089 it is unlikely to be called.
6092 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
6093 Build up a unique section name, expressed as a @code{STRING_CST} node,
6094 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
6095 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
6096 the initial value of @var{exp} requires link-time relocations.
6098 The default version of this function appends the symbol name to the
6099 ELF section name that would normally be used for the symbol. For
6100 example, the function @code{foo} would be placed in @code{.text.foo}.
6101 Whatever the actual target object format, this is often good enough.
6104 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
6105 Switches to a readonly data section associated with
6106 @samp{DECL_SECTION_NAME (@var{decl})}.
6107 The default version of this function switches to @code{.gnu.linkonce.r.name}
6108 section if function's section is @code{.gnu.linkonce.t.name}, to
6109 @code{.rodata.name} if function is in @code{.text.name} section
6110 and otherwise switches to the normal readonly data section.
6113 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
6114 Switches to the appropriate section for output of constant pool entry
6115 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
6116 constant in RTL@. The argument @var{mode} is redundant except in the
6117 case of a @code{const_int} rtx. Select the section by calling
6118 @code{readonly_data_section} or one of the alternatives for other
6119 sections. @var{align} is the constant alignment in bits.
6121 The default version of this function takes care of putting symbolic
6122 constants in @code{flag_pic} mode in @code{data_section} and everything
6123 else in @code{readonly_data_section}.
6126 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
6127 Define this hook if references to a symbol or a constant must be
6128 treated differently depending on something about the variable or
6129 function named by the symbol (such as what section it is in).
6131 The hook is executed immediately after rtl has been created for
6132 @var{decl}, which may be a variable or function declaration or
6133 an entry in the constant pool. In either case, @var{rtl} is the
6134 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6135 in this hook; that field may not have been initialized yet.
6137 In the case of a constant, it is safe to assume that the rtl is
6138 a @code{mem} whose address is a @code{symbol_ref}. Most decls
6139 will also have this form, but that is not guaranteed. Global
6140 register variables, for instance, will have a @code{reg} for their
6141 rtl. (Normally the right thing to do with such unusual rtl is
6144 The @var{new_decl_p} argument will be true if this is the first time
6145 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
6146 be false for subsequent invocations, which will happen for duplicate
6147 declarations. Whether or not anything must be done for the duplicate
6148 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6149 @var{new_decl_p} is always true when the hook is called for a constant.
6151 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6152 The usual thing for this hook to do is to record flags in the
6153 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
6154 Historically, the name string was modified if it was necessary to
6155 encode more than one bit of information, but this practice is now
6156 discouraged; use @code{SYMBOL_REF_FLAGS}.
6158 The default definition of this hook, @code{default_encode_section_info}
6159 in @file{varasm.c}, sets a number of commonly-useful bits in
6160 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6161 before overriding it.
6164 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6165 Decode @var{name} and return the real name part, sans
6166 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6170 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6171 Returns true if @var{exp} should be placed into a ``small data'' section.
6172 The default version of this hook always returns false.
6175 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6176 Contains the value true if the target places read-only
6177 ``small data'' into a separate section. The default value is false.
6180 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6181 Returns true if @var{exp} names an object for which name resolution
6182 rules must resolve to the current ``module'' (dynamic shared library
6183 or executable image).
6185 The default version of this hook implements the name resolution rules
6186 for ELF, which has a looser model of global name binding than other
6187 currently supported object file formats.
6190 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6191 Contains the value true if the target supports thread-local storage.
6192 The default value is false.
6197 @section Position Independent Code
6198 @cindex position independent code
6201 This section describes macros that help implement generation of position
6202 independent code. Simply defining these macros is not enough to
6203 generate valid PIC; you must also add support to the macros
6204 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6205 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6206 @samp{movsi} to do something appropriate when the source operand
6207 contains a symbolic address. You may also need to alter the handling of
6208 switch statements so that they use relative addresses.
6209 @c i rearranged the order of the macros above to try to force one of
6210 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6212 @defmac PIC_OFFSET_TABLE_REGNUM
6213 The register number of the register used to address a table of static
6214 data addresses in memory. In some cases this register is defined by a
6215 processor's ``application binary interface'' (ABI)@. When this macro
6216 is defined, RTL is generated for this register once, as with the stack
6217 pointer and frame pointer registers. If this macro is not defined, it
6218 is up to the machine-dependent files to allocate such a register (if
6219 necessary). Note that this register must be fixed when in use (e.g.@:
6220 when @code{flag_pic} is true).
6223 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6224 Define this macro if the register defined by
6225 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6226 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6229 @defmac FINALIZE_PIC
6230 By generating position-independent code, when two different programs (A
6231 and B) share a common library (libC.a), the text of the library can be
6232 shared whether or not the library is linked at the same address for both
6233 programs. In some of these environments, position-independent code
6234 requires not only the use of different addressing modes, but also
6235 special code to enable the use of these addressing modes.
6237 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
6238 codes once the function is being compiled into assembly code, but not
6239 before. (It is not done before, because in the case of compiling an
6240 inline function, it would lead to multiple PIC prologues being
6241 included in functions which used inline functions and were compiled to
6245 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6246 A C expression that is nonzero if @var{x} is a legitimate immediate
6247 operand on the target machine when generating position independent code.
6248 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6249 check this. You can also assume @var{flag_pic} is true, so you need not
6250 check it either. You need not define this macro if all constants
6251 (including @code{SYMBOL_REF}) can be immediate operands when generating
6252 position independent code.
6255 @node Assembler Format
6256 @section Defining the Output Assembler Language
6258 This section describes macros whose principal purpose is to describe how
6259 to write instructions in assembler language---rather than what the
6263 * File Framework:: Structural information for the assembler file.
6264 * Data Output:: Output of constants (numbers, strings, addresses).
6265 * Uninitialized Data:: Output of uninitialized variables.
6266 * Label Output:: Output and generation of labels.
6267 * Initialization:: General principles of initialization
6268 and termination routines.
6269 * Macros for Initialization::
6270 Specific macros that control the handling of
6271 initialization and termination routines.
6272 * Instruction Output:: Output of actual instructions.
6273 * Dispatch Tables:: Output of jump tables.
6274 * Exception Region Output:: Output of exception region code.
6275 * Alignment Output:: Pseudo ops for alignment and skipping data.
6278 @node File Framework
6279 @subsection The Overall Framework of an Assembler File
6280 @cindex assembler format
6281 @cindex output of assembler code
6283 @c prevent bad page break with this line
6284 This describes the overall framework of an assembly file.
6286 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6287 @findex default_file_start
6288 Output to @code{asm_out_file} any text which the assembler expects to
6289 find at the beginning of a file. The default behavior is controlled
6290 by two flags, documented below. Unless your target's assembler is
6291 quite unusual, if you override the default, you should call
6292 @code{default_file_start} at some point in your target hook. This
6293 lets other target files rely on these variables.
6296 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6297 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6298 printed as the very first line in the assembly file, unless
6299 @option{-fverbose-asm} is in effect. (If that macro has been defined
6300 to the empty string, this variable has no effect.) With the normal
6301 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6302 assembler that it need not bother stripping comments or extra
6303 whitespace from its input. This allows it to work a bit faster.
6305 The default is false. You should not set it to true unless you have
6306 verified that your port does not generate any extra whitespace or
6307 comments that will cause GAS to issue errors in NO_APP mode.
6310 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6311 If this flag is true, @code{output_file_directive} will be called
6312 for the primary source file, immediately after printing
6313 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6314 this to be done. The default is false.
6317 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6318 Output to @code{asm_out_file} any text which the assembler expects
6319 to find at the end of a file. The default is to output nothing.
6322 @deftypefun void file_end_indicate_exec_stack ()
6323 Some systems use a common convention, the @samp{.note.GNU-stack}
6324 special section, to indicate whether or not an object file relies on
6325 the stack being executable. If your system uses this convention, you
6326 should define @code{TARGET_ASM_FILE_END} to this function. If you
6327 need to do other things in that hook, have your hook function call
6331 @defmac ASM_COMMENT_START
6332 A C string constant describing how to begin a comment in the target
6333 assembler language. The compiler assumes that the comment will end at
6334 the end of the line.
6338 A C string constant for text to be output before each @code{asm}
6339 statement or group of consecutive ones. Normally this is
6340 @code{"#APP"}, which is a comment that has no effect on most
6341 assemblers but tells the GNU assembler that it must check the lines
6342 that follow for all valid assembler constructs.
6346 A C string constant for text to be output after each @code{asm}
6347 statement or group of consecutive ones. Normally this is
6348 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6349 time-saving assumptions that are valid for ordinary compiler output.
6352 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6353 A C statement to output COFF information or DWARF debugging information
6354 which indicates that filename @var{name} is the current source file to
6355 the stdio stream @var{stream}.
6357 This macro need not be defined if the standard form of output
6358 for the file format in use is appropriate.
6361 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6362 A C statement to output the string @var{string} to the stdio stream
6363 @var{stream}. If you do not call the function @code{output_quoted_string}
6364 in your config files, GCC will only call it to output filenames to
6365 the assembler source. So you can use it to canonicalize the format
6366 of the filename using this macro.
6369 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6370 A C statement to output something to the assembler file to handle a
6371 @samp{#ident} directive containing the text @var{string}. If this
6372 macro is not defined, nothing is output for a @samp{#ident} directive.
6375 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6376 Output assembly directives to switch to section @var{name}. The section
6377 should have attributes as specified by @var{flags}, which is a bit mask
6378 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6379 is nonzero, it contains an alignment in bytes to be used for the section,
6380 otherwise some target default should be used. Only targets that must
6381 specify an alignment within the section directive need pay attention to
6382 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6385 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6386 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6389 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6390 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6391 based on a variable or function decl, a section name, and whether or not the
6392 declaration's initializer may contain runtime relocations. @var{decl} may be
6393 null, in which case read-write data should be assumed.
6395 The default version if this function handles choosing code vs data,
6396 read-only vs read-write data, and @code{flag_pic}. You should only
6397 need to override this if your target has special flags that might be
6398 set via @code{__attribute__}.
6403 @subsection Output of Data
6406 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6407 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6408 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6409 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6410 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6411 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6412 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6413 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6414 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6415 These hooks specify assembly directives for creating certain kinds
6416 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6417 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6418 aligned two-byte object, and so on. Any of the hooks may be
6419 @code{NULL}, indicating that no suitable directive is available.
6421 The compiler will print these strings at the start of a new line,
6422 followed immediately by the object's initial value. In most cases,
6423 the string should contain a tab, a pseudo-op, and then another tab.
6426 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6427 The @code{assemble_integer} function uses this hook to output an
6428 integer object. @var{x} is the object's value, @var{size} is its size
6429 in bytes and @var{aligned_p} indicates whether it is aligned. The
6430 function should return @code{true} if it was able to output the
6431 object. If it returns false, @code{assemble_integer} will try to
6432 split the object into smaller parts.
6434 The default implementation of this hook will use the
6435 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6436 when the relevant string is @code{NULL}.
6439 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6440 A C statement to recognize @var{rtx} patterns that
6441 @code{output_addr_const} can't deal with, and output assembly code to
6442 @var{stream} corresponding to the pattern @var{x}. This may be used to
6443 allow machine-dependent @code{UNSPEC}s to appear within constants.
6445 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6446 @code{goto fail}, so that a standard error message is printed. If it
6447 prints an error message itself, by calling, for example,
6448 @code{output_operand_lossage}, it may just complete normally.
6451 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6452 A C statement to output to the stdio stream @var{stream} an assembler
6453 instruction to assemble a string constant containing the @var{len}
6454 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6455 @code{char *} and @var{len} a C expression of type @code{int}.
6457 If the assembler has a @code{.ascii} pseudo-op as found in the
6458 Berkeley Unix assembler, do not define the macro
6459 @code{ASM_OUTPUT_ASCII}.
6462 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6463 A C statement to output word @var{n} of a function descriptor for
6464 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6465 is defined, and is otherwise unused.
6468 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6469 You may define this macro as a C expression. You should define the
6470 expression to have a nonzero value if GCC should output the constant
6471 pool for a function before the code for the function, or a zero value if
6472 GCC should output the constant pool after the function. If you do
6473 not define this macro, the usual case, GCC will output the constant
6474 pool before the function.
6477 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6478 A C statement to output assembler commands to define the start of the
6479 constant pool for a function. @var{funname} is a string giving
6480 the name of the function. Should the return type of the function
6481 be required, it can be obtained via @var{fundecl}. @var{size}
6482 is the size, in bytes, of the constant pool that will be written
6483 immediately after this call.
6485 If no constant-pool prefix is required, the usual case, this macro need
6489 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6490 A C statement (with or without semicolon) to output a constant in the
6491 constant pool, if it needs special treatment. (This macro need not do
6492 anything for RTL expressions that can be output normally.)
6494 The argument @var{file} is the standard I/O stream to output the
6495 assembler code on. @var{x} is the RTL expression for the constant to
6496 output, and @var{mode} is the machine mode (in case @var{x} is a
6497 @samp{const_int}). @var{align} is the required alignment for the value
6498 @var{x}; you should output an assembler directive to force this much
6501 The argument @var{labelno} is a number to use in an internal label for
6502 the address of this pool entry. The definition of this macro is
6503 responsible for outputting the label definition at the proper place.
6504 Here is how to do this:
6507 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6510 When you output a pool entry specially, you should end with a
6511 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6512 entry from being output a second time in the usual manner.
6514 You need not define this macro if it would do nothing.
6517 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6518 A C statement to output assembler commands to at the end of the constant
6519 pool for a function. @var{funname} is a string giving the name of the
6520 function. Should the return type of the function be required, you can
6521 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6522 constant pool that GCC wrote immediately before this call.
6524 If no constant-pool epilogue is required, the usual case, you need not
6528 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6529 Define this macro as a C expression which is nonzero if @var{C} is
6530 used as a logical line separator by the assembler.
6532 If you do not define this macro, the default is that only
6533 the character @samp{;} is treated as a logical line separator.
