1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Target Description Macros and Functions
9 @cindex machine description macros
10 @cindex target description macros
11 @cindex macros, target description
12 @cindex @file{tm.h} macros
14 In addition to the file @file{@var{machine}.md}, a machine description
15 includes a C header file conventionally given the name
16 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
17 The header file defines numerous macros that convey the information
18 about the target machine that does not fit into the scheme of the
19 @file{.md} file. The file @file{tm.h} should be a link to
20 @file{@var{machine}.h}. The header file @file{config.h} includes
21 @file{tm.h} and most compiler source files include @file{config.h}. The
22 source file defines a variable @code{targetm}, which is a structure
23 containing pointers to functions and data relating to the target
24 machine. @file{@var{machine}.c} should also contain their definitions,
25 if they are not defined elsewhere in GCC, and other functions called
26 through the macros defined in the @file{.h} file.
29 * Target Structure:: The @code{targetm} variable.
30 * Driver:: Controlling how the driver runs the compilation passes.
31 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
32 * Per-Function Data:: Defining data structures for per-function information.
33 * Storage Layout:: Defining sizes and alignments of data.
34 * Type Layout:: Defining sizes and properties of basic user data types.
35 * Registers:: Naming and describing the hardware registers.
36 * Register Classes:: Defining the classes of hardware registers.
37 * Old Constraints:: The old way to define machine-specific constraints.
38 * Stack and Calling:: Defining which way the stack grows and by how much.
39 * Varargs:: Defining the varargs macros.
40 * Trampolines:: Code set up at run time to enter a nested function.
41 * Library Calls:: Controlling how library routines are implicitly called.
42 * Addressing Modes:: Defining addressing modes valid for memory operands.
43 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
44 * Condition Code:: Defining how insns update the condition code.
45 * Costs:: Defining relative costs of different operations.
46 * Scheduling:: Adjusting the behavior of the instruction scheduler.
47 * Sections:: Dividing storage into text, data, and other sections.
48 * PIC:: Macros for position independent code.
49 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
50 * Debugging Info:: Defining the format of debugging output.
51 * Floating Point:: Handling floating point for cross-compilers.
52 * Mode Switching:: Insertion of mode-switching instructions.
53 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
54 * Emulated TLS:: Emulated TLS support.
55 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
56 * PCH Target:: Validity checking for precompiled headers.
57 * C++ ABI:: Controlling C++ ABI changes.
58 * Named Address Spaces:: Adding support for named address spaces
59 * Misc:: Everything else.
62 @node Target Structure
63 @section The Global @code{targetm} Variable
65 @cindex target functions
67 @deftypevar {struct gcc_target} targetm
68 The target @file{.c} file must define the global @code{targetm} variable
69 which contains pointers to functions and data relating to the target
70 machine. The variable is declared in @file{target.h};
71 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
72 used to initialize the variable, and macros for the default initializers
73 for elements of the structure. The @file{.c} file should override those
74 macros for which the default definition is inappropriate. For example:
77 #include "target-def.h"
79 /* @r{Initialize the GCC target structure.} */
81 #undef TARGET_COMP_TYPE_ATTRIBUTES
82 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
84 struct gcc_target targetm = TARGET_INITIALIZER;
88 Where a macro should be defined in the @file{.c} file in this manner to
89 form part of the @code{targetm} structure, it is documented below as a
90 ``Target Hook'' with a prototype. Many macros will change in future
91 from being defined in the @file{.h} file to being part of the
92 @code{targetm} structure.
95 @section Controlling the Compilation Driver, @file{gcc}
97 @cindex controlling the compilation driver
99 @c prevent bad page break with this line
100 You can control the compilation driver.
102 @defmac SWITCH_TAKES_ARG (@var{char})
103 A C expression which determines whether the option @option{-@var{char}}
104 takes arguments. The value should be the number of arguments that
105 option takes--zero, for many options.
107 By default, this macro is defined as
108 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
109 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
110 wish to add additional options which take arguments. Any redefinition
111 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
115 @defmac WORD_SWITCH_TAKES_ARG (@var{name})
116 A C expression which determines whether the option @option{-@var{name}}
117 takes arguments. The value should be the number of arguments that
118 option takes--zero, for many options. This macro rather than
119 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
121 By default, this macro is defined as
122 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
123 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
124 wish to add additional options which take arguments. Any redefinition
125 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
129 @defmac TARGET_OPTION_TRANSLATE_TABLE
130 If defined, a list of pairs of strings, the first of which is a
131 potential command line target to the @file{gcc} driver program, and the
132 second of which is a space-separated (tabs and other whitespace are not
133 supported) list of options with which to replace the first option. The
134 target defining this list is responsible for assuring that the results
135 are valid. Replacement options may not be the @code{--opt} style, they
136 must be the @code{-opt} style. It is the intention of this macro to
137 provide a mechanism for substitution that affects the multilibs chosen,
138 such as one option that enables many options, some of which select
139 multilibs. Example nonsensical definition, where @option{-malt-abi},
140 @option{-EB}, and @option{-mspoo} cause different multilibs to be chosen:
143 #define TARGET_OPTION_TRANSLATE_TABLE \
144 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
145 @{ "-compat", "-EB -malign=4 -mspoo" @}
149 @defmac DRIVER_SELF_SPECS
150 A list of specs for the driver itself. It should be a suitable
151 initializer for an array of strings, with no surrounding braces.
153 The driver applies these specs to its own command line between loading
154 default @file{specs} files (but not command-line specified ones) and
155 choosing the multilib directory or running any subcommands. It
156 applies them in the order given, so each spec can depend on the
157 options added by earlier ones. It is also possible to remove options
158 using @samp{%<@var{option}} in the usual way.
160 This macro can be useful when a port has several interdependent target
161 options. It provides a way of standardizing the command line so
162 that the other specs are easier to write.
164 Do not define this macro if it does not need to do anything.
167 @defmac OPTION_DEFAULT_SPECS
168 A list of specs used to support configure-time default options (i.e.@:
169 @option{--with} options) in the driver. It should be a suitable initializer
170 for an array of structures, each containing two strings, without the
171 outermost pair of surrounding braces.
173 The first item in the pair is the name of the default. This must match
174 the code in @file{config.gcc} for the target. The second item is a spec
175 to apply if a default with this name was specified. The string
176 @samp{%(VALUE)} in the spec will be replaced by the value of the default
177 everywhere it occurs.
179 The driver will apply these specs to its own command line between loading
180 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
181 the same mechanism as @code{DRIVER_SELF_SPECS}.
183 Do not define this macro if it does not need to do anything.
187 A C string constant that tells the GCC driver program options to
188 pass to CPP@. It can also specify how to translate options you
189 give to GCC into options for GCC to pass to the CPP@.
191 Do not define this macro if it does not need to do anything.
194 @defmac CPLUSPLUS_CPP_SPEC
195 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
196 than C@. If you do not define this macro, then the value of
197 @code{CPP_SPEC} (if any) will be used instead.
201 A C string constant that tells the GCC driver program options to
202 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
204 It can also specify how to translate options you give to GCC into options
205 for GCC to pass to front ends.
207 Do not define this macro if it does not need to do anything.
211 A C string constant that tells the GCC driver program options to
212 pass to @code{cc1plus}. It can also specify how to translate options you
213 give to GCC into options for GCC to pass to the @code{cc1plus}.
215 Do not define this macro if it does not need to do anything.
216 Note that everything defined in CC1_SPEC is already passed to
217 @code{cc1plus} so there is no need to duplicate the contents of
218 CC1_SPEC in CC1PLUS_SPEC@.
222 A C string constant that tells the GCC driver program options to
223 pass to the assembler. It can also specify how to translate options
224 you give to GCC into options for GCC to pass to the assembler.
225 See the file @file{sun3.h} for an example of this.
227 Do not define this macro if it does not need to do anything.
230 @defmac ASM_FINAL_SPEC
231 A C string constant that tells the GCC driver program how to
232 run any programs which cleanup after the normal assembler.
233 Normally, this is not needed. See the file @file{mips.h} for
236 Do not define this macro if it does not need to do anything.
239 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
240 Define this macro, with no value, if the driver should give the assembler
241 an argument consisting of a single dash, @option{-}, to instruct it to
242 read from its standard input (which will be a pipe connected to the
243 output of the compiler proper). This argument is given after any
244 @option{-o} option specifying the name of the output file.
246 If you do not define this macro, the assembler is assumed to read its
247 standard input if given no non-option arguments. If your assembler
248 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
249 see @file{mips.h} for instance.
253 A C string constant that tells the GCC driver program options to
254 pass to the linker. It can also specify how to translate options you
255 give to GCC into options for GCC to pass to the linker.
257 Do not define this macro if it does not need to do anything.
261 Another C string constant used much like @code{LINK_SPEC}. The difference
262 between the two is that @code{LIB_SPEC} is used at the end of the
263 command given to the linker.
265 If this macro is not defined, a default is provided that
266 loads the standard C library from the usual place. See @file{gcc.c}.
270 Another C string constant that tells the GCC driver program
271 how and when to place a reference to @file{libgcc.a} into the
272 linker command line. This constant is placed both before and after
273 the value of @code{LIB_SPEC}.
275 If this macro is not defined, the GCC driver provides a default that
276 passes the string @option{-lgcc} to the linker.
279 @defmac REAL_LIBGCC_SPEC
280 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
281 @code{LIBGCC_SPEC} is not directly used by the driver program but is
282 instead modified to refer to different versions of @file{libgcc.a}
283 depending on the values of the command line flags @option{-static},
284 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
285 targets where these modifications are inappropriate, define
286 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
287 driver how to place a reference to @file{libgcc} on the link command
288 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
291 @defmac USE_LD_AS_NEEDED
292 A macro that controls the modifications to @code{LIBGCC_SPEC}
293 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
294 generated that uses --as-needed and the shared libgcc in place of the
295 static exception handler library, when linking without any of
296 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
300 If defined, this C string constant is added to @code{LINK_SPEC}.
301 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
302 the modifications to @code{LIBGCC_SPEC} mentioned in
303 @code{REAL_LIBGCC_SPEC}.
306 @defmac STARTFILE_SPEC
307 Another C string constant used much like @code{LINK_SPEC}. The
308 difference between the two is that @code{STARTFILE_SPEC} is used at
309 the very beginning of the command given to the linker.
311 If this macro is not defined, a default is provided that loads the
312 standard C startup file from the usual place. See @file{gcc.c}.
316 Another C string constant used much like @code{LINK_SPEC}. The
317 difference between the two is that @code{ENDFILE_SPEC} is used at
318 the very end of the command given to the linker.
320 Do not define this macro if it does not need to do anything.
323 @defmac THREAD_MODEL_SPEC
324 GCC @code{-v} will print the thread model GCC was configured to use.
325 However, this doesn't work on platforms that are multilibbed on thread
326 models, such as AIX 4.3. On such platforms, define
327 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
328 blanks that names one of the recognized thread models. @code{%*}, the
329 default value of this macro, will expand to the value of
330 @code{thread_file} set in @file{config.gcc}.
333 @defmac SYSROOT_SUFFIX_SPEC
334 Define this macro to add a suffix to the target sysroot when GCC is
335 configured with a sysroot. This will cause GCC to search for usr/lib,
336 et al, within sysroot+suffix.
339 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
340 Define this macro to add a headers_suffix to the target sysroot when
341 GCC is configured with a sysroot. This will cause GCC to pass the
342 updated sysroot+headers_suffix to CPP, causing it to search for
343 usr/include, et al, within sysroot+headers_suffix.
347 Define this macro to provide additional specifications to put in the
348 @file{specs} file that can be used in various specifications like
351 The definition should be an initializer for an array of structures,
352 containing a string constant, that defines the specification name, and a
353 string constant that provides the specification.
355 Do not define this macro if it does not need to do anything.
357 @code{EXTRA_SPECS} is useful when an architecture contains several
358 related targets, which have various @code{@dots{}_SPECS} which are similar
359 to each other, and the maintainer would like one central place to keep
362 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
363 define either @code{_CALL_SYSV} when the System V calling sequence is
364 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
367 The @file{config/rs6000/rs6000.h} target file defines:
370 #define EXTRA_SPECS \
371 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
373 #define CPP_SYS_DEFAULT ""
376 The @file{config/rs6000/sysv.h} target file defines:
380 "%@{posix: -D_POSIX_SOURCE @} \
381 %@{mcall-sysv: -D_CALL_SYSV @} \
382 %@{!mcall-sysv: %(cpp_sysv_default) @} \
383 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
385 #undef CPP_SYSV_DEFAULT
386 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
389 while the @file{config/rs6000/eabiaix.h} target file defines
390 @code{CPP_SYSV_DEFAULT} as:
393 #undef CPP_SYSV_DEFAULT
394 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
398 @defmac LINK_LIBGCC_SPECIAL_1
399 Define this macro if the driver program should find the library
400 @file{libgcc.a}. If you do not define this macro, the driver program will pass
401 the argument @option{-lgcc} to tell the linker to do the search.
404 @defmac LINK_GCC_C_SEQUENCE_SPEC
405 The sequence in which libgcc and libc are specified to the linker.
406 By default this is @code{%G %L %G}.
409 @defmac LINK_COMMAND_SPEC
410 A C string constant giving the complete command line need to execute the
411 linker. When you do this, you will need to update your port each time a
412 change is made to the link command line within @file{gcc.c}. Therefore,
413 define this macro only if you need to completely redefine the command
414 line for invoking the linker and there is no other way to accomplish
415 the effect you need. Overriding this macro may be avoidable by overriding
416 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
419 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
420 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
421 directories from linking commands. Do not give it a nonzero value if
422 removing duplicate search directories changes the linker's semantics.
425 @defmac MULTILIB_DEFAULTS
426 Define this macro as a C expression for the initializer of an array of
427 string to tell the driver program which options are defaults for this
428 target and thus do not need to be handled specially when using
429 @code{MULTILIB_OPTIONS}.
431 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
432 the target makefile fragment or if none of the options listed in
433 @code{MULTILIB_OPTIONS} are set by default.
434 @xref{Target Fragment}.
437 @defmac RELATIVE_PREFIX_NOT_LINKDIR
438 Define this macro to tell @command{gcc} that it should only translate
439 a @option{-B} prefix into a @option{-L} linker option if the prefix
440 indicates an absolute file name.
443 @defmac MD_EXEC_PREFIX
444 If defined, this macro is an additional prefix to try after
445 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
446 when the compiler is built as a cross
447 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
448 to the list of directories used to find the assembler in @file{configure.in}.
451 @defmac STANDARD_STARTFILE_PREFIX
452 Define this macro as a C string constant if you wish to override the
453 standard choice of @code{libdir} as the default prefix to
454 try when searching for startup files such as @file{crt0.o}.
455 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
456 is built as a cross compiler.
459 @defmac STANDARD_STARTFILE_PREFIX_1
460 Define this macro as a C string constant if you wish to override the
461 standard choice of @code{/lib} as a prefix to try after the default prefix
462 when searching for startup files such as @file{crt0.o}.
463 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
464 is built as a cross compiler.
467 @defmac STANDARD_STARTFILE_PREFIX_2
468 Define this macro as a C string constant if you wish to override the
469 standard choice of @code{/lib} as yet another prefix to try after the
470 default prefix when searching for startup files such as @file{crt0.o}.
471 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
472 is built as a cross compiler.
475 @defmac MD_STARTFILE_PREFIX
476 If defined, this macro supplies an additional prefix to try after the
477 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
478 compiler is built as a cross compiler.
481 @defmac MD_STARTFILE_PREFIX_1
482 If defined, this macro supplies yet another prefix to try after the
483 standard prefixes. It is not searched when the compiler is built as a
487 @defmac INIT_ENVIRONMENT
488 Define this macro as a C string constant if you wish to set environment
489 variables for programs called by the driver, such as the assembler and
490 loader. The driver passes the value of this macro to @code{putenv} to
491 initialize the necessary environment variables.
494 @defmac LOCAL_INCLUDE_DIR
495 Define this macro as a C string constant if you wish to override the
496 standard choice of @file{/usr/local/include} as the default prefix to
497 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
498 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
500 Cross compilers do not search either @file{/usr/local/include} or its
504 @defmac SYSTEM_INCLUDE_DIR
505 Define this macro as a C string constant if you wish to specify a
506 system-specific directory to search for header files before the standard
507 directory. @code{SYSTEM_INCLUDE_DIR} comes before
508 @code{STANDARD_INCLUDE_DIR} in the search order.
510 Cross compilers do not use this macro and do not search the directory
514 @defmac STANDARD_INCLUDE_DIR
515 Define this macro as a C string constant if you wish to override the
516 standard choice of @file{/usr/include} as the default prefix to
517 try when searching for header files.
519 Cross compilers ignore this macro and do not search either
520 @file{/usr/include} or its replacement.
523 @defmac STANDARD_INCLUDE_COMPONENT
524 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
525 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
526 If you do not define this macro, no component is used.
529 @defmac INCLUDE_DEFAULTS
530 Define this macro if you wish to override the entire default search path
531 for include files. For a native compiler, the default search path
532 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
533 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
534 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
535 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
536 and specify private search areas for GCC@. The directory
537 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
539 The definition should be an initializer for an array of structures.
540 Each array element should have four elements: the directory name (a
541 string constant), the component name (also a string constant), a flag
542 for C++-only directories,
543 and a flag showing that the includes in the directory don't need to be
544 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
545 the array with a null element.
547 The component name denotes what GNU package the include file is part of,
548 if any, in all uppercase letters. For example, it might be @samp{GCC}
549 or @samp{BINUTILS}. If the package is part of a vendor-supplied
550 operating system, code the component name as @samp{0}.
552 For example, here is the definition used for VAX/VMS:
555 #define INCLUDE_DEFAULTS \
557 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
558 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
559 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
566 Here is the order of prefixes tried for exec files:
570 Any prefixes specified by the user with @option{-B}.
573 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
574 is not set and the compiler has not been installed in the configure-time
575 @var{prefix}, the location in which the compiler has actually been installed.
578 The directories specified by the environment variable @code{COMPILER_PATH}.
581 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
582 in the configured-time @var{prefix}.
585 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
588 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
591 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
595 Here is the order of prefixes tried for startfiles:
599 Any prefixes specified by the user with @option{-B}.
602 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
603 value based on the installed toolchain location.
606 The directories specified by the environment variable @code{LIBRARY_PATH}
607 (or port-specific name; native only, cross compilers do not use this).
610 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
611 in the configured @var{prefix} or this is a native compiler.
614 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
617 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
621 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
622 native compiler, or we have a target system root.
625 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
626 native compiler, or we have a target system root.
629 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
630 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
631 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
634 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
635 compiler, or we have a target system root. The default for this macro is
639 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
640 compiler, or we have a target system root. The default for this macro is
644 @node Run-time Target
645 @section Run-time Target Specification
646 @cindex run-time target specification
647 @cindex predefined macros
648 @cindex target specifications
650 @c prevent bad page break with this line
651 Here are run-time target specifications.
653 @defmac TARGET_CPU_CPP_BUILTINS ()
654 This function-like macro expands to a block of code that defines
655 built-in preprocessor macros and assertions for the target CPU, using
656 the functions @code{builtin_define}, @code{builtin_define_std} and
657 @code{builtin_assert}. When the front end
658 calls this macro it provides a trailing semicolon, and since it has
659 finished command line option processing your code can use those
662 @code{builtin_assert} takes a string in the form you pass to the
663 command-line option @option{-A}, such as @code{cpu=mips}, and creates
664 the assertion. @code{builtin_define} takes a string in the form
665 accepted by option @option{-D} and unconditionally defines the macro.
667 @code{builtin_define_std} takes a string representing the name of an
668 object-like macro. If it doesn't lie in the user's namespace,
669 @code{builtin_define_std} defines it unconditionally. Otherwise, it
670 defines a version with two leading underscores, and another version
671 with two leading and trailing underscores, and defines the original
672 only if an ISO standard was not requested on the command line. For
673 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
674 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
675 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
676 defines only @code{_ABI64}.
678 You can also test for the C dialect being compiled. The variable
679 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
680 or @code{clk_objective_c}. Note that if we are preprocessing
681 assembler, this variable will be @code{clk_c} but the function-like
682 macro @code{preprocessing_asm_p()} will return true, so you might want
683 to check for that first. If you need to check for strict ANSI, the
684 variable @code{flag_iso} can be used. The function-like macro
685 @code{preprocessing_trad_p()} can be used to check for traditional
689 @defmac TARGET_OS_CPP_BUILTINS ()
690 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
691 and is used for the target operating system instead.
694 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
695 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
696 and is used for the target object format. @file{elfos.h} uses this
697 macro to define @code{__ELF__}, so you probably do not need to define
701 @deftypevar {extern int} target_flags
702 This variable is declared in @file{options.h}, which is included before
703 any target-specific headers.
706 @deftypevr {Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
707 This variable specifies the initial value of @code{target_flags}.
708 Its default setting is 0.
711 @cindex optional hardware or system features
712 @cindex features, optional, in system conventions
714 @deftypefn {Target Hook} bool TARGET_HANDLE_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
715 This hook is called whenever the user specifies one of the
716 target-specific options described by the @file{.opt} definition files
717 (@pxref{Options}). It has the opportunity to do some option-specific
718 processing and should return true if the option is valid. The default
719 definition does nothing but return true.
721 @var{code} specifies the @code{OPT_@var{name}} enumeration value
722 associated with the selected option; @var{name} is just a rendering of
723 the option name in which non-alphanumeric characters are replaced by
724 underscores. @var{arg} specifies the string argument and is null if
725 no argument was given. If the option is flagged as a @code{UInteger}
726 (@pxref{Option properties}), @var{value} is the numeric value of the
727 argument. Otherwise @var{value} is 1 if the positive form of the
728 option was used and 0 if the ``no-'' form was.
731 @deftypefn {Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
732 This target hook is called whenever the user specifies one of the
733 target-specific C language family options described by the @file{.opt}
734 definition files(@pxref{Options}). It has the opportunity to do some
735 option-specific processing and should return true if the option is
736 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
737 default definition does nothing but return false.
739 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
740 options. However, if processing an option requires routines that are
741 only available in the C (and related language) front ends, then you
742 should use @code{TARGET_HANDLE_C_OPTION} instead.
745 @deftypefn {Target Hook} tree TARGET_OBJC_CONSTRUCT_STRING (tree @var{string})
746 Construct a constant string representation for @var{string}
749 @defmac TARGET_VERSION
750 This macro is a C statement to print on @code{stderr} a string
751 describing the particular machine description choice. Every machine
752 description should define @code{TARGET_VERSION}. For example:
756 #define TARGET_VERSION \
757 fprintf (stderr, " (68k, Motorola syntax)");
759 #define TARGET_VERSION \
760 fprintf (stderr, " (68k, MIT syntax)");
765 @deftypefn {Target Hook} void TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE (void)
766 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
767 but is called when the optimize level is changed via an attribute or
768 pragma or when it is reset at the end of the code affected by the
769 attribute or pragma. It is not called at the beginning of compilation
770 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
771 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
772 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
775 @defmac C_COMMON_OVERRIDE_OPTIONS
776 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
777 but is only used in the C
778 language frontends (C, Objective-C, C++, Objective-C++) and so can be
779 used to alter option flag variables which only exist in those
783 @deftypevr {Target Hook} {const struct default_options *} TARGET_OPTION_OPTIMIZATION_TABLE
784 Some machines may desire to change what optimizations are performed for
785 various optimization levels. This variable, if defined, describes
786 options to enable at particular sets of optimization levels. These
787 options are processed once
788 just after the optimization level is determined and before the remainder
789 of the command options have been parsed, so may be overridden by other
790 options passed explicily.
792 This processing is run once at program startup and when the optimization
793 options are changed via @code{#pragma GCC optimize} or by using the
794 @code{optimize} attribute.
797 @deftypefn {Target Hook} void TARGET_OPTION_INIT_STRUCT (struct gcc_options *@var{opts})
798 Set target-dependent initial values of fields in @var{opts}.
801 @deftypefn {Target Hook} void TARGET_OPTION_DEFAULT_PARAMS (void)
802 Set target-dependent default values for @option{--param} settings, using calls to @code{set_default_param_value}.
805 @deftypefn {Target Hook} void TARGET_HELP (void)
806 This hook is called in response to the user invoking
807 @option{--target-help} on the command line. It gives the target a
808 chance to display extra information on the target specific command
809 line options found in its @file{.opt} file.
812 @defmac SWITCHABLE_TARGET
813 Some targets need to switch between substantially different subtargets
814 during compilation. For example, the MIPS target has one subtarget for
815 the traditional MIPS architecture and another for MIPS16. Source code
816 can switch between these two subarchitectures using the @code{mips16}
817 and @code{nomips16} attributes.
819 Such subtargets can differ in things like the set of available
820 registers, the set of available instructions, the costs of various
821 operations, and so on. GCC caches a lot of this type of information
822 in global variables, and recomputing them for each subtarget takes a
823 significant amount of time. The compiler therefore provides a facility
824 for maintaining several versions of the global variables and quickly
825 switching between them; see @file{target-globals.h} for details.
827 Define this macro to 1 if your target needs this facility. The default
831 @node Per-Function Data
832 @section Defining data structures for per-function information.
833 @cindex per-function data
834 @cindex data structures
836 If the target needs to store information on a per-function basis, GCC
837 provides a macro and a couple of variables to allow this. Note, just
838 using statics to store the information is a bad idea, since GCC supports
839 nested functions, so you can be halfway through encoding one function
840 when another one comes along.
842 GCC defines a data structure called @code{struct function} which
843 contains all of the data specific to an individual function. This
844 structure contains a field called @code{machine} whose type is
845 @code{struct machine_function *}, which can be used by targets to point
846 to their own specific data.
848 If a target needs per-function specific data it should define the type
849 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
850 This macro should be used to initialize the function pointer
851 @code{init_machine_status}. This pointer is explained below.
853 One typical use of per-function, target specific data is to create an
854 RTX to hold the register containing the function's return address. This
855 RTX can then be used to implement the @code{__builtin_return_address}
856 function, for level 0.
858 Note---earlier implementations of GCC used a single data area to hold
859 all of the per-function information. Thus when processing of a nested
860 function began the old per-function data had to be pushed onto a
861 stack, and when the processing was finished, it had to be popped off the
862 stack. GCC used to provide function pointers called
863 @code{save_machine_status} and @code{restore_machine_status} to handle
864 the saving and restoring of the target specific information. Since the
865 single data area approach is no longer used, these pointers are no
868 @defmac INIT_EXPANDERS
869 Macro called to initialize any target specific information. This macro
870 is called once per function, before generation of any RTL has begun.
871 The intention of this macro is to allow the initialization of the
872 function pointer @code{init_machine_status}.
875 @deftypevar {void (*)(struct function *)} init_machine_status
876 If this function pointer is non-@code{NULL} it will be called once per
877 function, before function compilation starts, in order to allow the
878 target to perform any target specific initialization of the
879 @code{struct function} structure. It is intended that this would be
880 used to initialize the @code{machine} of that structure.
882 @code{struct machine_function} structures are expected to be freed by GC@.
883 Generally, any memory that they reference must be allocated by using
884 GC allocation, including the structure itself.
888 @section Storage Layout
889 @cindex storage layout
891 Note that the definitions of the macros in this table which are sizes or
892 alignments measured in bits do not need to be constant. They can be C
893 expressions that refer to static variables, such as the @code{target_flags}.
894 @xref{Run-time Target}.
896 @defmac BITS_BIG_ENDIAN
897 Define this macro to have the value 1 if the most significant bit in a
898 byte has the lowest number; otherwise define it to have the value zero.
899 This means that bit-field instructions count from the most significant
900 bit. If the machine has no bit-field instructions, then this must still
901 be defined, but it doesn't matter which value it is defined to. This
902 macro need not be a constant.
904 This macro does not affect the way structure fields are packed into
905 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
908 @defmac BYTES_BIG_ENDIAN
909 Define this macro to have the value 1 if the most significant byte in a
910 word has the lowest number. This macro need not be a constant.
913 @defmac WORDS_BIG_ENDIAN
914 Define this macro to have the value 1 if, in a multiword object, the
915 most significant word has the lowest number. This applies to both
916 memory locations and registers; GCC fundamentally assumes that the
917 order of words in memory is the same as the order in registers. This
918 macro need not be a constant.
921 @defmac FLOAT_WORDS_BIG_ENDIAN
922 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
923 @code{TFmode} floating point numbers are stored in memory with the word
924 containing the sign bit at the lowest address; otherwise define it to
925 have the value 0. This macro need not be a constant.
927 You need not define this macro if the ordering is the same as for
931 @defmac BITS_PER_UNIT
932 Define this macro to be the number of bits in an addressable storage
933 unit (byte). If you do not define this macro the default is 8.
936 @defmac BITS_PER_WORD
937 Number of bits in a word. If you do not define this macro, the default
938 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
941 @defmac MAX_BITS_PER_WORD
942 Maximum number of bits in a word. If this is undefined, the default is
943 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
944 largest value that @code{BITS_PER_WORD} can have at run-time.
947 @defmac UNITS_PER_WORD
948 Number of storage units in a word; normally the size of a general-purpose
949 register, a power of two from 1 or 8.
952 @defmac MIN_UNITS_PER_WORD
953 Minimum number of units in a word. If this is undefined, the default is
954 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
955 smallest value that @code{UNITS_PER_WORD} can have at run-time.
959 Width of a pointer, in bits. You must specify a value no wider than the
960 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
961 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
962 a value the default is @code{BITS_PER_WORD}.
965 @defmac POINTERS_EXTEND_UNSIGNED
966 A C expression that determines how pointers should be extended from
967 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
968 greater than zero if pointers should be zero-extended, zero if they
969 should be sign-extended, and negative if some other sort of conversion
970 is needed. In the last case, the extension is done by the target's
971 @code{ptr_extend} instruction.
973 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
974 and @code{word_mode} are all the same width.
977 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
978 A macro to update @var{m} and @var{unsignedp} when an object whose type
979 is @var{type} and which has the specified mode and signedness is to be
980 stored in a register. This macro is only called when @var{type} is a
983 On most RISC machines, which only have operations that operate on a full
984 register, define this macro to set @var{m} to @code{word_mode} if
985 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
986 cases, only integer modes should be widened because wider-precision
987 floating-point operations are usually more expensive than their narrower
990 For most machines, the macro definition does not change @var{unsignedp}.
991 However, some machines, have instructions that preferentially handle
992 either signed or unsigned quantities of certain modes. For example, on
993 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
994 sign-extend the result to 64 bits. On such machines, set
995 @var{unsignedp} according to which kind of extension is more efficient.
997 Do not define this macro if it would never modify @var{m}.
1000 @deftypefn {Target Hook} {enum machine_mode} TARGET_PROMOTE_FUNCTION_MODE (const_tree @var{type}, enum machine_mode @var{mode}, int *@var{punsignedp}, const_tree @var{funtype}, int @var{for_return})
1001 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
1002 function return values. The target hook should return the new mode
1003 and possibly change @code{*@var{punsignedp}} if the promotion should
1004 change signedness. This function is called only for scalar @emph{or
1007 @var{for_return} allows to distinguish the promotion of arguments and
1008 return values. If it is @code{1}, a return value is being promoted and
1009 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
1010 If it is @code{2}, the returned mode should be that of the register in
1011 which an incoming parameter is copied, or the outgoing result is computed;
1012 then the hook should return the same mode as @code{promote_mode}, though
1013 the signedness may be different.
1015 The default is to not promote arguments and return values. You can
1016 also define the hook to @code{default_promote_function_mode_always_promote}
1017 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
1020 @defmac PARM_BOUNDARY
1021 Normal alignment required for function parameters on the stack, in
1022 bits. All stack parameters receive at least this much alignment
1023 regardless of data type. On most machines, this is the same as the
1027 @defmac STACK_BOUNDARY
1028 Define this macro to the minimum alignment enforced by hardware for the
1029 stack pointer on this machine. The definition is a C expression for the
1030 desired alignment (measured in bits). This value is used as a default
1031 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1032 this should be the same as @code{PARM_BOUNDARY}.
1035 @defmac PREFERRED_STACK_BOUNDARY
1036 Define this macro if you wish to preserve a certain alignment for the
1037 stack pointer, greater than what the hardware enforces. The definition
1038 is a C expression for the desired alignment (measured in bits). This
1039 macro must evaluate to a value equal to or larger than
1040 @code{STACK_BOUNDARY}.
1043 @defmac INCOMING_STACK_BOUNDARY
1044 Define this macro if the incoming stack boundary may be different
1045 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
1046 to a value equal to or larger than @code{STACK_BOUNDARY}.
1049 @defmac FUNCTION_BOUNDARY
1050 Alignment required for a function entry point, in bits.
1053 @defmac BIGGEST_ALIGNMENT
1054 Biggest alignment that any data type can require on this machine, in
1055 bits. Note that this is not the biggest alignment that is supported,
1056 just the biggest alignment that, when violated, may cause a fault.
1059 @defmac MALLOC_ABI_ALIGNMENT
1060 Alignment, in bits, a C conformant malloc implementation has to
1061 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1064 @defmac ATTRIBUTE_ALIGNED_VALUE
1065 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1066 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1069 @defmac MINIMUM_ATOMIC_ALIGNMENT
1070 If defined, the smallest alignment, in bits, that can be given to an
1071 object that can be referenced in one operation, without disturbing any
1072 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1073 on machines that don't have byte or half-word store operations.
1076 @defmac BIGGEST_FIELD_ALIGNMENT
1077 Biggest alignment that any structure or union field can require on this
1078 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1079 structure and union fields only, unless the field alignment has been set
1080 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1083 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1084 An expression for the alignment of a structure field @var{field} if the
1085 alignment computed in the usual way (including applying of
1086 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1087 alignment) is @var{computed}. It overrides alignment only if the
1088 field alignment has not been set by the
1089 @code{__attribute__ ((aligned (@var{n})))} construct.
1092 @defmac MAX_STACK_ALIGNMENT
1093 Biggest stack alignment guaranteed by the backend. Use this macro
1094 to specify the maximum alignment of a variable on stack.
1096 If not defined, the default value is @code{STACK_BOUNDARY}.
1098 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1099 @c But the fix for PR 32893 indicates that we can only guarantee
1100 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1101 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1104 @defmac MAX_OFILE_ALIGNMENT
1105 Biggest alignment supported by the object file format of this machine.
1106 Use this macro to limit the alignment which can be specified using the
1107 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1108 the default value is @code{BIGGEST_ALIGNMENT}.
1110 On systems that use ELF, the default (in @file{config/elfos.h}) is
1111 the largest supported 32-bit ELF section alignment representable on
1112 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1113 On 32-bit ELF the largest supported section alignment in bits is
1114 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1117 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1118 If defined, a C expression to compute the alignment for a variable in
1119 the static store. @var{type} is the data type, and @var{basic-align} is
1120 the alignment that the object would ordinarily have. The value of this
1121 macro is used instead of that alignment to align the object.
1123 If this macro is not defined, then @var{basic-align} is used.
1126 One use of this macro is to increase alignment of medium-size data to
1127 make it all fit in fewer cache lines. Another is to cause character
1128 arrays to be word-aligned so that @code{strcpy} calls that copy
1129 constants to character arrays can be done inline.
1132 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1133 If defined, a C expression to compute the alignment given to a constant
1134 that is being placed in memory. @var{constant} is the constant and
1135 @var{basic-align} is the alignment that the object would ordinarily
1136 have. The value of this macro is used instead of that alignment to
1139 If this macro is not defined, then @var{basic-align} is used.
1141 The typical use of this macro is to increase alignment for string
1142 constants to be word aligned so that @code{strcpy} calls that copy
1143 constants can be done inline.
1146 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1147 If defined, a C expression to compute the alignment for a variable in
1148 the local store. @var{type} is the data type, and @var{basic-align} is
1149 the alignment that the object would ordinarily have. The value of this
1150 macro is used instead of that alignment to align the object.
1152 If this macro is not defined, then @var{basic-align} is used.
1154 One use of this macro is to increase alignment of medium-size data to
1155 make it all fit in fewer cache lines.
1158 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1159 If defined, a C expression to compute the alignment for stack slot.
1160 @var{type} is the data type, @var{mode} is the widest mode available,
1161 and @var{basic-align} is the alignment that the slot would ordinarily
1162 have. The value of this macro is used instead of that alignment to
1165 If this macro is not defined, then @var{basic-align} is used when
1166 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1169 This macro is to set alignment of stack slot to the maximum alignment
1170 of all possible modes which the slot may have.
1173 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1174 If defined, a C expression to compute the alignment for a local
1175 variable @var{decl}.
1177 If this macro is not defined, then
1178 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1181 One use of this macro is to increase alignment of medium-size data to
1182 make it all fit in fewer cache lines.
1185 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1186 If defined, a C expression to compute the minimum required alignment
1187 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1188 @var{mode}, assuming normal alignment @var{align}.
1190 If this macro is not defined, then @var{align} will be used.
1193 @defmac EMPTY_FIELD_BOUNDARY
1194 Alignment in bits to be given to a structure bit-field that follows an
1195 empty field such as @code{int : 0;}.
1197 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1200 @defmac STRUCTURE_SIZE_BOUNDARY
1201 Number of bits which any structure or union's size must be a multiple of.
1202 Each structure or union's size is rounded up to a multiple of this.
1204 If you do not define this macro, the default is the same as
1205 @code{BITS_PER_UNIT}.
1208 @defmac STRICT_ALIGNMENT
1209 Define this macro to be the value 1 if instructions will fail to work
1210 if given data not on the nominal alignment. If instructions will merely
1211 go slower in that case, define this macro as 0.
1214 @defmac PCC_BITFIELD_TYPE_MATTERS
1215 Define this if you wish to imitate the way many other C compilers handle
1216 alignment of bit-fields and the structures that contain them.
1218 The behavior is that the type written for a named bit-field (@code{int},
1219 @code{short}, or other integer type) imposes an alignment for the entire
1220 structure, as if the structure really did contain an ordinary field of
1221 that type. In addition, the bit-field is placed within the structure so
1222 that it would fit within such a field, not crossing a boundary for it.
1224 Thus, on most machines, a named bit-field whose type is written as
1225 @code{int} would not cross a four-byte boundary, and would force
1226 four-byte alignment for the whole structure. (The alignment used may
1227 not be four bytes; it is controlled by the other alignment parameters.)
1229 An unnamed bit-field will not affect the alignment of the containing
1232 If the macro is defined, its definition should be a C expression;
1233 a nonzero value for the expression enables this behavior.
1235 Note that if this macro is not defined, or its value is zero, some
1236 bit-fields may cross more than one alignment boundary. The compiler can
1237 support such references if there are @samp{insv}, @samp{extv}, and
1238 @samp{extzv} insns that can directly reference memory.
1240 The other known way of making bit-fields work is to define
1241 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1242 Then every structure can be accessed with fullwords.