6536 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6537 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6538 These target hooks are C string constants, describing the syntax in the
6539 assembler for grouping arithmetic expressions. If not overridden, they
6540 default to normal parentheses, which is correct for most assemblers.
6543 These macros are provided by @file{real.h} for writing the definitions
6544 of @code{ASM_OUTPUT_DOUBLE} and the like:
6546 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6547 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6548 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6549 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6550 floating point representation, and store its bit pattern in the variable
6551 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6552 be a simple @code{long int}. For the others, it should be an array of
6553 @code{long int}. The number of elements in this array is determined by
6554 the size of the desired target floating point data type: 32 bits of it
6555 go in each @code{long int} array element. Each array element holds 32
6556 bits of the result, even if @code{long int} is wider than 32 bits on the
6559 The array element values are designed so that you can print them out
6560 using @code{fprintf} in the order they should appear in the target
6564 @node Uninitialized Data
6565 @subsection Output of Uninitialized Variables
6567 Each of the macros in this section is used to do the whole job of
6568 outputting a single uninitialized variable.
6570 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6571 A C statement (sans semicolon) to output to the stdio stream
6572 @var{stream} the assembler definition of a common-label named
6573 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6574 is the size rounded up to whatever alignment the caller wants.
6576 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6577 output the name itself; before and after that, output the additional
6578 assembler syntax for defining the name, and a newline.
6580 This macro controls how the assembler definitions of uninitialized
6581 common global variables are output.
6584 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6585 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6586 separate, explicit argument. If you define this macro, it is used in
6587 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6588 handling the required alignment of the variable. The alignment is specified
6589 as the number of bits.
6592 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6593 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6594 variable to be output, if there is one, or @code{NULL_TREE} if there
6595 is no corresponding variable. If you define this macro, GCC will use it
6596 in place of both @code{ASM_OUTPUT_COMMON} and
6597 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6598 the variable's decl in order to chose what to output.
6601 @defmac ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6602 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6603 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6607 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6608 A C statement (sans semicolon) to output to the stdio stream
6609 @var{stream} the assembler definition of uninitialized global @var{decl} named
6610 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6611 is the size rounded up to whatever alignment the caller wants.
6613 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6614 defining this macro. If unable, use the expression
6615 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6616 before and after that, output the additional assembler syntax for defining
6617 the name, and a newline.
6619 This macro controls how the assembler definitions of uninitialized global
6620 variables are output. This macro exists to properly support languages like
6621 C++ which do not have @code{common} data. However, this macro currently
6622 is not defined for all targets. If this macro and
6623 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6624 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6625 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6628 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6629 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6630 separate, explicit argument. If you define this macro, it is used in
6631 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6632 handling the required alignment of the variable. The alignment is specified
6633 as the number of bits.
6635 Try to use function @code{asm_output_aligned_bss} defined in file
6636 @file{varasm.c} when defining this macro.
6639 @defmac ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6640 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6641 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6645 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6646 A C statement (sans semicolon) to output to the stdio stream
6647 @var{stream} the assembler definition of a local-common-label named
6648 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6649 is the size rounded up to whatever alignment the caller wants.
6651 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6652 output the name itself; before and after that, output the additional
6653 assembler syntax for defining the name, and a newline.
6655 This macro controls how the assembler definitions of uninitialized
6656 static variables are output.
6659 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6660 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6661 separate, explicit argument. If you define this macro, it is used in
6662 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6663 handling the required alignment of the variable. The alignment is specified
6664 as the number of bits.
6667 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6668 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6669 variable to be output, if there is one, or @code{NULL_TREE} if there
6670 is no corresponding variable. If you define this macro, GCC will use it
6671 in place of both @code{ASM_OUTPUT_DECL} and
6672 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6673 the variable's decl in order to chose what to output.
6676 @defmac ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6677 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6678 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6683 @subsection Output and Generation of Labels
6685 @c prevent bad page break with this line
6686 This is about outputting labels.
6688 @findex assemble_name
6689 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6690 A C statement (sans semicolon) to output to the stdio stream
6691 @var{stream} the assembler definition of a label named @var{name}.
6692 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6693 output the name itself; before and after that, output the additional
6694 assembler syntax for defining the name, and a newline. A default
6695 definition of this macro is provided which is correct for most systems.
6698 @findex assemble_name_raw
6699 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
6700 Identical to @code{ASM_OUTPUT_lABEL}, except that @var{name} is known
6701 to refer to a compiler-generated label. The default definition uses
6702 @code{assemble_name_raw}, which is like @code{assemble_name} except
6703 that it is more efficient.
6707 A C string containing the appropriate assembler directive to specify the
6708 size of a symbol, without any arguments. On systems that use ELF, the
6709 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6710 systems, the default is not to define this macro.
6712 Define this macro only if it is correct to use the default definitions
6713 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6714 for your system. If you need your own custom definitions of those
6715 macros, or if you do not need explicit symbol sizes at all, do not
6719 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6720 A C statement (sans semicolon) to output to the stdio stream
6721 @var{stream} a directive telling the assembler that the size of the
6722 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6723 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6727 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6728 A C statement (sans semicolon) to output to the stdio stream
6729 @var{stream} a directive telling the assembler to calculate the size of
6730 the symbol @var{name} by subtracting its address from the current
6733 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6734 provided. The default assumes that the assembler recognizes a special
6735 @samp{.} symbol as referring to the current address, and can calculate
6736 the difference between this and another symbol. If your assembler does
6737 not recognize @samp{.} or cannot do calculations with it, you will need
6738 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6742 A C string containing the appropriate assembler directive to specify the
6743 type of a symbol, without any arguments. On systems that use ELF, the
6744 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6745 systems, the default is not to define this macro.
6747 Define this macro only if it is correct to use the default definition of
6748 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6749 custom definition of this macro, or if you do not need explicit symbol
6750 types at all, do not define this macro.
6753 @defmac TYPE_OPERAND_FMT
6754 A C string which specifies (using @code{printf} syntax) the format of
6755 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6756 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6757 the default is not to define this macro.
6759 Define this macro only if it is correct to use the default definition of
6760 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6761 custom definition of this macro, or if you do not need explicit symbol
6762 types at all, do not define this macro.
6765 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6766 A C statement (sans semicolon) to output to the stdio stream
6767 @var{stream} a directive telling the assembler that the type of the
6768 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6769 that string is always either @samp{"function"} or @samp{"object"}, but
6770 you should not count on this.
6772 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6773 definition of this macro is provided.
6776 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6777 A C statement (sans semicolon) to output to the stdio stream
6778 @var{stream} any text necessary for declaring the name @var{name} of a
6779 function which is being defined. This macro is responsible for
6780 outputting the label definition (perhaps using
6781 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6782 @code{FUNCTION_DECL} tree node representing the function.
6784 If this macro is not defined, then the function name is defined in the
6785 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6787 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6791 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6792 A C statement (sans semicolon) to output to the stdio stream
6793 @var{stream} any text necessary for declaring the size of a function
6794 which is being defined. The argument @var{name} is the name of the
6795 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6796 representing the function.
6798 If this macro is not defined, then the function size is not defined.
6800 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6804 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6805 A C statement (sans semicolon) to output to the stdio stream
6806 @var{stream} any text necessary for declaring the name @var{name} of an
6807 initialized variable which is being defined. This macro must output the
6808 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6809 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6811 If this macro is not defined, then the variable name is defined in the
6812 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6814 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6815 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6818 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
6819 A C statement (sans semicolon) to output to the stdio stream
6820 @var{stream} any text necessary for declaring the name @var{name} of a
6821 constant which is being defined. This macro is responsible for
6822 outputting the label definition (perhaps using
6823 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
6824 value of the constant, and @var{size} is the size of the constant
6825 in bytes. @var{name} will be an internal label.
6827 If this macro is not defined, then the @var{name} is defined in the
6828 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6830 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6834 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6835 A C statement (sans semicolon) to output to the stdio stream
6836 @var{stream} any text necessary for claiming a register @var{regno}
6837 for a global variable @var{decl} with name @var{name}.
6839 If you don't define this macro, that is equivalent to defining it to do
6843 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6844 A C statement (sans semicolon) to finish up declaring a variable name
6845 once the compiler has processed its initializer fully and thus has had a
6846 chance to determine the size of an array when controlled by an
6847 initializer. This is used on systems where it's necessary to declare
6848 something about the size of the object.
6850 If you don't define this macro, that is equivalent to defining it to do
6853 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6854 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6857 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6858 This target hook is a function to output to the stdio stream
6859 @var{stream} some commands that will make the label @var{name} global;
6860 that is, available for reference from other files.
6862 The default implementation relies on a proper definition of
6863 @code{GLOBAL_ASM_OP}.
6866 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6867 A C statement (sans semicolon) to output to the stdio stream
6868 @var{stream} some commands that will make the label @var{name} weak;
6869 that is, available for reference from other files but only used if
6870 no other definition is available. Use the expression
6871 @code{assemble_name (@var{stream}, @var{name})} to output the name
6872 itself; before and after that, output the additional assembler syntax
6873 for making that name weak, and a newline.
6875 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6876 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6880 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6881 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6882 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6883 or variable decl. If @var{value} is not @code{NULL}, this C statement
6884 should output to the stdio stream @var{stream} assembler code which
6885 defines (equates) the weak symbol @var{name} to have the value
6886 @var{value}. If @var{value} is @code{NULL}, it should output commands
6887 to make @var{name} weak.
6890 @defmac SUPPORTS_WEAK
6891 A C expression which evaluates to true if the target supports weak symbols.
6893 If you don't define this macro, @file{defaults.h} provides a default
6894 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6895 is defined, the default definition is @samp{1}; otherwise, it is
6896 @samp{0}. Define this macro if you want to control weak symbol support
6897 with a compiler flag such as @option{-melf}.
6900 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
6901 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6902 public symbol such that extra copies in multiple translation units will
6903 be discarded by the linker. Define this macro if your object file
6904 format provides support for this concept, such as the @samp{COMDAT}
6905 section flags in the Microsoft Windows PE/COFF format, and this support
6906 requires changes to @var{decl}, such as putting it in a separate section.
6909 @defmac SUPPORTS_ONE_ONLY
6910 A C expression which evaluates to true if the target supports one-only
6913 If you don't define this macro, @file{varasm.c} provides a default
6914 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6915 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6916 you want to control one-only symbol support with a compiler flag, or if
6917 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6918 be emitted as one-only.
6921 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6922 This target hook is a function to output to @var{asm_out_file} some
6923 commands that will make the symbol(s) associated with @var{decl} have
6924 hidden, protected or internal visibility as specified by @var{visibility}.
6927 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
6928 A C expression that evaluates to true if the target's linker expects
6929 that weak symbols do not appear in a static archive's table of contents.
6930 The default is @code{0}.
6932 Leaving weak symbols out of an archive's table of contents means that,
6933 if a symbol will only have a definition in one translation unit and
6934 will have undefined references from other translation units, that
6935 symbol should not be weak. Defining this macro to be nonzero will
6936 thus have the effect that certain symbols that would normally be weak
6937 (explicit template instantiations, and vtables for polymorphic classes
6938 with noninline key methods) will instead be nonweak.
6940 The C++ ABI requires this macro to be zero. Define this macro for
6941 targets where full C++ ABI compliance is impossible and where linker
6942 restrictions require weak symbols to be left out of a static archive's
6946 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6947 A C statement (sans semicolon) to output to the stdio stream
6948 @var{stream} any text necessary for declaring the name of an external
6949 symbol named @var{name} which is referenced in this compilation but
6950 not defined. The value of @var{decl} is the tree node for the
6953 This macro need not be defined if it does not need to output anything.
6954 The GNU assembler and most Unix assemblers don't require anything.
6957 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
6958 This target hook is a function to output to @var{asm_out_file} an assembler
6959 pseudo-op to declare a library function name external. The name of the
6960 library function is given by @var{symref}, which is a @code{symbol_ref}.
6963 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl})
6964 This target hook is a function to output to @var{asm_out_file} an assembler
6965 directive to annotate used symbol. Darwin target use .no_dead_code_strip
6969 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6970 A C statement (sans semicolon) to output to the stdio stream
6971 @var{stream} a reference in assembler syntax to a label named
6972 @var{name}. This should add @samp{_} to the front of the name, if that
6973 is customary on your operating system, as it is in most Berkeley Unix
6974 systems. This macro is used in @code{assemble_name}.
6977 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6978 A C statement (sans semicolon) to output a reference to
6979 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6980 will be used to output the name of the symbol. This macro may be used
6981 to modify the way a symbol is referenced depending on information
6982 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6985 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6986 A C statement (sans semicolon) to output a reference to @var{buf}, the
6987 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6988 @code{assemble_name} will be used to output the name of the symbol.
6989 This macro is not used by @code{output_asm_label}, or the @code{%l}
6990 specifier that calls it; the intention is that this macro should be set
6991 when it is necessary to output a label differently when its address is
6995 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6996 A function to output to the stdio stream @var{stream} a label whose
6997 name is made from the string @var{prefix} and the number @var{labelno}.
6999 It is absolutely essential that these labels be distinct from the labels
7000 used for user-level functions and variables. Otherwise, certain programs
7001 will have name conflicts with internal labels.
7003 It is desirable to exclude internal labels from the symbol table of the
7004 object file. Most assemblers have a naming convention for labels that
7005 should be excluded; on many systems, the letter @samp{L} at the
7006 beginning of a label has this effect. You should find out what
7007 convention your system uses, and follow it.
7009 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
7012 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
7013 A C statement to output to the stdio stream @var{stream} a debug info
7014 label whose name is made from the string @var{prefix} and the number
7015 @var{num}. This is useful for VLIW targets, where debug info labels
7016 may need to be treated differently than branch target labels. On some
7017 systems, branch target labels must be at the beginning of instruction
7018 bundles, but debug info labels can occur in the middle of instruction
7021 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
7025 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
7026 A C statement to store into the string @var{string} a label whose name
7027 is made from the string @var{prefix} and the number @var{num}.
7029 This string, when output subsequently by @code{assemble_name}, should
7030 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
7031 with the same @var{prefix} and @var{num}.