1244 Unless the machine has bit-field instructions or you define
1245 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1246 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1248 If your aim is to make GCC use the same conventions for laying out
1249 bit-fields as are used by another compiler, here is how to investigate
1250 what the other compiler does. Compile and run this program:
1269 printf ("Size of foo1 is %d\n",
1270 sizeof (struct foo1));
1271 printf ("Size of foo2 is %d\n",
1272 sizeof (struct foo2));
1277 If this prints 2 and 5, then the compiler's behavior is what you would
1278 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1281 @defmac BITFIELD_NBYTES_LIMITED
1282 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1283 to aligning a bit-field within the structure.
1286 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void)
1287 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1288 whether unnamed bitfields affect the alignment of the containing
1289 structure. The hook should return true if the structure should inherit
1290 the alignment requirements of an unnamed bitfield's type.
1293 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void)
1294 This target hook should return @code{true} if accesses to volatile bitfields
1295 should use the narrowest mode possible. It should return @code{false} if
1296 these accesses should use the bitfield container type.
1298 The default is @code{!TARGET_STRICT_ALIGN}.
1301 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1302 Return 1 if a structure or array containing @var{field} should be accessed using
1305 If @var{field} is the only field in the structure, @var{mode} is its
1306 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1307 case where structures of one field would require the structure's mode to
1308 retain the field's mode.
1310 Normally, this is not needed.
1313 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1314 Define this macro as an expression for the alignment of a type (given
1315 by @var{type} as a tree node) if the alignment computed in the usual
1316 way is @var{computed} and the alignment explicitly specified was
1319 The default is to use @var{specified} if it is larger; otherwise, use
1320 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1323 @defmac MAX_FIXED_MODE_SIZE
1324 An integer expression for the size in bits of the largest integer
1325 machine mode that should actually be used. All integer machine modes of
1326 this size or smaller can be used for structures and unions with the
1327 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1328 (DImode)} is assumed.
1331 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1332 If defined, an expression of type @code{enum machine_mode} that
1333 specifies the mode of the save area operand of a
1334 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1335 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1336 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1337 having its mode specified.
1339 You need not define this macro if it always returns @code{Pmode}. You
1340 would most commonly define this macro if the
1341 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1345 @defmac STACK_SIZE_MODE
1346 If defined, an expression of type @code{enum machine_mode} that
1347 specifies the mode of the size increment operand of an
1348 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1350 You need not define this macro if it always returns @code{word_mode}.
1351 You would most commonly define this macro if the @code{allocate_stack}
1352 pattern needs to support both a 32- and a 64-bit mode.
1355 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_CMP_RETURN_MODE (void)
1356 This target hook should return the mode to be used for the return value
1357 of compare instructions expanded to libgcc calls. If not defined
1358 @code{word_mode} is returned which is the right choice for a majority of
1362 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_SHIFT_COUNT_MODE (void)
1363 This target hook should return the mode to be used for the shift count operand
1364 of shift instructions expanded to libgcc calls. If not defined
1365 @code{word_mode} is returned which is the right choice for a majority of
1369 @deftypefn {Target Hook} {enum machine_mode} TARGET_UNWIND_WORD_MODE (void)
1370 Return machine mode to be used for @code{_Unwind_Word} type.
1371 The default is to use @code{word_mode}.
1374 @defmac ROUND_TOWARDS_ZERO
1375 If defined, this macro should be true if the prevailing rounding
1376 mode is towards zero.
1378 Defining this macro only affects the way @file{libgcc.a} emulates
1379 floating-point arithmetic.
1381 Not defining this macro is equivalent to returning zero.
1384 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1385 This macro should return true if floats with @var{size}
1386 bits do not have a NaN or infinity representation, but use the largest
1387 exponent for normal numbers instead.
1389 Defining this macro only affects the way @file{libgcc.a} emulates
1390 floating-point arithmetic.
1392 The default definition of this macro returns false for all sizes.
1395 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (const_tree @var{record_type})
1396 This target hook returns @code{true} if bit-fields in the given
1397 @var{record_type} are to be laid out following the rules of Microsoft
1398 Visual C/C++, namely: (i) a bit-field won't share the same storage
1399 unit with the previous bit-field if their underlying types have
1400 different sizes, and the bit-field will be aligned to the highest
1401 alignment of the underlying types of itself and of the previous
1402 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1403 the whole enclosing structure, even if it is unnamed; except that
1404 (iii) a zero-sized bit-field will be disregarded unless it follows
1405 another bit-field of nonzero size. If this hook returns @code{true},
1406 other macros that control bit-field layout are ignored.
1408 When a bit-field is inserted into a packed record, the whole size
1409 of the underlying type is used by one or more same-size adjacent
1410 bit-fields (that is, if its long:3, 32 bits is used in the record,
1411 and any additional adjacent long bit-fields are packed into the same
1412 chunk of 32 bits. However, if the size changes, a new field of that
1413 size is allocated). In an unpacked record, this is the same as using
1414 alignment, but not equivalent when packing.
1416 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1417 the latter will take precedence. If @samp{__attribute__((packed))} is
1418 used on a single field when MS bit-fields are in use, it will take
1419 precedence for that field, but the alignment of the rest of the structure
1420 may affect its placement.
1423 @deftypefn {Target Hook} bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1424 Returns true if the target supports decimal floating point.
1427 @deftypefn {Target Hook} bool TARGET_FIXED_POINT_SUPPORTED_P (void)
1428 Returns true if the target supports fixed-point arithmetic.
1431 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1432 This hook is called just before expansion into rtl, allowing the target
1433 to perform additional initializations or analysis before the expansion.
1434 For example, the rs6000 port uses it to allocate a scratch stack slot
1435 for use in copying SDmode values between memory and floating point
1436 registers whenever the function being expanded has any SDmode
1440 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1441 This hook allows the backend to perform additional instantiations on rtl
1442 that are not actually in any insns yet, but will be later.
1445 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (const_tree @var{type})
1446 If your target defines any fundamental types, or any types your target
1447 uses should be mangled differently from the default, define this hook
1448 to return the appropriate encoding for these types as part of a C++
1449 mangled name. The @var{type} argument is the tree structure representing
1450 the type to be mangled. The hook may be applied to trees which are
1451 not target-specific fundamental types; it should return @code{NULL}
1452 for all such types, as well as arguments it does not recognize. If the
1453 return value is not @code{NULL}, it must point to a statically-allocated
1456 Target-specific fundamental types might be new fundamental types or
1457 qualified versions of ordinary fundamental types. Encode new
1458 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1459 is the name used for the type in source code, and @var{n} is the
1460 length of @var{name} in decimal. Encode qualified versions of
1461 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1462 @var{name} is the name used for the type qualifier in source code,
1463 @var{n} is the length of @var{name} as above, and @var{code} is the
1464 code used to represent the unqualified version of this type. (See
1465 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1466 codes.) In both cases the spaces are for clarity; do not include any
1467 spaces in your string.
1469 This hook is applied to types prior to typedef resolution. If the mangled
1470 name for a particular type depends only on that type's main variant, you
1471 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1474 The default version of this hook always returns @code{NULL}, which is
1475 appropriate for a target that does not define any new fundamental
1480 @section Layout of Source Language Data Types
1482 These macros define the sizes and other characteristics of the standard
1483 basic data types used in programs being compiled. Unlike the macros in
1484 the previous section, these apply to specific features of C and related
1485 languages, rather than to fundamental aspects of storage layout.
1487 @defmac INT_TYPE_SIZE
1488 A C expression for the size in bits of the type @code{int} on the
1489 target machine. If you don't define this, the default is one word.
1492 @defmac SHORT_TYPE_SIZE
1493 A C expression for the size in bits of the type @code{short} on the
1494 target machine. If you don't define this, the default is half a word.
1495 (If this would be less than one storage unit, it is rounded up to one
1499 @defmac LONG_TYPE_SIZE
1500 A C expression for the size in bits of the type @code{long} on the
1501 target machine. If you don't define this, the default is one word.
1504 @defmac ADA_LONG_TYPE_SIZE
1505 On some machines, the size used for the Ada equivalent of the type
1506 @code{long} by a native Ada compiler differs from that used by C@. In
1507 that situation, define this macro to be a C expression to be used for
1508 the size of that type. If you don't define this, the default is the
1509 value of @code{LONG_TYPE_SIZE}.
1512 @defmac LONG_LONG_TYPE_SIZE
1513 A C expression for the size in bits of the type @code{long long} on the
1514 target machine. If you don't define this, the default is two
1515 words. If you want to support GNU Ada on your machine, the value of this
1516 macro must be at least 64.
1519 @defmac CHAR_TYPE_SIZE
1520 A C expression for the size in bits of the type @code{char} on the
1521 target machine. If you don't define this, the default is
1522 @code{BITS_PER_UNIT}.
1525 @defmac BOOL_TYPE_SIZE
1526 A C expression for the size in bits of the C++ type @code{bool} and
1527 C99 type @code{_Bool} on the target machine. If you don't define
1528 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1531 @defmac FLOAT_TYPE_SIZE
1532 A C expression for the size in bits of the type @code{float} on the
1533 target machine. If you don't define this, the default is one word.
1536 @defmac DOUBLE_TYPE_SIZE
1537 A C expression for the size in bits of the type @code{double} on the
1538 target machine. If you don't define this, the default is two
1542 @defmac LONG_DOUBLE_TYPE_SIZE
1543 A C expression for the size in bits of the type @code{long double} on
1544 the target machine. If you don't define this, the default is two
1548 @defmac SHORT_FRACT_TYPE_SIZE
1549 A C expression for the size in bits of the type @code{short _Fract} on
1550 the target machine. If you don't define this, the default is
1551 @code{BITS_PER_UNIT}.
1554 @defmac FRACT_TYPE_SIZE
1555 A C expression for the size in bits of the type @code{_Fract} on
1556 the target machine. If you don't define this, the default is
1557 @code{BITS_PER_UNIT * 2}.
1560 @defmac LONG_FRACT_TYPE_SIZE
1561 A C expression for the size in bits of the type @code{long _Fract} on
1562 the target machine. If you don't define this, the default is
1563 @code{BITS_PER_UNIT * 4}.
1566 @defmac LONG_LONG_FRACT_TYPE_SIZE
1567 A C expression for the size in bits of the type @code{long long _Fract} on
1568 the target machine. If you don't define this, the default is
1569 @code{BITS_PER_UNIT * 8}.
1572 @defmac SHORT_ACCUM_TYPE_SIZE
1573 A C expression for the size in bits of the type @code{short _Accum} on
1574 the target machine. If you don't define this, the default is
1575 @code{BITS_PER_UNIT * 2}.
1578 @defmac ACCUM_TYPE_SIZE
1579 A C expression for the size in bits of the type @code{_Accum} on
1580 the target machine. If you don't define this, the default is
1581 @code{BITS_PER_UNIT * 4}.
1584 @defmac LONG_ACCUM_TYPE_SIZE
1585 A C expression for the size in bits of the type @code{long _Accum} on
1586 the target machine. If you don't define this, the default is
1587 @code{BITS_PER_UNIT * 8}.
1590 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1591 A C expression for the size in bits of the type @code{long long _Accum} on
1592 the target machine. If you don't define this, the default is
1593 @code{BITS_PER_UNIT * 16}.
1596 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1597 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1598 if you want routines in @file{libgcc2.a} for a size other than
1599 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1600 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1603 @defmac LIBGCC2_HAS_DF_MODE
1604 Define this macro if neither @code{DOUBLE_TYPE_SIZE} nor
1605 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1606 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1607 anyway. If you don't define this and either @code{DOUBLE_TYPE_SIZE}
1608 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1612 @defmac LIBGCC2_HAS_XF_MODE
1613 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1614 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1615 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1616 is 80 then the default is 1, otherwise it is 0.
1619 @defmac LIBGCC2_HAS_TF_MODE
1620 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1621 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1622 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1623 is 128 then the default is 1, otherwise it is 0.
1630 Define these macros to be the size in bits of the mantissa of
1631 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1632 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1633 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1634 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1635 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1636 @code{DOUBLE_TYPE_SIZE} or
1637 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1640 @defmac TARGET_FLT_EVAL_METHOD
1641 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1642 assuming, if applicable, that the floating-point control word is in its
1643 default state. If you do not define this macro the value of
1644 @code{FLT_EVAL_METHOD} will be zero.
1647 @defmac WIDEST_HARDWARE_FP_SIZE
1648 A C expression for the size in bits of the widest floating-point format
1649 supported by the hardware. If you define this macro, you must specify a
1650 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1651 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1655 @defmac DEFAULT_SIGNED_CHAR
1656 An expression whose value is 1 or 0, according to whether the type
1657 @code{char} should be signed or unsigned by default. The user can
1658 always override this default with the options @option{-fsigned-char}
1659 and @option{-funsigned-char}.
1662 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1663 This target hook should return true if the compiler should give an
1664 @code{enum} type only as many bytes as it takes to represent the range
1665 of possible values of that type. It should return false if all
1666 @code{enum} types should be allocated like @code{int}.
1668 The default is to return false.
1672 A C expression for a string describing the name of the data type to use
1673 for size values. The typedef name @code{size_t} is defined using the
1674 contents of the string.
1676 The string can contain more than one keyword. If so, separate them with
1677 spaces, and write first any length keyword, then @code{unsigned} if
1678 appropriate, and finally @code{int}. The string must exactly match one
1679 of the data type names defined in the function
1680 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1681 omit @code{int} or change the order---that would cause the compiler to
1684 If you don't define this macro, the default is @code{"long unsigned
1688 @defmac PTRDIFF_TYPE
1689 A C expression for a string describing the name of the data type to use
1690 for the result of subtracting two pointers. The typedef name
1691 @code{ptrdiff_t} is defined using the contents of the string. See
1692 @code{SIZE_TYPE} above for more information.
1694 If you don't define this macro, the default is @code{"long int"}.
1698 A C expression for a string describing the name of the data type to use
1699 for wide characters. The typedef name @code{wchar_t} is defined using
1700 the contents of the string. See @code{SIZE_TYPE} above for more
1703 If you don't define this macro, the default is @code{"int"}.
1706 @defmac WCHAR_TYPE_SIZE
1707 A C expression for the size in bits of the data type for wide
1708 characters. This is used in @code{cpp}, which cannot make use of
1713 A C expression for a string describing the name of the data type to
1714 use for wide characters passed to @code{printf} and returned from
1715 @code{getwc}. The typedef name @code{wint_t} is defined using the
1716 contents of the string. See @code{SIZE_TYPE} above for more
1719 If you don't define this macro, the default is @code{"unsigned int"}.
1723 A C expression for a string describing the name of the data type that
1724 can represent any value of any standard or extended signed integer type.
1725 The typedef name @code{intmax_t} is defined using the contents of the
1726 string. See @code{SIZE_TYPE} above for more information.
1728 If you don't define this macro, the default is the first of
1729 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1730 much precision as @code{long long int}.
1733 @defmac UINTMAX_TYPE
1734 A C expression for a string describing the name of the data type that
1735 can represent any value of any standard or extended unsigned integer
1736 type. The typedef name @code{uintmax_t} is defined using the contents
1737 of the string. See @code{SIZE_TYPE} above for more information.
1739 If you don't define this macro, the default is the first of
1740 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1741 unsigned int"} that has as much precision as @code{long long unsigned
1745 @defmac SIG_ATOMIC_TYPE
1751 @defmacx UINT16_TYPE
1752 @defmacx UINT32_TYPE
1753 @defmacx UINT64_TYPE
1754 @defmacx INT_LEAST8_TYPE
1755 @defmacx INT_LEAST16_TYPE
1756 @defmacx INT_LEAST32_TYPE
1757 @defmacx INT_LEAST64_TYPE
1758 @defmacx UINT_LEAST8_TYPE
1759 @defmacx UINT_LEAST16_TYPE
1760 @defmacx UINT_LEAST32_TYPE
1761 @defmacx UINT_LEAST64_TYPE
1762 @defmacx INT_FAST8_TYPE
1763 @defmacx INT_FAST16_TYPE
1764 @defmacx INT_FAST32_TYPE
1765 @defmacx INT_FAST64_TYPE
1766 @defmacx UINT_FAST8_TYPE
1767 @defmacx UINT_FAST16_TYPE
1768 @defmacx UINT_FAST32_TYPE
1769 @defmacx UINT_FAST64_TYPE
1770 @defmacx INTPTR_TYPE
1771 @defmacx UINTPTR_TYPE
1772 C expressions for the standard types @code{sig_atomic_t},
1773 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1774 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1775 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1776 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1777 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1778 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1779 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1780 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1781 @code{SIZE_TYPE} above for more information.
1783 If any of these macros evaluates to a null pointer, the corresponding
1784 type is not supported; if GCC is configured to provide
1785 @code{<stdint.h>} in such a case, the header provided may not conform
1786 to C99, depending on the type in question. The defaults for all of
1787 these macros are null pointers.
1790 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1791 The C++ compiler represents a pointer-to-member-function with a struct
1798 ptrdiff_t vtable_index;
1805 The C++ compiler must use one bit to indicate whether the function that
1806 will be called through a pointer-to-member-function is virtual.
1807 Normally, we assume that the low-order bit of a function pointer must
1808 always be zero. Then, by ensuring that the vtable_index is odd, we can
1809 distinguish which variant of the union is in use. But, on some
1810 platforms function pointers can be odd, and so this doesn't work. In
1811 that case, we use the low-order bit of the @code{delta} field, and shift
1812 the remainder of the @code{delta} field to the left.
1814 GCC will automatically make the right selection about where to store
1815 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1816 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1817 set such that functions always start at even addresses, but the lowest
1818 bit of pointers to functions indicate whether the function at that
1819 address is in ARM or Thumb mode. If this is the case of your
1820 architecture, you should define this macro to
1821 @code{ptrmemfunc_vbit_in_delta}.
1823 In general, you should not have to define this macro. On architectures
1824 in which function addresses are always even, according to
1825 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1826 @code{ptrmemfunc_vbit_in_pfn}.
1829 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1830 Normally, the C++ compiler uses function pointers in vtables. This
1831 macro allows the target to change to use ``function descriptors''
1832 instead. Function descriptors are found on targets for whom a
1833 function pointer is actually a small data structure. Normally the
1834 data structure consists of the actual code address plus a data
1835 pointer to which the function's data is relative.
1837 If vtables are used, the value of this macro should be the number
1838 of words that the function descriptor occupies.
1841 @defmac TARGET_VTABLE_ENTRY_ALIGN
1842 By default, the vtable entries are void pointers, the so the alignment
1843 is the same as pointer alignment. The value of this macro specifies
1844 the alignment of the vtable entry in bits. It should be defined only
1845 when special alignment is necessary. */
1848 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1849 There are a few non-descriptor entries in the vtable at offsets below
1850 zero. If these entries must be padded (say, to preserve the alignment
1851 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1852 of words in each data entry.
1856 @section Register Usage
1857 @cindex register usage
1859 This section explains how to describe what registers the target machine
1860 has, and how (in general) they can be used.
1862 The description of which registers a specific instruction can use is
1863 done with register classes; see @ref{Register Classes}. For information
1864 on using registers to access a stack frame, see @ref{Frame Registers}.
1865 For passing values in registers, see @ref{Register Arguments}.
1866 For returning values in registers, see @ref{Scalar Return}.
1869 * Register Basics:: Number and kinds of registers.
1870 * Allocation Order:: Order in which registers are allocated.
1871 * Values in Registers:: What kinds of values each reg can hold.
1872 * Leaf Functions:: Renumbering registers for leaf functions.
1873 * Stack Registers:: Handling a register stack such as 80387.
1876 @node Register Basics
1877 @subsection Basic Characteristics of Registers
1879 @c prevent bad page break with this line
1880 Registers have various characteristics.
1882 @defmac FIRST_PSEUDO_REGISTER
1883 Number of hardware registers known to the compiler. They receive
1884 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1885 pseudo register's number really is assigned the number
1886 @code{FIRST_PSEUDO_REGISTER}.
1889 @defmac FIXED_REGISTERS
1890 @cindex fixed register
1891 An initializer that says which registers are used for fixed purposes
1892 all throughout the compiled code and are therefore not available for
1893 general allocation. These would include the stack pointer, the frame
1894 pointer (except on machines where that can be used as a general
1895 register when no frame pointer is needed), the program counter on
1896 machines where that is considered one of the addressable registers,
1897 and any other numbered register with a standard use.
1899 This information is expressed as a sequence of numbers, separated by
1900 commas and surrounded by braces. The @var{n}th number is 1 if
1901 register @var{n} is fixed, 0 otherwise.
1903 The table initialized from this macro, and the table initialized by
1904 the following one, may be overridden at run time either automatically,
1905 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1906 the user with the command options @option{-ffixed-@var{reg}},
1907 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1910 @defmac CALL_USED_REGISTERS
1911 @cindex call-used register
1912 @cindex call-clobbered register
1913 @cindex call-saved register
1914 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1915 clobbered (in general) by function calls as well as for fixed
1916 registers. This macro therefore identifies the registers that are not
1917 available for general allocation of values that must live across
1920 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1921 automatically saves it on function entry and restores it on function
1922 exit, if the register is used within the function.
1925 @defmac CALL_REALLY_USED_REGISTERS
1926 @cindex call-used register
1927 @cindex call-clobbered register
1928 @cindex call-saved register
1929 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1930 that the entire set of @code{FIXED_REGISTERS} be included.
1931 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1932 This macro is optional. If not specified, it defaults to the value
1933 of @code{CALL_USED_REGISTERS}.
1936 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1937 @cindex call-used register
1938 @cindex call-clobbered register
1939 @cindex call-saved register
1940 A C expression that is nonzero if it is not permissible to store a
1941 value of mode @var{mode} in hard register number @var{regno} across a
1942 call without some part of it being clobbered. For most machines this
1943 macro need not be defined. It is only required for machines that do not
1944 preserve the entire contents of a register across a call.
1948 @findex call_used_regs
1951 @findex reg_class_contents
1952 @defmac CONDITIONAL_REGISTER_USAGE
1953 Zero or more C statements that may conditionally modify five variables
1954 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1955 @code{reg_names}, and @code{reg_class_contents}, to take into account
1956 any dependence of these register sets on target flags. The first three
1957 of these are of type @code{char []} (interpreted as Boolean vectors).
1958 @code{global_regs} is a @code{const char *[]}, and
1959 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1960 called, @code{fixed_regs}, @code{call_used_regs},
1961 @code{reg_class_contents}, and @code{reg_names} have been initialized
1962 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1963 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1964 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1965 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1966 command options have been applied.
1968 You need not define this macro if it has no work to do.
1970 @cindex disabling certain registers
1971 @cindex controlling register usage
1972 If the usage of an entire class of registers depends on the target
1973 flags, you may indicate this to GCC by using this macro to modify
1974 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1975 registers in the classes which should not be used by GCC@. Also define
1976 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1977 to return @code{NO_REGS} if it
1978 is called with a letter for a class that shouldn't be used.
1980 (However, if this class is not included in @code{GENERAL_REGS} and all
1981 of the insn patterns whose constraints permit this class are
1982 controlled by target switches, then GCC will automatically avoid using
1983 these registers when the target switches are opposed to them.)
1986 @defmac INCOMING_REGNO (@var{out})
1987 Define this macro if the target machine has register windows. This C
1988 expression returns the register number as seen by the called function
1989 corresponding to the register number @var{out} as seen by the calling
1990 function. Return @var{out} if register number @var{out} is not an
1994 @defmac OUTGOING_REGNO (@var{in})
1995 Define this macro if the target machine has register windows. This C
1996 expression returns the register number as seen by the calling function
1997 corresponding to the register number @var{in} as seen by the called
1998 function. Return @var{in} if register number @var{in} is not an inbound
2002 @defmac LOCAL_REGNO (@var{regno})
2003 Define this macro if the target machine has register windows. This C
2004 expression returns true if the register is call-saved but is in the
2005 register window. Unlike most call-saved registers, such registers
2006 need not be explicitly restored on function exit or during non-local
2011 If the program counter has a register number, define this as that
2012 register number. Otherwise, do not define it.
2015 @node Allocation Order
2016 @subsection Order of Allocation of Registers
2017 @cindex order of register allocation
2018 @cindex register allocation order
2020 @c prevent bad page break with this line
2021 Registers are allocated in order.
2023 @defmac REG_ALLOC_ORDER
2024 If defined, an initializer for a vector of integers, containing the
2025 numbers of hard registers in the order in which GCC should prefer
2026 to use them (from most preferred to least).
2028 If this macro is not defined, registers are used lowest numbered first
2029 (all else being equal).
2031 One use of this macro is on machines where the highest numbered
2032 registers must always be saved and the save-multiple-registers
2033 instruction supports only sequences of consecutive registers. On such
2034 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2035 the highest numbered allocable register first.
2038 @defmac ADJUST_REG_ALLOC_ORDER
2039 A C statement (sans semicolon) to choose the order in which to allocate
2040 hard registers for pseudo-registers local to a basic block.
2042 Store the desired register order in the array @code{reg_alloc_order}.
2043 Element 0 should be the register to allocate first; element 1, the next
2044 register; and so on.
2046 The macro body should not assume anything about the contents of
2047 @code{reg_alloc_order} before execution of the macro.
2049 On most machines, it is not necessary to define this macro.
2052 @defmac HONOR_REG_ALLOC_ORDER
2053 Normally, IRA tries to estimate the costs for saving a register in the
2054 prologue and restoring it in the epilogue. This discourages it from
2055 using call-saved registers. If a machine wants to ensure that IRA
2056 allocates registers in the order given by REG_ALLOC_ORDER even if some
2057 call-saved registers appear earlier than call-used ones, this macro
2061 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2062 In some case register allocation order is not enough for the
2063 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2064 If this macro is defined, it should return a floating point value
2065 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2066 be increased by approximately the pseudo's usage frequency times the
2067 value returned by this macro. Not defining this macro is equivalent
2068 to having it always return @code{0.0}.
2070 On most machines, it is not necessary to define this macro.
2073 @node Values in Registers
2074 @subsection How Values Fit in Registers
2076 This section discusses the macros that describe which kinds of values
2077 (specifically, which machine modes) each register can hold, and how many
2078 consecutive registers are needed for a given mode.
2080 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2081 A C expression for the number of consecutive hard registers, starting
2082 at register number @var{regno}, required to hold a value of mode
2083 @var{mode}. This macro must never return zero, even if a register
2084 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2085 and/or CANNOT_CHANGE_MODE_CLASS instead.
2087 On a machine where all registers are exactly one word, a suitable
2088 definition of this macro is
2091 #define HARD_REGNO_NREGS(REGNO, MODE) \
2092 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2097 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2098 A C expression that is nonzero if a value of mode @var{mode}, stored
2099 in memory, ends with padding that causes it to take up more space than
2100 in registers starting at register number @var{regno} (as determined by
2101 multiplying GCC's notion of the size of the register when containing
2102 this mode by the number of registers returned by
2103 @code{HARD_REGNO_NREGS}). By default this is zero.
2105 For example, if a floating-point value is stored in three 32-bit
2106 registers but takes up 128 bits in memory, then this would be
2109 This macros only needs to be defined if there are cases where
2110 @code{subreg_get_info}
2111 would otherwise wrongly determine that a @code{subreg} can be
2112 represented by an offset to the register number, when in fact such a
2113 @code{subreg} would contain some of the padding not stored in
2114 registers and so not be representable.
2117 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2118 For values of @var{regno} and @var{mode} for which
2119 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2120 returning the greater number of registers required to hold the value
2121 including any padding. In the example above, the value would be four.
2124 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2125 Define this macro if the natural size of registers that hold values
2126 of mode @var{mode} is not the word size. It is a C expression that
2127 should give the natural size in bytes for the specified mode. It is
2128 used by the register allocator to try to optimize its results. This
2129 happens for example on SPARC 64-bit where the natural size of
2130 floating-point registers is still 32-bit.
2133 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2134 A C expression that is nonzero if it is permissible to store a value
2135 of mode @var{mode} in hard register number @var{regno} (or in several
2136 registers starting with that one). For a machine where all registers
2137 are equivalent, a suitable definition is
2140 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2143 You need not include code to check for the numbers of fixed registers,
2144 because the allocation mechanism considers them to be always occupied.
2146 @cindex register pairs
2147 On some machines, double-precision values must be kept in even/odd
2148 register pairs. You can implement that by defining this macro to reject
2149 odd register numbers for such modes.
2151 The minimum requirement for a mode to be OK in a register is that the
2152 @samp{mov@var{mode}} instruction pattern support moves between the
2153 register and other hard register in the same class and that moving a
2154 value into the register and back out not alter it.
2156 Since the same instruction used to move @code{word_mode} will work for
2157 all narrower integer modes, it is not necessary on any machine for
2158 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2159 you define patterns @samp{movhi}, etc., to take advantage of this. This
2160 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2161 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2164 Many machines have special registers for floating point arithmetic.
2165 Often people assume that floating point machine modes are allowed only
2166 in floating point registers. This is not true. Any registers that
2167 can hold integers can safely @emph{hold} a floating point machine
2168 mode, whether or not floating arithmetic can be done on it in those
2169 registers. Integer move instructions can be used to move the values.
2171 On some machines, though, the converse is true: fixed-point machine
2172 modes may not go in floating registers. This is true if the floating
2173 registers normalize any value stored in them, because storing a
2174 non-floating value there would garble it. In this case,
2175 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2176 floating registers. But if the floating registers do not automatically
2177 normalize, if you can store any bit pattern in one and retrieve it
2178 unchanged without a trap, then any machine mode may go in a floating
2179 register, so you can define this macro to say so.
2181 The primary significance of special floating registers is rather that
2182 they are the registers acceptable in floating point arithmetic
2183 instructions. However, this is of no concern to
2184 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2185 constraints for those instructions.
2187 On some machines, the floating registers are especially slow to access,
2188 so that it is better to store a value in a stack frame than in such a
2189 register if floating point arithmetic is not being done. As long as the
2190 floating registers are not in class @code{GENERAL_REGS}, they will not
2191 be used unless some pattern's constraint asks for one.
2194 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2195 A C expression that is nonzero if it is OK to rename a hard register
2196 @var{from} to another hard register @var{to}.
2198 One common use of this macro is to prevent renaming of a register to
2199 another register that is not saved by a prologue in an interrupt
2202 The default is always nonzero.
2205 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2206 A C expression that is nonzero if a value of mode
2207 @var{mode1} is accessible in mode @var{mode2} without copying.
2209 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2210 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2211 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2212 should be nonzero. If they differ for any @var{r}, you should define
2213 this macro to return zero unless some other mechanism ensures the
2214 accessibility of the value in a narrower mode.
2216 You should define this macro to return nonzero in as many cases as
2217 possible since doing so will allow GCC to perform better register
2221 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2222 This target hook should return @code{true} if it is OK to use a hard register
2223 @var{regno} as scratch reg in peephole2.
2225 One common use of this macro is to prevent using of a register that
2226 is not saved by a prologue in an interrupt handler.
2228 The default version of this hook always returns @code{true}.
2231 @defmac AVOID_CCMODE_COPIES
2232 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2233 registers. You should only define this macro if support for copying to/from
2234 @code{CCmode} is incomplete.
2237 @node Leaf Functions
2238 @subsection Handling Leaf Functions
2240 @cindex leaf functions
2241 @cindex functions, leaf
2242 On some machines, a leaf function (i.e., one which makes no calls) can run
2243 more efficiently if it does not make its own register window. Often this
2244 means it is required to receive its arguments in the registers where they
2245 are passed by the caller, instead of the registers where they would
2248 The special treatment for leaf functions generally applies only when
2249 other conditions are met; for example, often they may use only those
2250 registers for its own variables and temporaries. We use the term ``leaf
2251 function'' to mean a function that is suitable for this special
2252 handling, so that functions with no calls are not necessarily ``leaf
2255 GCC assigns register numbers before it knows whether the function is
2256 suitable for leaf function treatment. So it needs to renumber the
2257 registers in order to output a leaf function. The following macros
2260 @defmac LEAF_REGISTERS
2261 Name of a char vector, indexed by hard register number, which
2262 contains 1 for a register that is allowable in a candidate for leaf
2265 If leaf function treatment involves renumbering the registers, then the
2266 registers marked here should be the ones before renumbering---those that
2267 GCC would ordinarily allocate. The registers which will actually be
2268 used in the assembler code, after renumbering, should not be marked with 1
2271 Define this macro only if the target machine offers a way to optimize
2272 the treatment of leaf functions.
2275 @defmac LEAF_REG_REMAP (@var{regno})
2276 A C expression whose value is the register number to which @var{regno}
2277 should be renumbered, when a function is treated as a leaf function.
2279 If @var{regno} is a register number which should not appear in a leaf
2280 function before renumbering, then the expression should yield @minus{}1, which
2281 will cause the compiler to abort.
2283 Define this macro only if the target machine offers a way to optimize the
2284 treatment of leaf functions, and registers need to be renumbered to do
2288 @findex current_function_is_leaf
2289 @findex current_function_uses_only_leaf_regs
2290 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2291 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2292 specially. They can test the C variable @code{current_function_is_leaf}
2293 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2294 set prior to local register allocation and is valid for the remaining
2295 compiler passes. They can also test the C variable
2296 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2297 functions which only use leaf registers.
2298 @code{current_function_uses_only_leaf_regs} is valid after all passes
2299 that modify the instructions have been run and is only useful if
2300 @code{LEAF_REGISTERS} is defined.
2301 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2302 @c of the next paragraph?! --mew 2feb93
2304 @node Stack Registers
2305 @subsection Registers That Form a Stack
2307 There are special features to handle computers where some of the
2308 ``registers'' form a stack. Stack registers are normally written by
2309 pushing onto the stack, and are numbered relative to the top of the
2312 Currently, GCC can only handle one group of stack-like registers, and
2313 they must be consecutively numbered. Furthermore, the existing
2314 support for stack-like registers is specific to the 80387 floating
2315 point coprocessor. If you have a new architecture that uses
2316 stack-like registers, you will need to do substantial work on
2317 @file{reg-stack.c} and write your machine description to cooperate
2318 with it, as well as defining these macros.
2321 Define this if the machine has any stack-like registers.
2324 @defmac STACK_REG_COVER_CLASS
2325 This is a cover class containing the stack registers. Define this if
2326 the machine has any stack-like registers.
2329 @defmac FIRST_STACK_REG
2330 The number of the first stack-like register. This one is the top
2334 @defmac LAST_STACK_REG
2335 The number of the last stack-like register. This one is the bottom of
2339 @node Register Classes
2340 @section Register Classes
2341 @cindex register class definitions
2342 @cindex class definitions, register
2344 On many machines, the numbered registers are not all equivalent.
2345 For example, certain registers may not be allowed for indexed addressing;
2346 certain registers may not be allowed in some instructions. These machine
2347 restrictions are described to the compiler using @dfn{register classes}.
2349 You define a number of register classes, giving each one a name and saying
2350 which of the registers belong to it. Then you can specify register classes
2351 that are allowed as operands to particular instruction patterns.
2355 In general, each register will belong to several classes. In fact, one
2356 class must be named @code{ALL_REGS} and contain all the registers. Another
2357 class must be named @code{NO_REGS} and contain no registers. Often the
2358 union of two classes will be another class; however, this is not required.
2360 @findex GENERAL_REGS
2361 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2362 terribly special about the name, but the operand constraint letters
2363 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2364 the same as @code{ALL_REGS}, just define it as a macro which expands
2367 Order the classes so that if class @var{x} is contained in class @var{y}
2368 then @var{x} has a lower class number than @var{y}.
2370 The way classes other than @code{GENERAL_REGS} are specified in operand
2371 constraints is through machine-dependent operand constraint letters.
2372 You can define such letters to correspond to various classes, then use
2373 them in operand constraints.
2375 You should define a class for the union of two classes whenever some
2376 instruction allows both classes. For example, if an instruction allows
2377 either a floating point (coprocessor) register or a general register for a
2378 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2379 which includes both of them. Otherwise you will get suboptimal code.
2381 You must also specify certain redundant information about the register
2382 classes: for each class, which classes contain it and which ones are
2383 contained in it; for each pair of classes, the largest class contained
2386 When a value occupying several consecutive registers is expected in a
2387 certain class, all the registers used must belong to that class.
2388 Therefore, register classes cannot be used to enforce a requirement for
2389 a register pair to start with an even-numbered register. The way to
2390 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2392 Register classes used for input-operands of bitwise-and or shift
2393 instructions have a special requirement: each such class must have, for
2394 each fixed-point machine mode, a subclass whose registers can transfer that
2395 mode to or from memory. For example, on some machines, the operations for
2396 single-byte values (@code{QImode}) are limited to certain registers. When
2397 this is so, each register class that is used in a bitwise-and or shift
2398 instruction must have a subclass consisting of registers from which
2399 single-byte values can be loaded or stored. This is so that
2400 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2402 @deftp {Data type} {enum reg_class}
2403 An enumerated type that must be defined with all the register class names
2404 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2405 must be the last register class, followed by one more enumerated value,
2406 @code{LIM_REG_CLASSES}, which is not a register class but rather
2407 tells how many classes there are.
2409 Each register class has a number, which is the value of casting
2410 the class name to type @code{int}. The number serves as an index
2411 in many of the tables described below.
2414 @defmac N_REG_CLASSES
2415 The number of distinct register classes, defined as follows:
2418 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2422 @defmac REG_CLASS_NAMES
2423 An initializer containing the names of the register classes as C string
2424 constants. These names are used in writing some of the debugging dumps.
2427 @defmac REG_CLASS_CONTENTS
2428 An initializer containing the contents of the register classes, as integers
2429 which are bit masks. The @var{n}th integer specifies the contents of class
2430 @var{n}. The way the integer @var{mask} is interpreted is that
2431 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2433 When the machine has more than 32 registers, an integer does not suffice.
2434 Then the integers are replaced by sub-initializers, braced groupings containing
2435 several integers. Each sub-initializer must be suitable as an initializer
2436 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2437 In this situation, the first integer in each sub-initializer corresponds to
2438 registers 0 through 31, the second integer to registers 32 through 63, and
2442 @defmac REGNO_REG_CLASS (@var{regno})
2443 A C expression whose value is a register class containing hard register
2444 @var{regno}. In general there is more than one such class; choose a class
2445 which is @dfn{minimal}, meaning that no smaller class also contains the
2449 @defmac BASE_REG_CLASS
2450 A macro whose definition is the name of the class to which a valid
2451 base register must belong. A base register is one used in an address
2452 which is the register value plus a displacement.
2455 @defmac MODE_BASE_REG_CLASS (@var{mode})
2456 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2457 the selection of a base register in a mode dependent manner. If
2458 @var{mode} is VOIDmode then it should return the same value as
2459 @code{BASE_REG_CLASS}.