7033 If the string begins with @samp{*}, then @code{assemble_name} will
7034 output the rest of the string unchanged. It is often convenient for
7035 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
7036 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
7037 to output the string, and may change it. (Of course,
7038 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
7039 you should know what it does on your machine.)
7042 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
7043 A C expression to assign to @var{outvar} (which is a variable of type
7044 @code{char *}) a newly allocated string made from the string
7045 @var{name} and the number @var{number}, with some suitable punctuation
7046 added. Use @code{alloca} to get space for the string.
7048 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
7049 produce an assembler label for an internal static variable whose name is
7050 @var{name}. Therefore, the string must be such as to result in valid
7051 assembler code. The argument @var{number} is different each time this
7052 macro is executed; it prevents conflicts between similarly-named
7053 internal static variables in different scopes.
7055 Ideally this string should not be a valid C identifier, to prevent any
7056 conflict with the user's own symbols. Most assemblers allow periods
7057 or percent signs in assembler symbols; putting at least one of these
7058 between the name and the number will suffice.
7060 If this macro is not defined, a default definition will be provided
7061 which is correct for most systems.
7064 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
7065 A C statement to output to the stdio stream @var{stream} assembler code
7066 which defines (equates) the symbol @var{name} to have the value @var{value}.
7069 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7070 correct for most systems.
7073 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
7074 A C statement to output to the stdio stream @var{stream} assembler code
7075 which defines (equates) the symbol whose tree node is @var{decl_of_name}
7076 to have the value of the tree node @var{decl_of_value}. This macro will
7077 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
7078 the tree nodes are available.
7081 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7082 correct for most systems.
7085 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
7086 A C statement that evaluates to true if the assembler code which defines
7087 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
7088 of the tree node @var{decl_of_value} should be emitted near the end of the
7089 current compilation unit. The default is to not defer output of defines.
7090 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
7091 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
7094 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
7095 A C statement to output to the stdio stream @var{stream} assembler code
7096 which defines (equates) the weak symbol @var{name} to have the value
7097 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
7098 an undefined weak symbol.
7100 Define this macro if the target only supports weak aliases; define
7101 @code{ASM_OUTPUT_DEF} instead if possible.
7104 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
7105 Define this macro to override the default assembler names used for
7106 Objective-C methods.
7108 The default name is a unique method number followed by the name of the
7109 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
7110 the category is also included in the assembler name (e.g.@:
7113 These names are safe on most systems, but make debugging difficult since
7114 the method's selector is not present in the name. Therefore, particular
7115 systems define other ways of computing names.
7117 @var{buf} is an expression of type @code{char *} which gives you a
7118 buffer in which to store the name; its length is as long as
7119 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
7120 50 characters extra.
7122 The argument @var{is_inst} specifies whether the method is an instance
7123 method or a class method; @var{class_name} is the name of the class;
7124 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
7125 in a category); and @var{sel_name} is the name of the selector.
7127 On systems where the assembler can handle quoted names, you can use this
7128 macro to provide more human-readable names.
7131 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
7132 A C statement (sans semicolon) to output to the stdio stream
7133 @var{stream} commands to declare that the label @var{name} is an
7134 Objective-C class reference. This is only needed for targets whose
7135 linkers have special support for NeXT-style runtimes.
7138 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
7139 A C statement (sans semicolon) to output to the stdio stream
7140 @var{stream} commands to declare that the label @var{name} is an
7141 unresolved Objective-C class reference. This is only needed for targets
7142 whose linkers have special support for NeXT-style runtimes.
7145 @node Initialization
7146 @subsection How Initialization Functions Are Handled
7147 @cindex initialization routines
7148 @cindex termination routines
7149 @cindex constructors, output of
7150 @cindex destructors, output of
7152 The compiled code for certain languages includes @dfn{constructors}
7153 (also called @dfn{initialization routines})---functions to initialize
7154 data in the program when the program is started. These functions need
7155 to be called before the program is ``started''---that is to say, before
7156 @code{main} is called.
7158 Compiling some languages generates @dfn{destructors} (also called
7159 @dfn{termination routines}) that should be called when the program
7162 To make the initialization and termination functions work, the compiler
7163 must output something in the assembler code to cause those functions to
7164 be called at the appropriate time. When you port the compiler to a new
7165 system, you need to specify how to do this.
7167 There are two major ways that GCC currently supports the execution of
7168 initialization and termination functions. Each way has two variants.
7169 Much of the structure is common to all four variations.
7171 @findex __CTOR_LIST__
7172 @findex __DTOR_LIST__
7173 The linker must build two lists of these functions---a list of
7174 initialization functions, called @code{__CTOR_LIST__}, and a list of
7175 termination functions, called @code{__DTOR_LIST__}.
7177 Each list always begins with an ignored function pointer (which may hold
7178 0, @minus{}1, or a count of the function pointers after it, depending on
7179 the environment). This is followed by a series of zero or more function
7180 pointers to constructors (or destructors), followed by a function
7181 pointer containing zero.
7183 Depending on the operating system and its executable file format, either
7184 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7185 time and exit time. Constructors are called in reverse order of the
7186 list; destructors in forward order.
7188 The best way to handle static constructors works only for object file
7189 formats which provide arbitrarily-named sections. A section is set
7190 aside for a list of constructors, and another for a list of destructors.
7191 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7192 object file that defines an initialization function also puts a word in
7193 the constructor section to point to that function. The linker
7194 accumulates all these words into one contiguous @samp{.ctors} section.
7195 Termination functions are handled similarly.
7197 This method will be chosen as the default by @file{target-def.h} if
7198 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7199 support arbitrary sections, but does support special designated
7200 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7201 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7203 When arbitrary sections are available, there are two variants, depending
7204 upon how the code in @file{crtstuff.c} is called. On systems that
7205 support a @dfn{.init} section which is executed at program startup,
7206 parts of @file{crtstuff.c} are compiled into that section. The
7207 program is linked by the @command{gcc} driver like this:
7210 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7213 The prologue of a function (@code{__init}) appears in the @code{.init}
7214 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7215 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7216 files are provided by the operating system or by the GNU C library, but
7217 are provided by GCC for a few targets.
7219 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7220 compiled from @file{crtstuff.c}. They contain, among other things, code
7221 fragments within the @code{.init} and @code{.fini} sections that branch
7222 to routines in the @code{.text} section. The linker will pull all parts
7223 of a section together, which results in a complete @code{__init} function
7224 that invokes the routines we need at startup.
7226 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7229 If no init section is available, when GCC compiles any function called
7230 @code{main} (or more accurately, any function designated as a program
7231 entry point by the language front end calling @code{expand_main_function}),
7232 it inserts a procedure call to @code{__main} as the first executable code
7233 after the function prologue. The @code{__main} function is defined
7234 in @file{libgcc2.c} and runs the global constructors.
7236 In file formats that don't support arbitrary sections, there are again
7237 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7238 and an `a.out' format must be used. In this case,
7239 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7240 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7241 and with the address of the void function containing the initialization
7242 code as its value. The GNU linker recognizes this as a request to add
7243 the value to a @dfn{set}; the values are accumulated, and are eventually
7244 placed in the executable as a vector in the format described above, with
7245 a leading (ignored) count and a trailing zero element.
7246 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7247 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7248 the compilation of @code{main} to call @code{__main} as above, starting
7249 the initialization process.
7251 The last variant uses neither arbitrary sections nor the GNU linker.
7252 This is preferable when you want to do dynamic linking and when using
7253 file formats which the GNU linker does not support, such as `ECOFF'@. In
7254 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7255 termination functions are recognized simply by their names. This requires
7256 an extra program in the linkage step, called @command{collect2}. This program
7257 pretends to be the linker, for use with GCC; it does its job by running
7258 the ordinary linker, but also arranges to include the vectors of
7259 initialization and termination functions. These functions are called
7260 via @code{__main} as described above. In order to use this method,
7261 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7264 The following section describes the specific macros that control and
7265 customize the handling of initialization and termination functions.
7268 @node Macros for Initialization
7269 @subsection Macros Controlling Initialization Routines
7271 Here are the macros that control how the compiler handles initialization
7272 and termination functions:
7274 @defmac INIT_SECTION_ASM_OP
7275 If defined, a C string constant, including spacing, for the assembler
7276 operation to identify the following data as initialization code. If not
7277 defined, GCC will assume such a section does not exist. When you are
7278 using special sections for initialization and termination functions, this
7279 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7280 run the initialization functions.
7283 @defmac HAS_INIT_SECTION
7284 If defined, @code{main} will not call @code{__main} as described above.
7285 This macro should be defined for systems that control start-up code
7286 on a symbol-by-symbol basis, such as OSF/1, and should not
7287 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7290 @defmac LD_INIT_SWITCH
7291 If defined, a C string constant for a switch that tells the linker that
7292 the following symbol is an initialization routine.
7295 @defmac LD_FINI_SWITCH
7296 If defined, a C string constant for a switch that tells the linker that
7297 the following symbol is a finalization routine.
7300 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7301 If defined, a C statement that will write a function that can be
7302 automatically called when a shared library is loaded. The function
7303 should call @var{func}, which takes no arguments. If not defined, and
7304 the object format requires an explicit initialization function, then a
7305 function called @code{_GLOBAL__DI} will be generated.
7307 This function and the following one are used by collect2 when linking a
7308 shared library that needs constructors or destructors, or has DWARF2
7309 exception tables embedded in the code.
7312 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7313 If defined, a C statement that will write a function that can be
7314 automatically called when a shared library is unloaded. The function
7315 should call @var{func}, which takes no arguments. If not defined, and
7316 the object format requires an explicit finalization function, then a
7317 function called @code{_GLOBAL__DD} will be generated.
7320 @defmac INVOKE__main
7321 If defined, @code{main} will call @code{__main} despite the presence of
7322 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7323 where the init section is not actually run automatically, but is still
7324 useful for collecting the lists of constructors and destructors.
7327 @defmac SUPPORTS_INIT_PRIORITY
7328 If nonzero, the C++ @code{init_priority} attribute is supported and the
7329 compiler should emit instructions to control the order of initialization
7330 of objects. If zero, the compiler will issue an error message upon
7331 encountering an @code{init_priority} attribute.
7334 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7335 This value is true if the target supports some ``native'' method of
7336 collecting constructors and destructors to be run at startup and exit.
7337 It is false if we must use @command{collect2}.
7340 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7341 If defined, a function that outputs assembler code to arrange to call
7342 the function referenced by @var{symbol} at initialization time.
7344 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7345 no arguments and with no return value. If the target supports initialization
7346 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7347 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7349 If this macro is not defined by the target, a suitable default will
7350 be chosen if (1) the target supports arbitrary section names, (2) the
7351 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7355 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7356 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7357 functions rather than initialization functions.
7360 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7361 generated for the generated object file will have static linkage.
7363 If your system uses @command{collect2} as the means of processing
7364 constructors, then that program normally uses @command{nm} to scan
7365 an object file for constructor functions to be called.
7367 On certain kinds of systems, you can define this macro to make
7368 @command{collect2} work faster (and, in some cases, make it work at all):
7370 @defmac OBJECT_FORMAT_COFF
7371 Define this macro if the system uses COFF (Common Object File Format)
7372 object files, so that @command{collect2} can assume this format and scan
7373 object files directly for dynamic constructor/destructor functions.
7375 This macro is effective only in a native compiler; @command{collect2} as
7376 part of a cross compiler always uses @command{nm} for the target machine.
7379 @defmac REAL_NM_FILE_NAME
7380 Define this macro as a C string constant containing the file name to use
7381 to execute @command{nm}. The default is to search the path normally for
7384 If your system supports shared libraries and has a program to list the
7385 dynamic dependencies of a given library or executable, you can define
7386 these macros to enable support for running initialization and
7387 termination functions in shared libraries:
7391 Define this macro to a C string constant containing the name of the program
7392 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7395 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7396 Define this macro to be C code that extracts filenames from the output
7397 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7398 of type @code{char *} that points to the beginning of a line of output
7399 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7400 code must advance @var{ptr} to the beginning of the filename on that
7401 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7404 @node Instruction Output
7405 @subsection Output of Assembler Instructions
7407 @c prevent bad page break with this line
7408 This describes assembler instruction output.
7410 @defmac REGISTER_NAMES
7411 A C initializer containing the assembler's names for the machine
7412 registers, each one as a C string constant. This is what translates
7413 register numbers in the compiler into assembler language.
7416 @defmac ADDITIONAL_REGISTER_NAMES
7417 If defined, a C initializer for an array of structures containing a name
7418 and a register number. This macro defines additional names for hard
7419 registers, thus allowing the @code{asm} option in declarations to refer
7420 to registers using alternate names.
7423 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7424 Define this macro if you are using an unusual assembler that
7425 requires different names for the machine instructions.
7427 The definition is a C statement or statements which output an
7428 assembler instruction opcode to the stdio stream @var{stream}. The
7429 macro-operand @var{ptr} is a variable of type @code{char *} which
7430 points to the opcode name in its ``internal'' form---the form that is
7431 written in the machine description. The definition should output the
7432 opcode name to @var{stream}, performing any translation you desire, and
7433 increment the variable @var{ptr} to point at the end of the opcode
7434 so that it will not be output twice.
7436 In fact, your macro definition may process less than the entire opcode
7437 name, or more than the opcode name; but if you want to process text
7438 that includes @samp{%}-sequences to substitute operands, you must take
7439 care of the substitution yourself. Just be sure to increment
7440 @var{ptr} over whatever text should not be output normally.
7442 @findex recog_data.operand
7443 If you need to look at the operand values, they can be found as the
7444 elements of @code{recog_data.operand}.
7446 If the macro definition does nothing, the instruction is output
7450 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7451 If defined, a C statement to be executed just prior to the output of
7452 assembler code for @var{insn}, to modify the extracted operands so
7453 they will be output differently.
7455 Here the argument @var{opvec} is the vector containing the operands
7456 extracted from @var{insn}, and @var{noperands} is the number of
7457 elements of the vector which contain meaningful data for this insn.
7458 The contents of this vector are what will be used to convert the insn
7459 template into assembler code, so you can change the assembler output
7460 by changing the contents of the vector.
7462 This macro is useful when various assembler syntaxes share a single
7463 file of instruction patterns; by defining this macro differently, you
7464 can cause a large class of instructions to be output differently (such
7465 as with rearranged operands). Naturally, variations in assembler
7466 syntax affecting individual insn patterns ought to be handled by
7467 writing conditional output routines in those patterns.
7469 If this macro is not defined, it is equivalent to a null statement.
7472 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7473 A C compound statement to output to stdio stream @var{stream} the
7474 assembler syntax for an instruction operand @var{x}. @var{x} is an
7477 @var{code} is a value that can be used to specify one of several ways
7478 of printing the operand. It is used when identical operands must be
7479 printed differently depending on the context. @var{code} comes from
7480 the @samp{%} specification that was used to request printing of the
7481 operand. If the specification was just @samp{%@var{digit}} then
7482 @var{code} is 0; if the specification was @samp{%@var{ltr}
7483 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7486 If @var{x} is a register, this macro should print the register's name.
7487 The names can be found in an array @code{reg_names} whose type is
7488 @code{char *[]}. @code{reg_names} is initialized from
7489 @code{REGISTER_NAMES}.
7491 When the machine description has a specification @samp{%@var{punct}}
7492 (a @samp{%} followed by a punctuation character), this macro is called
7493 with a null pointer for @var{x} and the punctuation character for
7497 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7498 A C expression which evaluates to true if @var{code} is a valid
7499 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7500 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7501 punctuation characters (except for the standard one, @samp{%}) are used
7505 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7506 A C compound statement to output to stdio stream @var{stream} the
7507 assembler syntax for an instruction operand that is a memory reference
7508 whose address is @var{x}. @var{x} is an RTL expression.
7510 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7511 On some machines, the syntax for a symbolic address depends on the
7512 section that the address refers to. On these machines, define the hook
7513 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7514 @code{symbol_ref}, and then check for it here. @xref{Assembler
7518 @findex dbr_sequence_length
7519 @defmac DBR_OUTPUT_SEQEND (@var{file})
7520 A C statement, to be executed after all slot-filler instructions have
7521 been output. If necessary, call @code{dbr_sequence_length} to
7522 determine the number of slots filled in a sequence (zero if not
7523 currently outputting a sequence), to decide how many no-ops to output,
7526 Don't define this macro if it has nothing to do, but it is helpful in
7527 reading assembly output if the extent of the delay sequence is made
7528 explicit (e.g.@: with white space).
7531 @findex final_sequence
7532 Note that output routines for instructions with delay slots must be
7533 prepared to deal with not being output as part of a sequence
7534 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7535 found.) The variable @code{final_sequence} is null when not
7536 processing a sequence, otherwise it contains the @code{sequence} rtx
7540 @defmac REGISTER_PREFIX
7541 @defmacx LOCAL_LABEL_PREFIX
7542 @defmacx USER_LABEL_PREFIX
7543 @defmacx IMMEDIATE_PREFIX
7544 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7545 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7546 @file{final.c}). These are useful when a single @file{md} file must
7547 support multiple assembler formats. In that case, the various @file{tm.h}
7548 files can define these macros differently.
7551 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
7552 If defined this macro should expand to a series of @code{case}
7553 statements which will be parsed inside the @code{switch} statement of
7554 the @code{asm_fprintf} function. This allows targets to define extra
7555 printf formats which may useful when generating their assembler
7556 statements. Note that uppercase letters are reserved for future
7557 generic extensions to asm_fprintf, and so are not available to target
7558 specific code. The output file is given by the parameter @var{file}.
7559 The varargs input pointer is @var{argptr} and the rest of the format
7560 string, starting the character after the one that is being switched
7561 upon, is pointed to by @var{format}.
7564 @defmac ASSEMBLER_DIALECT
7565 If your target supports multiple dialects of assembler language (such as
7566 different opcodes), define this macro as a C expression that gives the
7567 numeric index of the assembler language dialect to use, with zero as the
7570 If this macro is defined, you may use constructs of the form
7572 @samp{@{option0|option1|option2@dots{}@}}
7575 in the output templates of patterns (@pxref{Output Template}) or in the
7576 first argument of @code{asm_fprintf}. This construct outputs
7577 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7578 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7579 within these strings retain their usual meaning. If there are fewer
7580 alternatives within the braces than the value of
7581 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7583 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7584 @samp{@}} do not have any special meaning when used in templates or
7585 operands to @code{asm_fprintf}.
7587 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7588 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7589 the variations in assembler language syntax with that mechanism. Define
7590 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7591 if the syntax variant are larger and involve such things as different
7592 opcodes or operand order.
7595 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7596 A C expression to output to @var{stream} some assembler code
7597 which will push hard register number @var{regno} onto the stack.
7598 The code need not be optimal, since this macro is used only when
7602 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7603 A C expression to output to @var{stream} some assembler code
7604 which will pop hard register number @var{regno} off of the stack.
7605 The code need not be optimal, since this macro is used only when
7609 @node Dispatch Tables
7610 @subsection Output of Dispatch Tables
7612 @c prevent bad page break with this line
7613 This concerns dispatch tables.
7615 @cindex dispatch table
7616 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7617 A C statement to output to the stdio stream @var{stream} an assembler
7618 pseudo-instruction to generate a difference between two labels.
7619 @var{value} and @var{rel} are the numbers of two internal labels. The
7620 definitions of these labels are output using
7621 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7622 way here. For example,
7625 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7626 @var{value}, @var{rel})
7629 You must provide this macro on machines where the addresses in a
7630 dispatch table are relative to the table's own address. If defined, GCC
7631 will also use this macro on all machines when producing PIC@.
7632 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7633 mode and flags can be read.
7636 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7637 This macro should be provided on machines where the addresses
7638 in a dispatch table are absolute.
7640 The definition should be a C statement to output to the stdio stream
7641 @var{stream} an assembler pseudo-instruction to generate a reference to
7642 a label. @var{value} is the number of an internal label whose
7643 definition is output using @code{(*targetm.asm_out.internal_label)}.
7647 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7651 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7652 Define this if the label before a jump-table needs to be output
7653 specially. The first three arguments are the same as for
7654 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7655 jump-table which follows (a @code{jump_insn} containing an
7656 @code{addr_vec} or @code{addr_diff_vec}).
7658 This feature is used on system V to output a @code{swbeg} statement
7661 If this macro is not defined, these labels are output with
7662 @code{(*targetm.asm_out.internal_label)}.
7665 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7666 Define this if something special must be output at the end of a
7667 jump-table. The definition should be a C statement to be executed
7668 after the assembler code for the table is written. It should write
7669 the appropriate code to stdio stream @var{stream}. The argument
7670 @var{table} is the jump-table insn, and @var{num} is the label-number
7671 of the preceding label.
7673 If this macro is not defined, nothing special is output at the end of
7677 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
7678 This target hook emits a label at the beginning of each FDE@. It
7679 should be defined on targets where FDEs need special labels, and it
7680 should write the appropriate label, for the FDE associated with the
7681 function declaration @var{decl}, to the stdio stream @var{stream}.
7682 The third argument, @var{for_eh}, is a boolean: true if this is for an
7683 exception table. The fourth argument, @var{empty}, is a boolean:
7684 true if this is a placeholder label for an omitted FDE@.
7686 The default is that FDEs are not given nonlocal labels.
7689 @deftypefn {Taget Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn})
7690 This target hook emits and assembly directives required to unwind the
7691 given instruction. This is only used when TARGET_UNWIND_INFO is set.
7694 @node Exception Region Output
7695 @subsection Assembler Commands for Exception Regions
7697 @c prevent bad page break with this line
7699 This describes commands marking the start and the end of an exception
7702 @defmac EH_FRAME_SECTION_NAME
7703 If defined, a C string constant for the name of the section containing
7704 exception handling frame unwind information. If not defined, GCC will
7705 provide a default definition if the target supports named sections.
7706 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7708 You should define this symbol if your target supports DWARF 2 frame
7709 unwind information and the default definition does not work.
7712 @defmac EH_FRAME_IN_DATA_SECTION
7713 If defined, DWARF 2 frame unwind information will be placed in the
7714 data section even though the target supports named sections. This
7715 might be necessary, for instance, if the system linker does garbage
7716 collection and sections cannot be marked as not to be collected.
7718 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7722 @defmac EH_TABLES_CAN_BE_READ_ONLY
7723 Define this macro to 1 if your target is such that no frame unwind
7724 information encoding used with non-PIC code will ever require a
7725 runtime relocation, but the linker may not support merging read-only
7726 and read-write sections into a single read-write section.
7729 @defmac MASK_RETURN_ADDR
7730 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7731 that it does not contain any extraneous set bits in it.
7734 @defmac DWARF2_UNWIND_INFO
7735 Define this macro to 0 if your target supports DWARF 2 frame unwind
7736 information, but it does not yet work with exception handling.
7737 Otherwise, if your target supports this information (if it defines
7738 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7739 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7742 If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder
7743 will be used in all cases. Defining this macro will enable the generation
7744 of DWARF 2 frame debugging information.
7746 If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1,
7747 the DWARF 2 unwinder will be the default exception handling mechanism;
7748 otherwise, @code{setjmp}/@code{longjmp} will be used by default.
7751 @defmac TARGET_UNWIND_INFO
7752 Define this macro if your target has ABI specified unwind tables. Usually
7753 these will be output by @code{TARGET_UNWIND_EMIT}.
7756 @defmac MUST_USE_SJLJ_EXCEPTIONS
7757 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
7758 runtime-variable. In that case, @file{except.h} cannot correctly
7759 determine the corresponding definition of
7760 @code{MUST_USE_SJLJ_EXCEPTIONS}, so the target must provide it directly.
7763 @defmac DWARF_CIE_DATA_ALIGNMENT
7764 This macro need only be defined if the target might save registers in the
7765 function prologue at an offset to the stack pointer that is not aligned to
7766 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7767 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7768 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7769 the target supports DWARF 2 frame unwind information.
7772 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7773 If defined, a function that switches to the section in which the main
7774 exception table is to be placed (@pxref{Sections}). The default is a
7775 function that switches to a section named @code{.gcc_except_table} on
7776 machines that support named sections via
7777 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7778 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7779 @code{readonly_data_section}.
7782 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7783 If defined, a function that switches to the section in which the DWARF 2
7784 frame unwind information to be placed (@pxref{Sections}). The default
7785 is a function that outputs a standard GAS section directive, if
7786 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7787 directive followed by a synthetic label.
7790 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7791 Contains the value true if the target should add a zero word onto the
7792 end of a Dwarf-2 frame info section when used for exception handling.
7793 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7797 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
7798 Given a register, this hook should return a parallel of registers to
7799 represent where to find the register pieces. Define this hook if the
7800 register and its mode are represented in Dwarf in non-contiguous
7801 locations, or if the register should be represented in more than one
7802 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
7803 If not defined, the default is to return @code{NULL_RTX}.
7806 @node Alignment Output
7807 @subsection Assembler Commands for Alignment
7809 @c prevent bad page break with this line
7810 This describes commands for alignment.
7812 @defmac JUMP_ALIGN (@var{label})
7813 The alignment (log base 2) to put in front of @var{label}, which is
7814 a common destination of jumps and has no fallthru incoming edge.
7816 This macro need not be defined if you don't want any special alignment
7817 to be done at such a time. Most machine descriptions do not currently
7820 Unless it's necessary to inspect the @var{label} parameter, it is better
7821 to set the variable @var{align_jumps} in the target's
7822 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7823 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7826 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
7827 The alignment (log base 2) to put in front of @var{label}, which follows
7830 This macro need not be defined if you don't want any special alignment
7831 to be done at such a time. Most machine descriptions do not currently
7835 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7836 The maximum number of bytes to skip when applying
7837 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7838 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7841 @defmac LOOP_ALIGN (@var{label})
7842 The alignment (log base 2) to put in front of @var{label}, which follows
7843 a @code{NOTE_INSN_LOOP_BEG} note.
7845 This macro need not be defined if you don't want any special alignment
7846 to be done at such a time. Most machine descriptions do not currently
7849 Unless it's necessary to inspect the @var{label} parameter, it is better
7850 to set the variable @code{align_loops} in the target's
7851 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7852 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7855 @defmac LOOP_ALIGN_MAX_SKIP
7856 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7857 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7860 @defmac LABEL_ALIGN (@var{label})
7861 The alignment (log base 2) to put in front of @var{label}.
7862 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7863 the maximum of the specified values is used.
7865 Unless it's necessary to inspect the @var{label} parameter, it is better
7866 to set the variable @code{align_labels} in the target's
7867 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7868 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7871 @defmac LABEL_ALIGN_MAX_SKIP
7872 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7873 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7876 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7877 A C statement to output to the stdio stream @var{stream} an assembler
7878 instruction to advance the location counter by @var{nbytes} bytes.
7879 Those bytes should be zero when loaded. @var{nbytes} will be a C
7880 expression of type @code{int}.
7883 @defmac ASM_NO_SKIP_IN_TEXT
7884 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7885 text section because it fails to put zeros in the bytes that are skipped.
7886 This is true on many Unix systems, where the pseudo--op to skip bytes
7887 produces no-op instructions rather than zeros when used in the text
7891 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7892 A C statement to output to the stdio stream @var{stream} an assembler
7893 command to advance the location counter to a multiple of 2 to the
7894 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7897 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7898 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7899 for padding, if necessary.
7902 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7903 A C statement to output to the stdio stream @var{stream} an assembler
7904 command to advance the location counter to a multiple of 2 to the
7905 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7906 satisfy the alignment request. @var{power} and @var{max_skip} will be
7907 a C expression of type @code{int}.
7911 @node Debugging Info
7912 @section Controlling Debugging Information Format
7914 @c prevent bad page break with this line
7915 This describes how to specify debugging information.
7918 * All Debuggers:: Macros that affect all debugging formats uniformly.
7919 * DBX Options:: Macros enabling specific options in DBX format.
7920 * DBX Hooks:: Hook macros for varying DBX format.