2462 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2463 A C expression whose value is the register class to which a valid
2464 base register must belong in order to be used in a base plus index
2465 register address. You should define this macro if base plus index
2466 addresses have different requirements than other base register uses.
2469 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{outer_code}, @var{index_code})
2470 A C expression whose value is the register class to which a valid
2471 base register must belong. @var{outer_code} and @var{index_code} define the
2472 context in which the base register occurs. @var{outer_code} is the code of
2473 the immediately enclosing expression (@code{MEM} for the top level of an
2474 address, @code{ADDRESS} for something that occurs in an
2475 @code{address_operand}). @var{index_code} is the code of the corresponding
2476 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2479 @defmac INDEX_REG_CLASS
2480 A macro whose definition is the name of the class to which a valid
2481 index register must belong. An index register is one used in an
2482 address where its value is either multiplied by a scale factor or
2483 added to another register (as well as added to a displacement).
2486 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2487 A C expression which is nonzero if register number @var{num} is
2488 suitable for use as a base register in operand addresses.
2489 Like @code{TARGET_LEGITIMATE_ADDRESS_P}, this macro should also
2490 define a strict and a non-strict variant. Both variants behave
2491 the same for hard register; for pseudos, the strict variant will
2492 pass only those that have been allocated to a valid hard registers,
2493 while the non-strict variant will pass all pseudos.
2495 @findex REG_OK_STRICT
2496 Compiler source files that want to use the strict variant of this and
2497 other macros define the macro @code{REG_OK_STRICT}. You should use an
2498 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
2499 that case and the non-strict variant otherwise.
2502 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2503 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2504 that expression may examine the mode of the memory reference in
2505 @var{mode}. You should define this macro if the mode of the memory
2506 reference affects whether a register may be used as a base register. If
2507 you define this macro, the compiler will use it instead of
2508 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2509 addresses that appear outside a @code{MEM}, i.e., as an
2510 @code{address_operand}.
2512 This macro also has strict and non-strict variants.
2515 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2516 A C expression which is nonzero if register number @var{num} is suitable for
2517 use as a base register in base plus index operand addresses, accessing
2518 memory in mode @var{mode}. It may be either a suitable hard register or a
2519 pseudo register that has been allocated such a hard register. You should
2520 define this macro if base plus index addresses have different requirements
2521 than other base register uses.
2523 Use of this macro is deprecated; please use the more general
2524 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2526 This macro also has strict and non-strict variants.
2529 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{outer_code}, @var{index_code})
2530 A C expression that is just like @code{REGNO_MODE_OK_FOR_BASE_P}, except
2531 that that expression may examine the context in which the register
2532 appears in the memory reference. @var{outer_code} is the code of the
2533 immediately enclosing expression (@code{MEM} if at the top level of the
2534 address, @code{ADDRESS} for something that occurs in an
2535 @code{address_operand}). @var{index_code} is the code of the
2536 corresponding index expression if @var{outer_code} is @code{PLUS};
2537 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2538 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2540 This macro also has strict and non-strict variants.
2543 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2544 A C expression which is nonzero if register number @var{num} is
2545 suitable for use as an index register in operand addresses. It may be
2546 either a suitable hard register or a pseudo register that has been
2547 allocated such a hard register.
2549 The difference between an index register and a base register is that
2550 the index register may be scaled. If an address involves the sum of
2551 two registers, neither one of them scaled, then either one may be
2552 labeled the ``base'' and the other the ``index''; but whichever
2553 labeling is used must fit the machine's constraints of which registers
2554 may serve in each capacity. The compiler will try both labelings,
2555 looking for one that is valid, and will reload one or both registers
2556 only if neither labeling works.
2558 This macro also has strict and non-strict variants.
2561 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2562 A target hook that places additional restrictions on the register class
2563 to use when it is necessary to copy value @var{x} into a register in class
2564 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2565 another, smaller class.
2567 The default version of this hook always returns value of @code{rclass} argument.
2569 Sometimes returning a more restrictive class makes better code. For
2570 example, on the 68000, when @var{x} is an integer constant that is in range
2571 for a @samp{moveq} instruction, the value of this macro is always
2572 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2573 Requiring a data register guarantees that a @samp{moveq} will be used.
2575 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2576 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2577 loaded into some register class. By returning @code{NO_REGS} you can
2578 force @var{x} into a memory location. For example, rs6000 can load
2579 immediate values into general-purpose registers, but does not have an
2580 instruction for loading an immediate value into a floating-point
2581 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2582 @var{x} is a floating-point constant. If the constant can't be loaded
2583 into any kind of register, code generation will be better if
2584 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2585 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2587 If an insn has pseudos in it after register allocation, reload will go
2588 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2589 to find the best one. Returning @code{NO_REGS}, in this case, makes
2590 reload add a @code{!} in front of the constraint: the x86 back-end uses
2591 this feature to discourage usage of 387 registers when math is done in
2592 the SSE registers (and vice versa).
2595 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2596 A C expression that places additional restrictions on the register class
2597 to use when it is necessary to copy value @var{x} into a register in class
2598 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2599 another, smaller class. On many machines, the following definition is
2603 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2606 Sometimes returning a more restrictive class makes better code. For
2607 example, on the 68000, when @var{x} is an integer constant that is in range
2608 for a @samp{moveq} instruction, the value of this macro is always
2609 @code{DATA_REGS} as long as @var{class} includes the data registers.
2610 Requiring a data register guarantees that a @samp{moveq} will be used.
2612 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2613 @var{class} is if @var{x} is a legitimate constant which cannot be
2614 loaded into some register class. By returning @code{NO_REGS} you can
2615 force @var{x} into a memory location. For example, rs6000 can load
2616 immediate values into general-purpose registers, but does not have an
2617 instruction for loading an immediate value into a floating-point
2618 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2619 @var{x} is a floating-point constant. If the constant can't be loaded
2620 into any kind of register, code generation will be better if
2621 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2622 of using @code{PREFERRED_RELOAD_CLASS}.
2624 If an insn has pseudos in it after register allocation, reload will go
2625 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2626 to find the best one. Returning @code{NO_REGS}, in this case, makes
2627 reload add a @code{!} in front of the constraint: the x86 back-end uses
2628 this feature to discourage usage of 387 registers when math is done in
2629 the SSE registers (and vice versa).
2632 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2633 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2634 input reloads. If you don't define this macro, the default is to use
2635 @var{class}, unchanged.
2637 You can also use @code{PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2638 reload from using some alternatives, like @code{PREFERRED_RELOAD_CLASS}.
2641 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_OUTPUT_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2642 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2645 The default version of this hook always returns value of @code{rclass}
2648 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2649 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2652 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2653 A C expression that places additional restrictions on the register class
2654 to use when it is necessary to be able to hold a value of mode
2655 @var{mode} in a reload register for which class @var{class} would
2658 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2659 there are certain modes that simply can't go in certain reload classes.
2661 The value is a register class; perhaps @var{class}, or perhaps another,
2664 Don't define this macro unless the target machine has limitations which
2665 require the macro to do something nontrivial.
2668 @deftypefn {Target Hook} reg_class_t TARGET_SECONDARY_RELOAD (bool @var{in_p}, rtx @var{x}, reg_class_t @var{reload_class}, enum machine_mode @var{reload_mode}, secondary_reload_info *@var{sri})
2669 Many machines have some registers that cannot be copied directly to or
2670 from memory or even from other types of registers. An example is the
2671 @samp{MQ} register, which on most machines, can only be copied to or
2672 from general registers, but not memory. Below, we shall be using the
2673 term 'intermediate register' when a move operation cannot be performed
2674 directly, but has to be done by copying the source into the intermediate
2675 register first, and then copying the intermediate register to the
2676 destination. An intermediate register always has the same mode as
2677 source and destination. Since it holds the actual value being copied,
2678 reload might apply optimizations to re-use an intermediate register
2679 and eliding the copy from the source when it can determine that the
2680 intermediate register still holds the required value.
2682 Another kind of secondary reload is required on some machines which
2683 allow copying all registers to and from memory, but require a scratch
2684 register for stores to some memory locations (e.g., those with symbolic
2685 address on the RT, and those with certain symbolic address on the SPARC
2686 when compiling PIC)@. Scratch registers need not have the same mode
2687 as the value being copied, and usually hold a different value than
2688 that being copied. Special patterns in the md file are needed to
2689 describe how the copy is performed with the help of the scratch register;
2690 these patterns also describe the number, register class(es) and mode(s)
2691 of the scratch register(s).
2693 In some cases, both an intermediate and a scratch register are required.
2695 For input reloads, this target hook is called with nonzero @var{in_p},
2696 and @var{x} is an rtx that needs to be copied to a register of class
2697 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2698 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2699 needs to be copied to rtx @var{x} in @var{reload_mode}.
2701 If copying a register of @var{reload_class} from/to @var{x} requires
2702 an intermediate register, the hook @code{secondary_reload} should
2703 return the register class required for this intermediate register.
2704 If no intermediate register is required, it should return NO_REGS.
2705 If more than one intermediate register is required, describe the one
2706 that is closest in the copy chain to the reload register.
2708 If scratch registers are needed, you also have to describe how to
2709 perform the copy from/to the reload register to/from this
2710 closest intermediate register. Or if no intermediate register is
2711 required, but still a scratch register is needed, describe the
2712 copy from/to the reload register to/from the reload operand @var{x}.
2714 You do this by setting @code{sri->icode} to the instruction code of a pattern
2715 in the md file which performs the move. Operands 0 and 1 are the output
2716 and input of this copy, respectively. Operands from operand 2 onward are
2717 for scratch operands. These scratch operands must have a mode, and a
2718 single-register-class
2719 @c [later: or memory]
2722 When an intermediate register is used, the @code{secondary_reload}
2723 hook will be called again to determine how to copy the intermediate
2724 register to/from the reload operand @var{x}, so your hook must also
2725 have code to handle the register class of the intermediate operand.
2727 @c [For later: maybe we'll allow multi-alternative reload patterns -
2728 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2729 @c and match the constraints of input and output to determine the required
2730 @c alternative. A restriction would be that constraints used to match
2731 @c against reloads registers would have to be written as register class
2732 @c constraints, or we need a new target macro / hook that tells us if an
2733 @c arbitrary constraint can match an unknown register of a given class.
2734 @c Such a macro / hook would also be useful in other places.]
2737 @var{x} might be a pseudo-register or a @code{subreg} of a
2738 pseudo-register, which could either be in a hard register or in memory.
2739 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2740 in memory and the hard register number if it is in a register.
2742 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2743 currently not supported. For the time being, you will have to continue
2744 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2746 @code{copy_cost} also uses this target hook to find out how values are
2747 copied. If you want it to include some extra cost for the need to allocate
2748 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2749 Or if two dependent moves are supposed to have a lower cost than the sum
2750 of the individual moves due to expected fortuitous scheduling and/or special
2751 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2754 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2755 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2756 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2757 These macros are obsolete, new ports should use the target hook
2758 @code{TARGET_SECONDARY_RELOAD} instead.
2760 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2761 target hook. Older ports still define these macros to indicate to the
2762 reload phase that it may
2763 need to allocate at least one register for a reload in addition to the
2764 register to contain the data. Specifically, if copying @var{x} to a
2765 register @var{class} in @var{mode} requires an intermediate register,
2766 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2767 largest register class all of whose registers can be used as
2768 intermediate registers or scratch registers.
2770 If copying a register @var{class} in @var{mode} to @var{x} requires an
2771 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2772 was supposed to be defined be defined to return the largest register
2773 class required. If the
2774 requirements for input and output reloads were the same, the macro
2775 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2778 The values returned by these macros are often @code{GENERAL_REGS}.
2779 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2780 can be directly copied to or from a register of @var{class} in
2781 @var{mode} without requiring a scratch register. Do not define this
2782 macro if it would always return @code{NO_REGS}.
2784 If a scratch register is required (either with or without an
2785 intermediate register), you were supposed to define patterns for
2786 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2787 (@pxref{Standard Names}. These patterns, which were normally
2788 implemented with a @code{define_expand}, should be similar to the
2789 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2792 These patterns need constraints for the reload register and scratch
2794 contain a single register class. If the original reload register (whose
2795 class is @var{class}) can meet the constraint given in the pattern, the
2796 value returned by these macros is used for the class of the scratch
2797 register. Otherwise, two additional reload registers are required.
2798 Their classes are obtained from the constraints in the insn pattern.
2800 @var{x} might be a pseudo-register or a @code{subreg} of a
2801 pseudo-register, which could either be in a hard register or in memory.
2802 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2803 in memory and the hard register number if it is in a register.
2805 These macros should not be used in the case where a particular class of
2806 registers can only be copied to memory and not to another class of
2807 registers. In that case, secondary reload registers are not needed and
2808 would not be helpful. Instead, a stack location must be used to perform
2809 the copy and the @code{mov@var{m}} pattern should use memory as an
2810 intermediate storage. This case often occurs between floating-point and
2814 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2815 Certain machines have the property that some registers cannot be copied
2816 to some other registers without using memory. Define this macro on
2817 those machines to be a C expression that is nonzero if objects of mode
2818 @var{m} in registers of @var{class1} can only be copied to registers of
2819 class @var{class2} by storing a register of @var{class1} into memory
2820 and loading that memory location into a register of @var{class2}.
2822 Do not define this macro if its value would always be zero.
2825 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2826 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2827 allocates a stack slot for a memory location needed for register copies.
2828 If this macro is defined, the compiler instead uses the memory location
2829 defined by this macro.
2831 Do not define this macro if you do not define
2832 @code{SECONDARY_MEMORY_NEEDED}.
2835 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2836 When the compiler needs a secondary memory location to copy between two
2837 registers of mode @var{mode}, it normally allocates sufficient memory to
2838 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2839 load operations in a mode that many bits wide and whose class is the
2840 same as that of @var{mode}.
2842 This is right thing to do on most machines because it ensures that all
2843 bits of the register are copied and prevents accesses to the registers
2844 in a narrower mode, which some machines prohibit for floating-point
2847 However, this default behavior is not correct on some machines, such as
2848 the DEC Alpha, that store short integers in floating-point registers
2849 differently than in integer registers. On those machines, the default
2850 widening will not work correctly and you must define this macro to
2851 suppress that widening in some cases. See the file @file{alpha.h} for
2854 Do not define this macro if you do not define
2855 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2856 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2859 @deftypefn {Target Hook} bool TARGET_CLASS_LIKELY_SPILLED_P (reg_class_t @var{rclass})
2860 A target hook which returns @code{true} if pseudos that have been assigned
2861 to registers of class @var{rclass} would likely be spilled because
2862 registers of @var{rclass} are needed for spill registers.
2864 The default version of this target hook returns @code{true} if @var{rclass}
2865 has exactly one register and @code{false} otherwise. On most machines, this
2866 default should be used. Only use this target hook to some other expression
2867 if pseudos allocated by @file{local-alloc.c} end up in memory because their
2868 hard registers were needed for spill registers. If this target hook returns
2869 @code{false} for those classes, those pseudos will only be allocated by
2870 @file{global.c}, which knows how to reallocate the pseudo to another
2871 register. If there would not be another register available for reallocation,
2872 you should not change the implementation of this target hook since
2873 the only effect of such implementation would be to slow down register
2877 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2878 A C expression for the maximum number of consecutive registers
2879 of class @var{class} needed to hold a value of mode @var{mode}.
2881 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2882 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2883 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2884 @var{mode})} for all @var{regno} values in the class @var{class}.
2886 This macro helps control the handling of multiple-word values
2890 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2891 If defined, a C expression that returns nonzero for a @var{class} for which
2892 a change from mode @var{from} to mode @var{to} is invalid.
2894 For the example, loading 32-bit integer or floating-point objects into
2895 floating-point registers on the Alpha extends them to 64 bits.
2896 Therefore loading a 64-bit object and then storing it as a 32-bit object
2897 does not store the low-order 32 bits, as would be the case for a normal
2898 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2902 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2903 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2904 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2908 @deftypefn {Target Hook} {const reg_class_t *} TARGET_IRA_COVER_CLASSES (void)
2909 Return an array of cover classes for the Integrated Register Allocator
2910 (@acronym{IRA}). Cover classes are a set of non-intersecting register
2911 classes covering all hard registers used for register allocation
2912 purposes. If a move between two registers in the same cover class is
2913 possible, it should be cheaper than a load or store of the registers.
2914 The array is terminated by a @code{LIM_REG_CLASSES} element.
2916 The order of cover classes in the array is important. If two classes
2917 have the same cost of usage for a pseudo, the class occurred first in
2918 the array is chosen for the pseudo.
2920 This hook is called once at compiler startup, after the command-line
2921 options have been processed. It is then re-examined by every call to
2922 @code{target_reinit}.
2924 The default implementation returns @code{IRA_COVER_CLASSES}, if defined,
2925 otherwise there is no default implementation. You must define either this
2926 macro or @code{IRA_COVER_CLASSES} in order to use the integrated register
2927 allocator with Chaitin-Briggs coloring. If the macro is not defined,
2928 the only available coloring algorithm is Chow's priority coloring.
2931 @defmac IRA_COVER_CLASSES
2932 See the documentation for @code{TARGET_IRA_COVER_CLASSES}.
2935 @node Old Constraints
2936 @section Obsolete Macros for Defining Constraints
2937 @cindex defining constraints, obsolete method
2938 @cindex constraints, defining, obsolete method
2940 Machine-specific constraints can be defined with these macros instead
2941 of the machine description constructs described in @ref{Define
2942 Constraints}. This mechanism is obsolete. New ports should not use
2943 it; old ports should convert to the new mechanism.
2945 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2946 For the constraint at the start of @var{str}, which starts with the letter
2947 @var{c}, return the length. This allows you to have register class /
2948 constant / extra constraints that are longer than a single letter;
2949 you don't need to define this macro if you can do with single-letter
2950 constraints only. The definition of this macro should use
2951 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2952 to handle specially.
2953 There are some sanity checks in genoutput.c that check the constraint lengths
2954 for the md file, so you can also use this macro to help you while you are
2955 transitioning from a byzantine single-letter-constraint scheme: when you
2956 return a negative length for a constraint you want to re-use, genoutput
2957 will complain about every instance where it is used in the md file.
2960 @defmac REG_CLASS_FROM_LETTER (@var{char})
2961 A C expression which defines the machine-dependent operand constraint
2962 letters for register classes. If @var{char} is such a letter, the
2963 value should be the register class corresponding to it. Otherwise,
2964 the value should be @code{NO_REGS}. The register letter @samp{r},
2965 corresponding to class @code{GENERAL_REGS}, will not be passed
2966 to this macro; you do not need to handle it.
2969 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2970 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2971 passed in @var{str}, so that you can use suffixes to distinguish between
2975 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2976 A C expression that defines the machine-dependent operand constraint
2977 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2978 particular ranges of integer values. If @var{c} is one of those
2979 letters, the expression should check that @var{value}, an integer, is in
2980 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2981 not one of those letters, the value should be 0 regardless of
2985 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2986 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2987 string passed in @var{str}, so that you can use suffixes to distinguish
2988 between different variants.
2991 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2992 A C expression that defines the machine-dependent operand constraint
2993 letters that specify particular ranges of @code{const_double} values
2994 (@samp{G} or @samp{H}).
2996 If @var{c} is one of those letters, the expression should check that
2997 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2998 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2999 letters, the value should be 0 regardless of @var{value}.
3001 @code{const_double} is used for all floating-point constants and for
3002 @code{DImode} fixed-point constants. A given letter can accept either
3003 or both kinds of values. It can use @code{GET_MODE} to distinguish
3004 between these kinds.
3007 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
3008 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
3009 string passed in @var{str}, so that you can use suffixes to distinguish
3010 between different variants.
3013 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
3014 A C expression that defines the optional machine-dependent constraint
3015 letters that can be used to segregate specific types of operands, usually
3016 memory references, for the target machine. Any letter that is not
3017 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
3018 @code{REG_CLASS_FROM_CONSTRAINT}
3019 may be used. Normally this macro will not be defined.
3021 If it is required for a particular target machine, it should return 1
3022 if @var{value} corresponds to the operand type represented by the
3023 constraint letter @var{c}. If @var{c} is not defined as an extra
3024 constraint, the value returned should be 0 regardless of @var{value}.
3026 For example, on the ROMP, load instructions cannot have their output
3027 in r0 if the memory reference contains a symbolic address. Constraint
3028 letter @samp{Q} is defined as representing a memory address that does
3029 @emph{not} contain a symbolic address. An alternative is specified with
3030 a @samp{Q} constraint on the input and @samp{r} on the output. The next
3031 alternative specifies @samp{m} on the input and a register class that
3032 does not include r0 on the output.
3035 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
3036 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
3037 in @var{str}, so that you can use suffixes to distinguish between different
3041 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
3042 A C expression that defines the optional machine-dependent constraint
3043 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
3044 be treated like memory constraints by the reload pass.
3046 It should return 1 if the operand type represented by the constraint
3047 at the start of @var{str}, the first letter of which is the letter @var{c},
3048 comprises a subset of all memory references including
3049 all those whose address is simply a base register. This allows the reload
3050 pass to reload an operand, if it does not directly correspond to the operand
3051 type of @var{c}, by copying its address into a base register.
3053 For example, on the S/390, some instructions do not accept arbitrary
3054 memory references, but only those that do not make use of an index
3055 register. The constraint letter @samp{Q} is defined via
3056 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
3057 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
3058 a @samp{Q} constraint can handle any memory operand, because the
3059 reload pass knows it can be reloaded by copying the memory address
3060 into a base register if required. This is analogous to the way
3061 an @samp{o} constraint can handle any memory operand.
3064 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
3065 A C expression that defines the optional machine-dependent constraint
3066 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
3067 @code{EXTRA_CONSTRAINT_STR}, that should
3068 be treated like address constraints by the reload pass.
3070 It should return 1 if the operand type represented by the constraint
3071 at the start of @var{str}, which starts with the letter @var{c}, comprises
3072 a subset of all memory addresses including
3073 all those that consist of just a base register. This allows the reload
3074 pass to reload an operand, if it does not directly correspond to the operand
3075 type of @var{str}, by copying it into a base register.
3077 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
3078 be used with the @code{address_operand} predicate. It is treated
3079 analogously to the @samp{p} constraint.
3082 @node Stack and Calling
3083 @section Stack Layout and Calling Conventions
3084 @cindex calling conventions
3086 @c prevent bad page break with this line
3087 This describes the stack layout and calling conventions.
3091 * Exception Handling::
3096 * Register Arguments::
3098 * Aggregate Return::
3103 * Stack Smashing Protection::
3107 @subsection Basic Stack Layout
3108 @cindex stack frame layout
3109 @cindex frame layout
3111 @c prevent bad page break with this line
3112 Here is the basic stack layout.
3114 @defmac STACK_GROWS_DOWNWARD
3115 Define this macro if pushing a word onto the stack moves the stack
3116 pointer to a smaller address.
3118 When we say, ``define this macro if @dots{}'', it means that the
3119 compiler checks this macro only with @code{#ifdef} so the precise
3120 definition used does not matter.
3123 @defmac STACK_PUSH_CODE
3124 This macro defines the operation used when something is pushed
3125 on the stack. In RTL, a push operation will be
3126 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3128 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3129 and @code{POST_INC}. Which of these is correct depends on
3130 the stack direction and on whether the stack pointer points
3131 to the last item on the stack or whether it points to the
3132 space for the next item on the stack.
3134 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3135 defined, which is almost always right, and @code{PRE_INC} otherwise,
3136 which is often wrong.
3139 @defmac FRAME_GROWS_DOWNWARD
3140 Define this macro to nonzero value if the addresses of local variable slots
3141 are at negative offsets from the frame pointer.
3144 @defmac ARGS_GROW_DOWNWARD
3145 Define this macro if successive arguments to a function occupy decreasing
3146 addresses on the stack.
3149 @defmac STARTING_FRAME_OFFSET
3150 Offset from the frame pointer to the first local variable slot to be allocated.
3152 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3153 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3154 Otherwise, it is found by adding the length of the first slot to the
3155 value @code{STARTING_FRAME_OFFSET}.
3156 @c i'm not sure if the above is still correct.. had to change it to get
3157 @c rid of an overfull. --mew 2feb93
3160 @defmac STACK_ALIGNMENT_NEEDED
3161 Define to zero to disable final alignment of the stack during reload.
3162 The nonzero default for this macro is suitable for most ports.
3164 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3165 is a register save block following the local block that doesn't require
3166 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3167 stack alignment and do it in the backend.
3170 @defmac STACK_POINTER_OFFSET
3171 Offset from the stack pointer register to the first location at which
3172 outgoing arguments are placed. If not specified, the default value of
3173 zero is used. This is the proper value for most machines.
3175 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3176 the first location at which outgoing arguments are placed.
3179 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3180 Offset from the argument pointer register to the first argument's
3181 address. On some machines it may depend on the data type of the
3184 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3185 the first argument's address.
3188 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3189 Offset from the stack pointer register to an item dynamically allocated
3190 on the stack, e.g., by @code{alloca}.
3192 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3193 length of the outgoing arguments. The default is correct for most
3194 machines. See @file{function.c} for details.
3197 @defmac INITIAL_FRAME_ADDRESS_RTX
3198 A C expression whose value is RTL representing the address of the initial
3199 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3200 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3201 default value will be used. Define this macro in order to make frame pointer
3202 elimination work in the presence of @code{__builtin_frame_address (count)} and
3203 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3206 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3207 A C expression whose value is RTL representing the address in a stack
3208 frame where the pointer to the caller's frame is stored. Assume that
3209 @var{frameaddr} is an RTL expression for the address of the stack frame
3212 If you don't define this macro, the default is to return the value
3213 of @var{frameaddr}---that is, the stack frame address is also the
3214 address of the stack word that points to the previous frame.
3217 @defmac SETUP_FRAME_ADDRESSES
3218 If defined, a C expression that produces the machine-specific code to
3219 setup the stack so that arbitrary frames can be accessed. For example,
3220 on the SPARC, we must flush all of the register windows to the stack
3221 before we can access arbitrary stack frames. You will seldom need to
3225 @deftypefn {Target Hook} rtx TARGET_BUILTIN_SETJMP_FRAME_VALUE (void)
3226 This target hook should return an rtx that is used to store
3227 the address of the current frame into the built in @code{setjmp} buffer.
3228 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3229 machines. One reason you may need to define this target hook is if
3230 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3233 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3234 A C expression whose value is RTL representing the value of the frame
3235 address for the current frame. @var{frameaddr} is the frame pointer
3236 of the current frame. This is used for __builtin_frame_address.
3237 You need only define this macro if the frame address is not the same
3238 as the frame pointer. Most machines do not need to define it.
3241 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3242 A C expression whose value is RTL representing the value of the return
3243 address for the frame @var{count} steps up from the current frame, after
3244 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3245 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3246 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3248 The value of the expression must always be the correct address when
3249 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3250 determine the return address of other frames.
3253 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3254 Define this if the return address of a particular stack frame is accessed
3255 from the frame pointer of the previous stack frame.
3258 @defmac INCOMING_RETURN_ADDR_RTX
3259 A C expression whose value is RTL representing the location of the
3260 incoming return address at the beginning of any function, before the
3261 prologue. This RTL is either a @code{REG}, indicating that the return
3262 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3265 You only need to define this macro if you want to support call frame
3266 debugging information like that provided by DWARF 2.
3268 If this RTL is a @code{REG}, you should also define
3269 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3272 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3273 A C expression whose value is an integer giving a DWARF 2 column
3274 number that may be used as an alternative return column. The column
3275 must not correspond to any gcc hard register (that is, it must not
3276 be in the range of @code{DWARF_FRAME_REGNUM}).
3278 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3279 general register, but an alternative column needs to be used for signal
3280 frames. Some targets have also used different frame return columns
3284 @defmac DWARF_ZERO_REG
3285 A C expression whose value is an integer giving a DWARF 2 register
3286 number that is considered to always have the value zero. This should
3287 only be defined if the target has an architected zero register, and
3288 someone decided it was a good idea to use that register number to
3289 terminate the stack backtrace. New ports should avoid this.
3292 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3293 This target hook allows the backend to emit frame-related insns that
3294 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3295 info engine will invoke it on insns of the form
3297 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3301 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3303 to let the backend emit the call frame instructions. @var{label} is
3304 the CFI label attached to the insn, @var{pattern} is the pattern of
3305 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3308 @defmac INCOMING_FRAME_SP_OFFSET
3309 A C expression whose value is an integer giving the offset, in bytes,
3310 from the value of the stack pointer register to the top of the stack
3311 frame at the beginning of any function, before the prologue. The top of
3312 the frame is defined to be the value of the stack pointer in the
3313 previous frame, just before the call instruction.
3315 You only need to define this macro if you want to support call frame
3316 debugging information like that provided by DWARF 2.
3319 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3320 A C expression whose value is an integer giving the offset, in bytes,
3321 from the argument pointer to the canonical frame address (cfa). The
3322 final value should coincide with that calculated by
3323 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3324 during virtual register instantiation.
3326 The default value for this macro is
3327 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3328 which is correct for most machines; in general, the arguments are found
3329 immediately before the stack frame. Note that this is not the case on
3330 some targets that save registers into the caller's frame, such as SPARC
3331 and rs6000, and so such targets need to define this macro.
3333 You only need to define this macro if the default is incorrect, and you
3334 want to support call frame debugging information like that provided by
3338 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3339 If defined, a C expression whose value is an integer giving the offset
3340 in bytes from the frame pointer to the canonical frame address (cfa).
3341 The final value should coincide with that calculated by
3342 @code{INCOMING_FRAME_SP_OFFSET}.
3344 Normally the CFA is calculated as an offset from the argument pointer,
3345 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3346 variable due to the ABI, this may not be possible. If this macro is
3347 defined, it implies that the virtual register instantiation should be
3348 based on the frame pointer instead of the argument pointer. Only one
3349 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3353 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3354 If defined, a C expression whose value is an integer giving the offset
3355 in bytes from the canonical frame address (cfa) to the frame base used
3356 in DWARF 2 debug information. The default is zero. A different value
3357 may reduce the size of debug information on some ports.
3360 @node Exception Handling
3361 @subsection Exception Handling Support
3362 @cindex exception handling
3364 @defmac EH_RETURN_DATA_REGNO (@var{N})
3365 A C expression whose value is the @var{N}th register number used for
3366 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3367 @var{N} registers are usable.
3369 The exception handling library routines communicate with the exception
3370 handlers via a set of agreed upon registers. Ideally these registers
3371 should be call-clobbered; it is possible to use call-saved registers,
3372 but may negatively impact code size. The target must support at least
3373 2 data registers, but should define 4 if there are enough free registers.
3375 You must define this macro if you want to support call frame exception
3376 handling like that provided by DWARF 2.
3379 @defmac EH_RETURN_STACKADJ_RTX
3380 A C expression whose value is RTL representing a location in which
3381 to store a stack adjustment to be applied before function return.
3382 This is used to unwind the stack to an exception handler's call frame.
3383 It will be assigned zero on code paths that return normally.
3385 Typically this is a call-clobbered hard register that is otherwise
3386 untouched by the epilogue, but could also be a stack slot.
3388 Do not define this macro if the stack pointer is saved and restored
3389 by the regular prolog and epilog code in the call frame itself; in
3390 this case, the exception handling library routines will update the
3391 stack location to be restored in place. Otherwise, you must define
3392 this macro if you want to support call frame exception handling like
3393 that provided by DWARF 2.
3396 @defmac EH_RETURN_HANDLER_RTX
3397 A C expression whose value is RTL representing a location in which
3398 to store the address of an exception handler to which we should
3399 return. It will not be assigned on code paths that return normally.
3401 Typically this is the location in the call frame at which the normal
3402 return address is stored. For targets that return by popping an
3403 address off the stack, this might be a memory address just below
3404 the @emph{target} call frame rather than inside the current call
3405 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3406 been assigned, so it may be used to calculate the location of the
3409 Some targets have more complex requirements than storing to an
3410 address calculable during initial code generation. In that case
3411 the @code{eh_return} instruction pattern should be used instead.
3413 If you want to support call frame exception handling, you must
3414 define either this macro or the @code{eh_return} instruction pattern.
3417 @defmac RETURN_ADDR_OFFSET
3418 If defined, an integer-valued C expression for which rtl will be generated
3419 to add it to the exception handler address before it is searched in the
3420 exception handling tables, and to subtract it again from the address before
3421 using it to return to the exception handler.
3424 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3425 This macro chooses the encoding of pointers embedded in the exception
3426 handling sections. If at all possible, this should be defined such
3427 that the exception handling section will not require dynamic relocations,
3428 and so may be read-only.
3430 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3431 @var{global} is true if the symbol may be affected by dynamic relocations.
3432 The macro should return a combination of the @code{DW_EH_PE_*} defines
3433 as found in @file{dwarf2.h}.
3435 If this macro is not defined, pointers will not be encoded but
3436 represented directly.
3439 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3440 This macro allows the target to emit whatever special magic is required
3441 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3442 Generic code takes care of pc-relative and indirect encodings; this must
3443 be defined if the target uses text-relative or data-relative encodings.
3445 This is a C statement that branches to @var{done} if the format was
3446 handled. @var{encoding} is the format chosen, @var{size} is the number
3447 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3451 @defmac MD_UNWIND_SUPPORT
3452 A string specifying a file to be #include'd in unwind-dw2.c. The file
3453 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3456 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3457 This macro allows the target to add CPU and operating system specific
3458 code to the call-frame unwinder for use when there is no unwind data
3459 available. The most common reason to implement this macro is to unwind
3460 through signal frames.
3462 This macro is called from @code{uw_frame_state_for} in
3463 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3464 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3465 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3466 for the address of the code being executed and @code{context->cfa} for
3467 the stack pointer value. If the frame can be decoded, the register
3468 save addresses should be updated in @var{fs} and the macro should
3469 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3470 the macro should evaluate to @code{_URC_END_OF_STACK}.
3472 For proper signal handling in Java this macro is accompanied by
3473 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3476 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3477 This macro allows the target to add operating system specific code to the
3478 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3479 usually used for signal or interrupt frames.
3481 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3482 @var{context} is an @code{_Unwind_Context};
3483 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3484 for the abi and context in the @code{.unwabi} directive. If the
3485 @code{.unwabi} directive can be handled, the register save addresses should
3486 be updated in @var{fs}.
3489 @defmac TARGET_USES_WEAK_UNWIND_INFO
3490 A C expression that evaluates to true if the target requires unwind
3491 info to be given comdat linkage. Define it to be @code{1} if comdat
3492 linkage is necessary. The default is @code{0}.
3495 @node Stack Checking
3496 @subsection Specifying How Stack Checking is Done
3498 GCC will check that stack references are within the boundaries of the
3499 stack, if the option @option{-fstack-check} is specified, in one of
3504 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3505 will assume that you have arranged for full stack checking to be done
3506 at appropriate places in the configuration files. GCC will not do
3507 other special processing.
3510 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3511 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3512 that you have arranged for static stack checking (checking of the
3513 static stack frame of functions) to be done at appropriate places
3514 in the configuration files. GCC will only emit code to do dynamic
3515 stack checking (checking on dynamic stack allocations) using the third
3519 If neither of the above are true, GCC will generate code to periodically
3520 ``probe'' the stack pointer using the values of the macros defined below.
3523 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3524 GCC will change its allocation strategy for large objects if the option
3525 @option{-fstack-check} is specified: they will always be allocated
3526 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3528 @defmac STACK_CHECK_BUILTIN
3529 A nonzero value if stack checking is done by the configuration files in a
3530 machine-dependent manner. You should define this macro if stack checking
3531 is required by the ABI of your machine or if you would like to do stack
3532 checking in some more efficient way than the generic approach. The default
3533 value of this macro is zero.
3536 @defmac STACK_CHECK_STATIC_BUILTIN
3537 A nonzero value if static stack checking is done by the configuration files
3538 in a machine-dependent manner. You should define this macro if you would
3539 like to do static stack checking in some more efficient way than the generic
3540 approach. The default value of this macro is zero.
3543 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3544 An integer specifying the interval at which GCC must generate stack probe
3545 instructions, defined as 2 raised to this integer. You will normally
3546 define this macro so that the interval be no larger than the size of
3547 the ``guard pages'' at the end of a stack area. The default value
3548 of 12 (4096-byte interval) is suitable for most systems.
3551 @defmac STACK_CHECK_MOVING_SP
3552 An integer which is nonzero if GCC should move the stack pointer page by page
3553 when doing probes. This can be necessary on systems where the stack pointer
3554 contains the bottom address of the memory area accessible to the executing
3555 thread at any point in time. In this situation an alternate signal stack
3556 is required in order to be able to recover from a stack overflow. The
3557 default value of this macro is zero.
3560 @defmac STACK_CHECK_PROTECT
3561 The number of bytes of stack needed to recover from a stack overflow, for
3562 languages where such a recovery is supported. The default value of 75 words
3563 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3564 8192 bytes with other exception handling mechanisms should be adequate for
3568 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3569 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3570 in the opposite case.
3572 @defmac STACK_CHECK_MAX_FRAME_SIZE
3573 The maximum size of a stack frame, in bytes. GCC will generate probe
3574 instructions in non-leaf functions to ensure at least this many bytes of
3575 stack are available. If a stack frame is larger than this size, stack
3576 checking will not be reliable and GCC will issue a warning. The
3577 default is chosen so that GCC only generates one instruction on most
3578 systems. You should normally not change the default value of this macro.
3581 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3582 GCC uses this value to generate the above warning message. It
3583 represents the amount of fixed frame used by a function, not including
3584 space for any callee-saved registers, temporaries and user variables.
3585 You need only specify an upper bound for this amount and will normally
3586 use the default of four words.
3589 @defmac STACK_CHECK_MAX_VAR_SIZE
3590 The maximum size, in bytes, of an object that GCC will place in the
3591 fixed area of the stack frame when the user specifies
3592 @option{-fstack-check}.
3593 GCC computed the default from the values of the above macros and you will
3594 normally not need to override that default.
3598 @node Frame Registers
3599 @subsection Registers That Address the Stack Frame
3601 @c prevent bad page break with this line
3602 This discusses registers that address the stack frame.
3604 @defmac STACK_POINTER_REGNUM
3605 The register number of the stack pointer register, which must also be a
3606 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3607 the hardware determines which register this is.
3610 @defmac FRAME_POINTER_REGNUM
3611 The register number of the frame pointer register, which is used to
3612 access automatic variables in the stack frame. On some machines, the
3613 hardware determines which register this is. On other machines, you can
3614 choose any register you wish for this purpose.
3617 @defmac HARD_FRAME_POINTER_REGNUM
3618 On some machines the offset between the frame pointer and starting
3619 offset of the automatic variables is not known until after register
3620 allocation has been done (for example, because the saved registers are
3621 between these two locations). On those machines, define
3622 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3623 be used internally until the offset is known, and define
3624 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3625 used for the frame pointer.
3627 You should define this macro only in the very rare circumstances when it
3628 is not possible to calculate the offset between the frame pointer and
3629 the automatic variables until after register allocation has been
3630 completed. When this macro is defined, you must also indicate in your
3631 definition of @code{ELIMINABLE_REGS} how to eliminate
3632 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3633 or @code{STACK_POINTER_REGNUM}.
3635 Do not define this macro if it would be the same as
3636 @code{FRAME_POINTER_REGNUM}.
3639 @defmac ARG_POINTER_REGNUM
3640 The register number of the arg pointer register, which is used to access
3641 the function's argument list. On some machines, this is the same as the
3642 frame pointer register. On some machines, the hardware determines which
3643 register this is. On other machines, you can choose any register you
3644 wish for this purpose. If this is not the same register as the frame
3645 pointer register, then you must mark it as a fixed register according to
3646 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3647 (@pxref{Elimination}).
3650 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3651 Define this to a preprocessor constant that is nonzero if
3652 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3653 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3654 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3655 definition is not suitable for use in preprocessor conditionals.
3658 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3659 Define this to a preprocessor constant that is nonzero if
3660 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3661 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3662 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3663 definition is not suitable for use in preprocessor conditionals.
3666 @defmac RETURN_ADDRESS_POINTER_REGNUM
3667 The register number of the return address pointer register, which is used to
3668 access the current function's return address from the stack. On some
3669 machines, the return address is not at a fixed offset from the frame
3670 pointer or stack pointer or argument pointer. This register can be defined
3671 to point to the return address on the stack, and then be converted by
3672 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3674 Do not define this macro unless there is no other way to get the return
3675 address from the stack.
3678 @defmac STATIC_CHAIN_REGNUM
3679 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3680 Register numbers used for passing a function's static chain pointer. If
3681 register windows are used, the register number as seen by the called
3682 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3683 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3684 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3687 The static chain register need not be a fixed register.
3689 If the static chain is passed in memory, these macros should not be
3690 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3693 @deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl}, bool @var{incoming_p})
3694 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3695 targets that may use different static chain locations for different
3696 nested functions. This may be required if the target has function
3697 attributes that affect the calling conventions of the function and
3698 those calling conventions use different static chain locations.
3700 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3702 If the static chain is passed in memory, this hook should be used to
3703 provide rtx giving @code{mem} expressions that denote where they are stored.
3704 Often the @code{mem} expression as seen by the caller will be at an offset
3705 from the stack pointer and the @code{mem} expression as seen by the callee
3706 will be at an offset from the frame pointer.
3707 @findex stack_pointer_rtx
3708 @findex frame_pointer_rtx
3709 @findex arg_pointer_rtx
3710 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3711 @code{arg_pointer_rtx} will have been initialized and should be used
3712 to refer to those items.
3715 @defmac DWARF_FRAME_REGISTERS
3716 This macro specifies the maximum number of hard registers that can be
3717 saved in a call frame. This is used to size data structures used in
3718 DWARF2 exception handling.
3720 Prior to GCC 3.0, this macro was needed in order to establish a stable
3721 exception handling ABI in the face of adding new hard registers for ISA
3722 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3723 in the number of hard registers. Nevertheless, this macro can still be
3724 used to reduce the runtime memory requirements of the exception handling
3725 routines, which can be substantial if the ISA contains a lot of
3726 registers that are not call-saved.
3728 If this macro is not defined, it defaults to
3729 @code{FIRST_PSEUDO_REGISTER}.
3732 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3734 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3735 for backward compatibility in pre GCC 3.0 compiled code.
3737 If this macro is not defined, it defaults to
3738 @code{DWARF_FRAME_REGISTERS}.
3741 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3743 Define this macro if the target's representation for dwarf registers
3744 is different than the internal representation for unwind column.
3745 Given a dwarf register, this macro should return the internal unwind
3746 column number to use instead.
3748 See the PowerPC's SPE target for an example.
3751 @defmac DWARF_FRAME_REGNUM (@var{regno})
3753 Define this macro if the target's representation for dwarf registers
3754 used in .eh_frame or .debug_frame is different from that used in other
3755 debug info sections. Given a GCC hard register number, this macro
3756 should return the .eh_frame register number. The default is
3757 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3761 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3763 Define this macro to map register numbers held in the call frame info
3764 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3765 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3766 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3767 return @code{@var{regno}}.
3772 @subsection Eliminating Frame Pointer and Arg Pointer
3774 @c prevent bad page break with this line
3775 This is about eliminating the frame pointer and arg pointer.
3777 @deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void)
3778 This target hook should return @code{true} if a function must have and use
3779 a frame pointer. This target hook is called in the reload pass. If its return
3780 value is @code{true} the function will have a frame pointer.
3782 This target hook can in principle examine the current function and decide
3783 according to the facts, but on most machines the constant @code{false} or the
3784 constant @code{true} suffices. Use @code{false} when the machine allows code
3785 to be generated with no frame pointer, and doing so saves some time or space.
3786 Use @code{true} when there is no possible advantage to avoiding a frame
3789 In certain cases, the compiler does not know how to produce valid code
3790 without a frame pointer. The compiler recognizes those cases and
3791 automatically gives the function a frame pointer regardless of what
3792 @code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about
3795 In a function that does not require a frame pointer, the frame pointer
3796 register can be allocated for ordinary usage, unless you mark it as a
3797 fixed register. See @code{FIXED_REGISTERS} for more information.
3799 Default return value is @code{false}.
3802 @findex get_frame_size
3803 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3804 A C statement to store in the variable @var{depth-var} the difference
3805 between the frame pointer and the stack pointer values immediately after
3806 the function prologue. The value would be computed from information
3807 such as the result of @code{get_frame_size ()} and the tables of
3808 registers @code{regs_ever_live} and @code{call_used_regs}.
3810 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3811 need not be defined. Otherwise, it must be defined even if
3812 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3813 case, you may set @var{depth-var} to anything.
3816 @defmac ELIMINABLE_REGS
3817 If defined, this macro specifies a table of register pairs used to
3818 eliminate unneeded registers that point into the stack frame. If it is not
3819 defined, the only elimination attempted by the compiler is to replace
3820 references to the frame pointer with references to the stack pointer.
3822 The definition of this macro is a list of structure initializations, each
3823 of which specifies an original and replacement register.
3825 On some machines, the position of the argument pointer is not known until
3826 the compilation is completed. In such a case, a separate hard register
3827 must be used for the argument pointer. This register can be eliminated by
3828 replacing it with either the frame pointer or the argument pointer,
3829 depending on whether or not the frame pointer has been eliminated.
3831 In this case, you might specify:
3833 #define ELIMINABLE_REGS \
3834 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3835 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3836 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3839 Note that the elimination of the argument pointer with the stack pointer is
3840 specified first since that is the preferred elimination.
3843 @deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from_reg}, const int @var{to_reg})
3844 This target hook should returns @code{true} if the compiler is allowed to
3845 try to replace register number @var{from_reg} with register number
3846 @var{to_reg}. This target hook need only be defined if @code{ELIMINABLE_REGS}
3847 is defined, and will usually be @code{true}, since most of the cases
3848 preventing register elimination are things that the compiler already
3851 Default return value is @code{true}.
3854 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3855 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3856 specifies the initial difference between the specified pair of
3857 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3861 @node Stack Arguments
3862 @subsection Passing Function Arguments on the Stack
3863 @cindex arguments on stack
3864 @cindex stack arguments
3866 The macros in this section control how arguments are passed
3867 on the stack. See the following section for other macros that
3868 control passing certain arguments in registers.
3870 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (const_tree @var{fntype})
3871 This target hook returns @code{true} if an argument declared in a
3872 prototype as an integral type smaller than @code{int} should actually be
3873 passed as an @code{int}. In addition to avoiding errors in certain
3874 cases of mismatch, it also makes for better code on certain machines.
3875 The default is to not promote prototypes.
3879 A C expression. If nonzero, push insns will be used to pass
3881 If the target machine does not have a push instruction, set it to zero.
3882 That directs GCC to use an alternate strategy: to
3883 allocate the entire argument block and then store the arguments into
3884 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3887 @defmac PUSH_ARGS_REVERSED
3888 A C expression. If nonzero, function arguments will be evaluated from
3889 last to first, rather than from first to last. If this macro is not
3890 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3891 and args grow in opposite directions, and 0 otherwise.
3894 @defmac PUSH_ROUNDING (@var{npushed})
3895 A C expression that is the number of bytes actually pushed onto the
3896 stack when an instruction attempts to push @var{npushed} bytes.
3898 On some machines, the definition
3901 #define PUSH_ROUNDING(BYTES) (BYTES)
3905 will suffice. But on other machines, instructions that appear
3906 to push one byte actually push two bytes in an attempt to maintain
3907 alignment. Then the definition should be
3910 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3914 @findex current_function_outgoing_args_size
3915 @defmac ACCUMULATE_OUTGOING_ARGS
3916 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3917 will be computed and placed into the variable
3918 @code{current_function_outgoing_args_size}. No space will be pushed
3919 onto the stack for each call; instead, the function prologue should
3920 increase the stack frame size by this amount.
3922 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3926 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3927 Define this macro if functions should assume that stack space has been
3928 allocated for arguments even when their values are passed in
3931 The value of this macro is the size, in bytes, of the area reserved for
3932 arguments passed in registers for the function represented by @var{fndecl},
3933 which can be zero if GCC is calling a library function.
3934 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3937 This space can be allocated by the caller, or be a part of the
3938 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3941 @c above is overfull. not sure what to do. --mew 5feb93 did
3942 @c something, not sure if it looks good. --mew 10feb93
3944 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3945 Define this to a nonzero value if it is the responsibility of the
3946 caller to allocate the area reserved for arguments passed in registers
3947 when calling a function of @var{fntype}. @var{fntype} may be NULL
3948 if the function called is a library function.
3950 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3951 whether the space for these arguments counts in the value of
3952 @code{current_function_outgoing_args_size}.
3955 @defmac STACK_PARMS_IN_REG_PARM_AREA
3956 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3957 stack parameters don't skip the area specified by it.
3958 @c i changed this, makes more sens and it should have taken care of the
3959 @c overfull.. not as specific, tho. --mew 5feb93
3961 Normally, when a parameter is not passed in registers, it is placed on the
3962 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3963 suppresses this behavior and causes the parameter to be passed on the
3964 stack in its natural location.
3967 @deftypefn {Target Hook} int TARGET_RETURN_POPS_ARGS (tree @var{fundecl}, tree @var{funtype}, int @var{size})
3968 This target hook returns the number of bytes of its own arguments that
3969 a function pops on returning, or 0 if the function pops no arguments
3970 and the caller must therefore pop them all after the function returns.
3972 @var{fundecl} is a C variable whose value is a tree node that describes
3973 the function in question. Normally it is a node of type
3974 @code{FUNCTION_DECL} that describes the declaration of the function.
3975 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3977 @var{funtype} is a C variable whose value is a tree node that
3978 describes the function in question. Normally it is a node of type
3979 @code{FUNCTION_TYPE} that describes the data type of the function.
3980 From this it is possible to obtain the data types of the value and
3981 arguments (if known).
3983 When a call to a library function is being considered, @var{fundecl}
3984 will contain an identifier node for the library function. Thus, if
3985 you need to distinguish among various library functions, you can do so
3986 by their names. Note that ``library function'' in this context means
3987 a function used to perform arithmetic, whose name is known specially
3988 in the compiler and was not mentioned in the C code being compiled.
3990 @var{size} is the number of bytes of arguments passed on the
3991 stack. If a variable number of bytes is passed, it is zero, and
3992 argument popping will always be the responsibility of the calling function.
3994 On the VAX, all functions always pop their arguments, so the definition
3995 of this macro is @var{size}. On the 68000, using the standard
3996 calling convention, no functions pop their arguments, so the value of
3997 the macro is always 0 in this case. But an alternative calling
3998 convention is available in which functions that take a fixed number of
3999 arguments pop them but other functions (such as @code{printf}) pop
4000 nothing (the caller pops all). When this convention is in use,
4001 @var{funtype} is examined to determine whether a function takes a fixed
4002 number of arguments.
4005 @defmac CALL_POPS_ARGS (@var{cum})
4006 A C expression that should indicate the number of bytes a call sequence
4007 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
4008 when compiling a function call.
4010 @var{cum} is the variable in which all arguments to the called function
4011 have been accumulated.
4013 On certain architectures, such as the SH5, a call trampoline is used
4014 that pops certain registers off the stack, depending on the arguments
4015 that have been passed to the function. Since this is a property of the
4016 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
4020 @node Register Arguments
4021 @subsection Passing Arguments in Registers
4022 @cindex arguments in registers
4023 @cindex registers arguments
4025 This section describes the macros which let you control how various
4026 types of arguments are passed in registers or how they are arranged in
4029 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
4030 A C expression that controls whether a function argument is passed
4031 in a register, and which register.
4033 The arguments are @var{cum}, which summarizes all the previous
4034 arguments; @var{mode}, the machine mode of the argument; @var{type},
4035 the data type of the argument as a tree node or 0 if that is not known
4036 (which happens for C support library functions); and @var{named},
4037 which is 1 for an ordinary argument and 0 for nameless arguments that
4038 correspond to @samp{@dots{}} in the called function's prototype.
4039 @var{type} can be an incomplete type if a syntax error has previously
4042 The value of the expression is usually either a @code{reg} RTX for the
4043 hard register in which to pass the argument, or zero to pass the
4044 argument on the stack.
4046 For machines like the VAX and 68000, where normally all arguments are
4047 pushed, zero suffices as a definition.
4049 The value of the expression can also be a @code{parallel} RTX@. This is
4050 used when an argument is passed in multiple locations. The mode of the
4051 @code{parallel} should be the mode of the entire argument. The
4052 @code{parallel} holds any number of @code{expr_list} pairs; each one
4053 describes where part of the argument is passed. In each
4054 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
4055 register in which to pass this part of the argument, and the mode of the
4056 register RTX indicates how large this part of the argument is. The
4057 second operand of the @code{expr_list} is a @code{const_int} which gives
4058 the offset in bytes into the entire argument of where this part starts.
4059 As a special exception the first @code{expr_list} in the @code{parallel}
4060 RTX may have a first operand of zero. This indicates that the entire
4061 argument is also stored on the stack.
4063 The last time this macro is called, it is called with @code{MODE ==
4064 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
4065 pattern as operands 2 and 3 respectively.
4067 @cindex @file{stdarg.h} and register arguments
4068 The usual way to make the ISO library @file{stdarg.h} work on a machine
4069 where some arguments are usually passed in registers, is to cause
4070 nameless arguments to be passed on the stack instead. This is done
4071 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
4073 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
4074 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
4075 You may use the hook @code{targetm.calls.must_pass_in_stack}
4076 in the definition of this macro to determine if this argument is of a
4077 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
4078 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
4079 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
4080 defined, the argument will be computed in the stack and then loaded into
4084 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, const_tree @var{type})
4085 This target hook should return @code{true} if we should not pass @var{type}
4086 solely in registers. The file @file{expr.h} defines a
4087 definition that is usually appropriate, refer to @file{expr.h} for additional
4091 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
4092 Define this macro if the target machine has ``register windows'', so
4093 that the register in which a function sees an arguments is not
4094 necessarily the same as the one in which the caller passed the
4097 For such machines, @code{FUNCTION_ARG} computes the register in which
4098 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
4099 be defined in a similar fashion to tell the function being called
4100 where the arguments will arrive.
4102 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
4103 serves both purposes.
4106 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4107 This target hook returns the number of bytes at the beginning of an
4108 argument that must be put in registers. The value must be zero for
4109 arguments that are passed entirely in registers or that are entirely
4110 pushed on the stack.
4112 On some machines, certain arguments must be passed partially in
4113 registers and partially in memory. On these machines, typically the
4114 first few words of arguments are passed in registers, and the rest
4115 on the stack. If a multi-word argument (a @code{double} or a
4116 structure) crosses that boundary, its first few words must be passed
4117 in registers and the rest must be pushed. This macro tells the
4118 compiler when this occurs, and how many bytes should go in registers.
4120 @code{FUNCTION_ARG} for these arguments should return the first
4121 register to be used by the caller for this argument; likewise
4122 @code{FUNCTION_INCOMING_ARG}, for the called function.
4125 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4126 This target hook should return @code{true} if an argument at the
4127 position indicated by @var{cum} should be passed by reference. This
4128 predicate is queried after target independent reasons for being
4129 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
4131 If the hook returns true, a copy of that argument is made in memory and a
4132 pointer to the argument is passed instead of the argument itself.
4133 The pointer is passed in whatever way is appropriate for passing a pointer
4137 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4138 The function argument described by the parameters to this hook is
4139 known to be passed by reference. The hook should return true if the
4140 function argument should be copied by the callee instead of copied
4143 For any argument for which the hook returns true, if it can be
4144 determined that the argument is not modified, then a copy need
4147 The default version of this hook always returns false.
4150 @defmac CUMULATIVE_ARGS
4151 A C type for declaring a variable that is used as the first argument of
4152 @code{FUNCTION_ARG} and other related values. For some target machines,
4153 the type @code{int} suffices and can hold the number of bytes of
4156 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4157 arguments that have been passed on the stack. The compiler has other
4158 variables to keep track of that. For target machines on which all
4159 arguments are passed on the stack, there is no need to store anything in
4160 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4161 should not be empty, so use @code{int}.
4164 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4165 If defined, this macro is called before generating any code for a
4166 function, but after the @var{cfun} descriptor for the function has been
4167 created. The back end may use this macro to update @var{cfun} to
4168 reflect an ABI other than that which would normally be used by default.
4169 If the compiler is generating code for a compiler-generated function,
4170 @var{fndecl} may be @code{NULL}.
4173 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4174 A C statement (sans semicolon) for initializing the variable
4175 @var{cum} for the state at the beginning of the argument list. The
4176 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4177 is the tree node for the data type of the function which will receive
4178 the args, or 0 if the args are to a compiler support library function.
4179 For direct calls that are not libcalls, @var{fndecl} contain the
4180 declaration node of the function. @var{fndecl} is also set when
4181 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4182 being compiled. @var{n_named_args} is set to the number of named
4183 arguments, including a structure return address if it is passed as a
4184 parameter, when making a call. When processing incoming arguments,
4185 @var{n_named_args} is set to @minus{}1.
4187 When processing a call to a compiler support library function,
4188 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4189 contains the name of the function, as a string. @var{libname} is 0 when
4190 an ordinary C function call is being processed. Thus, each time this
4191 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4192 never both of them at once.
4195 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4196 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4197 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4198 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4199 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4200 0)} is used instead.
4203 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4204 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4205 finding the arguments for the function being compiled. If this macro is
4206 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4208 The value passed for @var{libname} is always 0, since library routines
4209 with special calling conventions are never compiled with GCC@. The
4210 argument @var{libname} exists for symmetry with
4211 @code{INIT_CUMULATIVE_ARGS}.
4212 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4213 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4216 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
4217 A C statement (sans semicolon) to update the summarizer variable
4218 @var{cum} to advance past an argument in the argument list. The
4219 values @var{mode}, @var{type} and @var{named} describe that argument.
4220 Once this is done, the variable @var{cum} is suitable for analyzing
4221 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
4223 This macro need not do anything if the argument in question was passed
4224 on the stack. The compiler knows how to track the amount of stack space
4225 used for arguments without any special help.
4228 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4229 If defined, a C expression that is the number of bytes to add to the
4230 offset of the argument passed in memory. This is needed for the SPU,
4231 which passes @code{char} and @code{short} arguments in the preferred
4232 slot that is in the middle of the quad word instead of starting at the
4236 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4237 If defined, a C expression which determines whether, and in which direction,
4238 to pad out an argument with extra space. The value should be of type
4239 @code{enum direction}: either @code{upward} to pad above the argument,
4240 @code{downward} to pad below, or @code{none} to inhibit padding.
4242 The @emph{amount} of padding is always just enough to reach the next
4243 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
4246 This macro has a default definition which is right for most systems.
4247 For little-endian machines, the default is to pad upward. For
4248 big-endian machines, the default is to pad downward for an argument of
4249 constant size shorter than an @code{int}, and upward otherwise.
4252 @defmac PAD_VARARGS_DOWN
4253 If defined, a C expression which determines whether the default
4254 implementation of va_arg will attempt to pad down before reading the
4255 next argument, if that argument is smaller than its aligned space as
4256 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4257 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4260 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4261 Specify padding for the last element of a block move between registers and
4262 memory. @var{first} is nonzero if this is the only element. Defining this
4263 macro allows better control of register function parameters on big-endian
4264 machines, without using @code{PARALLEL} rtl. In particular,
4265 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4266 registers, as there is no longer a "wrong" part of a register; For example,
4267 a three byte aggregate may be passed in the high part of a register if so
4271 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
4272 If defined, a C expression that gives the alignment boundary, in bits,
4273 of an argument with the specified mode and type. If it is not defined,
4274 @code{PARM_BOUNDARY} is used for all arguments.
4277 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4278 A C expression that is nonzero if @var{regno} is the number of a hard
4279 register in which function arguments are sometimes passed. This does
4280 @emph{not} include implicit arguments such as the static chain and
4281 the structure-value address. On many machines, no registers can be
4282 used for this purpose since all function arguments are pushed on the
4286 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (const_tree @var{type})
4287 This hook should return true if parameter of type @var{type} are passed
4288 as two scalar parameters. By default, GCC will attempt to pack complex
4289 arguments into the target's word size. Some ABIs require complex arguments
4290 to be split and treated as their individual components. For example, on
4291 AIX64, complex floats should be passed in a pair of floating point
4292 registers, even though a complex float would fit in one 64-bit floating
4295 The default value of this hook is @code{NULL}, which is treated as always
4299 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4300 This hook returns a type node for @code{va_list} for the target.
4301 The default version of the hook returns @code{void*}.
4304 @deftypefn {Target Hook} int TARGET_ENUM_VA_LIST_P (int @var{idx}, const char **@var{pname}, tree *@var{ptree})
4305 This target hook is used in function @code{c_common_nodes_and_builtins}
4306 to iterate through the target specific builtin types for va_list. The
4307 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4308 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4310 The arguments @var{pname} and @var{ptree} are used to store the result of
4311 this macro and are set to the name of the va_list builtin type and its
4313 If the return value of this macro is zero, then there is no more element.
4314 Otherwise the @var{IDX} should be increased for the next call of this
4315 macro to iterate through all types.
4318 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4319 This hook returns the va_list type of the calling convention specified by
4321 The default version of this hook returns @code{va_list_type_node}.
4324 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4325 This hook returns the va_list type of the calling convention specified by the
4326 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4330 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, gimple_seq *@var{pre_p}, gimple_seq *@var{post_p})
4331 This hook performs target-specific gimplification of
4332 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4333 arguments to @code{va_arg}; the latter two are as in
4334 @code{gimplify.c:gimplify_expr}.
4337 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4338 Define this to return nonzero if the port can handle pointers
4339 with machine mode @var{mode}. The default version of this
4340 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4343 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4344 Define this to return nonzero if the port is prepared to handle
4345 insns involving scalar mode @var{mode}. For a scalar mode to be
4346 considered supported, all the basic arithmetic and comparisons
4349 The default version of this hook returns true for any mode
4350 required to handle the basic C types (as defined by the port).
4351 Included here are the double-word arithmetic supported by the
4352 code in @file{optabs.c}.
4355 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4356 Define this to return nonzero if the port is prepared to handle
4357 insns involving vector mode @var{mode}. At the very least, it
4358 must have move patterns for this mode.
4361 @deftypefn {Target Hook} bool TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P (enum machine_mode @var{mode})
4362 Define this to return nonzero for machine modes for which the port has
4363 small register classes. If this target hook returns nonzero for a given
4364 @var{mode}, the compiler will try to minimize the lifetime of registers
4365 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4366 In this case, the hook is expected to return nonzero if it returns nonzero
4369 On some machines, it is risky to let hard registers live across arbitrary
4370 insns. Typically, these machines have instructions that require values
4371 to be in specific registers (like an accumulator), and reload will fail
4372 if the required hard register is used for another purpose across such an
4375 Passes before reload do not know which hard registers will be used
4376 in an instruction, but the machine modes of the registers set or used in
4377 the instruction are already known. And for some machines, register
4378 classes are small for, say, integer registers but not for floating point
4379 registers. For example, the AMD x86-64 architecture requires specific
4380 registers for the legacy x86 integer instructions, but there are many
4381 SSE registers for floating point operations. On such targets, a good
4382 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4383 machine modes but zero for the SSE register classes.
4385 The default version of this hook retuns false for any mode. It is always
4386 safe to redefine this hook to return with a nonzero value. But if you
4387 unnecessarily define it, you will reduce the amount of optimizations
4388 that can be performed in some cases. If you do not define this hook
4389 to return a nonzero value when it is required, the compiler will run out
4390 of spill registers and print a fatal error message.
4394 @subsection How Scalar Function Values Are Returned
4395 @cindex return values in registers
4396 @cindex values, returned by functions
4397 @cindex scalars, returned as values
4399 This section discusses the macros that control returning scalars as
4400 values---values that can fit in registers.
4402 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
4404 Define this to return an RTX representing the place where a function
4405 returns or receives a value of data type @var{ret_type}, a tree node
4406 representing a data type. @var{fn_decl_or_type} is a tree node
4407 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4408 function being called. If @var{outgoing} is false, the hook should
4409 compute the register in which the caller will see the return value.
4410 Otherwise, the hook should return an RTX representing the place where
4411 a function returns a value.
4413 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4414 (Actually, on most machines, scalar values are returned in the same
4415 place regardless of mode.) The value of the expression is usually a
4416 @code{reg} RTX for the hard register where the return value is stored.
4417 The value can also be a @code{parallel} RTX, if the return value is in
4418 multiple places. See @code{FUNCTION_ARG} for an explanation of the
4419 @code{parallel} form. Note that the callee will populate every
4420 location specified in the @code{parallel}, but if the first element of
4421 the @code{parallel} contains the whole return value, callers will use
4422 that element as the canonical location and ignore the others. The m68k
4423 port uses this type of @code{parallel} to return pointers in both
4424 @samp{%a0} (the canonical location) and @samp{%d0}.
4426 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4427 the same promotion rules specified in @code{PROMOTE_MODE} if
4428 @var{valtype} is a scalar type.
4430 If the precise function being called is known, @var{func} is a tree
4431 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4432 pointer. This makes it possible to use a different value-returning
4433 convention for specific functions when all their calls are
4436 Some target machines have ``register windows'' so that the register in
4437 which a function returns its value is not the same as the one in which
4438 the caller sees the value. For such machines, you should return
4439 different RTX depending on @var{outgoing}.
4441 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4442 aggregate data types, because these are returned in another way. See
4443 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4446 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4447 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4448 a new target instead.
4451 @defmac LIBCALL_VALUE (@var{mode})
4452 A C expression to create an RTX representing the place where a library
4453 function returns a value of mode @var{mode}.
4455 Note that ``library function'' in this context means a compiler
4456 support routine, used to perform arithmetic, whose name is known
4457 specially by the compiler and was not mentioned in the C code being
4461 @deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (enum machine_mode @var{mode}, const_rtx @var{fun})
4462 Define this hook if the back-end needs to know the name of the libcall
4463 function in order to determine where the result should be returned.
4465 The mode of the result is given by @var{mode} and the name of the called
4466 library function is given by @var{fun}. The hook should return an RTX
4467 representing the place where the library function result will be returned.
4469 If this hook is not defined, then LIBCALL_VALUE will be used.
4472 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4473 A C expression that is nonzero if @var{regno} is the number of a hard
4474 register in which the values of called function may come back.
4476 A register whose use for returning values is limited to serving as the
4477 second of a pair (for a value of type @code{double}, say) need not be
4478 recognized by this macro. So for most machines, this definition
4482 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4485 If the machine has register windows, so that the caller and the called
4486 function use different registers for the return value, this macro
4487 should recognize only the caller's register numbers.
4489 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4490 for a new target instead.
4493 @deftypefn {Target Hook} bool TARGET_FUNCTION_VALUE_REGNO_P (const unsigned int @var{regno})
4494 A target hook that return @code{true} if @var{regno} is the number of a hard
4495 register in which the values of called function may come back.
4497 A register whose use for returning values is limited to serving as the
4498 second of a pair (for a value of type @code{double}, say) need not be
4499 recognized by this target hook.
4501 If the machine has register windows, so that the caller and the called
4502 function use different registers for the return value, this target hook
4503 should recognize only the caller's register numbers.
4505 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4508 @defmac APPLY_RESULT_SIZE
4509 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4510 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4511 saving and restoring an arbitrary return value.
4514 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (const_tree @var{type})
4515 This hook should return true if values of type @var{type} are returned
4516 at the most significant end of a register (in other words, if they are
4517 padded at the least significant end). You can assume that @var{type}
4518 is returned in a register; the caller is required to check this.
4520 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4521 be able to hold the complete return value. For example, if a 1-, 2-
4522 or 3-byte structure is returned at the most significant end of a
4523 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4527 @node Aggregate Return
4528 @subsection How Large Values Are Returned
4529 @cindex aggregates as return values
4530 @cindex large return values
4531 @cindex returning aggregate values
4532 @cindex structure value address
4534 When a function value's mode is @code{BLKmode} (and in some other
4535 cases), the value is not returned according to
4536 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4537 caller passes the address of a block of memory in which the value
4538 should be stored. This address is called the @dfn{structure value
4541 This section describes how to control returning structure values in
4544 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (const_tree @var{type}, const_tree @var{fntype})
4545 This target hook should return a nonzero value to say to return the
4546 function value in memory, just as large structures are always returned.
4547 Here @var{type} will be the data type of the value, and @var{fntype}
4548 will be the type of the function doing the returning, or @code{NULL} for
4551 Note that values of mode @code{BLKmode} must be explicitly handled
4552 by this function. Also, the option @option{-fpcc-struct-return}
4553 takes effect regardless of this macro. On most systems, it is
4554 possible to leave the hook undefined; this causes a default
4555 definition to be used, whose value is the constant 1 for @code{BLKmode}
4556 values, and 0 otherwise.
4558 Do not use this hook to indicate that structures and unions should always
4559 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4563 @defmac DEFAULT_PCC_STRUCT_RETURN
4564 Define this macro to be 1 if all structure and union return values must be
4565 in memory. Since this results in slower code, this should be defined
4566 only if needed for compatibility with other compilers or with an ABI@.
4567 If you define this macro to be 0, then the conventions used for structure
4568 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4571 If not defined, this defaults to the value 1.
4574 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4575 This target hook should return the location of the structure value
4576 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4577 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4578 be @code{NULL}, for libcalls. You do not need to define this target
4579 hook if the address is always passed as an ``invisible'' first
4582 On some architectures the place where the structure value address
4583 is found by the called function is not the same place that the
4584 caller put it. This can be due to register windows, or it could
4585 be because the function prologue moves it to a different place.
4586 @var{incoming} is @code{1} or @code{2} when the location is needed in
4587 the context of the called function, and @code{0} in the context of
4590 If @var{incoming} is nonzero and the address is to be found on the
4591 stack, return a @code{mem} which refers to the frame pointer. If
4592 @var{incoming} is @code{2}, the result is being used to fetch the
4593 structure value address at the beginning of a function. If you need
4594 to emit adjusting code, you should do it at this point.
4597 @defmac PCC_STATIC_STRUCT_RETURN
4598 Define this macro if the usual system convention on the target machine
4599 for returning structures and unions is for the called function to return
4600 the address of a static variable containing the value.
4602 Do not define this if the usual system convention is for the caller to
4603 pass an address to the subroutine.
4605 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4606 nothing when you use @option{-freg-struct-return} mode.
4610 @subsection Caller-Saves Register Allocation
4612 If you enable it, GCC can save registers around function calls. This
4613 makes it possible to use call-clobbered registers to hold variables that
4614 must live across calls.
4616 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4617 A C expression to determine whether it is worthwhile to consider placing
4618 a pseudo-register in a call-clobbered hard register and saving and
4619 restoring it around each function call. The expression should be 1 when
4620 this is worth doing, and 0 otherwise.
4622 If you don't define this macro, a default is used which is good on most
4623 machines: @code{4 * @var{calls} < @var{refs}}.
4626 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4627 A C expression specifying which mode is required for saving @var{nregs}
4628 of a pseudo-register in call-clobbered hard register @var{regno}. If
4629 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4630 returned. For most machines this macro need not be defined since GCC
4631 will select the smallest suitable mode.
4634 @node Function Entry
4635 @subsection Function Entry and Exit
4636 @cindex function entry and exit
4640 This section describes the macros that output function entry
4641 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4643 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4644 If defined, a function that outputs the assembler code for entry to a
4645 function. The prologue is responsible for setting up the stack frame,
4646 initializing the frame pointer register, saving registers that must be
4647 saved, and allocating @var{size} additional bytes of storage for the
4648 local variables. @var{size} is an integer. @var{file} is a stdio
4649 stream to which the assembler code should be output.
4651 The label for the beginning of the function need not be output by this
4652 macro. That has already been done when the macro is run.
4654 @findex regs_ever_live
4655 To determine which registers to save, the macro can refer to the array
4656 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4657 @var{r} is used anywhere within the function. This implies the function
4658 prologue should save register @var{r}, provided it is not one of the
4659 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4660 @code{regs_ever_live}.)
4662 On machines that have ``register windows'', the function entry code does
4663 not save on the stack the registers that are in the windows, even if
4664 they are supposed to be preserved by function calls; instead it takes
4665 appropriate steps to ``push'' the register stack, if any non-call-used
4666 registers are used in the function.
4668 @findex frame_pointer_needed
4669 On machines where functions may or may not have frame-pointers, the
4670 function entry code must vary accordingly; it must set up the frame
4671 pointer if one is wanted, and not otherwise. To determine whether a
4672 frame pointer is in wanted, the macro can refer to the variable
4673 @code{frame_pointer_needed}. The variable's value will be 1 at run
4674 time in a function that needs a frame pointer. @xref{Elimination}.
4676 The function entry code is responsible for allocating any stack space
4677 required for the function. This stack space consists of the regions
4678 listed below. In most cases, these regions are allocated in the
4679 order listed, with the last listed region closest to the top of the
4680 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4681 the highest address if it is not defined). You can use a different order
4682 for a machine if doing so is more convenient or required for
4683 compatibility reasons. Except in cases where required by standard
4684 or by a debugger, there is no reason why the stack layout used by GCC
4685 need agree with that used by other compilers for a machine.
4688 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4689 If defined, a function that outputs assembler code at the end of a
4690 prologue. This should be used when the function prologue is being
4691 emitted as RTL, and you have some extra assembler that needs to be
4692 emitted. @xref{prologue instruction pattern}.
4695 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4696 If defined, a function that outputs assembler code at the start of an
4697 epilogue. This should be used when the function epilogue is being
4698 emitted as RTL, and you have some extra assembler that needs to be
4699 emitted. @xref{epilogue instruction pattern}.
4702 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4703 If defined, a function that outputs the assembler code for exit from a
4704 function. The epilogue is responsible for restoring the saved
4705 registers and stack pointer to their values when the function was
4706 called, and returning control to the caller. This macro takes the
4707 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4708 registers to restore are determined from @code{regs_ever_live} and
4709 @code{CALL_USED_REGISTERS} in the same way.
4711 On some machines, there is a single instruction that does all the work
4712 of returning from the function. On these machines, give that
4713 instruction the name @samp{return} and do not define the macro
4714 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4716 Do not define a pattern named @samp{return} if you want the
4717 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4718 switches to control whether return instructions or epilogues are used,
4719 define a @samp{return} pattern with a validity condition that tests the
4720 target switches appropriately. If the @samp{return} pattern's validity
4721 condition is false, epilogues will be used.
4723 On machines where functions may or may not have frame-pointers, the
4724 function exit code must vary accordingly. Sometimes the code for these
4725 two cases is completely different. To determine whether a frame pointer
4726 is wanted, the macro can refer to the variable
4727 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4728 a function that needs a frame pointer.
4730 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4731 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4732 The C variable @code{current_function_is_leaf} is nonzero for such a
4733 function. @xref{Leaf Functions}.
4735 On some machines, some functions pop their arguments on exit while
4736 others leave that for the caller to do. For example, the 68020 when
4737 given @option{-mrtd} pops arguments in functions that take a fixed
4738 number of arguments.
4740 @findex current_function_pops_args
4741 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4742 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4743 needs to know what was decided. The number of bytes of the current
4744 function's arguments that this function should pop is available in
4745 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4750 @findex current_function_pretend_args_size
4751 A region of @code{current_function_pretend_args_size} bytes of
4752 uninitialized space just underneath the first argument arriving on the
4753 stack. (This may not be at the very start of the allocated stack region
4754 if the calling sequence has pushed anything else since pushing the stack
4755 arguments. But usually, on such machines, nothing else has been pushed
4756 yet, because the function prologue itself does all the pushing.) This
4757 region is used on machines where an argument may be passed partly in
4758 registers and partly in memory, and, in some cases to support the
4759 features in @code{<stdarg.h>}.
4762 An area of memory used to save certain registers used by the function.
4763 The size of this area, which may also include space for such things as
4764 the return address and pointers to previous stack frames, is
4765 machine-specific and usually depends on which registers have been used
4766 in the function. Machines with register windows often do not require
4770 A region of at least @var{size} bytes, possibly rounded up to an allocation
4771 boundary, to contain the local variables of the function. On some machines,
4772 this region and the save area may occur in the opposite order, with the
4773 save area closer to the top of the stack.
4776 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4777 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4778 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4779 argument lists of the function. @xref{Stack Arguments}.
4782 @defmac EXIT_IGNORE_STACK
4783 Define this macro as a C expression that is nonzero if the return
4784 instruction or the function epilogue ignores the value of the stack
4785 pointer; in other words, if it is safe to delete an instruction to
4786 adjust the stack pointer before a return from the function. The
4789 Note that this macro's value is relevant only for functions for which
4790 frame pointers are maintained. It is never safe to delete a final
4791 stack adjustment in a function that has no frame pointer, and the
4792 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4795 @defmac EPILOGUE_USES (@var{regno})
4796 Define this macro as a C expression that is nonzero for registers that are
4797 used by the epilogue or the @samp{return} pattern. The stack and frame
4798 pointer registers are already assumed to be used as needed.
4801 @defmac EH_USES (@var{regno})
4802 Define this macro as a C expression that is nonzero for registers that are
4803 used by the exception handling mechanism, and so should be considered live
4804 on entry to an exception edge.
4807 @defmac DELAY_SLOTS_FOR_EPILOGUE
4808 Define this macro if the function epilogue contains delay slots to which
4809 instructions from the rest of the function can be ``moved''. The
4810 definition should be a C expression whose value is an integer
4811 representing the number of delay slots there.