7921 * File Names and DBX:: Macros controlling output of file names in DBX format.
7922 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7923 * VMS Debug:: Macros for VMS debug format.
7927 @subsection Macros Affecting All Debugging Formats
7929 @c prevent bad page break with this line
7930 These macros affect all debugging formats.
7932 @defmac DBX_REGISTER_NUMBER (@var{regno})
7933 A C expression that returns the DBX register number for the compiler
7934 register number @var{regno}. In the default macro provided, the value
7935 of this expression will be @var{regno} itself. But sometimes there are
7936 some registers that the compiler knows about and DBX does not, or vice
7937 versa. In such cases, some register may need to have one number in the
7938 compiler and another for DBX@.
7940 If two registers have consecutive numbers inside GCC, and they can be
7941 used as a pair to hold a multiword value, then they @emph{must} have
7942 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7943 Otherwise, debuggers will be unable to access such a pair, because they
7944 expect register pairs to be consecutive in their own numbering scheme.
7946 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7947 does not preserve register pairs, then what you must do instead is
7948 redefine the actual register numbering scheme.
7951 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
7952 A C expression that returns the integer offset value for an automatic
7953 variable having address @var{x} (an RTL expression). The default
7954 computation assumes that @var{x} is based on the frame-pointer and
7955 gives the offset from the frame-pointer. This is required for targets
7956 that produce debugging output for DBX or COFF-style debugging output
7957 for SDB and allow the frame-pointer to be eliminated when the
7958 @option{-g} options is used.
7961 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7962 A C expression that returns the integer offset value for an argument
7963 having address @var{x} (an RTL expression). The nominal offset is
7967 @defmac PREFERRED_DEBUGGING_TYPE
7968 A C expression that returns the type of debugging output GCC should
7969 produce when the user specifies just @option{-g}. Define
7970 this if you have arranged for GCC to support more than one format of
7971 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7972 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7973 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7975 When the user specifies @option{-ggdb}, GCC normally also uses the
7976 value of this macro to select the debugging output format, but with two
7977 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
7978 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7979 defined, GCC uses @code{DBX_DEBUG}.
7981 The value of this macro only affects the default debugging output; the
7982 user can always get a specific type of output by using @option{-gstabs},
7983 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
7987 @subsection Specific Options for DBX Output
7989 @c prevent bad page break with this line
7990 These are specific options for DBX output.
7992 @defmac DBX_DEBUGGING_INFO
7993 Define this macro if GCC should produce debugging output for DBX
7994 in response to the @option{-g} option.
7997 @defmac XCOFF_DEBUGGING_INFO
7998 Define this macro if GCC should produce XCOFF format debugging output
7999 in response to the @option{-g} option. This is a variant of DBX format.
8002 @defmac DEFAULT_GDB_EXTENSIONS
8003 Define this macro to control whether GCC should by default generate
8004 GDB's extended version of DBX debugging information (assuming DBX-format
8005 debugging information is enabled at all). If you don't define the
8006 macro, the default is 1: always generate the extended information
8007 if there is any occasion to.
8010 @defmac DEBUG_SYMS_TEXT
8011 Define this macro if all @code{.stabs} commands should be output while
8012 in the text section.
8015 @defmac ASM_STABS_OP
8016 A C string constant, including spacing, naming the assembler pseudo op to
8017 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
8018 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
8019 applies only to DBX debugging information format.
8022 @defmac ASM_STABD_OP
8023 A C string constant, including spacing, naming the assembler pseudo op to
8024 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
8025 value is the current location. If you don't define this macro,
8026 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
8030 @defmac ASM_STABN_OP
8031 A C string constant, including spacing, naming the assembler pseudo op to
8032 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
8033 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
8034 macro applies only to DBX debugging information format.
8037 @defmac DBX_NO_XREFS
8038 Define this macro if DBX on your system does not support the construct
8039 @samp{xs@var{tagname}}. On some systems, this construct is used to
8040 describe a forward reference to a structure named @var{tagname}.
8041 On other systems, this construct is not supported at all.
8044 @defmac DBX_CONTIN_LENGTH
8045 A symbol name in DBX-format debugging information is normally
8046 continued (split into two separate @code{.stabs} directives) when it
8047 exceeds a certain length (by default, 80 characters). On some
8048 operating systems, DBX requires this splitting; on others, splitting
8049 must not be done. You can inhibit splitting by defining this macro
8050 with the value zero. You can override the default splitting-length by
8051 defining this macro as an expression for the length you desire.
8054 @defmac DBX_CONTIN_CHAR
8055 Normally continuation is indicated by adding a @samp{\} character to
8056 the end of a @code{.stabs} string when a continuation follows. To use
8057 a different character instead, define this macro as a character
8058 constant for the character you want to use. Do not define this macro
8059 if backslash is correct for your system.
8062 @defmac DBX_STATIC_STAB_DATA_SECTION
8063 Define this macro if it is necessary to go to the data section before
8064 outputting the @samp{.stabs} pseudo-op for a non-global static
8068 @defmac DBX_TYPE_DECL_STABS_CODE
8069 The value to use in the ``code'' field of the @code{.stabs} directive
8070 for a typedef. The default is @code{N_LSYM}.
8073 @defmac DBX_STATIC_CONST_VAR_CODE
8074 The value to use in the ``code'' field of the @code{.stabs} directive
8075 for a static variable located in the text section. DBX format does not
8076 provide any ``right'' way to do this. The default is @code{N_FUN}.
8079 @defmac DBX_REGPARM_STABS_CODE
8080 The value to use in the ``code'' field of the @code{.stabs} directive
8081 for a parameter passed in registers. DBX format does not provide any
8082 ``right'' way to do this. The default is @code{N_RSYM}.
8085 @defmac DBX_REGPARM_STABS_LETTER
8086 The letter to use in DBX symbol data to identify a symbol as a parameter
8087 passed in registers. DBX format does not customarily provide any way to
8088 do this. The default is @code{'P'}.
8091 @defmac DBX_FUNCTION_FIRST
8092 Define this macro if the DBX information for a function and its
8093 arguments should precede the assembler code for the function. Normally,
8094 in DBX format, the debugging information entirely follows the assembler
8098 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
8099 Define this macro, with value 1, if the value of a symbol describing
8100 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
8101 relative to the start of the enclosing function. Normally, GCC uses
8102 an absolute address.
8105 @defmac DBX_LINES_FUNCTION_RELATIVE
8106 Define this macro, with value 1, if the value of a symbol indicating
8107 the current line number (@code{N_SLINE}) should be relative to the
8108 start of the enclosing function. Normally, GCC uses an absolute address.
8111 @defmac DBX_USE_BINCL
8112 Define this macro if GCC should generate @code{N_BINCL} and
8113 @code{N_EINCL} stabs for included header files, as on Sun systems. This
8114 macro also directs GCC to output a type number as a pair of a file
8115 number and a type number within the file. Normally, GCC does not
8116 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
8117 number for a type number.
8121 @subsection Open-Ended Hooks for DBX Format
8123 @c prevent bad page break with this line
8124 These are hooks for DBX format.
8126 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
8127 Define this macro to say how to output to @var{stream} the debugging
8128 information for the start of a scope level for variable names. The
8129 argument @var{name} is the name of an assembler symbol (for use with
8130 @code{assemble_name}) whose value is the address where the scope begins.
8133 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
8134 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
8137 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
8138 Define this macro if the target machine requires special handling to
8139 output an @code{N_FUN} entry for the function @var{decl}.
8142 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
8143 A C statement to output DBX debugging information before code for line
8144 number @var{line} of the current source file to the stdio stream
8145 @var{stream}. @var{counter} is the number of time the macro was
8146 invoked, including the current invocation; it is intended to generate
8147 unique labels in the assembly output.
8149 This macro should not be defined if the default output is correct, or
8150 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
8153 @defmac NO_DBX_FUNCTION_END
8154 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8155 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8156 On those machines, define this macro to turn this feature off without
8157 disturbing the rest of the gdb extensions.
8160 @defmac NO_DBX_BNSYM_ENSYM
8161 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
8162 extension construct. On those machines, define this macro to turn this
8163 feature off without disturbing the rest of the gdb extensions.
8166 @node File Names and DBX
8167 @subsection File Names in DBX Format
8169 @c prevent bad page break with this line
8170 This describes file names in DBX format.
8172 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8173 A C statement to output DBX debugging information to the stdio stream
8174 @var{stream}, which indicates that file @var{name} is the main source
8175 file---the file specified as the input file for compilation.
8176 This macro is called only once, at the beginning of compilation.
8178 This macro need not be defined if the standard form of output
8179 for DBX debugging information is appropriate.
8181 It may be necessary to refer to a label equal to the beginning of the
8182 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
8183 to do so. If you do this, you must also set the variable
8184 @var{used_ltext_label_name} to @code{true}.
8187 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
8188 Define this macro, with value 1, if GCC should not emit an indication
8189 of the current directory for compilation and current source language at
8190 the beginning of the file.
8193 @defmac NO_DBX_GCC_MARKER
8194 Define this macro, with value 1, if GCC should not emit an indication
8195 that this object file was compiled by GCC@. The default is to emit
8196 an @code{N_OPT} stab at the beginning of every source file, with
8197 @samp{gcc2_compiled.} for the string and value 0.
8200 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8201 A C statement to output DBX debugging information at the end of
8202 compilation of the main source file @var{name}. Output should be
8203 written to the stdio stream @var{stream}.
8205 If you don't define this macro, nothing special is output at the end
8206 of compilation, which is correct for most machines.
8209 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
8210 Define this macro @emph{instead of} defining
8211 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
8212 the end of compilation is a @code{N_SO} stab with an empty string,
8213 whose value is the highest absolute text address in the file.
8218 @subsection Macros for SDB and DWARF Output
8220 @c prevent bad page break with this line
8221 Here are macros for SDB and DWARF output.
8223 @defmac SDB_DEBUGGING_INFO
8224 Define this macro if GCC should produce COFF-style debugging output
8225 for SDB in response to the @option{-g} option.
8228 @defmac DWARF2_DEBUGGING_INFO
8229 Define this macro if GCC should produce dwarf version 2 format
8230 debugging output in response to the @option{-g} option.
8232 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (tree @var{function})
8233 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
8234 be emitted for each function. Instead of an integer return the enum
8235 value for the @code{DW_CC_} tag.
8238 To support optional call frame debugging information, you must also
8239 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8240 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8241 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8242 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8245 @defmac DWARF2_FRAME_INFO
8246 Define this macro to a nonzero value if GCC should always output
8247 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8248 (@pxref{Exception Region Output} is nonzero, GCC will output this
8249 information not matter how you define @code{DWARF2_FRAME_INFO}.
8252 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8253 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8254 line debug info sections. This will result in much more compact line number
8255 tables, and hence is desirable if it works.
8258 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8259 A C statement to issue assembly directives that create a difference
8260 between the two given labels, using an integer of the given size.
8263 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8264 A C statement to issue assembly directives that create a
8265 section-relative reference to the given label, using an integer of the
8269 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8270 A C statement to issue assembly directives that create a self-relative
8271 reference to the given label, using an integer of the given size.
8274 @defmac PUT_SDB_@dots{}
8275 Define these macros to override the assembler syntax for the special
8276 SDB assembler directives. See @file{sdbout.c} for a list of these
8277 macros and their arguments. If the standard syntax is used, you need
8278 not define them yourself.
8282 Some assemblers do not support a semicolon as a delimiter, even between
8283 SDB assembler directives. In that case, define this macro to be the
8284 delimiter to use (usually @samp{\n}). It is not necessary to define
8285 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8289 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8290 Define this macro to allow references to unknown structure,
8291 union, or enumeration tags to be emitted. Standard COFF does not
8292 allow handling of unknown references, MIPS ECOFF has support for
8296 @defmac SDB_ALLOW_FORWARD_REFERENCES
8297 Define this macro to allow references to structure, union, or
8298 enumeration tags that have not yet been seen to be handled. Some
8299 assemblers choke if forward tags are used, while some require it.
8302 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
8303 A C statement to output SDB debugging information before code for line
8304 number @var{line} of the current source file to the stdio stream
8305 @var{stream}. The default is to emit an @code{.ln} directive.
8310 @subsection Macros for VMS Debug Format
8312 @c prevent bad page break with this line
8313 Here are macros for VMS debug format.
8315 @defmac VMS_DEBUGGING_INFO
8316 Define this macro if GCC should produce debugging output for VMS
8317 in response to the @option{-g} option. The default behavior for VMS
8318 is to generate minimal debug info for a traceback in the absence of
8319 @option{-g} unless explicitly overridden with @option{-g0}. This
8320 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8321 @code{OVERRIDE_OPTIONS}.
8324 @node Floating Point
8325 @section Cross Compilation and Floating Point
8326 @cindex cross compilation and floating point
8327 @cindex floating point and cross compilation
8329 While all modern machines use twos-complement representation for integers,
8330 there are a variety of representations for floating point numbers. This
8331 means that in a cross-compiler the representation of floating point numbers
8332 in the compiled program may be different from that used in the machine
8333 doing the compilation.
8335 Because different representation systems may offer different amounts of
8336 range and precision, all floating point constants must be represented in
8337 the target machine's format. Therefore, the cross compiler cannot
8338 safely use the host machine's floating point arithmetic; it must emulate
8339 the target's arithmetic. To ensure consistency, GCC always uses
8340 emulation to work with floating point values, even when the host and
8341 target floating point formats are identical.
8343 The following macros are provided by @file{real.h} for the compiler to
8344 use. All parts of the compiler which generate or optimize
8345 floating-point calculations must use these macros. They may evaluate
8346 their operands more than once, so operands must not have side effects.
8348 @defmac REAL_VALUE_TYPE
8349 The C data type to be used to hold a floating point value in the target
8350 machine's format. Typically this is a @code{struct} containing an
8351 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8355 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8356 Compares for equality the two values, @var{x} and @var{y}. If the target
8357 floating point format supports negative zeroes and/or NaNs,
8358 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8359 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8362 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8363 Tests whether @var{x} is less than @var{y}.