4814 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4815 A C expression that returns 1 if @var{insn} can be placed in delay
4816 slot number @var{n} of the epilogue.
4818 The argument @var{n} is an integer which identifies the delay slot now
4819 being considered (since different slots may have different rules of
4820 eligibility). It is never negative and is always less than the number
4821 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4822 If you reject a particular insn for a given delay slot, in principle, it
4823 may be reconsidered for a subsequent delay slot. Also, other insns may
4824 (at least in principle) be considered for the so far unfilled delay
4827 @findex current_function_epilogue_delay_list
4828 @findex final_scan_insn
4829 The insns accepted to fill the epilogue delay slots are put in an RTL
4830 list made with @code{insn_list} objects, stored in the variable
4831 @code{current_function_epilogue_delay_list}. The insn for the first
4832 delay slot comes first in the list. Your definition of the macro
4833 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4834 outputting the insns in this list, usually by calling
4835 @code{final_scan_insn}.
4837 You need not define this macro if you did not define
4838 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4841 @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})
4842 A function that outputs the assembler code for a thunk
4843 function, used to implement C++ virtual function calls with multiple
4844 inheritance. The thunk acts as a wrapper around a virtual function,
4845 adjusting the implicit object parameter before handing control off to
4848 First, emit code to add the integer @var{delta} to the location that
4849 contains the incoming first argument. Assume that this argument
4850 contains a pointer, and is the one used to pass the @code{this} pointer
4851 in C++. This is the incoming argument @emph{before} the function prologue,
4852 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4853 all other incoming arguments.
4855 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4856 made after adding @code{delta}. In particular, if @var{p} is the
4857 adjusted pointer, the following adjustment should be made:
4860 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4863 After the additions, emit code to jump to @var{function}, which is a
4864 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4865 not touch the return address. Hence returning from @var{FUNCTION} will
4866 return to whoever called the current @samp{thunk}.
4868 The effect must be as if @var{function} had been called directly with
4869 the adjusted first argument. This macro is responsible for emitting all
4870 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4871 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4873 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4874 have already been extracted from it.) It might possibly be useful on
4875 some targets, but probably not.
4877 If you do not define this macro, the target-independent code in the C++
4878 front end will generate a less efficient heavyweight thunk that calls
4879 @var{function} instead of jumping to it. The generic approach does
4880 not support varargs.
4883 @deftypefn {Target Hook} bool TARGET_ASM_CAN_OUTPUT_MI_THUNK (const_tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, const_tree @var{function})
4884 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4885 to output the assembler code for the thunk function specified by the
4886 arguments it is passed, and false otherwise. In the latter case, the
4887 generic approach will be used by the C++ front end, with the limitations
4892 @subsection Generating Code for Profiling
4893 @cindex profiling, code generation
4895 These macros will help you generate code for profiling.
4897 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4898 A C statement or compound statement to output to @var{file} some
4899 assembler code to call the profiling subroutine @code{mcount}.
4902 The details of how @code{mcount} expects to be called are determined by
4903 your operating system environment, not by GCC@. To figure them out,
4904 compile a small program for profiling using the system's installed C
4905 compiler and look at the assembler code that results.
4907 Older implementations of @code{mcount} expect the address of a counter
4908 variable to be loaded into some register. The name of this variable is
4909 @samp{LP} followed by the number @var{labelno}, so you would generate
4910 the name using @samp{LP%d} in a @code{fprintf}.
4913 @defmac PROFILE_HOOK
4914 A C statement or compound statement to output to @var{file} some assembly
4915 code to call the profiling subroutine @code{mcount} even the target does
4916 not support profiling.
4919 @defmac NO_PROFILE_COUNTERS
4920 Define this macro to be an expression with a nonzero value if the
4921 @code{mcount} subroutine on your system does not need a counter variable
4922 allocated for each function. This is true for almost all modern
4923 implementations. If you define this macro, you must not use the
4924 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4927 @defmac PROFILE_BEFORE_PROLOGUE
4928 Define this macro if the code for function profiling should come before
4929 the function prologue. Normally, the profiling code comes after.
4933 @subsection Permitting tail calls
4936 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4937 True if it is ok to do sibling call optimization for the specified
4938 call expression @var{exp}. @var{decl} will be the called function,
4939 or @code{NULL} if this is an indirect call.
4941 It is not uncommon for limitations of calling conventions to prevent
4942 tail calls to functions outside the current unit of translation, or
4943 during PIC compilation. The hook is used to enforce these restrictions,
4944 as the @code{sibcall} md pattern can not fail, or fall over to a
4945 ``normal'' call. The criteria for successful sibling call optimization
4946 may vary greatly between different architectures.
4949 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap @var{regs})
4950 Add any hard registers to @var{regs} that are live on entry to the
4951 function. This hook only needs to be defined to provide registers that
4952 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4953 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4954 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4955 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4958 @node Stack Smashing Protection
4959 @subsection Stack smashing protection
4960 @cindex stack smashing protection
4962 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4963 This hook returns a @code{DECL} node for the external variable to use
4964 for the stack protection guard. This variable is initialized by the
4965 runtime to some random value and is used to initialize the guard value
4966 that is placed at the top of the local stack frame. The type of this
4967 variable must be @code{ptr_type_node}.
4969 The default version of this hook creates a variable called
4970 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4973 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4974 This hook returns a tree expression that alerts the runtime that the
4975 stack protect guard variable has been modified. This expression should
4976 involve a call to a @code{noreturn} function.
4978 The default version of this hook invokes a function called
4979 @samp{__stack_chk_fail}, taking no arguments. This function is
4980 normally defined in @file{libgcc2.c}.
4983 @deftypefn {Target Hook} bool TARGET_SUPPORTS_SPLIT_STACK (bool)
4984 Whether this target supports splitting the stack. This is called after options have been parsed, so the target may reject splitting the stack in some configurations. The default version of this hook returns false. If @var{report} is true, this function may issue a warning or error; if @var{report} is false, it must simply return a value
4988 @section Implementing the Varargs Macros
4989 @cindex varargs implementation
4991 GCC comes with an implementation of @code{<varargs.h>} and
4992 @code{<stdarg.h>} that work without change on machines that pass arguments
4993 on the stack. Other machines require their own implementations of
4994 varargs, and the two machine independent header files must have
4995 conditionals to include it.
4997 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4998 the calling convention for @code{va_start}. The traditional
4999 implementation takes just one argument, which is the variable in which
5000 to store the argument pointer. The ISO implementation of
5001 @code{va_start} takes an additional second argument. The user is
5002 supposed to write the last named argument of the function here.
5004 However, @code{va_start} should not use this argument. The way to find
5005 the end of the named arguments is with the built-in functions described
5008 @defmac __builtin_saveregs ()
5009 Use this built-in function to save the argument registers in memory so
5010 that the varargs mechanism can access them. Both ISO and traditional
5011 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
5012 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
5014 On some machines, @code{__builtin_saveregs} is open-coded under the
5015 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
5016 other machines, it calls a routine written in assembler language,
5017 found in @file{libgcc2.c}.
5019 Code generated for the call to @code{__builtin_saveregs} appears at the
5020 beginning of the function, as opposed to where the call to
5021 @code{__builtin_saveregs} is written, regardless of what the code is.
5022 This is because the registers must be saved before the function starts
5023 to use them for its own purposes.
5024 @c i rewrote the first sentence above to fix an overfull hbox. --mew
5028 @defmac __builtin_next_arg (@var{lastarg})
5029 This builtin returns the address of the first anonymous stack
5030 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
5031 returns the address of the location above the first anonymous stack
5032 argument. Use it in @code{va_start} to initialize the pointer for
5033 fetching arguments from the stack. Also use it in @code{va_start} to
5034 verify that the second parameter @var{lastarg} is the last named argument
5035 of the current function.
5038 @defmac __builtin_classify_type (@var{object})
5039 Since each machine has its own conventions for which data types are
5040 passed in which kind of register, your implementation of @code{va_arg}
5041 has to embody these conventions. The easiest way to categorize the
5042 specified data type is to use @code{__builtin_classify_type} together
5043 with @code{sizeof} and @code{__alignof__}.
5045 @code{__builtin_classify_type} ignores the value of @var{object},
5046 considering only its data type. It returns an integer describing what
5047 kind of type that is---integer, floating, pointer, structure, and so on.
5049 The file @file{typeclass.h} defines an enumeration that you can use to
5050 interpret the values of @code{__builtin_classify_type}.
5053 These machine description macros help implement varargs:
5055 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
5056 If defined, this hook produces the machine-specific code for a call to
5057 @code{__builtin_saveregs}. This code will be moved to the very
5058 beginning of the function, before any parameter access are made. The
5059 return value of this function should be an RTX that contains the value
5060 to use as the return of @code{__builtin_saveregs}.
5063 @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})
5064 This target hook offers an alternative to using
5065 @code{__builtin_saveregs} and defining the hook
5066 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5067 register arguments into the stack so that all the arguments appear to
5068 have been passed consecutively on the stack. Once this is done, you can
5069 use the standard implementation of varargs that works for machines that
5070 pass all their arguments on the stack.
5072 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5073 structure, containing the values that are obtained after processing the
5074 named arguments. The arguments @var{mode} and @var{type} describe the
5075 last named argument---its machine mode and its data type as a tree node.
5077 The target hook should do two things: first, push onto the stack all the
5078 argument registers @emph{not} used for the named arguments, and second,
5079 store the size of the data thus pushed into the @code{int}-valued
5080 variable pointed to by @var{pretend_args_size}. The value that you
5081 store here will serve as additional offset for setting up the stack
5084 Because you must generate code to push the anonymous arguments at
5085 compile time without knowing their data types,
5086 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5087 have just a single category of argument register and use it uniformly
5090 If the argument @var{second_time} is nonzero, it means that the
5091 arguments of the function are being analyzed for the second time. This
5092 happens for an inline function, which is not actually compiled until the
5093 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5094 not generate any instructions in this case.
5097 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
5098 Define this hook to return @code{true} if the location where a function
5099 argument is passed depends on whether or not it is a named argument.
5101 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
5102 is set for varargs and stdarg functions. If this hook returns
5103 @code{true}, the @var{named} argument is always true for named
5104 arguments, and false for unnamed arguments. If it returns @code{false},
5105 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5106 then all arguments are treated as named. Otherwise, all named arguments
5107 except the last are treated as named.
5109 You need not define this hook if it always returns @code{false}.
5112 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED (CUMULATIVE_ARGS *@var{ca})
5113 If you need to conditionally change ABIs so that one works with
5114 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5115 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5116 defined, then define this hook to return @code{true} if
5117 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5118 Otherwise, you should not define this hook.
5122 @section Trampolines for Nested Functions
5123 @cindex trampolines for nested functions
5124 @cindex nested functions, trampolines for
5126 A @dfn{trampoline} is a small piece of code that is created at run time
5127 when the address of a nested function is taken. It normally resides on
5128 the stack, in the stack frame of the containing function. These macros
5129 tell GCC how to generate code to allocate and initialize a
5132 The instructions in the trampoline must do two things: load a constant
5133 address into the static chain register, and jump to the real address of
5134 the nested function. On CISC machines such as the m68k, this requires
5135 two instructions, a move immediate and a jump. Then the two addresses
5136 exist in the trampoline as word-long immediate operands. On RISC
5137 machines, it is often necessary to load each address into a register in
5138 two parts. Then pieces of each address form separate immediate
5141 The code generated to initialize the trampoline must store the variable
5142 parts---the static chain value and the function address---into the
5143 immediate operands of the instructions. On a CISC machine, this is
5144 simply a matter of copying each address to a memory reference at the
5145 proper offset from the start of the trampoline. On a RISC machine, it
5146 may be necessary to take out pieces of the address and store them
5149 @deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f})
5150 This hook is called by @code{assemble_trampoline_template} to output,
5151 on the stream @var{f}, assembler code for a block of data that contains
5152 the constant parts of a trampoline. This code should not include a
5153 label---the label is taken care of automatically.
5155 If you do not define this hook, it means no template is needed
5156 for the target. Do not define this hook on systems where the block move
5157 code to copy the trampoline into place would be larger than the code
5158 to generate it on the spot.
5161 @defmac TRAMPOLINE_SECTION
5162 Return the section into which the trampoline template is to be placed
5163 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5166 @defmac TRAMPOLINE_SIZE
5167 A C expression for the size in bytes of the trampoline, as an integer.
5170 @defmac TRAMPOLINE_ALIGNMENT
5171 Alignment required for trampolines, in bits.
5173 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5174 is used for aligning trampolines.
5177 @deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain})
5178 This hook is called to initialize a trampoline.
5179 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5180 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5181 RTX for the static chain value that should be passed to the function
5184 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5185 first thing this hook should do is emit a block move into @var{m_tramp}
5186 from the memory block returned by @code{assemble_trampoline_template}.
5187 Note that the block move need only cover the constant parts of the
5188 trampoline. If the target isolates the variable parts of the trampoline
5189 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5191 If the target requires any other actions, such as flushing caches or
5192 enabling stack execution, these actions should be performed after
5193 initializing the trampoline proper.
5196 @deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr})
5197 This hook should perform any machine-specific adjustment in
5198 the address of the trampoline. Its argument contains the address of the
5199 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5200 the address to be used for a function call should be different from the
5201 address at which the template was stored, the different address should
5202 be returned; otherwise @var{addr} should be returned unchanged.
5203 If this hook is not defined, @var{addr} will be used for function calls.
5206 Implementing trampolines is difficult on many machines because they have
5207 separate instruction and data caches. Writing into a stack location
5208 fails to clear the memory in the instruction cache, so when the program
5209 jumps to that location, it executes the old contents.
5211 Here are two possible solutions. One is to clear the relevant parts of
5212 the instruction cache whenever a trampoline is set up. The other is to
5213 make all trampolines identical, by having them jump to a standard
5214 subroutine. The former technique makes trampoline execution faster; the
5215 latter makes initialization faster.
5217 To clear the instruction cache when a trampoline is initialized, define
5218 the following macro.
5220 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5221 If defined, expands to a C expression clearing the @emph{instruction
5222 cache} in the specified interval. The definition of this macro would
5223 typically be a series of @code{asm} statements. Both @var{beg} and
5224 @var{end} are both pointer expressions.
5227 The operating system may also require the stack to be made executable
5228 before calling the trampoline. To implement this requirement, define
5229 the following macro.
5231 @defmac ENABLE_EXECUTE_STACK
5232 Define this macro if certain operations must be performed before executing
5233 code located on the stack. The macro should expand to a series of C
5234 file-scope constructs (e.g.@: functions) and provide a unique entry point
5235 named @code{__enable_execute_stack}. The target is responsible for
5236 emitting calls to the entry point in the code, for example from the
5237 @code{TARGET_TRAMPOLINE_INIT} hook.
5240 To use a standard subroutine, define the following macro. In addition,
5241 you must make sure that the instructions in a trampoline fill an entire
5242 cache line with identical instructions, or else ensure that the
5243 beginning of the trampoline code is always aligned at the same point in
5244 its cache line. Look in @file{m68k.h} as a guide.
5246 @defmac TRANSFER_FROM_TRAMPOLINE
5247 Define this macro if trampolines need a special subroutine to do their
5248 work. The macro should expand to a series of @code{asm} statements
5249 which will be compiled with GCC@. They go in a library function named
5250 @code{__transfer_from_trampoline}.
5252 If you need to avoid executing the ordinary prologue code of a compiled
5253 C function when you jump to the subroutine, you can do so by placing a
5254 special label of your own in the assembler code. Use one @code{asm}
5255 statement to generate an assembler label, and another to make the label
5256 global. Then trampolines can use that label to jump directly to your
5257 special assembler code.
5261 @section Implicit Calls to Library Routines
5262 @cindex library subroutine names
5263 @cindex @file{libgcc.a}
5265 @c prevent bad page break with this line
5266 Here is an explanation of implicit calls to library routines.
5268 @defmac DECLARE_LIBRARY_RENAMES
5269 This macro, if defined, should expand to a piece of C code that will get
5270 expanded when compiling functions for libgcc.a. It can be used to
5271 provide alternate names for GCC's internal library functions if there
5272 are ABI-mandated names that the compiler should provide.
5275 @findex set_optab_libfunc
5276 @findex init_one_libfunc
5277 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5278 This hook should declare additional library routines or rename
5279 existing ones, using the functions @code{set_optab_libfunc} and
5280 @code{init_one_libfunc} defined in @file{optabs.c}.
5281 @code{init_optabs} calls this macro after initializing all the normal
5284 The default is to do nothing. Most ports don't need to define this hook.
5287 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5288 This macro should return @code{true} if the library routine that
5289 implements the floating point comparison operator @var{comparison} in
5290 mode @var{mode} will return a boolean, and @var{false} if it will
5293 GCC's own floating point libraries return tristates from the
5294 comparison operators, so the default returns false always. Most ports
5295 don't need to define this macro.
5298 @defmac TARGET_LIB_INT_CMP_BIASED
5299 This macro should evaluate to @code{true} if the integer comparison
5300 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5301 operand is smaller than the second, 1 to indicate that they are equal,
5302 and 2 to indicate that the first operand is greater than the second.
5303 If this macro evaluates to @code{false} the comparison functions return
5304 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5305 in @file{libgcc.a}, you do not need to define this macro.
5308 @cindex US Software GOFAST, floating point emulation library
5309 @cindex floating point emulation library, US Software GOFAST
5310 @cindex GOFAST, floating point emulation library
5311 @findex gofast_maybe_init_libfuncs
5312 @defmac US_SOFTWARE_GOFAST
5313 Define this macro if your system C library uses the US Software GOFAST
5314 library to provide floating point emulation.
5316 In addition to defining this macro, your architecture must set
5317 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
5318 else call that function from its version of that hook. It is defined
5319 in @file{config/gofast.h}, which must be included by your
5320 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
5323 If this macro is defined, the
5324 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
5325 false for @code{SFmode} and @code{DFmode} comparisons.
5328 @cindex @code{EDOM}, implicit usage
5331 The value of @code{EDOM} on the target machine, as a C integer constant
5332 expression. If you don't define this macro, GCC does not attempt to
5333 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5334 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5337 If you do not define @code{TARGET_EDOM}, then compiled code reports
5338 domain errors by calling the library function and letting it report the
5339 error. If mathematical functions on your system use @code{matherr} when
5340 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5341 that @code{matherr} is used normally.
5344 @cindex @code{errno}, implicit usage
5345 @defmac GEN_ERRNO_RTX
5346 Define this macro as a C expression to create an rtl expression that
5347 refers to the global ``variable'' @code{errno}. (On certain systems,
5348 @code{errno} may not actually be a variable.) If you don't define this
5349 macro, a reasonable default is used.
5352 @cindex C99 math functions, implicit usage
5353 @defmac TARGET_C99_FUNCTIONS
5354 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5355 @code{sinf} and similarly for other functions defined by C99 standard. The
5356 default is zero because a number of existing systems lack support for these
5357 functions in their runtime so this macro needs to be redefined to one on
5358 systems that do support the C99 runtime.
5361 @cindex sincos math function, implicit usage
5362 @defmac TARGET_HAS_SINCOS
5363 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5364 and @code{cos} with the same argument to a call to @code{sincos}. The
5365 default is zero. The target has to provide the following functions:
5367 void sincos(double x, double *sin, double *cos);
5368 void sincosf(float x, float *sin, float *cos);
5369 void sincosl(long double x, long double *sin, long double *cos);
5373 @defmac NEXT_OBJC_RUNTIME
5374 Define this macro to generate code for Objective-C message sending using
5375 the calling convention of the NeXT system. This calling convention
5376 involves passing the object, the selector and the method arguments all
5377 at once to the method-lookup library function.
5379 The default calling convention passes just the object and the selector
5380 to the lookup function, which returns a pointer to the method.
5383 @node Addressing Modes
5384 @section Addressing Modes
5385 @cindex addressing modes
5387 @c prevent bad page break with this line
5388 This is about addressing modes.
5390 @defmac HAVE_PRE_INCREMENT
5391 @defmacx HAVE_PRE_DECREMENT
5392 @defmacx HAVE_POST_INCREMENT
5393 @defmacx HAVE_POST_DECREMENT
5394 A C expression that is nonzero if the machine supports pre-increment,
5395 pre-decrement, post-increment, or post-decrement addressing respectively.
5398 @defmac HAVE_PRE_MODIFY_DISP
5399 @defmacx HAVE_POST_MODIFY_DISP
5400 A C expression that is nonzero if the machine supports pre- or
5401 post-address side-effect generation involving constants other than
5402 the size of the memory operand.
5405 @defmac HAVE_PRE_MODIFY_REG
5406 @defmacx HAVE_POST_MODIFY_REG
5407 A C expression that is nonzero if the machine supports pre- or
5408 post-address side-effect generation involving a register displacement.
5411 @defmac CONSTANT_ADDRESS_P (@var{x})
5412 A C expression that is 1 if the RTX @var{x} is a constant which
5413 is a valid address. On most machines the default definition of
5414 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5415 is acceptable, but a few machines are more restrictive as to which
5416 constant addresses are supported.
5419 @defmac CONSTANT_P (@var{x})
5420 @code{CONSTANT_P}, which is defined by target-independent code,
5421 accepts integer-values expressions whose values are not explicitly
5422 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5423 expressions and @code{const} arithmetic expressions, in addition to
5424 @code{const_int} and @code{const_double} expressions.
5427 @defmac MAX_REGS_PER_ADDRESS
5428 A number, the maximum number of registers that can appear in a valid
5429 memory address. Note that it is up to you to specify a value equal to
5430 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5434 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_ADDRESS_P (enum machine_mode @var{mode}, rtx @var{x}, bool @var{strict})
5435 A function that returns whether @var{x} (an RTX) is a legitimate memory
5436 address on the target machine for a memory operand of mode @var{mode}.
5438 Legitimate addresses are defined in two variants: a strict variant and a
5439 non-strict one. The @var{strict} parameter chooses which variant is
5440 desired by the caller.
5442 The strict variant is used in the reload pass. It must be defined so
5443 that any pseudo-register that has not been allocated a hard register is
5444 considered a memory reference. This is because in contexts where some
5445 kind of register is required, a pseudo-register with no hard register
5446 must be rejected. For non-hard registers, the strict variant should look
5447 up the @code{reg_renumber} array; it should then proceed using the hard
5448 register number in the array, or treat the pseudo as a memory reference
5449 if the array holds @code{-1}.
5451 The non-strict variant is used in other passes. It must be defined to
5452 accept all pseudo-registers in every context where some kind of
5453 register is required.
5455 Normally, constant addresses which are the sum of a @code{symbol_ref}
5456 and an integer are stored inside a @code{const} RTX to mark them as
5457 constant. Therefore, there is no need to recognize such sums
5458 specifically as legitimate addresses. Normally you would simply
5459 recognize any @code{const} as legitimate.
5461 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5462 sums that are not marked with @code{const}. It assumes that a naked
5463 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5464 naked constant sums as illegitimate addresses, so that none of them will
5465 be given to @code{PRINT_OPERAND_ADDRESS}.
5467 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5468 On some machines, whether a symbolic address is legitimate depends on
5469 the section that the address refers to. On these machines, define the
5470 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5471 into the @code{symbol_ref}, and then check for it here. When you see a
5472 @code{const}, you will have to look inside it to find the
5473 @code{symbol_ref} in order to determine the section. @xref{Assembler
5476 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5477 Some ports are still using a deprecated legacy substitute for
5478 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5482 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5486 and should @code{goto @var{label}} if the address @var{x} is a valid
5487 address on the target machine for a memory operand of mode @var{mode}.
5488 Whether the strict or non-strict variants are desired is defined by
5489 the @code{REG_OK_STRICT} macro introduced earlier in this section.
5490 Using the hook is usually simpler because it limits the number of
5491 files that are recompiled when changes are made.
5494 @defmac TARGET_MEM_CONSTRAINT
5495 A single character to be used instead of the default @code{'m'}
5496 character for general memory addresses. This defines the constraint
5497 letter which matches the memory addresses accepted by
5498 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5499 support new address formats in your back end without changing the
5500 semantics of the @code{'m'} constraint. This is necessary in order to
5501 preserve functionality of inline assembly constructs using the
5502 @code{'m'} constraint.
5505 @defmac FIND_BASE_TERM (@var{x})
5506 A C expression to determine the base term of address @var{x},
5507 or to provide a simplified version of @var{x} from which @file{alias.c}
5508 can easily find the base term. This macro is used in only two places:
5509 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5511 It is always safe for this macro to not be defined. It exists so
5512 that alias analysis can understand machine-dependent addresses.
5514 The typical use of this macro is to handle addresses containing
5515 a label_ref or symbol_ref within an UNSPEC@.
5518 @deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, enum machine_mode @var{mode})
5519 This hook is given an invalid memory address @var{x} for an
5520 operand of mode @var{mode} and should try to return a valid memory
5523 @findex break_out_memory_refs
5524 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5525 and @var{oldx} will be the operand that was given to that function to produce
5528 The code of the hook should not alter the substructure of
5529 @var{x}. If it transforms @var{x} into a more legitimate form, it
5530 should return the new @var{x}.
5532 It is not necessary for this hook to come up with a legitimate address.
5533 The compiler has standard ways of doing so in all cases. In fact, it
5534 is safe to omit this hook or make it return @var{x} if it cannot find
5535 a valid way to legitimize the address. But often a machine-dependent
5536 strategy can generate better code.
5539 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5540 A C compound statement that attempts to replace @var{x}, which is an address
5541 that needs reloading, with a valid memory address for an operand of mode
5542 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5543 It is not necessary to define this macro, but it might be useful for
5544 performance reasons.
5546 For example, on the i386, it is sometimes possible to use a single
5547 reload register instead of two by reloading a sum of two pseudo
5548 registers into a register. On the other hand, for number of RISC
5549 processors offsets are limited so that often an intermediate address
5550 needs to be generated in order to address a stack slot. By defining
5551 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5552 generated for adjacent some stack slots can be made identical, and thus
5555 @emph{Note}: This macro should be used with caution. It is necessary
5556 to know something of how reload works in order to effectively use this,
5557 and it is quite easy to produce macros that build in too much knowledge
5558 of reload internals.
5560 @emph{Note}: This macro must be able to reload an address created by a
5561 previous invocation of this macro. If it fails to handle such addresses
5562 then the compiler may generate incorrect code or abort.
5565 The macro definition should use @code{push_reload} to indicate parts that
5566 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5567 suitable to be passed unaltered to @code{push_reload}.
5569 The code generated by this macro must not alter the substructure of
5570 @var{x}. If it transforms @var{x} into a more legitimate form, it
5571 should assign @var{x} (which will always be a C variable) a new value.
5572 This also applies to parts that you change indirectly by calling
5575 @findex strict_memory_address_p
5576 The macro definition may use @code{strict_memory_address_p} to test if
5577 the address has become legitimate.
5580 If you want to change only a part of @var{x}, one standard way of doing
5581 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5582 single level of rtl. Thus, if the part to be changed is not at the
5583 top level, you'll need to replace first the top level.
5584 It is not necessary for this macro to come up with a legitimate
5585 address; but often a machine-dependent strategy can generate better code.
5588 @deftypefn {Target Hook} bool TARGET_MODE_DEPENDENT_ADDRESS_P (const_rtx @var{addr})
5589 This hook returns @code{true} if memory address @var{addr} can have
5590 different meanings depending on the machine mode of the memory
5591 reference it is used for or if the address is valid for some modes
5594 Autoincrement and autodecrement addresses typically have mode-dependent
5595 effects because the amount of the increment or decrement is the size
5596 of the operand being addressed. Some machines have other mode-dependent
5597 addresses. Many RISC machines have no mode-dependent addresses.
5599 You may assume that @var{addr} is a valid address for the machine.
5601 The default version of this hook returns @code{false}.
5604 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5605 A C statement or compound statement with a conditional @code{goto
5606 @var{label};} executed if memory address @var{x} (an RTX) can have
5607 different meanings depending on the machine mode of the memory
5608 reference it is used for or if the address is valid for some modes
5611 Autoincrement and autodecrement addresses typically have mode-dependent
5612 effects because the amount of the increment or decrement is the size
5613 of the operand being addressed. Some machines have other mode-dependent
5614 addresses. Many RISC machines have no mode-dependent addresses.
5616 You may assume that @var{addr} is a valid address for the machine.
5618 These are obsolete macros, replaced by the
5619 @code{TARGET_MODE_DEPENDENT_ADDRESS_P} target hook.
5622 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5623 A C expression that is nonzero if @var{x} is a legitimate constant for
5624 an immediate operand on the target machine. You can assume that
5625 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5626 @samp{1} is a suitable definition for this macro on machines where
5627 anything @code{CONSTANT_P} is valid.
5630 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5631 This hook is used to undo the possibly obfuscating effects of the
5632 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5633 macros. Some backend implementations of these macros wrap symbol
5634 references inside an @code{UNSPEC} rtx to represent PIC or similar
5635 addressing modes. This target hook allows GCC's optimizers to understand
5636 the semantics of these opaque @code{UNSPEC}s by converting them back
5637 into their original form.
5640 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x})
5641 This hook should return true if @var{x} is of a form that cannot (or
5642 should not) be spilled to the constant pool. The default version of
5643 this hook returns false.
5645 The primary reason to define this hook is to prevent reload from
5646 deciding that a non-legitimate constant would be better reloaded
5647 from the constant pool instead of spilling and reloading a register
5648 holding the constant. This restriction is often true of addresses
5649 of TLS symbols for various targets.
5652 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, const_rtx @var{x})
5653 This hook should return true if pool entries for constant @var{x} can
5654 be placed in an @code{object_block} structure. @var{mode} is the mode
5657 The default version returns false for all constants.
5660 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (unsigned @var{fn}, bool @var{md_fn}, bool @var{sqrt})
5661 This hook should return the DECL of a function that implements reciprocal of
5662 the builtin function with builtin function code @var{fn}, or
5663 @code{NULL_TREE} if such a function is not available. @var{md_fn} is true
5664 when @var{fn} is a code of a machine-dependent builtin function. When
5665 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5666 of a square root function are performed, and only reciprocals of @code{sqrt}
5670 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5671 This hook should return the DECL of a function @var{f} that given an
5672 address @var{addr} as an argument returns a mask @var{m} that can be
5673 used to extract from two vectors the relevant data that resides in
5674 @var{addr} in case @var{addr} is not properly aligned.
5676 The autovectorizer, when vectorizing a load operation from an address
5677 @var{addr} that may be unaligned, will generate two vector loads from
5678 the two aligned addresses around @var{addr}. It then generates a
5679 @code{REALIGN_LOAD} operation to extract the relevant data from the
5680 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5681 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5682 the third argument, @var{OFF}, defines how the data will be extracted
5683 from these two vectors: if @var{OFF} is 0, then the returned vector is
5684 @var{v2}; otherwise, the returned vector is composed from the last
5685 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5686 @var{OFF} elements of @var{v2}.
5688 If this hook is defined, the autovectorizer will generate a call
5689 to @var{f} (using the DECL tree that this hook returns) and will
5690 use the return value of @var{f} as the argument @var{OFF} to
5691 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5692 should comply with the semantics expected by @code{REALIGN_LOAD}
5694 If this hook is not defined, then @var{addr} will be used as
5695 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5696 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5699 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree @var{x})
5700 This hook should return the DECL of a function @var{f} that implements
5701 widening multiplication of the even elements of two input vectors of type @var{x}.
5703 If this hook is defined, the autovectorizer will use it along with the
5704 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5705 widening multiplication in cases that the order of the results does not have to be
5706 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5707 @code{widen_mult_hi/lo} idioms will be used.
5710 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree @var{x})
5711 This hook should return the DECL of a function @var{f} that implements
5712 widening multiplication of the odd elements of two input vectors of type @var{x}.
5714 If this hook is defined, the autovectorizer will use it along with the
5715 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5716 widening multiplication in cases that the order of the results does not have to be
5717 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5718 @code{widen_mult_hi/lo} idioms will be used.
5721 @deftypefn {Target Hook} int TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST (enum vect_cost_for_stmt @var{type_of_cost}, tree @var{vectype}, int @var{misalign})
5722 Returns cost of different scalar or vector statements for vectorization cost model.
5723 For vector memory operations the cost may depend on type (@var{vectype}) and
5724 misalignment value (@var{misalign}).
5727 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE (const_tree @var{type}, bool @var{is_packed})
5728 Return true if vector alignment is reachable (by peeling N iterations) for the given type.
5731 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VEC_PERM (tree @var{type}, tree *@var{mask_element_type})
5732 Target builtin that implements vector permute.
5735 @deftypefn {Target Hook} bool TARGET_VECTORIZE_BUILTIN_VEC_PERM_OK (tree @var{vec_type}, tree @var{mask})
5736 Return true if a vector created for @code{builtin_vec_perm} is valid.
5739 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (unsigned @var{code}, tree @var{dest_type}, tree @var{src_type})
5740 This hook should return the DECL of a function that implements conversion of the
5741 input vector of type @var{src_type} to type @var{dest_type}.
5742 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5743 specifies how the conversion is to be applied
5744 (truncation, rounding, etc.).
5746 If this hook is defined, the autovectorizer will use the
5747 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5748 conversion. Otherwise, it will return @code{NULL_TREE}.
5751 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (tree @var{fndecl}, tree @var{vec_type_out}, tree @var{vec_type_in})
5752 This hook should return the decl of a function that implements the
5753 vectorized variant of the builtin function with builtin function code
5754 @var{code} or @code{NULL_TREE} if such a function is not available.
5755 The value of @var{fndecl} is the builtin function declaration. The
5756 return type of the vectorized function shall be of vector type
5757 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5760 @deftypefn {Target Hook} bool TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT (enum machine_mode @var{mode}, const_tree @var{type}, int @var{misalignment}, bool @var{is_packed})
5761 This hook should return true if the target supports misaligned vector
5762 store/load of a specific factor denoted in the @var{misalignment}
5763 parameter. The vector store/load should be of machine mode @var{mode} and
5764 the elements in the vectors should be of type @var{type}. @var{is_packed}
5765 parameter is true if the memory access is defined in a packed struct.
5768 @deftypefn {Target Hook} {enum machine_mode} TARGET_VECTORIZE_PREFERRED_SIMD_MODE (enum machine_mode @var{mode})
5769 This hook should return the preferred mode for vectorizing scalar
5770 mode @var{mode}. The default is
5771 equal to @code{word_mode}, because the vectorizer can do some
5772 transformations even in absence of specialized @acronym{SIMD} hardware.
5775 @deftypefn {Target Hook} {unsigned int} TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES (void)
5776 This hook should return a mask of sizes that should be iterated over
5777 after trying to autovectorize using the vector size derived from the
5778 mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
5779 The default is zero which means to not iterate over other vector sizes.
5782 @node Anchored Addresses
5783 @section Anchored Addresses
5784 @cindex anchored addresses
5785 @cindex @option{-fsection-anchors}
5787 GCC usually addresses every static object as a separate entity.
5788 For example, if we have:
5792 int foo (void) @{ return a + b + c; @}
5795 the code for @code{foo} will usually calculate three separate symbolic
5796 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5797 it would be better to calculate just one symbolic address and access
5798 the three variables relative to it. The equivalent pseudocode would
5804 register int *xr = &x;
5805 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5809 (which isn't valid C). We refer to shared addresses like @code{x} as
5810 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5812 The hooks below describe the target properties that GCC needs to know
5813 in order to make effective use of section anchors. It won't use
5814 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5815 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5817 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5818 The minimum offset that should be applied to a section anchor.
5819 On most targets, it should be the smallest offset that can be
5820 applied to a base register while still giving a legitimate address
5821 for every mode. The default value is 0.
5824 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5825 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5826 offset that should be applied to section anchors. The default
5830 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5831 Write the assembly code to define section anchor @var{x}, which is a
5832 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5833 The hook is called with the assembly output position set to the beginning
5834 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5836 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5837 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5838 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5839 is @code{NULL}, which disables the use of section anchors altogether.
5842 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (const_rtx @var{x})
5843 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5844 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5845 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5847 The default version is correct for most targets, but you might need to
5848 intercept this hook to handle things like target-specific attributes
5849 or target-specific sections.
5852 @node Condition Code
5853 @section Condition Code Status
5854 @cindex condition code status
5856 The macros in this section can be split in two families, according to the
5857 two ways of representing condition codes in GCC.
5859 The first representation is the so called @code{(cc0)} representation
5860 (@pxref{Jump Patterns}), where all instructions can have an implicit
5861 clobber of the condition codes. The second is the condition code
5862 register representation, which provides better schedulability for
5863 architectures that do have a condition code register, but on which
5864 most instructions do not affect it. The latter category includes
5867 The implicit clobbering poses a strong restriction on the placement of
5868 the definition and use of the condition code, which need to be in adjacent
5869 insns for machines using @code{(cc0)}. This can prevent important
5870 optimizations on some machines. For example, on the IBM RS/6000, there
5871 is a delay for taken branches unless the condition code register is set
5872 three instructions earlier than the conditional branch. The instruction
5873 scheduler cannot perform this optimization if it is not permitted to
5874 separate the definition and use of the condition code register.
5876 For this reason, it is possible and suggested to use a register to
5877 represent the condition code for new ports. If there is a specific
5878 condition code register in the machine, use a hard register. If the
5879 condition code or comparison result can be placed in any general register,
5880 or if there are multiple condition registers, use a pseudo register.
5881 Registers used to store the condition code value will usually have a mode
5882 that is in class @code{MODE_CC}.
5884 Alternatively, you can use @code{BImode} if the comparison operator is
5885 specified already in the compare instruction. In this case, you are not
5886 interested in most macros in this section.
5889 * CC0 Condition Codes:: Old style representation of condition codes.
5890 * MODE_CC Condition Codes:: Modern representation of condition codes.
5891 * Cond. Exec. Macros:: Macros to control conditional execution.
5894 @node CC0 Condition Codes
5895 @subsection Representation of condition codes using @code{(cc0)}
5899 The file @file{conditions.h} defines a variable @code{cc_status} to
5900 describe how the condition code was computed (in case the interpretation of
5901 the condition code depends on the instruction that it was set by). This
5902 variable contains the RTL expressions on which the condition code is
5903 currently based, and several standard flags.
5905 Sometimes additional machine-specific flags must be defined in the machine
5906 description header file. It can also add additional machine-specific
5907 information by defining @code{CC_STATUS_MDEP}.
5909 @defmac CC_STATUS_MDEP
5910 C code for a data type which is used for declaring the @code{mdep}
5911 component of @code{cc_status}. It defaults to @code{int}.
5913 This macro is not used on machines that do not use @code{cc0}.
5916 @defmac CC_STATUS_MDEP_INIT
5917 A C expression to initialize the @code{mdep} field to ``empty''.
5918 The default definition does nothing, since most machines don't use
5919 the field anyway. If you want to use the field, you should probably
5920 define this macro to initialize it.
5922 This macro is not used on machines that do not use @code{cc0}.
5925 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5926 A C compound statement to set the components of @code{cc_status}
5927 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5928 this macro's responsibility to recognize insns that set the condition
5929 code as a byproduct of other activity as well as those that explicitly
5932 This macro is not used on machines that do not use @code{cc0}.
5934 If there are insns that do not set the condition code but do alter
5935 other machine registers, this macro must check to see whether they
5936 invalidate the expressions that the condition code is recorded as
5937 reflecting. For example, on the 68000, insns that store in address
5938 registers do not set the condition code, which means that usually
5939 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5940 insns. But suppose that the previous insn set the condition code
5941 based on location @samp{a4@@(102)} and the current insn stores a new
5942 value in @samp{a4}. Although the condition code is not changed by
5943 this, it will no longer be true that it reflects the contents of
5944 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5945 @code{cc_status} in this case to say that nothing is known about the
5946 condition code value.
5948 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5949 with the results of peephole optimization: insns whose patterns are
5950 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5951 constants which are just the operands. The RTL structure of these
5952 insns is not sufficient to indicate what the insns actually do. What
5953 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5954 @code{CC_STATUS_INIT}.
5956 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5957 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5958 @samp{cc}. This avoids having detailed information about patterns in
5959 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5962 @node MODE_CC Condition Codes
5963 @subsection Representation of condition codes using registers
5967 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5968 On many machines, the condition code may be produced by other instructions
5969 than compares, for example the branch can use directly the condition
5970 code set by a subtract instruction. However, on some machines
5971 when the condition code is set this way some bits (such as the overflow
5972 bit) are not set in the same way as a test instruction, so that a different
5973 branch instruction must be used for some conditional branches. When
5974 this happens, use the machine mode of the condition code register to
5975 record different formats of the condition code register. Modes can
5976 also be used to record which compare instruction (e.g. a signed or an
5977 unsigned comparison) produced the condition codes.
5979 If other modes than @code{CCmode} are required, add them to
5980 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
5981 a mode given an operand of a compare. This is needed because the modes
5982 have to be chosen not only during RTL generation but also, for example,
5983 by instruction combination. The result of @code{SELECT_CC_MODE} should
5984 be consistent with the mode used in the patterns; for example to support
5985 the case of the add on the SPARC discussed above, we have the pattern
5989 [(set (reg:CC_NOOV 0)
5991 (plus:SI (match_operand:SI 0 "register_operand" "%r")
5992 (match_operand:SI 1 "arith_operand" "rI"))
5999 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
6000 for comparisons whose argument is a @code{plus}:
6003 #define SELECT_CC_MODE(OP,X,Y) \
6004 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
6005 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
6006 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
6007 || GET_CODE (X) == NEG) \
6008 ? CC_NOOVmode : CCmode))
6011 Another reason to use modes is to retain information on which operands
6012 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
6015 You should define this macro if and only if you define extra CC modes
6016 in @file{@var{machine}-modes.def}.
6019 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
6020 On some machines not all possible comparisons are defined, but you can
6021 convert an invalid comparison into a valid one. For example, the Alpha
6022 does not have a @code{GT} comparison, but you can use an @code{LT}
6023 comparison instead and swap the order of the operands.
6025 On such machines, define this macro to be a C statement to do any
6026 required conversions. @var{code} is the initial comparison code
6027 and @var{op0} and @var{op1} are the left and right operands of the
6028 comparison, respectively. You should modify @var{code}, @var{op0}, and
6029 @var{op1} as required.
6031 GCC will not assume that the comparison resulting from this macro is
6032 valid but will see if the resulting insn matches a pattern in the
6035 You need not define this macro if it would never change the comparison
6039 @defmac REVERSIBLE_CC_MODE (@var{mode})
6040 A C expression whose value is one if it is always safe to reverse a
6041 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
6042 can ever return @var{mode} for a floating-point inequality comparison,
6043 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
6045 You need not define this macro if it would always returns zero or if the
6046 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
6047 For example, here is the definition used on the SPARC, where floating-point
6048 inequality comparisons are always given @code{CCFPEmode}:
6051 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
6055 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
6056 A C expression whose value is reversed condition code of the @var{code} for
6057 comparison done in CC_MODE @var{mode}. The macro is used only in case
6058 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
6059 machine has some non-standard way how to reverse certain conditionals. For
6060 instance in case all floating point conditions are non-trapping, compiler may
6061 freely convert unordered compares to ordered one. Then definition may look
6065 #define REVERSE_CONDITION(CODE, MODE) \
6066 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6067 : reverse_condition_maybe_unordered (CODE))
6071 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *@var{p1}, unsigned int *@var{p2})
6072 On targets which do not use @code{(cc0)}, and which use a hard
6073 register rather than a pseudo-register to hold condition codes, the
6074 regular CSE passes are often not able to identify cases in which the
6075 hard register is set to a common value. Use this hook to enable a
6076 small pass which optimizes such cases. This hook should return true
6077 to enable this pass, and it should set the integers to which its
6078 arguments point to the hard register numbers used for condition codes.
6079 When there is only one such register, as is true on most systems, the
6080 integer pointed to by @var{p2} should be set to
6081 @code{INVALID_REGNUM}.
6083 The default version of this hook returns false.
6086 @deftypefn {Target Hook} {enum machine_mode} TARGET_CC_MODES_COMPATIBLE (enum machine_mode @var{m1}, enum machine_mode @var{m2})
6087 On targets which use multiple condition code modes in class
6088 @code{MODE_CC}, it is sometimes the case that a comparison can be
6089 validly done in more than one mode. On such a system, define this
6090 target hook to take two mode arguments and to return a mode in which
6091 both comparisons may be validly done. If there is no such mode,
6092 return @code{VOIDmode}.
6094 The default version of this hook checks whether the modes are the
6095 same. If they are, it returns that mode. If they are different, it
6096 returns @code{VOIDmode}.
6099 @node Cond. Exec. Macros
6100 @subsection Macros to control conditional execution
6101 @findex conditional execution
6104 There is one macro that may need to be defined for targets
6105 supporting conditional execution, independent of how they
6106 represent conditional branches.
6108 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
6109 A C expression that returns true if the conditional execution predicate
6110 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
6111 versa. Define this to return 0 if the target has conditional execution
6112 predicates that cannot be reversed safely. There is no need to validate
6113 that the arguments of op1 and op2 are the same, this is done separately.
6114 If no expansion is specified, this macro is defined as follows:
6117 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
6118 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
6123 @section Describing Relative Costs of Operations
6124 @cindex costs of instructions
6125 @cindex relative costs
6126 @cindex speed of instructions
6128 These macros let you describe the relative speed of various operations
6129 on the target machine.
6131 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6132 A C expression for the cost of moving data of mode @var{mode} from a
6133 register in class @var{from} to one in class @var{to}. The classes are
6134 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6135 value of 2 is the default; other values are interpreted relative to
6138 It is not required that the cost always equal 2 when @var{from} is the
6139 same as @var{to}; on some machines it is expensive to move between
6140 registers if they are not general registers.
6142 If reload sees an insn consisting of a single @code{set} between two
6143 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6144 classes returns a value of 2, reload does not check to ensure that the
6145 constraints of the insn are met. Setting a cost of other than 2 will
6146 allow reload to verify that the constraints are met. You should do this
6147 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6149 These macros are obsolete, new ports should use the target hook
6150 @code{TARGET_REGISTER_MOVE_COST} instead.
6153 @deftypefn {Target Hook} int TARGET_REGISTER_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{from}, reg_class_t @var{to})
6154 This target hook should return the cost of moving data of mode @var{mode}
6155 from a register in class @var{from} to one in class @var{to}. The classes
6156 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6157 A value of 2 is the default; other values are interpreted relative to
6160 It is not required that the cost always equal 2 when @var{from} is the
6161 same as @var{to}; on some machines it is expensive to move between
6162 registers if they are not general registers.
6164 If reload sees an insn consisting of a single @code{set} between two
6165 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6166 classes returns a value of 2, reload does not check to ensure that the
6167 constraints of the insn are met. Setting a cost of other than 2 will
6168 allow reload to verify that the constraints are met. You should do this
6169 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6171 The default version of this function returns 2.
6174 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6175 A C expression for the cost of moving data of mode @var{mode} between a
6176 register of class @var{class} and memory; @var{in} is zero if the value
6177 is to be written to memory, nonzero if it is to be read in. This cost
6178 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6179 registers and memory is more expensive than between two registers, you
6180 should define this macro to express the relative cost.
6182 If you do not define this macro, GCC uses a default cost of 4 plus
6183 the cost of copying via a secondary reload register, if one is
6184 needed. If your machine requires a secondary reload register to copy
6185 between memory and a register of @var{class} but the reload mechanism is
6186 more complex than copying via an intermediate, define this macro to
6187 reflect the actual cost of the move.
6189 GCC defines the function @code{memory_move_secondary_cost} if
6190 secondary reloads are needed. It computes the costs due to copying via
6191 a secondary register. If your machine copies from memory using a
6192 secondary register in the conventional way but the default base value of
6193 4 is not correct for your machine, define this macro to add some other
6194 value to the result of that function. The arguments to that function
6195 are the same as to this macro.
6197 These macros are obsolete, new ports should use the target hook
6198 @code{TARGET_MEMORY_MOVE_COST} instead.
6201 @deftypefn {Target Hook} int TARGET_MEMORY_MOVE_COST (enum machine_mode @var{mode}, reg_class_t @var{rclass}, bool @var{in})
6202 This target hook should return the cost of moving data of mode @var{mode}
6203 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6204 if the value is to be written to memory, @code{true} if it is to be read in.
6205 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6206 If moving between registers and memory is more expensive than between two
6207 registers, you should add this target hook to express the relative cost.
6209 If you do not add this target hook, GCC uses a default cost of 4 plus
6210 the cost of copying via a secondary reload register, if one is
6211 needed. If your machine requires a secondary reload register to copy
6212 between memory and a register of @var{rclass} but the reload mechanism is
6213 more complex than copying via an intermediate, use this target hook to
6214 reflect the actual cost of the move.
6216 GCC defines the function @code{memory_move_secondary_cost} if
6217 secondary reloads are needed. It computes the costs due to copying via
6218 a secondary register. If your machine copies from memory using a
6219 secondary register in the conventional way but the default base value of
6220 4 is not correct for your machine, use this target hook to add some other
6221 value to the result of that function. The arguments to that function
6222 are the same as to this target hook.
6225 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6226 A C expression for the cost of a branch instruction. A value of 1 is the
6227 default; other values are interpreted relative to that. Parameter @var{speed_p}
6228 is true when the branch in question should be optimized for speed. When
6229 it is false, @code{BRANCH_COST} should be returning value optimal for code size
6230 rather then performance considerations. @var{predictable_p} is true for well
6231 predictable branches. On many architectures the @code{BRANCH_COST} can be
6235 Here are additional macros which do not specify precise relative costs,
6236 but only that certain actions are more expensive than GCC would
6239 @defmac SLOW_BYTE_ACCESS
6240 Define this macro as a C expression which is nonzero if accessing less
6241 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6242 faster than accessing a word of memory, i.e., if such access
6243 require more than one instruction or if there is no difference in cost
6244 between byte and (aligned) word loads.
6246 When this macro is not defined, the compiler will access a field by
6247 finding the smallest containing object; when it is defined, a fullword
6248 load will be used if alignment permits. Unless bytes accesses are
6249 faster than word accesses, using word accesses is preferable since it
6250 may eliminate subsequent memory access if subsequent accesses occur to
6251 other fields in the same word of the structure, but to different bytes.
6254 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
6255 Define this macro to be the value 1 if memory accesses described by the
6256 @var{mode} and @var{alignment} parameters have a cost many times greater
6257 than aligned accesses, for example if they are emulated in a trap
6260 When this macro is nonzero, the compiler will act as if
6261 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
6262 moves. This can cause significantly more instructions to be produced.
6263 Therefore, do not set this macro nonzero if unaligned accesses only add a
6264 cycle or two to the time for a memory access.
6266 If the value of this macro is always zero, it need not be defined. If
6267 this macro is defined, it should produce a nonzero value when
6268 @code{STRICT_ALIGNMENT} is nonzero.
6271 @defmac MOVE_RATIO (@var{speed})
6272 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6273 which a sequence of insns should be generated instead of a
6274 string move insn or a library call. Increasing the value will always
6275 make code faster, but eventually incurs high cost in increased code size.
6277 Note that on machines where the corresponding move insn is a
6278 @code{define_expand} that emits a sequence of insns, this macro counts
6279 the number of such sequences.
6281 The parameter @var{speed} is true if the code is currently being
6282 optimized for speed rather than size.
6284 If you don't define this, a reasonable default is used.
6287 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
6288 A C expression used to determine whether @code{move_by_pieces} will be used to
6289 copy a chunk of memory, or whether some other block move mechanism
6290 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6291 than @code{MOVE_RATIO}.
6294 @defmac MOVE_MAX_PIECES
6295 A C expression used by @code{move_by_pieces} to determine the largest unit
6296 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6299 @defmac CLEAR_RATIO (@var{speed})
6300 The threshold of number of scalar move insns, @emph{below} which a sequence
6301 of insns should be generated to clear memory instead of a string clear insn
6302 or a library call. Increasing the value will always make code faster, but
6303 eventually incurs high cost in increased code size.
6305 The parameter @var{speed} is true if the code is currently being
6306 optimized for speed rather than size.
6308 If you don't define this, a reasonable default is used.
6311 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
6312 A C expression used to determine whether @code{clear_by_pieces} will be used
6313 to clear a chunk of memory, or whether some other block clear mechanism
6314 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6315 than @code{CLEAR_RATIO}.
6318 @defmac SET_RATIO (@var{speed})
6319 The threshold of number of scalar move insns, @emph{below} which a sequence
6320 of insns should be generated to set memory to a constant value, instead of
6321 a block set insn or a library call.
6322 Increasing the value will always make code faster, but
6323 eventually incurs high cost in increased code size.
6325 The parameter @var{speed} is true if the code is currently being
6326 optimized for speed rather than size.
6328 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6331 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
6332 A C expression used to determine whether @code{store_by_pieces} will be
6333 used to set a chunk of memory to a constant value, or whether some
6334 other mechanism will be used. Used by @code{__builtin_memset} when
6335 storing values other than constant zero.
6336 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6337 than @code{SET_RATIO}.
6340 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
6341 A C expression used to determine whether @code{store_by_pieces} will be
6342 used to set a chunk of memory to a constant string value, or whether some
6343 other mechanism will be used. Used by @code{__builtin_strcpy} when
6344 called with a constant source string.
6345 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6346 than @code{MOVE_RATIO}.
6349 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6350 A C expression used to determine whether a load postincrement is a good
6351 thing to use for a given mode. Defaults to the value of
6352 @code{HAVE_POST_INCREMENT}.
6355 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6356 A C expression used to determine whether a load postdecrement is a good
6357 thing to use for a given mode. Defaults to the value of
6358 @code{HAVE_POST_DECREMENT}.
6361 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6362 A C expression used to determine whether a load preincrement is a good
6363 thing to use for a given mode. Defaults to the value of
6364 @code{HAVE_PRE_INCREMENT}.
6367 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6368 A C expression used to determine whether a load predecrement is a good
6369 thing to use for a given mode. Defaults to the value of
6370 @code{HAVE_PRE_DECREMENT}.
6373 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6374 A C expression used to determine whether a store postincrement is a good
6375 thing to use for a given mode. Defaults to the value of
6376 @code{HAVE_POST_INCREMENT}.
6379 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6380 A C expression used to determine whether a store postdecrement is a good
6381 thing to use for a given mode. Defaults to the value of
6382 @code{HAVE_POST_DECREMENT}.
6385 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6386 This macro is used to determine whether a store preincrement is a good
6387 thing to use for a given mode. Defaults to the value of
6388 @code{HAVE_PRE_INCREMENT}.
6391 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6392 This macro is used to determine whether a store predecrement is a good
6393 thing to use for a given mode. Defaults to the value of
6394 @code{HAVE_PRE_DECREMENT}.
6397 @defmac NO_FUNCTION_CSE
6398 Define this macro if it is as good or better to call a constant
6399 function address than to call an address kept in a register.
6402 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
6403 Define this macro if a non-short-circuit operation produced by
6404 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6405 @code{BRANCH_COST} is greater than or equal to the value 2.
6408 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total}, bool @var{speed})
6409 This target hook describes the relative costs of RTL expressions.
6411 The cost may depend on the precise form of the expression, which is
6412 available for examination in @var{x}, and the rtx code of the expression
6413 in which it is contained, found in @var{outer_code}. @var{code} is the
6414 expression code---redundant, since it can be obtained with
6415 @code{GET_CODE (@var{x})}.
6417 In implementing this hook, you can use the construct
6418 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6421 On entry to the hook, @code{*@var{total}} contains a default estimate
6422 for the cost of the expression. The hook should modify this value as
6423 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6424 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6425 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6427 When optimizing for code size, i.e.@: when @code{speed} is
6428 false, this target hook should be used to estimate the relative
6429 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6431 The hook returns true when all subexpressions of @var{x} have been
6432 processed, and false when @code{rtx_cost} should recurse.
6435 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}, bool @var{speed})
6436 This hook computes the cost of an addressing mode that contains
6437 @var{address}. If not defined, the cost is computed from
6438 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6440 For most CISC machines, the default cost is a good approximation of the
6441 true cost of the addressing mode. However, on RISC machines, all
6442 instructions normally have the same length and execution time. Hence
6443 all addresses will have equal costs.
6445 In cases where more than one form of an address is known, the form with
6446 the lowest cost will be used. If multiple forms have the same, lowest,
6447 cost, the one that is the most complex will be used.
6449 For example, suppose an address that is equal to the sum of a register
6450 and a constant is used twice in the same basic block. When this macro
6451 is not defined, the address will be computed in a register and memory
6452 references will be indirect through that register. On machines where
6453 the cost of the addressing mode containing the sum is no higher than
6454 that of a simple indirect reference, this will produce an additional
6455 instruction and possibly require an additional register. Proper
6456 specification of this macro eliminates this overhead for such machines.
6458 This hook is never called with an invalid address.
6460 On machines where an address involving more than one register is as
6461 cheap as an address computation involving only one register, defining
6462 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6463 be live over a region of code where only one would have been if
6464 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6465 should be considered in the definition of this macro. Equivalent costs
6466 should probably only be given to addresses with different numbers of
6467 registers on machines with lots of registers.
6471 @section Adjusting the Instruction Scheduler
6473 The instruction scheduler may need a fair amount of machine-specific
6474 adjustment in order to produce good code. GCC provides several target
6475 hooks for this purpose. It is usually enough to define just a few of
6476 them: try the first ones in this list first.
6478 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6479 This hook returns the maximum number of instructions that can ever
6480 issue at the same time on the target machine. The default is one.
6481 Although the insn scheduler can define itself the possibility of issue
6482 an insn on the same cycle, the value can serve as an additional
6483 constraint to issue insns on the same simulated processor cycle (see
6484 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6485 This value must be constant over the entire compilation. If you need
6486 it to vary depending on what the instructions are, you must use
6487 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6490 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
6491 This hook is executed by the scheduler after it has scheduled an insn
6492 from the ready list. It should return the number of insns which can
6493 still be issued in the current cycle. The default is
6494 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6495 @code{USE}, which normally are not counted against the issue rate.
6496 You should define this hook if some insns take more machine resources
6497 than others, so that fewer insns can follow them in the same cycle.
6498 @var{file} is either a null pointer, or a stdio stream to write any
6499 debug output to. @var{verbose} is the verbose level provided by
6500 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6504 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
6505 This function corrects the value of @var{cost} based on the
6506 relationship between @var{insn} and @var{dep_insn} through the
6507 dependence @var{link}. It should return the new value. The default
6508 is to make no adjustment to @var{cost}. This can be used for example
6509 to specify to the scheduler using the traditional pipeline description
6510 that an output- or anti-dependence does not incur the same cost as a
6511 data-dependence. If the scheduler using the automaton based pipeline
6512 description, the cost of anti-dependence is zero and the cost of
6513 output-dependence is maximum of one and the difference of latency
6514 times of the first and the second insns. If these values are not
6515 acceptable, you could use the hook to modify them too. See also
6516 @pxref{Processor pipeline description}.
6519 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
6520 This hook adjusts the integer scheduling priority @var{priority} of
6521 @var{insn}. It should return the new priority. Increase the priority to
6522 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6523 later. Do not define this hook if you do not need to adjust the
6524 scheduling priorities of insns.
6527 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6528 This hook is executed by the scheduler after it has scheduled the ready
6529 list, to allow the machine description to reorder it (for example to
6530 combine two small instructions together on @samp{VLIW} machines).
6531 @var{file} is either a null pointer, or a stdio stream to write any
6532 debug output to. @var{verbose} is the verbose level provided by
6533 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6534 list of instructions that are ready to be scheduled. @var{n_readyp} is
6535 a pointer to the number of elements in the ready list. The scheduler
6536 reads the ready list in reverse order, starting with
6537 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6538 is the timer tick of the scheduler. You may modify the ready list and
6539 the number of ready insns. The return value is the number of insns that
6540 can issue this cycle; normally this is just @code{issue_rate}. See also
6541 @samp{TARGET_SCHED_REORDER2}.
6544 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6545 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6546 function is called whenever the scheduler starts a new cycle. This one
6547 is called once per iteration over a cycle, immediately after
6548 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6549 return the number of insns to be scheduled in the same cycle. Defining
6550 this hook can be useful if there are frequent situations where
6551 scheduling one insn causes other insns to become ready in the same
6552 cycle. These other insns can then be taken into account properly.
6555 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6556 This hook is called after evaluation forward dependencies of insns in
6557 chain given by two parameter values (@var{head} and @var{tail}
6558 correspondingly) but before insns scheduling of the insn chain. For
6559 example, it can be used for better insn classification if it requires
6560 analysis of dependencies. This hook can use backward and forward
6561 dependencies of the insn scheduler because they are already
6565 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6566 This hook is executed by the scheduler at the beginning of each block of
6567 instructions that are to be scheduled. @var{file} is either a null
6568 pointer, or a stdio stream to write any debug output to. @var{verbose}
6569 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6570 @var{max_ready} is the maximum number of insns in the current scheduling
6571 region that can be live at the same time. This can be used to allocate
6572 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6575 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6576 This hook is executed by the scheduler at the end of each block of
6577 instructions that are to be scheduled. It can be used to perform
6578 cleanup of any actions done by the other scheduling hooks. @var{file}
6579 is either a null pointer, or a stdio stream to write any debug output
6580 to. @var{verbose} is the verbose level provided by
6581 @option{-fsched-verbose-@var{n}}.
6584 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6585 This hook is executed by the scheduler after function level initializations.
6586 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6587 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6588 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6591 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6592 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6593 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6594 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6597 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6598 The hook returns an RTL insn. The automaton state used in the
6599 pipeline hazard recognizer is changed as if the insn were scheduled
6600 when the new simulated processor cycle starts. Usage of the hook may
6601 simplify the automaton pipeline description for some @acronym{VLIW}
6602 processors. If the hook is defined, it is used only for the automaton
6603 based pipeline description. The default is not to change the state
6604 when the new simulated processor cycle starts.
6607 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6608 The hook can be used to initialize data used by the previous hook.
6611 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6612 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6613 to changed the state as if the insn were scheduled when the new
6614 simulated processor cycle finishes.
6617 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6618 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6619 used to initialize data used by the previous hook.
6622 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE (void)
6623 The hook to notify target that the current simulated cycle is about to finish.
6624 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6625 to change the state in more complicated situations - e.g., when advancing
6626 state on a single insn is not enough.
6629 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_ADVANCE_CYCLE (void)
6630 The hook to notify target that new simulated cycle has just started.
6631 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6632 to change the state in more complicated situations - e.g., when advancing
6633 state on a single insn is not enough.
6636 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6637 This hook controls better choosing an insn from the ready insn queue
6638 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6639 chooses the first insn from the queue. If the hook returns a positive
6640 value, an additional scheduler code tries all permutations of
6641 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6642 subsequent ready insns to choose an insn whose issue will result in
6643 maximal number of issued insns on the same cycle. For the
6644 @acronym{VLIW} processor, the code could actually solve the problem of
6645 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6646 rules of @acronym{VLIW} packing are described in the automaton.
6648 This code also could be used for superscalar @acronym{RISC}
6649 processors. Let us consider a superscalar @acronym{RISC} processor
6650 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6651 @var{B}, some insns can be executed only in pipelines @var{B} or
6652 @var{C}, and one insn can be executed in pipeline @var{B}. The
6653 processor may issue the 1st insn into @var{A} and the 2nd one into
6654 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6655 until the next cycle. If the scheduler issues the 3rd insn the first,
6656 the processor could issue all 3 insns per cycle.
6658 Actually this code demonstrates advantages of the automaton based
6659 pipeline hazard recognizer. We try quickly and easy many insn
6660 schedules to choose the best one.
6662 The default is no multipass scheduling.
6665 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx @var{insn})
6667 This hook controls what insns from the ready insn queue will be
6668 considered for the multipass insn scheduling. If the hook returns
6669 zero for @var{insn}, the insn will be not chosen to
6672 The default is that any ready insns can be chosen to be issued.
6675 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *@var{dump}, int @var{verbose}, rtx @var{insn}, int @var{last_clock}, int @var{clock}, int *@var{sort_p})
6676 This hook is called by the insn scheduler before issuing @var{insn}
6677 on cycle @var{clock}. If the hook returns nonzero,
6678 @var{insn} is not issued on this processor cycle. Instead,
6679 the processor cycle is advanced. If *@var{sort_p}
6680 is zero, the insn ready queue is not sorted on the new cycle
6681 start as usually. @var{dump} and @var{verbose} specify the file and
6682 verbosity level to use for debugging output.
6683 @var{last_clock} and @var{clock} are, respectively, the
6684 processor cycle on which the previous insn has been issued,
6685 and the current processor cycle.
6688 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct _dep *@var{_dep}, int @var{cost}, int @var{distance})
6689 This hook is used to define which dependences are considered costly by
6690 the target, so costly that it is not advisable to schedule the insns that
6691 are involved in the dependence too close to one another. The parameters
6692 to this hook are as follows: The first parameter @var{_dep} is the dependence
6693 being evaluated. The second parameter @var{cost} is the cost of the
6694 dependence as estimated by the scheduler, and the third
6695 parameter @var{distance} is the distance in cycles between the two insns.
6696 The hook returns @code{true} if considering the distance between the two
6697 insns the dependence between them is considered costly by the target,
6698 and @code{false} otherwise.
6700 Defining this hook can be useful in multiple-issue out-of-order machines,
6701 where (a) it's practically hopeless to predict the actual data/resource
6702 delays, however: (b) there's a better chance to predict the actual grouping
6703 that will be formed, and (c) correctly emulating the grouping can be very
6704 important. In such targets one may want to allow issuing dependent insns
6705 closer to one another---i.e., closer than the dependence distance; however,
6706 not in cases of ``costly dependences'', which this hooks allows to define.
6709 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6710 This hook is called by the insn scheduler after emitting a new instruction to
6711 the instruction stream. The hook notifies a target backend to extend its
6712 per instruction data structures.
6715 @deftypefn {Target Hook} {void *} TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
6716 Return a pointer to a store large enough to hold target scheduling context.
6719 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
6720 Initialize store pointed to by @var{tc} to hold target scheduling context.
6721 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6722 beginning of the block. Otherwise, copy the current context into @var{tc}.
6725 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
6726 Copy target scheduling context pointed to by @var{tc} to the current context.
6729 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
6730 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6733 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
6734 Deallocate a store for target scheduling context pointed to by @var{tc}.
6737 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6738 This hook is called by the insn scheduler when @var{insn} has only
6739 speculative dependencies and therefore can be scheduled speculatively.
6740 The hook is used to check if the pattern of @var{insn} has a speculative
6741 version and, in case of successful check, to generate that speculative
6742 pattern. The hook should return 1, if the instruction has a speculative form,
6743 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6744 speculation. If the return value equals 1 then @var{new_pat} is assigned
6745 the generated speculative pattern.
6748 @deftypefn {Target Hook} bool TARGET_SCHED_NEEDS_BLOCK_P (int @var{dep_status})
6749 This hook is called by the insn scheduler during generation of recovery code
6750 for @var{insn}. It should return @code{true}, if the corresponding check
6751 instruction should branch to recovery code, or @code{false} otherwise.
6754 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_SPEC_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6755 This hook is called by the insn scheduler to generate a pattern for recovery
6756 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6757 speculative instruction for which the check should be generated.
6758 @var{label} is either a label of a basic block, where recovery code should
6759 be emitted, or a null pointer, when requested check doesn't branch to
6760 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6761 a pattern for a branchy check corresponding to a simple check denoted by
6762 @var{insn} should be generated. In this case @var{label} can't be null.
6765 @deftypefn {Target Hook} bool TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (const_rtx @var{insn})
6766 This hook is used as a workaround for
6767 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6768 called on the first instruction of the ready list. The hook is used to
6769 discard speculative instructions that stand first in the ready list from
6770 being scheduled on the current cycle. If the hook returns @code{false},
6771 @var{insn} will not be chosen to be issued.
6772 For non-speculative instructions,
6773 the hook should always return @code{true}. For example, in the ia64 backend
6774 the hook is used to cancel data speculative insns when the ALAT table
6778 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (struct spec_info_def *@var{spec_info})
6779 This hook is used by the insn scheduler to find out what features should be
6781 The structure *@var{spec_info} should be filled in by the target.
6782 The structure describes speculation types that can be used in the scheduler.
6785 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6786 This hook is called by the swing modulo scheduler to calculate a
6787 resource-based lower bound which is based on the resources available in
6788 the machine and the resources required by each instruction. The target
6789 backend can use @var{g} to calculate such bound. A very simple lower
6790 bound will be used in case this hook is not implemented: the total number
6791 of instructions divided by the issue rate.
6794 @deftypefn {Target Hook} bool TARGET_SCHED_DISPATCH (rtx @var{insn}, int @var{x})
6795 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
6796 is supported in hardware and the condition specified in the parameter is true.
6799 @deftypefn {Target Hook} void TARGET_SCHED_DISPATCH_DO (rtx @var{insn}, int @var{x})
6800 This hook is called by Haifa Scheduler. It performs the operation specified
6801 in its second parameter.
6805 @section Dividing the Output into Sections (Texts, Data, @dots{})
6806 @c the above section title is WAY too long. maybe cut the part between
6807 @c the (...)? --mew 10feb93
6809 An object file is divided into sections containing different types of
6810 data. In the most common case, there are three sections: the @dfn{text
6811 section}, which holds instructions and read-only data; the @dfn{data
6812 section}, which holds initialized writable data; and the @dfn{bss
6813 section}, which holds uninitialized data. Some systems have other kinds
6816 @file{varasm.c} provides several well-known sections, such as
6817 @code{text_section}, @code{data_section} and @code{bss_section}.
6818 The normal way of controlling a @code{@var{foo}_section} variable
6819 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6820 as described below. The macros are only read once, when @file{varasm.c}
6821 initializes itself, so their values must be run-time constants.
6822 They may however depend on command-line flags.
6824 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6825 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6826 to be string literals.
6828 Some assemblers require a different string to be written every time a
6829 section is selected. If your assembler falls into this category, you
6830 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6831 @code{get_unnamed_section} to set up the sections.
6833 You must always create a @code{text_section}, either by defining
6834 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6835 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6836 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6837 create a distinct @code{readonly_data_section}, the default is to
6838 reuse @code{text_section}.
6840 All the other @file{varasm.c} sections are optional, and are null
6841 if the target does not provide them.
6843 @defmac TEXT_SECTION_ASM_OP
6844 A C expression whose value is a string, including spacing, containing the
6845 assembler operation that should precede instructions and read-only data.
6846 Normally @code{"\t.text"} is right.
6849 @defmac HOT_TEXT_SECTION_NAME
6850 If defined, a C string constant for the name of the section containing most
6851 frequently executed functions of the program. If not defined, GCC will provide
6852 a default definition if the target supports named sections.
6855 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6856 If defined, a C string constant for the name of the section containing unlikely
6857 executed functions in the program.
6860 @defmac DATA_SECTION_ASM_OP
6861 A C expression whose value is a string, including spacing, containing the
6862 assembler operation to identify the following data as writable initialized
6863 data. Normally @code{"\t.data"} is right.
6866 @defmac SDATA_SECTION_ASM_OP
6867 If defined, a C expression whose value is a string, including spacing,
6868 containing the assembler operation to identify the following data as
6869 initialized, writable small data.
6872 @defmac READONLY_DATA_SECTION_ASM_OP
6873 A C expression whose value is a string, including spacing, containing the
6874 assembler operation to identify the following data as read-only initialized
6878 @defmac BSS_SECTION_ASM_OP
6879 If defined, a C expression whose value is a string, including spacing,
6880 containing the assembler operation to identify the following data as
6881 uninitialized global data. If not defined, and neither
6882 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
6883 uninitialized global data will be output in the data section if
6884 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6888 @defmac SBSS_SECTION_ASM_OP
6889 If defined, a C expression whose value is a string, including spacing,
6890 containing the assembler operation to identify the following data as
6891 uninitialized, writable small data.
6894 @defmac TLS_COMMON_ASM_OP
6895 If defined, a C expression whose value is a string containing the
6896 assembler operation to identify the following data as thread-local
6897 common data. The default is @code{".tls_common"}.
6900 @defmac TLS_SECTION_ASM_FLAG
6901 If defined, a C expression whose value is a character constant
6902 containing the flag used to mark a section as a TLS section. The
6903 default is @code{'T'}.
6906 @defmac INIT_SECTION_ASM_OP
6907 If defined, a C expression whose value is a string, including spacing,
6908 containing the assembler operation to identify the following data as
6909 initialization code. If not defined, GCC will assume such a section does
6910 not exist. This section has no corresponding @code{init_section}
6911 variable; it is used entirely in runtime code.
6914 @defmac FINI_SECTION_ASM_OP
6915 If defined, a C expression whose value is a string, including spacing,
6916 containing the assembler operation to identify the following data as
6917 finalization code. If not defined, GCC will assume such a section does
6918 not exist. This section has no corresponding @code{fini_section}
6919 variable; it is used entirely in runtime code.
6922 @defmac INIT_ARRAY_SECTION_ASM_OP
6923 If defined, a C expression whose value is a string, including spacing,
6924 containing the assembler operation to identify the following data as
6925 part of the @code{.init_array} (or equivalent) section. If not
6926 defined, GCC will assume such a section does not exist. Do not define
6927 both this macro and @code{INIT_SECTION_ASM_OP}.
6930 @defmac FINI_ARRAY_SECTION_ASM_OP
6931 If defined, a C expression whose value is a string, including spacing,
6932 containing the assembler operation to identify the following data as
6933 part of the @code{.fini_array} (or equivalent) section. If not
6934 defined, GCC will assume such a section does not exist. Do not define
6935 both this macro and @code{FINI_SECTION_ASM_OP}.
6938 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6939 If defined, an ASM statement that switches to a different section
6940 via @var{section_op}, calls @var{function}, and switches back to
6941 the text section. This is used in @file{crtstuff.c} if
6942 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6943 to initialization and finalization functions from the init and fini
6944 sections. By default, this macro uses a simple function call. Some
6945 ports need hand-crafted assembly code to avoid dependencies on
6946 registers initialized in the function prologue or to ensure that
6947 constant pools don't end up too far way in the text section.
6950 @defmac TARGET_LIBGCC_SDATA_SECTION
6951 If defined, a string which names the section into which small
6952 variables defined in crtstuff and libgcc should go. This is useful
6953 when the target has options for optimizing access to small data, and
6954 you want the crtstuff and libgcc routines to be conservative in what
6955 they expect of your application yet liberal in what your application
6956 expects. For example, for targets with a @code{.sdata} section (like
6957 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6958 require small data support from your application, but use this macro
6959 to put small data into @code{.sdata} so that your application can
6960 access these variables whether it uses small data or not.
6963 @defmac FORCE_CODE_SECTION_ALIGN
6964 If defined, an ASM statement that aligns a code section to some
6965 arbitrary boundary. This is used to force all fragments of the
6966 @code{.init} and @code{.fini} sections to have to same alignment
6967 and thus prevent the linker from having to add any padding.
6970 @defmac JUMP_TABLES_IN_TEXT_SECTION
6971 Define this macro to be an expression with a nonzero value if jump
6972 tables (for @code{tablejump} insns) should be output in the text
6973 section, along with the assembler instructions. Otherwise, the
6974 readonly data section is used.
6976 This macro is irrelevant if there is no separate readonly data section.
6979 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
6980 Define this hook if you need to do something special to set up the
6981 @file{varasm.c} sections, or if your target has some special sections
6982 of its own that you need to create.
6984 GCC calls this hook after processing the command line, but before writing
6985 any assembly code, and before calling any of the section-returning hooks
6989 @deftypefn {Target Hook} int TARGET_ASM_RELOC_RW_MASK (void)
6990 Return a mask describing how relocations should be treated when
6991 selecting sections. Bit 1 should be set if global relocations
6992 should be placed in a read-write section; bit 0 should be set if
6993 local relocations should be placed in a read-write section.
6995 The default version of this function returns 3 when @option{-fpic}
6996 is in effect, and 0 otherwise. The hook is typically redefined
6997 when the target cannot support (some kinds of) dynamic relocations
6998 in read-only sections even in executables.
7001 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
7002 Return the section into which @var{exp} should be placed. You can
7003 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
7004 some sort. @var{reloc} indicates whether the initial value of @var{exp}
7005 requires link-time relocations. Bit 0 is set when variable contains
7006 local relocations only, while bit 1 is set for global relocations.
7007 @var{align} is the constant alignment in bits.
7009 The default version of this function takes care of putting read-only
7010 variables in @code{readonly_data_section}.
7012 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
7015 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
7016 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
7017 for @code{FUNCTION_DECL}s as well as for variables and constants.
7019 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
7020 function has been determined to be likely to be called, and nonzero if
7021 it is unlikely to be called.
7024 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
7025 Build up a unique section name, expressed as a @code{STRING_CST} node,
7026 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7027 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7028 the initial value of @var{exp} requires link-time relocations.
7030 The default version of this function appends the symbol name to the
7031 ELF section name that would normally be used for the symbol. For
7032 example, the function @code{foo} would be placed in @code{.text.foo}.
7033 Whatever the actual target object format, this is often good enough.
7036 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
7037 Return the readonly data section associated with
7038 @samp{DECL_SECTION_NAME (@var{decl})}.