8366 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8367 Truncates @var{x} to a signed integer, rounding toward zero.
8370 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8371 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8372 @var{x} is negative, returns zero.
8375 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8376 Converts @var{string} into a floating point number in the target machine's
8377 representation for mode @var{mode}. This routine can handle both
8378 decimal and hexadecimal floating point constants, using the syntax
8379 defined by the C language for both.
8382 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8383 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8386 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8387 Determines whether @var{x} represents infinity (positive or negative).
8390 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8391 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8394 @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})
8395 Calculates an arithmetic operation on the two floating point values
8396 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8399 The operation to be performed is specified by @var{code}. Only the
8400 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8401 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8403 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8404 target's floating point format cannot represent infinity, it will call
8405 @code{abort}. Callers should check for this situation first, using
8406 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8409 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8410 Returns the negative of the floating point value @var{x}.
8413 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8414 Returns the absolute value of @var{x}.
8417 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8418 Truncates the floating point value @var{x} to fit in @var{mode}. The
8419 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8420 appropriate bit pattern to be output asa floating constant whose
8421 precision accords with mode @var{mode}.
8424 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8425 Converts a floating point value @var{x} into a double-precision integer
8426 which is then stored into @var{low} and @var{high}. If the value is not
8427 integral, it is truncated.
8430 @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})
8431 Converts a double-precision integer found in @var{low} and @var{high},
8432 into a floating point value which is then stored into @var{x}. The
8433 value is truncated to fit in mode @var{mode}.
8436 @node Mode Switching
8437 @section Mode Switching Instructions
8438 @cindex mode switching
8439 The following macros control mode switching optimizations:
8441 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8442 Define this macro if the port needs extra instructions inserted for mode
8443 switching in an optimizing compilation.
8445 For an example, the SH4 can perform both single and double precision
8446 floating point operations, but to perform a single precision operation,
8447 the FPSCR PR bit has to be cleared, while for a double precision
8448 operation, this bit has to be set. Changing the PR bit requires a general
8449 purpose register as a scratch register, hence these FPSCR sets have to
8450 be inserted before reload, i.e.@: you can't put this into instruction emitting
8451 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
8453 You can have multiple entities that are mode-switched, and select at run time
8454 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8455 return nonzero for any @var{entity} that needs mode-switching.
8456 If you define this macro, you also have to define
8457 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8458 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8459 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
8463 @defmac NUM_MODES_FOR_MODE_SWITCHING
8464 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8465 initializer for an array of integers. Each initializer element
8466 N refers to an entity that needs mode switching, and specifies the number
8467 of different modes that might need to be set for this entity.
8468 The position of the initializer in the initializer---starting counting at
8469 zero---determines the integer that is used to refer to the mode-switched
8471 In macros that take mode arguments / yield a mode result, modes are
8472 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8473 switch is needed / supplied.
8476 @defmac MODE_NEEDED (@var{entity}, @var{insn})
8477 @var{entity} is an integer specifying a mode-switched entity. If
8478 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8479 return an integer value not larger than the corresponding element in
8480 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8481 be switched into prior to the execution of @var{insn}.
8484 @defmac MODE_AFTER (@var{mode}, @var{insn})
8485 If this macro is defined, it is evaluated for every @var{insn} during
8486 mode switching. It determines the mode that an insn results in (if
8487 different from the incoming mode).
8490 @defmac MODE_ENTRY (@var{entity})
8491 If this macro is defined, it is evaluated for every @var{entity} that needs
8492 mode switching. It should evaluate to an integer, which is a mode that
8493 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
8494 is defined then @code{MODE_EXIT} must be defined.
8497 @defmac MODE_EXIT (@var{entity})
8498 If this macro is defined, it is evaluated for every @var{entity} that needs
8499 mode switching. It should evaluate to an integer, which is a mode that
8500 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
8501 is defined then @code{MODE_ENTRY} must be defined.
8504 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8505 This macro specifies the order in which modes for @var{entity} are processed.
8506 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8507 lowest. The value of the macro should be an integer designating a mode
8508 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8509 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8510 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8513 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8514 Generate one or more insns to set @var{entity} to @var{mode}.
8515 @var{hard_reg_live} is the set of hard registers live at the point where
8516 the insn(s) are to be inserted.
8519 @node Target Attributes
8520 @section Defining target-specific uses of @code{__attribute__}
8521 @cindex target attributes
8522 @cindex machine attributes
8523 @cindex attributes, target-specific
8525 Target-specific attributes may be defined for functions, data and types.
8526 These are described using the following target hooks; they also need to
8527 be documented in @file{extend.texi}.
8529 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8530 If defined, this target hook points to an array of @samp{struct
8531 attribute_spec} (defined in @file{tree.h}) specifying the machine
8532 specific attributes for this target and some of the restrictions on the
8533 entities to which these attributes are applied and the arguments they
8537 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8538 If defined, this target hook is a function which returns zero if the attributes on
8539 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8540 and two if they are nearly compatible (which causes a warning to be
8541 generated). If this is not defined, machine-specific attributes are
8542 supposed always to be compatible.
8545 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8546 If defined, this target hook is a function which assigns default attributes to
8547 newly defined @var{type}.
8550 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8551 Define this target hook if the merging of type attributes needs special
8552 handling. If defined, the result is a list of the combined
8553 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8554 that @code{comptypes} has already been called and returned 1. This
8555 function may call @code{merge_attributes} to handle machine-independent
8559 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8560 Define this target hook if the merging of decl attributes needs special
8561 handling. If defined, the result is a list of the combined
8562 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8563 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8564 when this is needed are when one attribute overrides another, or when an
8565 attribute is nullified by a subsequent definition. This function may
8566 call @code{merge_attributes} to handle machine-independent merging.
8568 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8569 If the only target-specific handling you require is @samp{dllimport}
8570 for Microsoft Windows targets, you should define the macro
8571 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
8572 will then define a function called
8573 @code{merge_dllimport_decl_attributes} which can then be defined as
8574 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
8575 add @code{handle_dll_attribute} in the attribute table for your port
8576 to perform initial processing of the @samp{dllimport} and
8577 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
8578 @file{i386/i386.c}, for example.
8581 @defmac TARGET_DECLSPEC
8582 Define this macro to a nonzero value if you want to treat
8583 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
8584 default, this behavior is enabled only for targets that define
8585 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
8586 of @code{__declspec} is via a built-in macro, but you should not rely
8587 on this implementation detail.
8590 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8591 Define this target hook if you want to be able to add attributes to a decl
8592 when it is being created. This is normally useful for back ends which
8593 wish to implement a pragma by using the attributes which correspond to
8594 the pragma's effect. The @var{node} argument is the decl which is being
8595 created. The @var{attr_ptr} argument is a pointer to the attribute list
8596 for this decl. The list itself should not be modified, since it may be
8597 shared with other decls, but attributes may be chained on the head of
8598 the list and @code{*@var{attr_ptr}} modified to point to the new
8599 attributes, or a copy of the list may be made if further changes are
8603 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8605 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8606 into the current function, despite its having target-specific
8607 attributes, @code{false} otherwise. By default, if a function has a
8608 target specific attribute attached to it, it will not be inlined.
8611 @node MIPS Coprocessors
8612 @section Defining coprocessor specifics for MIPS targets.
8613 @cindex MIPS coprocessor-definition macros
8615 The MIPS specification allows MIPS implementations to have as many as 4
8616 coprocessors, each with as many as 32 private registers. GCC supports
8617 accessing these registers and transferring values between the registers
8618 and memory using asm-ized variables. For example:
8621 register unsigned int cp0count asm ("c0r1");
8627 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8628 names may be added as described below, or the default names may be
8629 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8631 Coprocessor registers are assumed to be epilogue-used; sets to them will
8632 be preserved even if it does not appear that the register is used again
8633 later in the function.
8635 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8636 the FPU@. One accesses COP1 registers through standard mips
8637 floating-point support; they are not included in this mechanism.
8639 There is one macro used in defining the MIPS coprocessor interface which
8640 you may want to override in subtargets; it is described below.
8642 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
8643 A comma-separated list (with leading comma) of pairs describing the
8644 alternate names of coprocessor registers. The format of each entry should be
8646 @{ @var{alternatename}, @var{register_number}@}
8652 @section Parameters for Precompiled Header Validity Checking
8653 @cindex parameters, precompiled headers
8655 @deftypefn {Target Hook} void * TARGET_GET_PCH_VALIDITY (size_t * @var{sz})
8656 Define this hook if your target needs to check a different collection
8657 of flags than the default, which is every flag defined by
8658 @code{TARGET_SWITCHES} and @code{TARGET_OPTIONS}. It should return
8659 some data which will be saved in the PCH file and presented to
8660 @code{TARGET_PCH_VALID_P} later; it should set @code{SZ} to the size
8664 @deftypefn {Target Hook} const char * TARGET_PCH_VALID_P (const void * @var{data}, size_t @var{sz})
8665 Define this hook if your target needs to check a different collection of
8666 flags than the default, which is every flag defined by @code{TARGET_SWITCHES}
8667 and @code{TARGET_OPTIONS}. It is given data which came from
8668 @code{TARGET_GET_PCH_VALIDITY} (in this version of this compiler, so there
8669 is no need for extensive validity checking). It returns @code{NULL} if
8670 it is safe to load a PCH file with this data, or a suitable error message
8671 if not. The error message will be presented to the user, so it should
8676 @section C++ ABI parameters
8677 @cindex parameters, c++ abi
8679 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
8680 Define this hook to override the integer type used for guard variables.
8681 These are used to implement one-time construction of static objects. The
8682 default is long_long_integer_type_node.
8685 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
8686 This hook determines how guard variables are used. It should return
8687 @code{false} (the default) if first byte should be used. A return value of
8688 @code{true} indicates the least significant bit should be used.
8691 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
8692 This hook returns the size of the cookie to use when allocating an array
8693 whose elements have the indicated @var{type}. Assumes that it is already
8694 known that a cookie is needed. The default is
8695 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
8696 IA64/Generic C++ ABI@.
8699 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
8700 This hook should return @code{true} if the element size should be stored in
8701 array cookies. The default is to return @code{false}.
8704 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
8705 If defined by a backend this hook allows the decision made to export
8706 class @var{type} to be overruled. Upon entry @var{import_export}
8707 will contain 1 if the class is going to be exported, @minus{}1 if it is going
8708 to be imported and 0 otherwise. This function should return the
8709 modified value and perform any other actions necessary to support the
8710 backend's targeted operating system.
8713 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
8714 This hook should return @code{true} if constructors and destructors return
8715 the address of the object created/destroyed. The default is to return
8719 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
8720 This hook returns true if the key method for a class (i.e., the method
8721 which, if defined in the current translation unit, causes the virtual
8722 table to be emitted) may be an inline function. Under the standard
8723 Itanium C++ ABI the key method may be an inline function so long as
8724 the function is not declared inline in the class definition. Under
8725 some variants of the ABI, an inline function can never be the key
8726 method. The default is to return @code{true}.
8729 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
8730 @var{decl} is a virtual table, virtual table table, typeinfo object,
8731 or other similar implicit class data object that will be emitted with
8732 external linkage in this translation unit. No ELF visibility has been
8733 explicitly specified. If the target needs to specify a visibility
8734 other than that of the containing class, use this hook to set
8735 @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
8738 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
8739 This hook returns true (the default) if virtual tables and other
8740 similar implicit class data objects are always COMDAT if they have
8741 external linkage. If this hook returns false, then class data for
8742 classes whose virtual table will be emitted in only one translation
8743 unit will not be COMDAT.
8746 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
8747 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
8748 should be used to register static destructors when @option{-fuse-cxa-atexit}
8749 is in effect. The default is to return false to use @code{__cxa_atexit}.
8753 @section Miscellaneous Parameters
8754 @cindex parameters, miscellaneous
8756 @c prevent bad page break with this line
8757 Here are several miscellaneous parameters.
8759 @defmac PREDICATE_CODES
8760 Define this if you have defined special-purpose predicates in the file
8761 @file{@var{machine}.c}. This macro is called within an initializer of an
8762 array of structures. The first field in the structure is the name of a
8763 predicate and the second field is an array of rtl codes. For each
8764 predicate, list all rtl codes that can be in expressions matched by the
8765 predicate. The list should have a trailing comma. Here is an example
8766 of two entries in the list for a typical RISC machine:
8769 #define PREDICATE_CODES \
8770 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8771 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8774 Defining this macro does not affect the generated code (however,
8775 incorrect definitions that omit an rtl code that may be matched by the
8776 predicate can cause the compiler to malfunction). Instead, it allows
8777 the table built by @file{genrecog} to be more compact and efficient,
8778 thus speeding up the compiler. The most important predicates to include
8779 in the list specified by this macro are those used in the most insn
8782 For each predicate function named in @code{PREDICATE_CODES}, a
8783 declaration will be generated in @file{insn-codes.h}.
8785 Use of this macro is deprecated; use @code{define_predicate} instead.
8786 @xref{Defining Predicates}.
8789 @defmac SPECIAL_MODE_PREDICATES
8790 Define this if you have special predicates that know special things
8791 about modes. Genrecog will warn about certain forms of
8792 @code{match_operand} without a mode; if the operand predicate is
8793 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8796 Here is an example from the IA-32 port (@code{ext_register_operand}
8797 specially checks for @code{HImode} or @code{SImode} in preparation
8798 for a byte extraction from @code{%ah} etc.).
8801 #define SPECIAL_MODE_PREDICATES \
8802 "ext_register_operand",
8805 Use of this macro is deprecated; use @code{define_special_predicate}
8806 instead. @xref{Defining Predicates}.
8809 @defmac HAS_LONG_COND_BRANCH
8810 Define this boolean macro to indicate whether or not your architecture
8811 has conditional branches that can span all of memory. It is used in
8812 conjunction with an optimization that partitions hot and cold basic
8813 blocks into separate sections of the executable. If this macro is
8814 set to false, gcc will convert any conditional branches that attempt
8815 to cross between sections into unconditional branches or indirect jumps.