7039 The default version of this function selects @code{.gnu.linkonce.r.name} if
7040 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7041 if function is in @code{.text.name}, and the normal readonly-data section
7045 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
7046 Return the section into which a constant @var{x}, of mode @var{mode},
7047 should be placed. You can assume that @var{x} is some kind of
7048 constant in RTL@. The argument @var{mode} is redundant except in the
7049 case of a @code{const_int} rtx. @var{align} is the constant alignment
7052 The default version of this function takes care of putting symbolic
7053 constants in @code{flag_pic} mode in @code{data_section} and everything
7054 else in @code{readonly_data_section}.
7057 @deftypefn {Target Hook} tree TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
7058 Define this hook if you need to postprocess the assembler name generated
7059 by target-independent code. The @var{id} provided to this hook will be
7060 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7061 or the mangled name of the @var{decl} in C++). The return value of the
7062 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7063 your target system. The default implementation of this hook just
7064 returns the @var{id} provided.
7067 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
7068 Define this hook if references to a symbol or a constant must be
7069 treated differently depending on something about the variable or
7070 function named by the symbol (such as what section it is in).
7072 The hook is executed immediately after rtl has been created for
7073 @var{decl}, which may be a variable or function declaration or
7074 an entry in the constant pool. In either case, @var{rtl} is the
7075 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7076 in this hook; that field may not have been initialized yet.
7078 In the case of a constant, it is safe to assume that the rtl is
7079 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7080 will also have this form, but that is not guaranteed. Global
7081 register variables, for instance, will have a @code{reg} for their
7082 rtl. (Normally the right thing to do with such unusual rtl is
7085 The @var{new_decl_p} argument will be true if this is the first time
7086 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7087 be false for subsequent invocations, which will happen for duplicate
7088 declarations. Whether or not anything must be done for the duplicate
7089 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7090 @var{new_decl_p} is always true when the hook is called for a constant.
7092 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7093 The usual thing for this hook to do is to record flags in the
7094 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7095 Historically, the name string was modified if it was necessary to
7096 encode more than one bit of information, but this practice is now
7097 discouraged; use @code{SYMBOL_REF_FLAGS}.
7099 The default definition of this hook, @code{default_encode_section_info}
7100 in @file{varasm.c}, sets a number of commonly-useful bits in
7101 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7102 before overriding it.
7105 @deftypefn {Target Hook} {const char *} TARGET_STRIP_NAME_ENCODING (const char *@var{name})
7106 Decode @var{name} and return the real name part, sans
7107 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7111 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (const_tree @var{exp})
7112 Returns true if @var{exp} should be placed into a ``small data'' section.
7113 The default version of this hook always returns false.
7116 @deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
7117 Contains the value true if the target places read-only
7118 ``small data'' into a separate section. The default value is false.
7121 @deftypefn {Target Hook} bool TARGET_PROFILE_BEFORE_PROLOGUE (void)
7122 It returns true if target wants profile code emitted before prologue.
7124 The default version of this hook use the target macro
7125 @code{PROFILE_BEFORE_PROLOGUE}.
7128 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (const_tree @var{exp})
7129 Returns true if @var{exp} names an object for which name resolution
7130 rules must resolve to the current ``module'' (dynamic shared library
7131 or executable image).
7133 The default version of this hook implements the name resolution rules
7134 for ELF, which has a looser model of global name binding than other
7135 currently supported object file formats.
7138 @deftypevr {Target Hook} bool TARGET_HAVE_TLS
7139 Contains the value true if the target supports thread-local storage.
7140 The default value is false.
7145 @section Position Independent Code
7146 @cindex position independent code
7149 This section describes macros that help implement generation of position
7150 independent code. Simply defining these macros is not enough to
7151 generate valid PIC; you must also add support to the hook
7152 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7153 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7154 must modify the definition of @samp{movsi} to do something appropriate
7155 when the source operand contains a symbolic address. You may also
7156 need to alter the handling of switch statements so that they use
7158 @c i rearranged the order of the macros above to try to force one of
7159 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7161 @defmac PIC_OFFSET_TABLE_REGNUM
7162 The register number of the register used to address a table of static
7163 data addresses in memory. In some cases this register is defined by a
7164 processor's ``application binary interface'' (ABI)@. When this macro
7165 is defined, RTL is generated for this register once, as with the stack
7166 pointer and frame pointer registers. If this macro is not defined, it
7167 is up to the machine-dependent files to allocate such a register (if
7168 necessary). Note that this register must be fixed when in use (e.g.@:
7169 when @code{flag_pic} is true).
7172 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7173 A C expression that is nonzero if the register defined by
7174 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7175 the default is zero. Do not define
7176 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7179 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7180 A C expression that is nonzero if @var{x} is a legitimate immediate
7181 operand on the target machine when generating position independent code.
7182 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7183 check this. You can also assume @var{flag_pic} is true, so you need not
7184 check it either. You need not define this macro if all constants
7185 (including @code{SYMBOL_REF}) can be immediate operands when generating
7186 position independent code.
7189 @node Assembler Format
7190 @section Defining the Output Assembler Language
7192 This section describes macros whose principal purpose is to describe how
7193 to write instructions in assembler language---rather than what the
7197 * File Framework:: Structural information for the assembler file.
7198 * Data Output:: Output of constants (numbers, strings, addresses).
7199 * Uninitialized Data:: Output of uninitialized variables.
7200 * Label Output:: Output and generation of labels.
7201 * Initialization:: General principles of initialization
7202 and termination routines.
7203 * Macros for Initialization::
7204 Specific macros that control the handling of
7205 initialization and termination routines.
7206 * Instruction Output:: Output of actual instructions.
7207 * Dispatch Tables:: Output of jump tables.
7208 * Exception Region Output:: Output of exception region code.
7209 * Alignment Output:: Pseudo ops for alignment and skipping data.
7212 @node File Framework
7213 @subsection The Overall Framework of an Assembler File
7214 @cindex assembler format
7215 @cindex output of assembler code
7217 @c prevent bad page break with this line
7218 This describes the overall framework of an assembly file.
7220 @findex default_file_start
7221 @deftypefn {Target Hook} void TARGET_ASM_FILE_START (void)
7222 Output to @code{asm_out_file} any text which the assembler expects to
7223 find at the beginning of a file. The default behavior is controlled
7224 by two flags, documented below. Unless your target's assembler is
7225 quite unusual, if you override the default, you should call
7226 @code{default_file_start} at some point in your target hook. This
7227 lets other target files rely on these variables.
7230 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
7231 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7232 printed as the very first line in the assembly file, unless
7233 @option{-fverbose-asm} is in effect. (If that macro has been defined
7234 to the empty string, this variable has no effect.) With the normal
7235 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7236 assembler that it need not bother stripping comments or extra
7237 whitespace from its input. This allows it to work a bit faster.
7239 The default is false. You should not set it to true unless you have
7240 verified that your port does not generate any extra whitespace or
7241 comments that will cause GAS to issue errors in NO_APP mode.
7244 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
7245 If this flag is true, @code{output_file_directive} will be called
7246 for the primary source file, immediately after printing
7247 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7248 this to be done. The default is false.
7251 @deftypefn {Target Hook} void TARGET_ASM_FILE_END (void)
7252 Output to @code{asm_out_file} any text which the assembler expects
7253 to find at the end of a file. The default is to output nothing.
7256 @deftypefun void file_end_indicate_exec_stack ()
7257 Some systems use a common convention, the @samp{.note.GNU-stack}
7258 special section, to indicate whether or not an object file relies on
7259 the stack being executable. If your system uses this convention, you
7260 should define @code{TARGET_ASM_FILE_END} to this function. If you
7261 need to do other things in that hook, have your hook function call
7265 @deftypefn {Target Hook} void TARGET_ASM_LTO_START (void)
7266 Output to @code{asm_out_file} any text which the assembler expects
7267 to find at the start of an LTO section. The default is to output
7271 @deftypefn {Target Hook} void TARGET_ASM_LTO_END (void)
7272 Output to @code{asm_out_file} any text which the assembler expects
7273 to find at the end of an LTO section. The default is to output
7277 @deftypefn {Target Hook} void TARGET_ASM_CODE_END (void)
7278 Output to @code{asm_out_file} any text which is needed before emitting
7279 unwind info and debug info at the end of a file. Some targets emit
7280 here PIC setup thunks that cannot be emitted at the end of file,
7281 because they couldn't have unwind info then. The default is to output
7285 @defmac ASM_COMMENT_START
7286 A C string constant describing how to begin a comment in the target
7287 assembler language. The compiler assumes that the comment will end at
7288 the end of the line.
7292 A C string constant for text to be output before each @code{asm}
7293 statement or group of consecutive ones. Normally this is
7294 @code{"#APP"}, which is a comment that has no effect on most
7295 assemblers but tells the GNU assembler that it must check the lines
7296 that follow for all valid assembler constructs.
7300 A C string constant for text to be output after each @code{asm}
7301 statement or group of consecutive ones. Normally this is
7302 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7303 time-saving assumptions that are valid for ordinary compiler output.
7306 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7307 A C statement to output COFF information or DWARF debugging information
7308 which indicates that filename @var{name} is the current source file to
7309 the stdio stream @var{stream}.
7311 This macro need not be defined if the standard form of output
7312 for the file format in use is appropriate.
7315 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_SOURCE_FILENAME (FILE *@var{file}, const char *@var{name})
7316 Output COFF information or DWARF debugging information which indicates that filename @var{name} is the current source file to the stdio stream @var{file}.
7318 This target hook need not be defined if the standard form of output for the file format in use is appropriate.
7321 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7322 A C statement to output the string @var{string} to the stdio stream
7323 @var{stream}. If you do not call the function @code{output_quoted_string}
7324 in your config files, GCC will only call it to output filenames to
7325 the assembler source. So you can use it to canonicalize the format
7326 of the filename using this macro.
7329 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
7330 A C statement to output something to the assembler file to handle a
7331 @samp{#ident} directive containing the text @var{string}. If this
7332 macro is not defined, nothing is output for a @samp{#ident} directive.
7335 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, tree @var{decl})
7336 Output assembly directives to switch to section @var{name}. The section
7337 should have attributes as specified by @var{flags}, which is a bit mask
7338 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7339 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7340 this section is associated.
7343 @deftypevr {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
7344 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7347 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7348 @deftypevr {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7349 This flag is true if we can create zeroed data by switching to a BSS
7350 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7351 This is true on most ELF targets.
7354 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
7355 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7356 based on a variable or function decl, a section name, and whether or not the
7357 declaration's initializer may contain runtime relocations. @var{decl} may be
7358 null, in which case read-write data should be assumed.
7360 The default version of this function handles choosing code vs data,
7361 read-only vs read-write data, and @code{flag_pic}. You should only
7362 need to override this if your target has special flags that might be
7363 set via @code{__attribute__}.
7366 @deftypefn {Target Hook} int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char *@var{text})
7367 Provides the target with the ability to record the gcc command line
7368 switches that have been passed to the compiler, and options that are
7369 enabled. The @var{type} argument specifies what is being recorded.
7370 It can take the following values:
7373 @item SWITCH_TYPE_PASSED
7374 @var{text} is a command line switch that has been set by the user.
7376 @item SWITCH_TYPE_ENABLED
7377 @var{text} is an option which has been enabled. This might be as a
7378 direct result of a command line switch, or because it is enabled by
7379 default or because it has been enabled as a side effect of a different
7380 command line switch. For example, the @option{-O2} switch enables
7381 various different individual optimization passes.
7383 @item SWITCH_TYPE_DESCRIPTIVE
7384 @var{text} is either NULL or some descriptive text which should be
7385 ignored. If @var{text} is NULL then it is being used to warn the
7386 target hook that either recording is starting or ending. The first
7387 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7388 warning is for start up and the second time the warning is for
7389 wind down. This feature is to allow the target hook to make any
7390 necessary preparations before it starts to record switches and to
7391 perform any necessary tidying up after it has finished recording
7394 @item SWITCH_TYPE_LINE_START
7395 This option can be ignored by this target hook.
7397 @item SWITCH_TYPE_LINE_END
7398 This option can be ignored by this target hook.
7401 The hook's return value must be zero. Other return values may be
7402 supported in the future.
7404 By default this hook is set to NULL, but an example implementation is
7405 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7406 it records the switches as ASCII text inside a new, string mergeable
7407 section in the assembler output file. The name of the new section is
7408 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7412 @deftypevr {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7413 This is the name of the section that will be created by the example
7414 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7420 @subsection Output of Data
7423 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
7424 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7425 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7426 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7427 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7428 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7429 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7430 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7431 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7432 These hooks specify assembly directives for creating certain kinds
7433 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7434 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7435 aligned two-byte object, and so on. Any of the hooks may be
7436 @code{NULL}, indicating that no suitable directive is available.
7438 The compiler will print these strings at the start of a new line,
7439 followed immediately by the object's initial value. In most cases,
7440 the string should contain a tab, a pseudo-op, and then another tab.
7443 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
7444 The @code{assemble_integer} function uses this hook to output an
7445 integer object. @var{x} is the object's value, @var{size} is its size
7446 in bytes and @var{aligned_p} indicates whether it is aligned. The
7447 function should return @code{true} if it was able to output the
7448 object. If it returns false, @code{assemble_integer} will try to
7449 split the object into smaller parts.
7451 The default implementation of this hook will use the
7452 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7453 when the relevant string is @code{NULL}.
7456 @deftypefn {Target Hook} bool TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA (FILE *@var{file}, rtx @var{x})
7457 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
7458 can't deal with, and output assembly code to @var{file} corresponding to
7459 the pattern @var{x}. This may be used to allow machine-dependent
7460 @code{UNSPEC}s to appear within constants.
7462 If target hook fails to recognize a pattern, it must return @code{false},
7463 so that a standard error message is printed. If it prints an error message
7464 itself, by calling, for example, @code{output_operand_lossage}, it may just
7468 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
7469 A C statement to recognize @var{rtx} patterns that
7470 @code{output_addr_const} can't deal with, and output assembly code to
7471 @var{stream} corresponding to the pattern @var{x}. This may be used to
7472 allow machine-dependent @code{UNSPEC}s to appear within constants.
7474 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
7475 @code{goto fail}, so that a standard error message is printed. If it
7476 prints an error message itself, by calling, for example,
7477 @code{output_operand_lossage}, it may just complete normally.
7480 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7481 A C statement to output to the stdio stream @var{stream} an assembler
7482 instruction to assemble a string constant containing the @var{len}
7483 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7484 @code{char *} and @var{len} a C expression of type @code{int}.
7486 If the assembler has a @code{.ascii} pseudo-op as found in the
7487 Berkeley Unix assembler, do not define the macro
7488 @code{ASM_OUTPUT_ASCII}.
7491 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7492 A C statement to output word @var{n} of a function descriptor for
7493 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7494 is defined, and is otherwise unused.
7497 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7498 You may define this macro as a C expression. You should define the
7499 expression to have a nonzero value if GCC should output the constant
7500 pool for a function before the code for the function, or a zero value if
7501 GCC should output the constant pool after the function. If you do
7502 not define this macro, the usual case, GCC will output the constant
7503 pool before the function.
7506 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7507 A C statement to output assembler commands to define the start of the
7508 constant pool for a function. @var{funname} is a string giving
7509 the name of the function. Should the return type of the function
7510 be required, it can be obtained via @var{fundecl}. @var{size}
7511 is the size, in bytes, of the constant pool that will be written
7512 immediately after this call.
7514 If no constant-pool prefix is required, the usual case, this macro need
7518 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7519 A C statement (with or without semicolon) to output a constant in the
7520 constant pool, if it needs special treatment. (This macro need not do
7521 anything for RTL expressions that can be output normally.)
7523 The argument @var{file} is the standard I/O stream to output the
7524 assembler code on. @var{x} is the RTL expression for the constant to
7525 output, and @var{mode} is the machine mode (in case @var{x} is a
7526 @samp{const_int}). @var{align} is the required alignment for the value
7527 @var{x}; you should output an assembler directive to force this much
7530 The argument @var{labelno} is a number to use in an internal label for
7531 the address of this pool entry. The definition of this macro is
7532 responsible for outputting the label definition at the proper place.
7533 Here is how to do this:
7536 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7539 When you output a pool entry specially, you should end with a
7540 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7541 entry from being output a second time in the usual manner.
7543 You need not define this macro if it would do nothing.
7546 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7547 A C statement to output assembler commands to at the end of the constant
7548 pool for a function. @var{funname} is a string giving the name of the
7549 function. Should the return type of the function be required, you can
7550 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7551 constant pool that GCC wrote immediately before this call.
7553 If no constant-pool epilogue is required, the usual case, you need not
7557 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7558 Define this macro as a C expression which is nonzero if @var{C} is
7559 used as a logical line separator by the assembler. @var{STR} points
7560 to the position in the string where @var{C} was found; this can be used if
7561 a line separator uses multiple characters.
7563 If you do not define this macro, the default is that only
7564 the character @samp{;} is treated as a logical line separator.
7567 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7568 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7569 These target hooks are C string constants, describing the syntax in the
7570 assembler for grouping arithmetic expressions. If not overridden, they
7571 default to normal parentheses, which is correct for most assemblers.
7574 These macros are provided by @file{real.h} for writing the definitions
7575 of @code{ASM_OUTPUT_DOUBLE} and the like:
7577 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7578 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7579 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7580 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7581 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7582 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7583 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7584 target's floating point representation, and store its bit pattern in
7585 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7586 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7587 simple @code{long int}. For the others, it should be an array of
7588 @code{long int}. The number of elements in this array is determined
7589 by the size of the desired target floating point data type: 32 bits of
7590 it go in each @code{long int} array element. Each array element holds
7591 32 bits of the result, even if @code{long int} is wider than 32 bits
7592 on the host machine.
7594 The array element values are designed so that you can print them out
7595 using @code{fprintf} in the order they should appear in the target
7599 @node Uninitialized Data
7600 @subsection Output of Uninitialized Variables
7602 Each of the macros in this section is used to do the whole job of
7603 outputting a single uninitialized variable.
7605 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7606 A C statement (sans semicolon) to output to the stdio stream
7607 @var{stream} the assembler definition of a common-label named
7608 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7609 is the size rounded up to whatever alignment the caller wants. It is
7610 possible that @var{size} may be zero, for instance if a struct with no
7611 other member than a zero-length array is defined. In this case, the
7612 backend must output a symbol definition that allocates at least one
7613 byte, both so that the address of the resulting object does not compare
7614 equal to any other, and because some object formats cannot even express
7615 the concept of a zero-sized common symbol, as that is how they represent
7616 an ordinary undefined external.
7618 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7619 output the name itself; before and after that, output the additional
7620 assembler syntax for defining the name, and a newline.
7622 This macro controls how the assembler definitions of uninitialized
7623 common global variables are output.
7626 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7627 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7628 separate, explicit argument. If you define this macro, it is used in
7629 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7630 handling the required alignment of the variable. The alignment is specified
7631 as the number of bits.
7634 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7635 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7636 variable to be output, if there is one, or @code{NULL_TREE} if there
7637 is no corresponding variable. If you define this macro, GCC will use it
7638 in place of both @code{ASM_OUTPUT_COMMON} and
7639 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7640 the variable's decl in order to chose what to output.
7643 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
7644 A C statement (sans semicolon) to output to the stdio stream
7645 @var{stream} the assembler definition of uninitialized global @var{decl} named
7646 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7647 is the size rounded up to whatever alignment the caller wants.
7649 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
7650 defining this macro. If unable, use the expression
7651 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7652 before and after that, output the additional assembler syntax for defining
7653 the name, and a newline.
7655 There are two ways of handling global BSS@. One is to define either
7656 this macro or its aligned counterpart, @code{ASM_OUTPUT_ALIGNED_BSS}.
7657 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7658 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7659 You do not need to do both.
7661 Some languages do not have @code{common} data, and require a
7662 non-common form of global BSS in order to handle uninitialized globals
7663 efficiently. C++ is one example of this. However, if the target does
7664 not support global BSS, the front end may choose to make globals
7665 common in order to save space in the object file.
7668 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7669 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
7670 separate, explicit argument. If you define this macro, it is used in
7671 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
7672 handling the required alignment of the variable. The alignment is specified
7673 as the number of bits.
7675 Try to use function @code{asm_output_aligned_bss} defined in file
7676 @file{varasm.c} when defining this macro.
7679 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7680 A C statement (sans semicolon) to output to the stdio stream
7681 @var{stream} the assembler definition of a local-common-label named
7682 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7683 is the size rounded up to whatever alignment the caller wants.
7685 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7686 output the name itself; before and after that, output the additional
7687 assembler syntax for defining the name, and a newline.
7689 This macro controls how the assembler definitions of uninitialized
7690 static variables are output.
7693 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7694 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7695 separate, explicit argument. If you define this macro, it is used in
7696 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7697 handling the required alignment of the variable. The alignment is specified
7698 as the number of bits.
7701 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7702 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7703 variable to be output, if there is one, or @code{NULL_TREE} if there
7704 is no corresponding variable. If you define this macro, GCC will use it
7705 in place of both @code{ASM_OUTPUT_DECL} and
7706 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7707 the variable's decl in order to chose what to output.
7711 @subsection Output and Generation of Labels
7713 @c prevent bad page break with this line
7714 This is about outputting labels.
7716 @findex assemble_name
7717 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7718 A C statement (sans semicolon) to output to the stdio stream
7719 @var{stream} the assembler definition of a label named @var{name}.
7720 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7721 output the name itself; before and after that, output the additional
7722 assembler syntax for defining the name, and a newline. A default
7723 definition of this macro is provided which is correct for most systems.
7726 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
7727 A C statement (sans semicolon) to output to the stdio stream
7728 @var{stream} the assembler definition of a label named @var{name} of
7730 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7731 output the name itself; before and after that, output the additional
7732 assembler syntax for defining the name, and a newline. A default
7733 definition of this macro is provided which is correct for most systems.
7735 If this macro is not defined, then the function name is defined in the
7736 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7739 @findex assemble_name_raw
7740 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7741 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7742 to refer to a compiler-generated label. The default definition uses
7743 @code{assemble_name_raw}, which is like @code{assemble_name} except
7744 that it is more efficient.
7748 A C string containing the appropriate assembler directive to specify the
7749 size of a symbol, without any arguments. On systems that use ELF, the
7750 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7751 systems, the default is not to define this macro.
7753 Define this macro only if it is correct to use the default definitions
7754 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7755 for your system. If you need your own custom definitions of those
7756 macros, or if you do not need explicit symbol sizes at all, do not
7760 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7761 A C statement (sans semicolon) to output to the stdio stream
7762 @var{stream} a directive telling the assembler that the size of the
7763 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7764 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7768 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7769 A C statement (sans semicolon) to output to the stdio stream
7770 @var{stream} a directive telling the assembler to calculate the size of
7771 the symbol @var{name} by subtracting its address from the current
7774 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7775 provided. The default assumes that the assembler recognizes a special
7776 @samp{.} symbol as referring to the current address, and can calculate
7777 the difference between this and another symbol. If your assembler does
7778 not recognize @samp{.} or cannot do calculations with it, you will need
7779 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7783 A C string containing the appropriate assembler directive to specify the
7784 type of a symbol, without any arguments. On systems that use ELF, the
7785 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7786 systems, the default is not to define this macro.
7788 Define this macro only if it is correct to use the default definition of
7789 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7790 custom definition of this macro, or if you do not need explicit symbol
7791 types at all, do not define this macro.
7794 @defmac TYPE_OPERAND_FMT
7795 A C string which specifies (using @code{printf} syntax) the format of
7796 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7797 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7798 the default is not to define this macro.
7800 Define this macro only if it is correct to use the default definition of
7801 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7802 custom definition of this macro, or if you do not need explicit symbol
7803 types at all, do not define this macro.
7806 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7807 A C statement (sans semicolon) to output to the stdio stream
7808 @var{stream} a directive telling the assembler that the type of the
7809 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7810 that string is always either @samp{"function"} or @samp{"object"}, but
7811 you should not count on this.
7813 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7814 definition of this macro is provided.
7817 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7818 A C statement (sans semicolon) to output to the stdio stream
7819 @var{stream} any text necessary for declaring the name @var{name} of a
7820 function which is being defined. This macro is responsible for
7821 outputting the label definition (perhaps using
7822 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
7823 @code{FUNCTION_DECL} tree node representing the function.
7825 If this macro is not defined, then the function name is defined in the
7826 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
7828 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7832 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7833 A C statement (sans semicolon) to output to the stdio stream
7834 @var{stream} any text necessary for declaring the size of a function
7835 which is being defined. The argument @var{name} is the name of the
7836 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7837 representing the function.
7839 If this macro is not defined, then the function size is not defined.
7841 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7845 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7846 A C statement (sans semicolon) to output to the stdio stream
7847 @var{stream} any text necessary for declaring the name @var{name} of an
7848 initialized variable which is being defined. This macro must output the
7849 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7850 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7852 If this macro is not defined, then the variable name is defined in the
7853 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7855 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7856 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7859 @deftypefn {Target Hook} void TARGET_ASM_DECLARE_CONSTANT_NAME (FILE *@var{file}, const char *@var{name}, const_tree @var{expr}, HOST_WIDE_INT @var{size})
7860 A target hook to output to the stdio stream @var{file} any text necessary
7861 for declaring the name @var{name} of a constant which is being defined. This
7862 target hook is responsible for outputting the label definition (perhaps using
7863 @code{assemble_label}). The argument @var{exp} is the value of the constant,
7864 and @var{size} is the size of the constant in bytes. The @var{name}
7865 will be an internal label.
7867 The default version of this target hook, define the @var{name} in the
7868 usual manner as a label (by means of @code{assemble_label}).
7870 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
7873 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7874 A C statement (sans semicolon) to output to the stdio stream
7875 @var{stream} any text necessary for claiming a register @var{regno}
7876 for a global variable @var{decl} with name @var{name}.
7878 If you don't define this macro, that is equivalent to defining it to do
7882 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7883 A C statement (sans semicolon) to finish up declaring a variable name
7884 once the compiler has processed its initializer fully and thus has had a
7885 chance to determine the size of an array when controlled by an
7886 initializer. This is used on systems where it's necessary to declare
7887 something about the size of the object.
7889 If you don't define this macro, that is equivalent to defining it to do
7892 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7893 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7896 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7897 This target hook is a function to output to the stdio stream
7898 @var{stream} some commands that will make the label @var{name} global;
7899 that is, available for reference from other files.
7901 The default implementation relies on a proper definition of
7902 @code{GLOBAL_ASM_OP}.
7905 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7906 This target hook is a function to output to the stdio stream
7907 @var{stream} some commands that will make the name associated with @var{decl}
7908 global; that is, available for reference from other files.
7910 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7913 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7914 A C statement (sans semicolon) to output to the stdio stream
7915 @var{stream} some commands that will make the label @var{name} weak;
7916 that is, available for reference from other files but only used if
7917 no other definition is available. Use the expression
7918 @code{assemble_name (@var{stream}, @var{name})} to output the name
7919 itself; before and after that, output the additional assembler syntax
7920 for making that name weak, and a newline.
7922 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7923 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7927 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7928 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7929 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7930 or variable decl. If @var{value} is not @code{NULL}, this C statement
7931 should output to the stdio stream @var{stream} assembler code which
7932 defines (equates) the weak symbol @var{name} to have the value
7933 @var{value}. If @var{value} is @code{NULL}, it should output commands
7934 to make @var{name} weak.
7937 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7938 Outputs a directive that enables @var{name} to be used to refer to
7939 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7940 declaration of @code{name}.
7943 @defmac SUPPORTS_WEAK
7944 A preprocessor constant expression which evaluates to true if the target
7945 supports weak symbols.
7947 If you don't define this macro, @file{defaults.h} provides a default
7948 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7949 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
7952 @defmac TARGET_SUPPORTS_WEAK
7953 A C expression which evaluates to true if the target supports weak symbols.
7955 If you don't define this macro, @file{defaults.h} provides a default
7956 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
7957 this macro if you want to control weak symbol support with a compiler
7958 flag such as @option{-melf}.
7961 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7962 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7963 public symbol such that extra copies in multiple translation units will
7964 be discarded by the linker. Define this macro if your object file
7965 format provides support for this concept, such as the @samp{COMDAT}
7966 section flags in the Microsoft Windows PE/COFF format, and this support
7967 requires changes to @var{decl}, such as putting it in a separate section.
7970 @defmac SUPPORTS_ONE_ONLY
7971 A C expression which evaluates to true if the target supports one-only
7974 If you don't define this macro, @file{varasm.c} provides a default
7975 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7976 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7977 you want to control one-only symbol support with a compiler flag, or if
7978 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7979 be emitted as one-only.
7982 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, int @var{visibility})
7983 This target hook is a function to output to @var{asm_out_file} some
7984 commands that will make the symbol(s) associated with @var{decl} have
7985 hidden, protected or internal visibility as specified by @var{visibility}.
7988 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7989 A C expression that evaluates to true if the target's linker expects
7990 that weak symbols do not appear in a static archive's table of contents.
7991 The default is @code{0}.
7993 Leaving weak symbols out of an archive's table of contents means that,
7994 if a symbol will only have a definition in one translation unit and
7995 will have undefined references from other translation units, that
7996 symbol should not be weak. Defining this macro to be nonzero will
7997 thus have the effect that certain symbols that would normally be weak
7998 (explicit template instantiations, and vtables for polymorphic classes
7999 with noninline key methods) will instead be nonweak.
8001 The C++ ABI requires this macro to be zero. Define this macro for
8002 targets where full C++ ABI compliance is impossible and where linker
8003 restrictions require weak symbols to be left out of a static archive's
8007 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
8008 A C statement (sans semicolon) to output to the stdio stream
8009 @var{stream} any text necessary for declaring the name of an external
8010 symbol named @var{name} which is referenced in this compilation but
8011 not defined. The value of @var{decl} is the tree node for the
8014 This macro need not be defined if it does not need to output anything.
8015 The GNU assembler and most Unix assemblers don't require anything.
8018 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
8019 This target hook is a function to output to @var{asm_out_file} an assembler
8020 pseudo-op to declare a library function name external. The name of the
8021 library function is given by @var{symref}, which is a @code{symbol_ref}.
8024 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (const char *@var{symbol})
8025 This target hook is a function to output to @var{asm_out_file} an assembler
8026 directive to annotate @var{symbol} as used. The Darwin target uses the
8027 .no_dead_code_strip directive.
8030 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
8031 A C statement (sans semicolon) to output to the stdio stream
8032 @var{stream} a reference in assembler syntax to a label named
8033 @var{name}. This should add @samp{_} to the front of the name, if that
8034 is customary on your operating system, as it is in most Berkeley Unix
8035 systems. This macro is used in @code{assemble_name}.
8038 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
8039 A C statement (sans semicolon) to output a reference to
8040 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8041 will be used to output the name of the symbol. This macro may be used
8042 to modify the way a symbol is referenced depending on information
8043 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8046 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8047 A C statement (sans semicolon) to output a reference to @var{buf}, the
8048 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8049 @code{assemble_name} will be used to output the name of the symbol.
8050 This macro is not used by @code{output_asm_label}, or the @code{%l}
8051 specifier that calls it; the intention is that this macro should be set
8052 when it is necessary to output a label differently when its address is
8056 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
8057 A function to output to the stdio stream @var{stream} a label whose
8058 name is made from the string @var{prefix} and the number @var{labelno}.
8060 It is absolutely essential that these labels be distinct from the labels
8061 used for user-level functions and variables. Otherwise, certain programs
8062 will have name conflicts with internal labels.
8064 It is desirable to exclude internal labels from the symbol table of the
8065 object file. Most assemblers have a naming convention for labels that
8066 should be excluded; on many systems, the letter @samp{L} at the
8067 beginning of a label has this effect. You should find out what
8068 convention your system uses, and follow it.
8070 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8073 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8074 A C statement to output to the stdio stream @var{stream} a debug info
8075 label whose name is made from the string @var{prefix} and the number
8076 @var{num}. This is useful for VLIW targets, where debug info labels
8077 may need to be treated differently than branch target labels. On some
8078 systems, branch target labels must be at the beginning of instruction
8079 bundles, but debug info labels can occur in the middle of instruction
8082 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8086 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8087 A C statement to store into the string @var{string} a label whose name
8088 is made from the string @var{prefix} and the number @var{num}.
8090 This string, when output subsequently by @code{assemble_name}, should
8091 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8092 with the same @var{prefix} and @var{num}.
8094 If the string begins with @samp{*}, then @code{assemble_name} will
8095 output the rest of the string unchanged. It is often convenient for
8096 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8097 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8098 to output the string, and may change it. (Of course,
8099 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8100 you should know what it does on your machine.)
8103 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8104 A C expression to assign to @var{outvar} (which is a variable of type
8105 @code{char *}) a newly allocated string made from the string
8106 @var{name} and the number @var{number}, with some suitable punctuation
8107 added. Use @code{alloca} to get space for the string.
8109 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8110 produce an assembler label for an internal static variable whose name is
8111 @var{name}. Therefore, the string must be such as to result in valid
8112 assembler code. The argument @var{number} is different each time this
8113 macro is executed; it prevents conflicts between similarly-named
8114 internal static variables in different scopes.
8116 Ideally this string should not be a valid C identifier, to prevent any
8117 conflict with the user's own symbols. Most assemblers allow periods
8118 or percent signs in assembler symbols; putting at least one of these
8119 between the name and the number will suffice.
8121 If this macro is not defined, a default definition will be provided
8122 which is correct for most systems.
8125 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8126 A C statement to output to the stdio stream @var{stream} assembler code
8127 which defines (equates) the symbol @var{name} to have the value @var{value}.
8130 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8131 correct for most systems.
8134 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8135 A C statement to output to the stdio stream @var{stream} assembler code
8136 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8137 to have the value of the tree node @var{decl_of_value}. This macro will
8138 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8139 the tree nodes are available.
8142 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8143 correct for most systems.
8146 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8147 A C statement that evaluates to true if the assembler code which defines
8148 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8149 of the tree node @var{decl_of_value} should be emitted near the end of the
8150 current compilation unit. The default is to not defer output of defines.
8151 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8152 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8155 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8156 A C statement to output to the stdio stream @var{stream} assembler code
8157 which defines (equates) the weak symbol @var{name} to have the value
8158 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8159 an undefined weak symbol.
8161 Define this macro if the target only supports weak aliases; define
8162 @code{ASM_OUTPUT_DEF} instead if possible.
8165 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8166 Define this macro to override the default assembler names used for
8167 Objective-C methods.
8169 The default name is a unique method number followed by the name of the
8170 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8171 the category is also included in the assembler name (e.g.@:
8174 These names are safe on most systems, but make debugging difficult since
8175 the method's selector is not present in the name. Therefore, particular
8176 systems define other ways of computing names.
8178 @var{buf} is an expression of type @code{char *} which gives you a
8179 buffer in which to store the name; its length is as long as
8180 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8181 50 characters extra.
8183 The argument @var{is_inst} specifies whether the method is an instance
8184 method or a class method; @var{class_name} is the name of the class;
8185 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8186 in a category); and @var{sel_name} is the name of the selector.
8188 On systems where the assembler can handle quoted names, you can use this
8189 macro to provide more human-readable names.
8192 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
8193 A C statement (sans semicolon) to output to the stdio stream
8194 @var{stream} commands to declare that the label @var{name} is an
8195 Objective-C class reference. This is only needed for targets whose
8196 linkers have special support for NeXT-style runtimes.
8199 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
8200 A C statement (sans semicolon) to output to the stdio stream
8201 @var{stream} commands to declare that the label @var{name} is an
8202 unresolved Objective-C class reference. This is only needed for targets
8203 whose linkers have special support for NeXT-style runtimes.
8206 @node Initialization
8207 @subsection How Initialization Functions Are Handled
8208 @cindex initialization routines
8209 @cindex termination routines
8210 @cindex constructors, output of
8211 @cindex destructors, output of
8213 The compiled code for certain languages includes @dfn{constructors}
8214 (also called @dfn{initialization routines})---functions to initialize
8215 data in the program when the program is started. These functions need
8216 to be called before the program is ``started''---that is to say, before
8217 @code{main} is called.
8219 Compiling some languages generates @dfn{destructors} (also called
8220 @dfn{termination routines}) that should be called when the program
8223 To make the initialization and termination functions work, the compiler
8224 must output something in the assembler code to cause those functions to
8225 be called at the appropriate time. When you port the compiler to a new
8226 system, you need to specify how to do this.
8228 There are two major ways that GCC currently supports the execution of
8229 initialization and termination functions. Each way has two variants.
8230 Much of the structure is common to all four variations.
8232 @findex __CTOR_LIST__
8233 @findex __DTOR_LIST__
8234 The linker must build two lists of these functions---a list of
8235 initialization functions, called @code{__CTOR_LIST__}, and a list of
8236 termination functions, called @code{__DTOR_LIST__}.
8238 Each list always begins with an ignored function pointer (which may hold
8239 0, @minus{}1, or a count of the function pointers after it, depending on
8240 the environment). This is followed by a series of zero or more function
8241 pointers to constructors (or destructors), followed by a function
8242 pointer containing zero.
8244 Depending on the operating system and its executable file format, either
8245 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8246 time and exit time. Constructors are called in reverse order of the
8247 list; destructors in forward order.
8249 The best way to handle static constructors works only for object file
8250 formats which provide arbitrarily-named sections. A section is set
8251 aside for a list of constructors, and another for a list of destructors.
8252 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8253 object file that defines an initialization function also puts a word in
8254 the constructor section to point to that function. The linker
8255 accumulates all these words into one contiguous @samp{.ctors} section.
8256 Termination functions are handled similarly.
8258 This method will be chosen as the default by @file{target-def.h} if
8259 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8260 support arbitrary sections, but does support special designated
8261 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8262 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8264 When arbitrary sections are available, there are two variants, depending
8265 upon how the code in @file{crtstuff.c} is called. On systems that
8266 support a @dfn{.init} section which is executed at program startup,
8267 parts of @file{crtstuff.c} are compiled into that section. The
8268 program is linked by the @command{gcc} driver like this:
8271 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8274 The prologue of a function (@code{__init}) appears in the @code{.init}
8275 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8276 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8277 files are provided by the operating system or by the GNU C library, but
8278 are provided by GCC for a few targets.
8280 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8281 compiled from @file{crtstuff.c}. They contain, among other things, code
8282 fragments within the @code{.init} and @code{.fini} sections that branch
8283 to routines in the @code{.text} section. The linker will pull all parts
8284 of a section together, which results in a complete @code{__init} function
8285 that invokes the routines we need at startup.
8287 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8290 If no init section is available, when GCC compiles any function called
8291 @code{main} (or more accurately, any function designated as a program
8292 entry point by the language front end calling @code{expand_main_function}),
8293 it inserts a procedure call to @code{__main} as the first executable code
8294 after the function prologue. The @code{__main} function is defined
8295 in @file{libgcc2.c} and runs the global constructors.