8818 @defmac HAS_LONG_UNCOND_BRANCH
8819 Define this boolean macro to indicate whether or not your architecture
8820 has unconditional branches that can span all of memory. It is used in
8821 conjunction with an optimization that partitions hot and cold basic
8822 blocks into separate sections of the executable. If this macro is
8823 set to false, gcc will convert any unconditional branches that attempt
8824 to cross between sections into indirect jumps.
8827 @defmac CASE_VECTOR_MODE
8828 An alias for a machine mode name. This is the machine mode that
8829 elements of a jump-table should have.
8832 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8833 Optional: return the preferred mode for an @code{addr_diff_vec}
8834 when the minimum and maximum offset are known. If you define this,
8835 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8836 To make this work, you also have to define @code{INSN_ALIGN} and
8837 make the alignment for @code{addr_diff_vec} explicit.
8838 The @var{body} argument is provided so that the offset_unsigned and scale
8839 flags can be updated.
8842 @defmac CASE_VECTOR_PC_RELATIVE
8843 Define this macro to be a C expression to indicate when jump-tables
8844 should contain relative addresses. You need not define this macro if
8845 jump-tables never contain relative addresses, or jump-tables should
8846 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
8850 @defmac CASE_VALUES_THRESHOLD
8851 Define this to be the smallest number of different values for which it
8852 is best to use a jump-table instead of a tree of conditional branches.
8853 The default is four for machines with a @code{casesi} instruction and
8854 five otherwise. This is best for most machines.
8857 @defmac CASE_USE_BIT_TESTS
8858 Define this macro to be a C expression to indicate whether C switch
8859 statements may be implemented by a sequence of bit tests. This is
8860 advantageous on processors that can efficiently implement left shift
8861 of 1 by the number of bits held in a register, but inappropriate on
8862 targets that would require a loop. By default, this macro returns
8863 @code{true} if the target defines an @code{ashlsi3} pattern, and
8864 @code{false} otherwise.
8867 @defmac WORD_REGISTER_OPERATIONS
8868 Define this macro if operations between registers with integral mode
8869 smaller than a word are always performed on the entire register.
8870 Most RISC machines have this property and most CISC machines do not.
8873 @defmac LOAD_EXTEND_OP (@var{mem_mode})
8874 Define this macro to be a C expression indicating when insns that read
8875 memory in @var{mem_mode}, an integral mode narrower than a word, set the
8876 bits outside of @var{mem_mode} to be either the sign-extension or the
8877 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8878 of @var{mem_mode} for which the
8879 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8880 @code{UNKNOWN} for other modes.
8882 This macro is not called with @var{mem_mode} non-integral or with a width
8883 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8884 value in this case. Do not define this macro if it would always return
8885 @code{UNKNOWN}. On machines where this macro is defined, you will normally
8886 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8888 You may return a non-@code{UNKNOWN} value even if for some hard registers
8889 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
8890 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
8891 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
8892 integral mode larger than this but not larger than @code{word_mode}.
8894 You must return @code{UNKNOWN} if for some hard registers that allow this
8895 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
8896 @code{word_mode}, but that they can change to another integral mode that
8897 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
8900 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
8901 Define this macro if loading short immediate values into registers sign
8905 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
8906 Define this macro if the same instructions that convert a floating
8907 point number to a signed fixed point number also convert validly to an
8912 The maximum number of bytes that a single instruction can move quickly
8913 between memory and registers or between two memory locations.
8916 @defmac MAX_MOVE_MAX
8917 The maximum number of bytes that a single instruction can move quickly
8918 between memory and registers or between two memory locations. If this
8919 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8920 constant value that is the largest value that @code{MOVE_MAX} can have
8924 @defmac SHIFT_COUNT_TRUNCATED
8925 A C expression that is nonzero if on this machine the number of bits
8926 actually used for the count of a shift operation is equal to the number
8927 of bits needed to represent the size of the object being shifted. When
8928 this macro is nonzero, the compiler will assume that it is safe to omit
8929 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8930 truncates the count of a shift operation. On machines that have
8931 instructions that act on bit-fields at variable positions, which may
8932 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8933 also enables deletion of truncations of the values that serve as
8934 arguments to bit-field instructions.
8936 If both types of instructions truncate the count (for shifts) and
8937 position (for bit-field operations), or if no variable-position bit-field
8938 instructions exist, you should define this macro.
8940 However, on some machines, such as the 80386 and the 680x0, truncation
8941 only applies to shift operations and not the (real or pretended)
8942 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8943 such machines. Instead, add patterns to the @file{md} file that include
8944 the implied truncation of the shift instructions.
8946 You need not define this macro if it would always have the value of zero.
8949 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
8950 @deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
8951 This function describes how the standard shift patterns for @var{mode}
8952 deal with shifts by negative amounts or by more than the width of the mode.
8953 @xref{shift patterns}.
8955 On many machines, the shift patterns will apply a mask @var{m} to the
8956 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
8957 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
8958 this is true for mode @var{mode}, the function should return @var{m},
8959 otherwise it should return 0. A return value of 0 indicates that no
8960 particular behavior is guaranteed.
8962 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
8963 @emph{not} apply to general shift rtxes; it applies only to instructions
8964 that are generated by the named shift patterns.
8966 The default implementation of this function returns
8967 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
8968 and 0 otherwise. This definition is always safe, but if
8969 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
8970 nevertheless truncate the shift count, you may get better code
8974 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8975 A C expression which is nonzero if on this machine it is safe to
8976 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8977 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8978 operating on it as if it had only @var{outprec} bits.
8980 On many machines, this expression can be 1.
8982 @c rearranged this, removed the phrase "it is reported that". this was
8983 @c to fix an overfull hbox. --mew 10feb93
8984 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8985 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8986 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8987 such cases may improve things.
8990 @defmac STORE_FLAG_VALUE
8991 A C expression describing the value returned by a comparison operator
8992 with an integral mode and stored by a store-flag instruction
8993 (@samp{s@var{cond}}) when the condition is true. This description must
8994 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8995 comparison operators whose results have a @code{MODE_INT} mode.
8997 A value of 1 or @minus{}1 means that the instruction implementing the
8998 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8999 and 0 when the comparison is false. Otherwise, the value indicates
9000 which bits of the result are guaranteed to be 1 when the comparison is
9001 true. This value is interpreted in the mode of the comparison
9002 operation, which is given by the mode of the first operand in the
9003 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
9004 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
9007 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
9008 generate code that depends only on the specified bits. It can also
9009 replace comparison operators with equivalent operations if they cause
9010 the required bits to be set, even if the remaining bits are undefined.
9011 For example, on a machine whose comparison operators return an
9012 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
9013 @samp{0x80000000}, saying that just the sign bit is relevant, the
9017 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
9024 (ashift:SI @var{x} (const_int @var{n}))
9028 where @var{n} is the appropriate shift count to move the bit being
9029 tested into the sign bit.
9031 There is no way to describe a machine that always sets the low-order bit
9032 for a true value, but does not guarantee the value of any other bits,
9033 but we do not know of any machine that has such an instruction. If you
9034 are trying to port GCC to such a machine, include an instruction to
9035 perform a logical-and of the result with 1 in the pattern for the
9036 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
9038 Often, a machine will have multiple instructions that obtain a value
9039 from a comparison (or the condition codes). Here are rules to guide the
9040 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
9045 Use the shortest sequence that yields a valid definition for
9046 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
9047 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
9048 comparison operators to do so because there may be opportunities to
9049 combine the normalization with other operations.
9052 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
9053 slightly preferred on machines with expensive jumps and 1 preferred on
9057 As a second choice, choose a value of @samp{0x80000001} if instructions
9058 exist that set both the sign and low-order bits but do not define the
9062 Otherwise, use a value of @samp{0x80000000}.
9065 Many machines can produce both the value chosen for
9066 @code{STORE_FLAG_VALUE} and its negation in the same number of
9067 instructions. On those machines, you should also define a pattern for
9068 those cases, e.g., one matching
9071 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
9074 Some machines can also perform @code{and} or @code{plus} operations on
9075 condition code values with less instructions than the corresponding
9076 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
9077 machines, define the appropriate patterns. Use the names @code{incscc}
9078 and @code{decscc}, respectively, for the patterns which perform
9079 @code{plus} or @code{minus} operations on condition code values. See
9080 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
9081 find such instruction sequences on other machines.
9083 If this macro is not defined, the default value, 1, is used. You need
9084 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
9085 instructions, or if the value generated by these instructions is 1.
9088 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
9089 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
9090 returned when comparison operators with floating-point results are true.
9091 Define this macro on machines that have comparison operations that return
9092 floating-point values. If there are no such operations, do not define
9096 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
9097 A C expression that gives a rtx representing the non-zero true element
9098 for vector comparisons. The returned rtx should be valid for the inner
9099 mode of @var{mode} which is guaranteed to be a vector mode. Define
9100 this macro on machines that have vector comparison operations that
9101 return a vector result. If there are no such operations, do not define
9102 this macro. Typically, this macro is defined as @code{const1_rtx} or
9103 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
9104 the compiler optimizing such vector comparison operations for the
9108 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9109 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9110 A C expression that evaluates to true if the architecture defines a value
9111 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
9112 should be set to this value. If this macro is not defined, the value of
9113 @code{clz} or @code{ctz} is assumed to be undefined.
9115 This macro must be defined if the target's expansion for @code{ffs}
9116 relies on a particular value to get correct results. Otherwise it
9117 is not necessary, though it may be used to optimize some corner cases.
9119 Note that regardless of this macro the ``definedness'' of @code{clz}
9120 and @code{ctz} at zero do @emph{not} extend to the builtin functions
9121 visible to the user. Thus one may be free to adjust the value at will
9122 to match the target expansion of these operations without fear of
9127 An alias for the machine mode for pointers. On most machines, define
9128 this to be the integer mode corresponding to the width of a hardware
9129 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
9130 On some machines you must define this to be one of the partial integer
9131 modes, such as @code{PSImode}.
9133 The width of @code{Pmode} must be at least as large as the value of
9134 @code{POINTER_SIZE}. If it is not equal, you must define the macro
9135 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
9139 @defmac FUNCTION_MODE
9140 An alias for the machine mode used for memory references to functions
9141 being called, in @code{call} RTL expressions. On most machines this
9142 should be @code{QImode}.
9145 @defmac STDC_0_IN_SYSTEM_HEADERS
9146 In normal operation, the preprocessor expands @code{__STDC__} to the
9147 constant 1, to signify that GCC conforms to ISO Standard C@. On some
9148 hosts, like Solaris, the system compiler uses a different convention,
9149 where @code{__STDC__} is normally 0, but is 1 if the user specifies
9150 strict conformance to the C Standard.
9152 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
9153 convention when processing system header files, but when processing user
9154 files @code{__STDC__} will always expand to 1.
9157 @defmac NO_IMPLICIT_EXTERN_C
9158 Define this macro if the system header files support C++ as well as C@.
9159 This macro inhibits the usual method of using system header files in
9160 C++, which is to pretend that the file's contents are enclosed in
9161 @samp{extern "C" @{@dots{}@}}.
9166 @defmac REGISTER_TARGET_PRAGMAS ()
9167 Define this macro if you want to implement any target-specific pragmas.
9168 If defined, it is a C expression which makes a series of calls to
9169 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
9170 for each pragma. The macro may also do any
9171 setup required for the pragmas.
9173 The primary reason to define this macro is to provide compatibility with
9174 other compilers for the same target. In general, we discourage
9175 definition of target-specific pragmas for GCC@.
9177 If the pragma can be implemented by attributes then you should consider
9178 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
9180 Preprocessor macros that appear on pragma lines are not expanded. All
9181 @samp{#pragma} directives that do not match any registered pragma are
9182 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
9185 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9186 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9188 Each call to @code{c_register_pragma} or
9189 @code{c_register_pragma_with_expansion} establishes one pragma. The
9190 @var{callback} routine will be called when the preprocessor encounters a
9194 #pragma [@var{space}] @var{name} @dots{}
9197 @var{space} is the case-sensitive namespace of the pragma, or
9198 @code{NULL} to put the pragma in the global namespace. The callback
9199 routine receives @var{pfile} as its first argument, which can be passed
9200 on to cpplib's functions if necessary. You can lex tokens after the
9201 @var{name} by calling @code{c_lex}. Tokens that are not read by the
9202 callback will be silently ignored. The end of the line is indicated by
9203 a token of type @code{CPP_EOF}. Macro expansion occurs on the
9204 arguments of pragmas registered with
9205 @code{c_register_pragma_with_expansion} but not on the arguments of
9206 pragmas registered with @code{c_register_pragma}.
9208 For an example use of this routine, see @file{c4x.h} and the callback
9209 routines defined in @file{c4x-c.c}.
9211 Note that the use of @code{c_lex} is specific to the C and C++
9212 compilers. It will not work in the Java or Fortran compilers, or any
9213 other language compilers for that matter. Thus if @code{c_lex} is going
9214 to be called from target-specific code, it must only be done so when
9215 building the C and C++ compilers. This can be done by defining the
9216 variables @code{c_target_objs} and @code{cxx_target_objs} in the
9217 target entry in the @file{config.gcc} file. These variables should name
9218 the target-specific, language-specific object file which contains the
9219 code that uses @code{c_lex}. Note it will also be necessary to add a
9220 rule to the makefile fragment pointed to by @code{tmake_file} that shows
9221 how to build this object file.
9226 @defmac HANDLE_SYSV_PRAGMA
9227 Define this macro (to a value of 1) if you want the System V style
9228 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
9229 [=<value>]} to be supported by gcc.
9231 The pack pragma specifies the maximum alignment (in bytes) of fields
9232 within a structure, in much the same way as the @samp{__aligned__} and
9233 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
9234 the behavior to the default.
9236 A subtlety for Microsoft Visual C/C++ style bit-field packing
9237 (e.g.@: -mms-bitfields) for targets that support it:
9238 When a bit-field is inserted into a packed record, the whole size
9239 of the underlying type is used by one or more same-size adjacent
9240 bit-fields (that is, if its long:3, 32 bits is used in the record,
9241 and any additional adjacent long bit-fields are packed into the same
9242 chunk of 32 bits. However, if the size changes, a new field of that
9245 If both MS bit-fields and @samp{__attribute__((packed))} are used,
9246 the latter will take precedence. If @samp{__attribute__((packed))} is
9247 used on a single field when MS bit-fields are in use, it will take
9248 precedence for that field, but the alignment of the rest of the structure
9249 may affect its placement.
9251 The weak pragma only works if @code{SUPPORTS_WEAK} and
9252 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
9253 of specifically named weak labels, optionally with a value.
9258 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
9259 Define this macro (to a value of 1) if you want to support the Win32
9260 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
9261 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
9262 alignment (in bytes) of fields within a structure, in much the same way as
9263 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
9264 pack value of zero resets the behavior to the default. Successive
9265 invocations of this pragma cause the previous values to be stacked, so
9266 that invocations of @samp{#pragma pack(pop)} will return to the previous
9270 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
9271 Define this macro, as well as
9272 @code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the
9273 arguments of @samp{#pragma pack}.
9276 @defmac TARGET_DEFAULT_PACK_STRUCT
9277 If your target requires a structure packing default other than 0 (meaning
9278 the machine default), define this macro to the necessary value (in bytes).
9279 This must be a value that would also valid to be used with
9280 @samp{#pragma pack()} (that is, a small power of two).
9283 @defmac DOLLARS_IN_IDENTIFIERS
9284 Define this macro to control use of the character @samp{$} in
9285 identifier names for the C family of languages. 0 means @samp{$} is
9286 not allowed by default; 1 means it is allowed. 1 is the default;
9287 there is no need to define this macro in that case.
9290 @defmac NO_DOLLAR_IN_LABEL
9291 Define this macro if the assembler does not accept the character
9292 @samp{$} in label names. By default constructors and destructors in
9293 G++ have @samp{$} in the identifiers. If this macro is defined,
9294 @samp{.} is used instead.
9297 @defmac NO_DOT_IN_LABEL
9298 Define this macro if the assembler does not accept the character
9299 @samp{.} in label names. By default constructors and destructors in G++
9300 have names that use @samp{.}. If this macro is defined, these names
9301 are rewritten to avoid @samp{.}.
9304 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
9305 Define this macro as a C expression that is nonzero if it is safe for the
9306 delay slot scheduler to place instructions in the delay slot of @var{insn},
9307 even if they appear to use a resource set or clobbered in @var{insn}.
9308 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
9309 every @code{call_insn} has this behavior. On machines where some @code{insn}
9310 or @code{jump_insn} is really a function call and hence has this behavior,
9311 you should define this macro.
9313 You need not define this macro if it would always return zero.
9316 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
9317 Define this macro as a C expression that is nonzero if it is safe for the
9318 delay slot scheduler to place instructions in the delay slot of @var{insn},
9319 even if they appear to set or clobber a resource referenced in @var{insn}.
9320 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
9321 some @code{insn} or @code{jump_insn} is really a function call and its operands
9322 are registers whose use is actually in the subroutine it calls, you should
9323 define this macro. Doing so allows the delay slot scheduler to move
9324 instructions which copy arguments into the argument registers into the delay
9327 You need not define this macro if it would always return zero.
9330 @defmac MULTIPLE_SYMBOL_SPACES
9331 Define this macro as a C expression that is nonzero if, in some cases,
9332 global symbols from one translation unit may not be bound to undefined
9333 symbols in another translation unit without user intervention. For
9334 instance, under Microsoft Windows symbols must be explicitly imported
9335 from shared libraries (DLLs).
9337 You need not define this macro if it would always evaluate to zero.
9340 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
9341 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
9342 any hard regs the port wishes to automatically clobber for an asm.
9343 It should return the result of the last @code{tree_cons} used to add a
9344 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
9345 corresponding parameters to the asm and may be inspected to avoid
9346 clobbering a register that is an input or output of the asm. You can use
9347 @code{decl_overlaps_hard_reg_set_p}, declared in @file{tree.h}, to test
9348 for overlap with regards to asm-declared registers.
9351 @defmac MATH_LIBRARY
9352 Define this macro as a C string constant for the linker argument to link
9353 in the system math library, or @samp{""} if the target does not have a
9354 separate math library.
9356 You need only define this macro if the default of @samp{"-lm"} is wrong.
9359 @defmac LIBRARY_PATH_ENV
9360 Define this macro as a C string constant for the environment variable that
9361 specifies where the linker should look for libraries.
9363 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9367 @defmac TARGET_HAS_F_SETLKW
9368 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
9369 Note that this functionality is part of POSIX@.
9370 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
9371 to use file locking when exiting a program, which avoids race conditions
9372 if the program has forked.
9375 @defmac MAX_CONDITIONAL_EXECUTE
9377 A C expression for the maximum number of instructions to execute via
9378 conditional execution instructions instead of a branch. A value of
9379 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9380 1 if it does use cc0.
9383 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
9384 Used if the target needs to perform machine-dependent modifications on the
9385 conditionals used for turning basic blocks into conditionally executed code.
9386 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9387 contains information about the currently processed blocks. @var{true_expr}
9388 and @var{false_expr} are the tests that are used for converting the
9389 then-block and the else-block, respectively. Set either @var{true_expr} or
9390 @var{false_expr} to a null pointer if the tests cannot be converted.
9393 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9394 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9395 if-statements into conditions combined by @code{and} and @code{or} operations.
9396 @var{bb} contains the basic block that contains the test that is currently
9397 being processed and about to be turned into a condition.
9400 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
9401 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9402 be converted to conditional execution format. @var{ce_info} points to
9403 a data structure, @code{struct ce_if_block}, which contains information
9404 about the currently processed blocks.
9407 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
9408 A C expression to perform any final machine dependent modifications in
9409 converting code to conditional execution. The involved basic blocks
9410 can be found in the @code{struct ce_if_block} structure that is pointed
9411 to by @var{ce_info}.
9414 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
9415 A C expression to cancel any machine dependent modifications in
9416 converting code to conditional execution. The involved basic blocks
9417 can be found in the @code{struct ce_if_block} structure that is pointed
9418 to by @var{ce_info}.
9421 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
9422 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9423 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9426 @defmac IFCVT_EXTRA_FIELDS
9427 If defined, it should expand to a set of field declarations that will be
9428 added to the @code{struct ce_if_block} structure. These should be initialized
9429 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9432 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
9433 If non-null, this hook performs a target-specific pass over the
9434 instruction stream. The compiler will run it at all optimization levels,
9435 just before the point at which it normally does delayed-branch scheduling.
9437 The exact purpose of the hook varies from target to target. Some use
9438 it to do transformations that are necessary for correctness, such as
9439 laying out in-function constant pools or avoiding hardware hazards.
9440 Others use it as an opportunity to do some machine-dependent optimizations.
9442 You need not implement the hook if it has nothing to do. The default
9446 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9447 Define this hook if you have any machine-specific built-in functions
9448 that need to be defined. It should be a function that performs the
9451 Machine specific built-in functions can be useful to expand special machine
9452 instructions that would otherwise not normally be generated because
9453 they have no equivalent in the source language (for example, SIMD vector
9454 instructions or prefetch instructions).
9456 To create a built-in function, call the function
9457 @code{lang_hooks.builtin_function}
9458 which is defined by the language front end. You can use any type nodes set
9459 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9460 only language front ends that use those two functions will call
9461 @samp{TARGET_INIT_BUILTINS}.
9464 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9466 Expand a call to a machine specific built-in function that was set up by
9467 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9468 function call; the result should go to @var{target} if that is
9469 convenient, and have mode @var{mode} if that is convenient.
9470 @var{subtarget} may be used as the target for computing one of
9471 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9472 ignored. This function should return the result of the call to the
9476 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, tree @var{arglist}, bool @var{ignore})
9478 Fold a call to a machine specific built-in function that was set up by
9479 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
9480 built-in function. @var{arglist} is the list of arguments passed to
9481 the built-in function. The result is another tree containing a
9482 simplified expression for the call's result. If @var{ignore} is true
9483 the value will be ignored.
9486 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
9488 Take a branch insn in @var{branch1} and another in @var{branch2}.
9489 Return true if redirecting @var{branch1} to the destination of
9490 @var{branch2} is possible.
9492 On some targets, branches may have a limited range. Optimizing the
9493 filling of delay slots can result in branches being redirected, and this
9494 may in turn cause a branch offset to overflow.
9497 @defmac ALLOCATE_INITIAL_VALUE (@var{hard_reg})
9499 When the initial value of a hard register has been copied in a pseudo
9500 register, it is often not necessary to actually allocate another register
9501 to this pseudo register, because the original hard register or a stack slot
9502 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9503 defined, is called at the start of register allocation once for each
9504 hard register that had its initial value copied by using
9505 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9506 Possible values are @code{NULL_RTX}, if you don't want
9507 to do any special allocation, a @code{REG} rtx---that would typically be
9508 the hard register itself, if it is known not to be clobbered---or a
9510 If you are returning a @code{MEM}, this is only a hint for the allocator;
9511 it might decide to use another register anyways.
9512 You may use @code{current_function_leaf_function} in the definition of the
9513 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9514 register in question will not be clobbered.
9517 @defmac TARGET_OBJECT_SUFFIX
9518 Define this macro to be a C string representing the suffix for object
9519 files on your target machine. If you do not define this macro, GCC will
9520 use @samp{.o} as the suffix for object files.
9523 @defmac TARGET_EXECUTABLE_SUFFIX
9524 Define this macro to be a C string representing the suffix to be
9525 automatically added to executable files on your target machine. If you
9526 do not define this macro, GCC will use the null string as the suffix for
9530 @defmac COLLECT_EXPORT_LIST
9531 If defined, @code{collect2} will scan the individual object files
9532 specified on its command line and create an export list for the linker.
9533 Define this macro for systems like AIX, where the linker discards
9534 object files that are not referenced from @code{main} and uses export
9538 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
9539 Define this macro to a C expression representing a variant of the
9540 method call @var{mdecl}, if Java Native Interface (JNI) methods
9541 must be invoked differently from other methods on your target.
9542 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
9543 the @code{stdcall} calling convention and this macro is then
9544 defined as this expression:
9547 build_type_attribute_variant (@var{mdecl},
9549 (get_identifier ("stdcall"),
9554 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9555 This target hook returns @code{true} past the point in which new jump
9556 instructions could be created. On machines that require a register for
9557 every jump such as the SHmedia ISA of SH5, this point would typically be
9558 reload, so this target hook should be defined to a function such as:
9562 cannot_modify_jumps_past_reload_p ()
9564 return (reload_completed || reload_in_progress);
9569 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
9570 This target hook returns a register class for which branch target register
9571 optimizations should be applied. All registers in this class should be
9572 usable interchangeably. After reload, registers in this class will be
9573 re-allocated and loads will be hoisted out of loops and be subjected
9574 to inter-block scheduling.
9577 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
9578 Branch target register optimization will by default exclude callee-saved
9580 that are not already live during the current function; if this target hook
9581 returns true, they will be included. The target code must than make sure
9582 that all target registers in the class returned by
9583 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
9584 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
9585 epilogues have already been generated. Note, even if you only return
9586 true when @var{after_prologue_epilogue_gen} is false, you still are likely
9587 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
9588 to reserve space for caller-saved target registers.
9591 @defmac POWI_MAX_MULTS
9592 If defined, this macro is interpreted as a signed integer C expression
9593 that specifies the maximum number of floating point multiplications
9594 that should be emitted when expanding exponentiation by an integer
9595 constant inline. When this value is defined, exponentiation requiring
9596 more than this number of multiplications is implemented by calling the
9597 system library's @code{pow}, @code{powf} or @code{powl} routines.
9598 The default value places no upper bound on the multiplication count.
9601 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
9602 This target hook should register any extra include files for the
9603 target. The parameter @var{stdinc} indicates if normal include files
9604 are present. The parameter @var{sysroot} is the system root directory.
9605 The parameter @var{iprefix} is the prefix for the gcc directory.
9608 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
9609 This target hook should register any extra include files for the
9610 target before any standard headers. The parameter @var{stdinc}
9611 indicates if normal include files are present. The parameter
9612 @var{sysroot} is the system root directory. The parameter
9613 @var{iprefix} is the prefix for the gcc directory.
9616 @deftypefn Macro void TARGET_OPTF (char *@var{path})
9617 This target hook should register special include paths for the target.
9618 The parameter @var{path} is the include to register. On Darwin
9619 systems, this is used for Framework includes, which have semantics
9620 that are different from @option{-I}.
9623 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
9624 This target hook returns @code{true} if it is safe to use a local alias
9625 for a virtual function @var{fndecl} when constructing thunks,
9626 @code{false} otherwise. By default, the hook returns @code{true} for all
9627 functions, if a target supports aliases (i.e.@: defines
9628 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
9631 @defmac TARGET_FORMAT_TYPES
9632 If defined, this macro is the name of a global variable containing
9633 target-specific format checking information for the @option{-Wformat}
9634 option. The default is to have no target-specific format checks.
9637 @defmac TARGET_N_FORMAT_TYPES
9638 If defined, this macro is the number of entries in
9639 @code{TARGET_FORMAT_TYPES}.
9642 @deftypefn {Target Hook} bool TARGET_RELAXED_ORDERING
9643 If set to @code{true}, means that the target's memory model does not
9644 guarantee that loads which do not depend on one another will access
9645 main memory in the order of the instruction stream; if ordering is
9646 important, an explicit memory barrier must be used. This is true of
9647 many recent processors which implement a policy of ``relaxed,''
9648 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
9649 and ia64. The default is @code{false}.
9652 @deftypefn {Target Hook} const char *TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (tree @var{typelist}, tree @var{funcdecl}, tree @var{val})
9653 If defined, this macro returns the diagnostic message when it is
9654 illegal to pass argument @var{val} to function @var{funcdecl}
9655 with prototype @var{typelist}.
9658 @defmac TARGET_USE_JCR_SECTION
9659 This macro determines whether to use the JCR section to register Java
9660 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
9661 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.