8297 In file formats that don't support arbitrary sections, there are again
8298 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8299 and an `a.out' format must be used. In this case,
8300 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8301 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8302 and with the address of the void function containing the initialization
8303 code as its value. The GNU linker recognizes this as a request to add
8304 the value to a @dfn{set}; the values are accumulated, and are eventually
8305 placed in the executable as a vector in the format described above, with
8306 a leading (ignored) count and a trailing zero element.
8307 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8308 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8309 the compilation of @code{main} to call @code{__main} as above, starting
8310 the initialization process.
8312 The last variant uses neither arbitrary sections nor the GNU linker.
8313 This is preferable when you want to do dynamic linking and when using
8314 file formats which the GNU linker does not support, such as `ECOFF'@. In
8315 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8316 termination functions are recognized simply by their names. This requires
8317 an extra program in the linkage step, called @command{collect2}. This program
8318 pretends to be the linker, for use with GCC; it does its job by running
8319 the ordinary linker, but also arranges to include the vectors of
8320 initialization and termination functions. These functions are called
8321 via @code{__main} as described above. In order to use this method,
8322 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8325 The following section describes the specific macros that control and
8326 customize the handling of initialization and termination functions.
8329 @node Macros for Initialization
8330 @subsection Macros Controlling Initialization Routines
8332 Here are the macros that control how the compiler handles initialization
8333 and termination functions:
8335 @defmac INIT_SECTION_ASM_OP
8336 If defined, a C string constant, including spacing, for the assembler
8337 operation to identify the following data as initialization code. If not
8338 defined, GCC will assume such a section does not exist. When you are
8339 using special sections for initialization and termination functions, this
8340 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8341 run the initialization functions.
8344 @defmac HAS_INIT_SECTION
8345 If defined, @code{main} will not call @code{__main} as described above.
8346 This macro should be defined for systems that control start-up code
8347 on a symbol-by-symbol basis, such as OSF/1, and should not
8348 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8351 @defmac LD_INIT_SWITCH
8352 If defined, a C string constant for a switch that tells the linker that
8353 the following symbol is an initialization routine.
8356 @defmac LD_FINI_SWITCH
8357 If defined, a C string constant for a switch that tells the linker that
8358 the following symbol is a finalization routine.
8361 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8362 If defined, a C statement that will write a function that can be
8363 automatically called when a shared library is loaded. The function
8364 should call @var{func}, which takes no arguments. If not defined, and
8365 the object format requires an explicit initialization function, then a
8366 function called @code{_GLOBAL__DI} will be generated.
8368 This function and the following one are used by collect2 when linking a
8369 shared library that needs constructors or destructors, or has DWARF2
8370 exception tables embedded in the code.
8373 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8374 If defined, a C statement that will write a function that can be
8375 automatically called when a shared library is unloaded. The function
8376 should call @var{func}, which takes no arguments. If not defined, and
8377 the object format requires an explicit finalization function, then a
8378 function called @code{_GLOBAL__DD} will be generated.
8381 @defmac INVOKE__main
8382 If defined, @code{main} will call @code{__main} despite the presence of
8383 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8384 where the init section is not actually run automatically, but is still
8385 useful for collecting the lists of constructors and destructors.
8388 @defmac SUPPORTS_INIT_PRIORITY
8389 If nonzero, the C++ @code{init_priority} attribute is supported and the
8390 compiler should emit instructions to control the order of initialization
8391 of objects. If zero, the compiler will issue an error message upon
8392 encountering an @code{init_priority} attribute.
8395 @deftypevr {Target Hook} bool TARGET_HAVE_CTORS_DTORS
8396 This value is true if the target supports some ``native'' method of
8397 collecting constructors and destructors to be run at startup and exit.
8398 It is false if we must use @command{collect2}.
8401 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
8402 If defined, a function that outputs assembler code to arrange to call
8403 the function referenced by @var{symbol} at initialization time.
8405 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8406 no arguments and with no return value. If the target supports initialization
8407 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8408 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8410 If this macro is not defined by the target, a suitable default will
8411 be chosen if (1) the target supports arbitrary section names, (2) the
8412 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8416 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
8417 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8418 functions rather than initialization functions.
8421 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8422 generated for the generated object file will have static linkage.
8424 If your system uses @command{collect2} as the means of processing
8425 constructors, then that program normally uses @command{nm} to scan
8426 an object file for constructor functions to be called.
8428 On certain kinds of systems, you can define this macro to make
8429 @command{collect2} work faster (and, in some cases, make it work at all):
8431 @defmac OBJECT_FORMAT_COFF
8432 Define this macro if the system uses COFF (Common Object File Format)
8433 object files, so that @command{collect2} can assume this format and scan
8434 object files directly for dynamic constructor/destructor functions.
8436 This macro is effective only in a native compiler; @command{collect2} as
8437 part of a cross compiler always uses @command{nm} for the target machine.
8440 @defmac REAL_NM_FILE_NAME
8441 Define this macro as a C string constant containing the file name to use
8442 to execute @command{nm}. The default is to search the path normally for
8445 If your system supports shared libraries and has a program to list the
8446 dynamic dependencies of a given library or executable, you can define
8447 these macros to enable support for running initialization and
8448 termination functions in shared libraries:
8452 Define this macro to a C string constant containing the name of the program
8453 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
8456 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8457 Define this macro to be C code that extracts filenames from the output
8458 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8459 of type @code{char *} that points to the beginning of a line of output
8460 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8461 code must advance @var{ptr} to the beginning of the filename on that
8462 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8465 @defmac SHLIB_SUFFIX
8466 Define this macro to a C string constant containing the default shared
8467 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8468 strips version information after this suffix when generating global
8469 constructor and destructor names. This define is only needed on targets
8470 that use @command{collect2} to process constructors and destructors.
8473 @node Instruction Output
8474 @subsection Output of Assembler Instructions
8476 @c prevent bad page break with this line
8477 This describes assembler instruction output.
8479 @defmac REGISTER_NAMES
8480 A C initializer containing the assembler's names for the machine
8481 registers, each one as a C string constant. This is what translates
8482 register numbers in the compiler into assembler language.
8485 @defmac ADDITIONAL_REGISTER_NAMES
8486 If defined, a C initializer for an array of structures containing a name
8487 and a register number. This macro defines additional names for hard
8488 registers, thus allowing the @code{asm} option in declarations to refer
8489 to registers using alternate names.
8492 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8493 Define this macro if you are using an unusual assembler that
8494 requires different names for the machine instructions.
8496 The definition is a C statement or statements which output an
8497 assembler instruction opcode to the stdio stream @var{stream}. The
8498 macro-operand @var{ptr} is a variable of type @code{char *} which
8499 points to the opcode name in its ``internal'' form---the form that is
8500 written in the machine description. The definition should output the
8501 opcode name to @var{stream}, performing any translation you desire, and
8502 increment the variable @var{ptr} to point at the end of the opcode
8503 so that it will not be output twice.
8505 In fact, your macro definition may process less than the entire opcode
8506 name, or more than the opcode name; but if you want to process text
8507 that includes @samp{%}-sequences to substitute operands, you must take
8508 care of the substitution yourself. Just be sure to increment
8509 @var{ptr} over whatever text should not be output normally.
8511 @findex recog_data.operand
8512 If you need to look at the operand values, they can be found as the
8513 elements of @code{recog_data.operand}.
8515 If the macro definition does nothing, the instruction is output
8519 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8520 If defined, a C statement to be executed just prior to the output of
8521 assembler code for @var{insn}, to modify the extracted operands so
8522 they will be output differently.
8524 Here the argument @var{opvec} is the vector containing the operands
8525 extracted from @var{insn}, and @var{noperands} is the number of
8526 elements of the vector which contain meaningful data for this insn.
8527 The contents of this vector are what will be used to convert the insn
8528 template into assembler code, so you can change the assembler output
8529 by changing the contents of the vector.
8531 This macro is useful when various assembler syntaxes share a single
8532 file of instruction patterns; by defining this macro differently, you
8533 can cause a large class of instructions to be output differently (such
8534 as with rearranged operands). Naturally, variations in assembler
8535 syntax affecting individual insn patterns ought to be handled by
8536 writing conditional output routines in those patterns.
8538 If this macro is not defined, it is equivalent to a null statement.
8541 @deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{file}, rtx @var{insn}, rtx *@var{opvec}, int @var{noperands})
8542 If defined, this target hook is a function which is executed just after the
8543 output of assembler code for @var{insn}, to change the mode of the assembler
8546 Here the argument @var{opvec} is the vector containing the operands
8547 extracted from @var{insn}, and @var{noperands} is the number of
8548 elements of the vector which contain meaningful data for this insn.
8549 The contents of this vector are what was used to convert the insn
8550 template into assembler code, so you can change the assembler mode
8551 by checking the contents of the vector.
8554 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8555 A C compound statement to output to stdio stream @var{stream} the
8556 assembler syntax for an instruction operand @var{x}. @var{x} is an
8559 @var{code} is a value that can be used to specify one of several ways
8560 of printing the operand. It is used when identical operands must be
8561 printed differently depending on the context. @var{code} comes from
8562 the @samp{%} specification that was used to request printing of the
8563 operand. If the specification was just @samp{%@var{digit}} then
8564 @var{code} is 0; if the specification was @samp{%@var{ltr}
8565 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8568 If @var{x} is a register, this macro should print the register's name.
8569 The names can be found in an array @code{reg_names} whose type is
8570 @code{char *[]}. @code{reg_names} is initialized from
8571 @code{REGISTER_NAMES}.
8573 When the machine description has a specification @samp{%@var{punct}}
8574 (a @samp{%} followed by a punctuation character), this macro is called
8575 with a null pointer for @var{x} and the punctuation character for
8579 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8580 A C expression which evaluates to true if @var{code} is a valid
8581 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8582 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8583 punctuation characters (except for the standard one, @samp{%}) are used
8587 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8588 A C compound statement to output to stdio stream @var{stream} the
8589 assembler syntax for an instruction operand that is a memory reference
8590 whose address is @var{x}. @var{x} is an RTL expression.
8592 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8593 On some machines, the syntax for a symbolic address depends on the
8594 section that the address refers to. On these machines, define the hook
8595 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8596 @code{symbol_ref}, and then check for it here. @xref{Assembler
8600 @findex dbr_sequence_length
8601 @defmac DBR_OUTPUT_SEQEND (@var{file})
8602 A C statement, to be executed after all slot-filler instructions have
8603 been output. If necessary, call @code{dbr_sequence_length} to
8604 determine the number of slots filled in a sequence (zero if not
8605 currently outputting a sequence), to decide how many no-ops to output,
8608 Don't define this macro if it has nothing to do, but it is helpful in
8609 reading assembly output if the extent of the delay sequence is made
8610 explicit (e.g.@: with white space).
8613 @findex final_sequence
8614 Note that output routines for instructions with delay slots must be
8615 prepared to deal with not being output as part of a sequence
8616 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8617 found.) The variable @code{final_sequence} is null when not
8618 processing a sequence, otherwise it contains the @code{sequence} rtx
8622 @defmac REGISTER_PREFIX
8623 @defmacx LOCAL_LABEL_PREFIX
8624 @defmacx USER_LABEL_PREFIX
8625 @defmacx IMMEDIATE_PREFIX
8626 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8627 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8628 @file{final.c}). These are useful when a single @file{md} file must
8629 support multiple assembler formats. In that case, the various @file{tm.h}
8630 files can define these macros differently.
8633 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8634 If defined this macro should expand to a series of @code{case}
8635 statements which will be parsed inside the @code{switch} statement of
8636 the @code{asm_fprintf} function. This allows targets to define extra
8637 printf formats which may useful when generating their assembler
8638 statements. Note that uppercase letters are reserved for future
8639 generic extensions to asm_fprintf, and so are not available to target
8640 specific code. The output file is given by the parameter @var{file}.
8641 The varargs input pointer is @var{argptr} and the rest of the format
8642 string, starting the character after the one that is being switched
8643 upon, is pointed to by @var{format}.
8646 @defmac ASSEMBLER_DIALECT
8647 If your target supports multiple dialects of assembler language (such as
8648 different opcodes), define this macro as a C expression that gives the
8649 numeric index of the assembler language dialect to use, with zero as the
8652 If this macro is defined, you may use constructs of the form
8654 @samp{@{option0|option1|option2@dots{}@}}
8657 in the output templates of patterns (@pxref{Output Template}) or in the
8658 first argument of @code{asm_fprintf}. This construct outputs
8659 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8660 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8661 within these strings retain their usual meaning. If there are fewer
8662 alternatives within the braces than the value of
8663 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8665 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8666 @samp{@}} do not have any special meaning when used in templates or
8667 operands to @code{asm_fprintf}.
8669 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8670 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8671 the variations in assembler language syntax with that mechanism. Define
8672 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8673 if the syntax variant are larger and involve such things as different
8674 opcodes or operand order.
8677 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8678 A C expression to output to @var{stream} some assembler code
8679 which will push hard register number @var{regno} onto the stack.
8680 The code need not be optimal, since this macro is used only when
8684 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8685 A C expression to output to @var{stream} some assembler code
8686 which will pop hard register number @var{regno} off of the stack.
8687 The code need not be optimal, since this macro is used only when
8691 @node Dispatch Tables
8692 @subsection Output of Dispatch Tables
8694 @c prevent bad page break with this line
8695 This concerns dispatch tables.
8697 @cindex dispatch table
8698 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8699 A C statement to output to the stdio stream @var{stream} an assembler
8700 pseudo-instruction to generate a difference between two labels.
8701 @var{value} and @var{rel} are the numbers of two internal labels. The
8702 definitions of these labels are output using
8703 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8704 way here. For example,
8707 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8708 @var{value}, @var{rel})
8711 You must provide this macro on machines where the addresses in a
8712 dispatch table are relative to the table's own address. If defined, GCC
8713 will also use this macro on all machines when producing PIC@.
8714 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8715 mode and flags can be read.
8718 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8719 This macro should be provided on machines where the addresses
8720 in a dispatch table are absolute.
8722 The definition should be a C statement to output to the stdio stream
8723 @var{stream} an assembler pseudo-instruction to generate a reference to
8724 a label. @var{value} is the number of an internal label whose
8725 definition is output using @code{(*targetm.asm_out.internal_label)}.
8729 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8733 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8734 Define this if the label before a jump-table needs to be output
8735 specially. The first three arguments are the same as for
8736 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8737 jump-table which follows (a @code{jump_insn} containing an
8738 @code{addr_vec} or @code{addr_diff_vec}).
8740 This feature is used on system V to output a @code{swbeg} statement
8743 If this macro is not defined, these labels are output with
8744 @code{(*targetm.asm_out.internal_label)}.
8747 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8748 Define this if something special must be output at the end of a
8749 jump-table. The definition should be a C statement to be executed
8750 after the assembler code for the table is written. It should write
8751 the appropriate code to stdio stream @var{stream}. The argument
8752 @var{table} is the jump-table insn, and @var{num} is the label-number
8753 of the preceding label.
8755 If this macro is not defined, nothing special is output at the end of
8759 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (FILE *@var{stream}, tree @var{decl}, int @var{for_eh}, int @var{empty})
8760 This target hook emits a label at the beginning of each FDE@. It
8761 should be defined on targets where FDEs need special labels, and it
8762 should write the appropriate label, for the FDE associated with the
8763 function declaration @var{decl}, to the stdio stream @var{stream}.
8764 The third argument, @var{for_eh}, is a boolean: true if this is for an
8765 exception table. The fourth argument, @var{empty}, is a boolean:
8766 true if this is a placeholder label for an omitted FDE@.
8768 The default is that FDEs are not given nonlocal labels.
8771 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (FILE *@var{stream})
8772 This target hook emits a label at the beginning of the exception table.
8773 It should be defined on targets where it is desirable for the table
8774 to be broken up according to function.
8776 The default is that no label is emitted.
8779 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_PERSONALITY (rtx @var{personality})
8780 If the target implements @code{TARGET_ASM_UNWIND_EMIT}, this hook may be used to emit a directive to install a personality hook into the unwind info. This hook should not be used if dwarf2 unwind info is used.
8783 @deftypefn {Target Hook} void TARGET_ASM_UNWIND_EMIT (FILE *@var{stream}, rtx @var{insn})
8784 This target hook emits assembly directives required to unwind the
8785 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
8786 returns @code{UI_TARGET}.
8789 @deftypevr {Target Hook} bool TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
8790 True if the @code{TARGET_ASM_UNWIND_EMIT} hook should be called before the assembly for @var{insn} has been emitted, false if the hook should be called afterward.
8793 @node Exception Region Output
8794 @subsection Assembler Commands for Exception Regions
8796 @c prevent bad page break with this line
8798 This describes commands marking the start and the end of an exception
8801 @defmac EH_FRAME_SECTION_NAME
8802 If defined, a C string constant for the name of the section containing
8803 exception handling frame unwind information. If not defined, GCC will
8804 provide a default definition if the target supports named sections.
8805 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8807 You should define this symbol if your target supports DWARF 2 frame
8808 unwind information and the default definition does not work.
8811 @defmac EH_FRAME_IN_DATA_SECTION
8812 If defined, DWARF 2 frame unwind information will be placed in the
8813 data section even though the target supports named sections. This
8814 might be necessary, for instance, if the system linker does garbage
8815 collection and sections cannot be marked as not to be collected.
8817 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8821 @defmac EH_TABLES_CAN_BE_READ_ONLY
8822 Define this macro to 1 if your target is such that no frame unwind
8823 information encoding used with non-PIC code will ever require a
8824 runtime relocation, but the linker may not support merging read-only
8825 and read-write sections into a single read-write section.
8828 @defmac MASK_RETURN_ADDR
8829 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8830 that it does not contain any extraneous set bits in it.
8833 @defmac DWARF2_UNWIND_INFO
8834 Define this macro to 0 if your target supports DWARF 2 frame unwind
8835 information, but it does not yet work with exception handling.
8836 Otherwise, if your target supports this information (if it defines
8837 @code{INCOMING_RETURN_ADDR_RTX} and either @code{UNALIGNED_INT_ASM_OP}
8838 or @code{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8841 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_EXCEPT_UNWIND_INFO (void)
8842 This hook defines the mechanism that will be used for exception handling
8843 by the target. If the target has ABI specified unwind tables, the hook
8844 should return @code{UI_TARGET}. If the target is to use the
8845 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
8846 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
8847 information, the hook should return @code{UI_DWARF2}.
8849 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
8850 This may end up simplifying other parts of target-specific code. The
8851 default implementation of this hook never returns @code{UI_NONE}.
8853 Note that the value returned by this hook should be constant. It should
8854 not depend on anything except command-line switches. In particular, the
8855 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
8856 macros and builtin functions related to exception handling are set up
8857 depending on this setting.
8859 The default implementation of the hook first honors the
8860 @option{--enable-sjlj-exceptions} configure option, then
8861 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}.
8864 @deftypevr {Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8865 This variable should be set to @code{true} if the target ABI requires unwinding
8866 tables even when exceptions are not used.
8869 @defmac MUST_USE_SJLJ_EXCEPTIONS
8870 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
8871 runtime-variable. In that case, @file{except.h} cannot correctly
8872 determine the corresponding definition of @code{MUST_USE_SJLJ_EXCEPTIONS},
8873 so the target must provide it directly.
8876 @defmac DONT_USE_BUILTIN_SETJMP
8877 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8878 should use the @code{setjmp}/@code{longjmp} functions from the C library
8879 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8882 @defmac DWARF_CIE_DATA_ALIGNMENT
8883 This macro need only be defined if the target might save registers in the
8884 function prologue at an offset to the stack pointer that is not aligned to
8885 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8886 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8887 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8888 the target supports DWARF 2 frame unwind information.
8891 @deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8892 Contains the value true if the target should add a zero word onto the
8893 end of a Dwarf-2 frame info section when used for exception handling.
8894 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8898 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
8899 Given a register, this hook should return a parallel of registers to
8900 represent where to find the register pieces. Define this hook if the
8901 register and its mode are represented in Dwarf in non-contiguous
8902 locations, or if the register should be represented in more than one
8903 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8904 If not defined, the default is to return @code{NULL_RTX}.
8907 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
8908 If some registers are represented in Dwarf-2 unwind information in
8909 multiple pieces, define this hook to fill in information about the
8910 sizes of those pieces in the table used by the unwinder at runtime.
8911 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8912 filling in a single size corresponding to each hard register;
8913 @var{address} is the address of the table.
8916 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
8917 This hook is used to output a reference from a frame unwinding table to
8918 the type_info object identified by @var{sym}. It should return @code{true}
8919 if the reference was output. Returning @code{false} will cause the
8920 reference to be output using the normal Dwarf2 routines.
8923 @deftypevr {Target Hook} bool TARGET_ARM_EABI_UNWINDER
8924 This flag should be set to @code{true} on targets that use an ARM EABI
8925 based unwinding library, and @code{false} on other targets. This effects
8926 the format of unwinding tables, and how the unwinder in entered after
8927 running a cleanup. The default is @code{false}.
8930 @node Alignment Output
8931 @subsection Assembler Commands for Alignment
8933 @c prevent bad page break with this line
8934 This describes commands for alignment.
8936 @defmac JUMP_ALIGN (@var{label})
8937 The alignment (log base 2) to put in front of @var{label}, which is
8938 a common destination of jumps and has no fallthru incoming edge.
8940 This macro need not be defined if you don't want any special alignment
8941 to be done at such a time. Most machine descriptions do not currently
8944 Unless it's necessary to inspect the @var{label} parameter, it is better
8945 to set the variable @var{align_jumps} in the target's
8946 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8947 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8950 @deftypefn {Target Hook} int TARGET_ASM_JUMP_ALIGN_MAX_SKIP (rtx @var{label})
8951 The maximum number of bytes to skip before @var{label} when applying
8952 @code{JUMP_ALIGN}. This works only if
8953 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8956 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8957 The alignment (log base 2) to put in front of @var{label}, which follows
8960 This macro need not be defined if you don't want any special alignment
8961 to be done at such a time. Most machine descriptions do not currently
8965 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP (rtx @var{label})
8966 The maximum number of bytes to skip before @var{label} when applying
8967 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8968 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8971 @defmac LOOP_ALIGN (@var{label})
8972 The alignment (log base 2) to put in front of @var{label}, which follows
8973 a @code{NOTE_INSN_LOOP_BEG} note.
8975 This macro need not be defined if you don't want any special alignment
8976 to be done at such a time. Most machine descriptions do not currently
8979 Unless it's necessary to inspect the @var{label} parameter, it is better
8980 to set the variable @code{align_loops} in the target's
8981 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8982 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
8985 @deftypefn {Target Hook} int TARGET_ASM_LOOP_ALIGN_MAX_SKIP (rtx @var{label})
8986 The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to
8987 @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is
8991 @defmac LABEL_ALIGN (@var{label})
8992 The alignment (log base 2) to put in front of @var{label}.
8993 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
8994 the maximum of the specified values is used.
8996 Unless it's necessary to inspect the @var{label} parameter, it is better
8997 to set the variable @code{align_labels} in the target's
8998 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
8999 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
9002 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_MAX_SKIP (rtx @var{label})
9003 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}
9004 to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN}
9008 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
9009 A C statement to output to the stdio stream @var{stream} an assembler
9010 instruction to advance the location counter by @var{nbytes} bytes.
9011 Those bytes should be zero when loaded. @var{nbytes} will be a C
9012 expression of type @code{unsigned HOST_WIDE_INT}.
9015 @defmac ASM_NO_SKIP_IN_TEXT
9016 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
9017 text section because it fails to put zeros in the bytes that are skipped.
9018 This is true on many Unix systems, where the pseudo--op to skip bytes
9019 produces no-op instructions rather than zeros when used in the text
9023 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9024 A C statement to output to the stdio stream @var{stream} an assembler
9025 command to advance the location counter to a multiple of 2 to the
9026 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9029 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9030 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9031 for padding, if necessary.
9034 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9035 A C statement to output to the stdio stream @var{stream} an assembler
9036 command to advance the location counter to a multiple of 2 to the
9037 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9038 satisfy the alignment request. @var{power} and @var{max_skip} will be
9039 a C expression of type @code{int}.
9043 @node Debugging Info
9044 @section Controlling Debugging Information Format
9046 @c prevent bad page break with this line
9047 This describes how to specify debugging information.
9050 * All Debuggers:: Macros that affect all debugging formats uniformly.
9051 * DBX Options:: Macros enabling specific options in DBX format.
9052 * DBX Hooks:: Hook macros for varying DBX format.
9053 * File Names and DBX:: Macros controlling output of file names in DBX format.
9054 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
9055 * VMS Debug:: Macros for VMS debug format.
9059 @subsection Macros Affecting All Debugging Formats
9061 @c prevent bad page break with this line
9062 These macros affect all debugging formats.
9064 @defmac DBX_REGISTER_NUMBER (@var{regno})
9065 A C expression that returns the DBX register number for the compiler
9066 register number @var{regno}. In the default macro provided, the value
9067 of this expression will be @var{regno} itself. But sometimes there are
9068 some registers that the compiler knows about and DBX does not, or vice
9069 versa. In such cases, some register may need to have one number in the
9070 compiler and another for DBX@.
9072 If two registers have consecutive numbers inside GCC, and they can be
9073 used as a pair to hold a multiword value, then they @emph{must} have
9074 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9075 Otherwise, debuggers will be unable to access such a pair, because they
9076 expect register pairs to be consecutive in their own numbering scheme.
9078 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9079 does not preserve register pairs, then what you must do instead is
9080 redefine the actual register numbering scheme.
9083 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9084 A C expression that returns the integer offset value for an automatic
9085 variable having address @var{x} (an RTL expression). The default
9086 computation assumes that @var{x} is based on the frame-pointer and
9087 gives the offset from the frame-pointer. This is required for targets
9088 that produce debugging output for DBX or COFF-style debugging output
9089 for SDB and allow the frame-pointer to be eliminated when the
9090 @option{-g} options is used.
9093 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9094 A C expression that returns the integer offset value for an argument
9095 having address @var{x} (an RTL expression). The nominal offset is
9099 @defmac PREFERRED_DEBUGGING_TYPE
9100 A C expression that returns the type of debugging output GCC should
9101 produce when the user specifies just @option{-g}. Define
9102 this if you have arranged for GCC to support more than one format of
9103 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9104 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
9105 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
9107 When the user specifies @option{-ggdb}, GCC normally also uses the
9108 value of this macro to select the debugging output format, but with two
9109 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9110 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9111 defined, GCC uses @code{DBX_DEBUG}.
9113 The value of this macro only affects the default debugging output; the
9114 user can always get a specific type of output by using @option{-gstabs},
9115 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9119 @subsection Specific Options for DBX Output
9121 @c prevent bad page break with this line
9122 These are specific options for DBX output.
9124 @defmac DBX_DEBUGGING_INFO
9125 Define this macro if GCC should produce debugging output for DBX
9126 in response to the @option{-g} option.
9129 @defmac XCOFF_DEBUGGING_INFO
9130 Define this macro if GCC should produce XCOFF format debugging output
9131 in response to the @option{-g} option. This is a variant of DBX format.
9134 @defmac DEFAULT_GDB_EXTENSIONS
9135 Define this macro to control whether GCC should by default generate
9136 GDB's extended version of DBX debugging information (assuming DBX-format
9137 debugging information is enabled at all). If you don't define the
9138 macro, the default is 1: always generate the extended information
9139 if there is any occasion to.
9142 @defmac DEBUG_SYMS_TEXT
9143 Define this macro if all @code{.stabs} commands should be output while
9144 in the text section.
9147 @defmac ASM_STABS_OP
9148 A C string constant, including spacing, naming the assembler pseudo op to
9149 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9150 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9151 applies only to DBX debugging information format.
9154 @defmac ASM_STABD_OP
9155 A C string constant, including spacing, naming the assembler pseudo op to
9156 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9157 value is the current location. If you don't define this macro,
9158 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9162 @defmac ASM_STABN_OP
9163 A C string constant, including spacing, naming the assembler pseudo op to
9164 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9165 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9166 macro applies only to DBX debugging information format.
9169 @defmac DBX_NO_XREFS
9170 Define this macro if DBX on your system does not support the construct
9171 @samp{xs@var{tagname}}. On some systems, this construct is used to
9172 describe a forward reference to a structure named @var{tagname}.
9173 On other systems, this construct is not supported at all.
9176 @defmac DBX_CONTIN_LENGTH
9177 A symbol name in DBX-format debugging information is normally
9178 continued (split into two separate @code{.stabs} directives) when it
9179 exceeds a certain length (by default, 80 characters). On some
9180 operating systems, DBX requires this splitting; on others, splitting
9181 must not be done. You can inhibit splitting by defining this macro
9182 with the value zero. You can override the default splitting-length by
9183 defining this macro as an expression for the length you desire.
9186 @defmac DBX_CONTIN_CHAR
9187 Normally continuation is indicated by adding a @samp{\} character to
9188 the end of a @code{.stabs} string when a continuation follows. To use
9189 a different character instead, define this macro as a character
9190 constant for the character you want to use. Do not define this macro
9191 if backslash is correct for your system.
9194 @defmac DBX_STATIC_STAB_DATA_SECTION
9195 Define this macro if it is necessary to go to the data section before
9196 outputting the @samp{.stabs} pseudo-op for a non-global static
9200 @defmac DBX_TYPE_DECL_STABS_CODE
9201 The value to use in the ``code'' field of the @code{.stabs} directive
9202 for a typedef. The default is @code{N_LSYM}.
9205 @defmac DBX_STATIC_CONST_VAR_CODE
9206 The value to use in the ``code'' field of the @code{.stabs} directive
9207 for a static variable located in the text section. DBX format does not
9208 provide any ``right'' way to do this. The default is @code{N_FUN}.
9211 @defmac DBX_REGPARM_STABS_CODE
9212 The value to use in the ``code'' field of the @code{.stabs} directive
9213 for a parameter passed in registers. DBX format does not provide any
9214 ``right'' way to do this. The default is @code{N_RSYM}.
9217 @defmac DBX_REGPARM_STABS_LETTER
9218 The letter to use in DBX symbol data to identify a symbol as a parameter
9219 passed in registers. DBX format does not customarily provide any way to
9220 do this. The default is @code{'P'}.
9223 @defmac DBX_FUNCTION_FIRST
9224 Define this macro if the DBX information for a function and its
9225 arguments should precede the assembler code for the function. Normally,
9226 in DBX format, the debugging information entirely follows the assembler
9230 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9231 Define this macro, with value 1, if the value of a symbol describing
9232 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9233 relative to the start of the enclosing function. Normally, GCC uses
9234 an absolute address.
9237 @defmac DBX_LINES_FUNCTION_RELATIVE
9238 Define this macro, with value 1, if the value of a symbol indicating
9239 the current line number (@code{N_SLINE}) should be relative to the
9240 start of the enclosing function. Normally, GCC uses an absolute address.
9243 @defmac DBX_USE_BINCL
9244 Define this macro if GCC should generate @code{N_BINCL} and
9245 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9246 macro also directs GCC to output a type number as a pair of a file
9247 number and a type number within the file. Normally, GCC does not
9248 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9249 number for a type number.
9253 @subsection Open-Ended Hooks for DBX Format
9255 @c prevent bad page break with this line
9256 These are hooks for DBX format.
9258 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
9259 Define this macro to say how to output to @var{stream} the debugging
9260 information for the start of a scope level for variable names. The
9261 argument @var{name} is the name of an assembler symbol (for use with
9262 @code{assemble_name}) whose value is the address where the scope begins.
9265 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
9266 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
9269 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
9270 Define this macro if the target machine requires special handling to
9271 output an @code{N_FUN} entry for the function @var{decl}.
9274 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9275 A C statement to output DBX debugging information before code for line
9276 number @var{line} of the current source file to the stdio stream
9277 @var{stream}. @var{counter} is the number of time the macro was
9278 invoked, including the current invocation; it is intended to generate
9279 unique labels in the assembly output.
9281 This macro should not be defined if the default output is correct, or
9282 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9285 @defmac NO_DBX_FUNCTION_END
9286 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9287 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9288 On those machines, define this macro to turn this feature off without
9289 disturbing the rest of the gdb extensions.
9292 @defmac NO_DBX_BNSYM_ENSYM
9293 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9294 extension construct. On those machines, define this macro to turn this
9295 feature off without disturbing the rest of the gdb extensions.
9298 @node File Names and DBX
9299 @subsection File Names in DBX Format
9301 @c prevent bad page break with this line
9302 This describes file names in DBX format.
9304 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9305 A C statement to output DBX debugging information to the stdio stream
9306 @var{stream}, which indicates that file @var{name} is the main source
9307 file---the file specified as the input file for compilation.
9308 This macro is called only once, at the beginning of compilation.
9310 This macro need not be defined if the standard form of output
9311 for DBX debugging information is appropriate.
9313 It may be necessary to refer to a label equal to the beginning of the
9314 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9315 to do so. If you do this, you must also set the variable
9316 @var{used_ltext_label_name} to @code{true}.
9319 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9320 Define this macro, with value 1, if GCC should not emit an indication
9321 of the current directory for compilation and current source language at
9322 the beginning of the file.
9325 @defmac NO_DBX_GCC_MARKER
9326 Define this macro, with value 1, if GCC should not emit an indication
9327 that this object file was compiled by GCC@. The default is to emit
9328 an @code{N_OPT} stab at the beginning of every source file, with
9329 @samp{gcc2_compiled.} for the string and value 0.
9332 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9333 A C statement to output DBX debugging information at the end of
9334 compilation of the main source file @var{name}. Output should be
9335 written to the stdio stream @var{stream}.
9337 If you don't define this macro, nothing special is output at the end
9338 of compilation, which is correct for most machines.
9341 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9342 Define this macro @emph{instead of} defining
9343 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9344 the end of compilation is an @code{N_SO} stab with an empty string,
9345 whose value is the highest absolute text address in the file.
9350 @subsection Macros for SDB and DWARF Output
9352 @c prevent bad page break with this line
9353 Here are macros for SDB and DWARF output.
9355 @defmac SDB_DEBUGGING_INFO
9356 Define this macro if GCC should produce COFF-style debugging output
9357 for SDB in response to the @option{-g} option.
9360 @defmac DWARF2_DEBUGGING_INFO
9361 Define this macro if GCC should produce dwarf version 2 format
9362 debugging output in response to the @option{-g} option.
9364 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (const_tree @var{function})
9365 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9366 be emitted for each function. Instead of an integer return the enum
9367 value for the @code{DW_CC_} tag.
9370 To support optional call frame debugging information, you must also
9371 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9372 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9373 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9374 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9377 @defmac DWARF2_FRAME_INFO
9378 Define this macro to a nonzero value if GCC should always output
9379 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
9380 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
9381 exceptions are enabled, GCC will output this information not matter
9382 how you define @code{DWARF2_FRAME_INFO}.
9385 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_DEBUG_UNWIND_INFO (void)
9386 This hook defines the mechanism that will be used for describing frame
9387 unwind information to the debugger. Normally the hook will return
9388 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
9389 return @code{UI_NONE} otherwise.
9391 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
9392 is disabled in order to always output DWARF 2 frame information.
9394 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
9395 This will suppress generation of the normal debug frame unwind information.
9398 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9399 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9400 line debug info sections. This will result in much more compact line number
9401 tables, and hence is desirable if it works.
9404 @deftypevr {Target Hook} bool TARGET_WANT_DEBUG_PUB_SECTIONS
9405 True if the @code{.debug_pubtypes} and @code{.debug_pubnames} sections should be emitted. These sections are not used on most platforms, and in particular GDB does not use them.
9408 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9409 A C statement to issue assembly directives that create a difference
9410 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9413 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9414 A C statement to issue assembly directives that create a difference
9415 between the two given labels in system defined units, e.g. instruction
9416 slots on IA64 VMS, using an integer of the given size.
9419 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
9420 A C statement to issue assembly directives that create a
9421 section-relative reference to the given @var{label}, using an integer of the
9422 given @var{size}. The label is known to be defined in the given @var{section}.
9425 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9426 A C statement to issue assembly directives that create a self-relative
9427 reference to the given @var{label}, using an integer of the given @var{size}.
9430 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9431 A C statement to issue assembly directives that create a reference to
9432 the DWARF table identifier @var{label} from the current section. This
9433 is used on some systems to avoid garbage collecting a DWARF table which
9434 is referenced by a function.
9437 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{file}, int @var{size}, rtx @var{x})
9438 If defined, this target hook is a function which outputs a DTP-relative
9439 reference to the given TLS symbol of the specified size.
9442 @defmac PUT_SDB_@dots{}
9443 Define these macros to override the assembler syntax for the special
9444 SDB assembler directives. See @file{sdbout.c} for a list of these
9445 macros and their arguments. If the standard syntax is used, you need
9446 not define them yourself.
9450 Some assemblers do not support a semicolon as a delimiter, even between
9451 SDB assembler directives. In that case, define this macro to be the
9452 delimiter to use (usually @samp{\n}). It is not necessary to define
9453 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9457 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9458 Define this macro to allow references to unknown structure,
9459 union, or enumeration tags to be emitted. Standard COFF does not
9460 allow handling of unknown references, MIPS ECOFF has support for
9464 @defmac SDB_ALLOW_FORWARD_REFERENCES
9465 Define this macro to allow references to structure, union, or
9466 enumeration tags that have not yet been seen to be handled. Some
9467 assemblers choke if forward tags are used, while some require it.
9470 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9471 A C statement to output SDB debugging information before code for line
9472 number @var{line} of the current source file to the stdio stream
9473 @var{stream}. The default is to emit an @code{.ln} directive.
9478 @subsection Macros for VMS Debug Format
9480 @c prevent bad page break with this line
9481 Here are macros for VMS debug format.
9483 @defmac VMS_DEBUGGING_INFO
9484 Define this macro if GCC should produce debugging output for VMS
9485 in response to the @option{-g} option. The default behavior for VMS
9486 is to generate minimal debug info for a traceback in the absence of
9487 @option{-g} unless explicitly overridden with @option{-g0}. This
9488 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
9489 @code{TARGET_OPTION_OVERRIDE}.
9492 @node Floating Point
9493 @section Cross Compilation and Floating Point
9494 @cindex cross compilation and floating point
9495 @cindex floating point and cross compilation
9497 While all modern machines use twos-complement representation for integers,
9498 there are a variety of representations for floating point numbers. This
9499 means that in a cross-compiler the representation of floating point numbers
9500 in the compiled program may be different from that used in the machine
9501 doing the compilation.
9503 Because different representation systems may offer different amounts of
9504 range and precision, all floating point constants must be represented in
9505 the target machine's format. Therefore, the cross compiler cannot
9506 safely use the host machine's floating point arithmetic; it must emulate
9507 the target's arithmetic. To ensure consistency, GCC always uses
9508 emulation to work with floating point values, even when the host and
9509 target floating point formats are identical.
9511 The following macros are provided by @file{real.h} for the compiler to
9512 use. All parts of the compiler which generate or optimize
9513 floating-point calculations must use these macros. They may evaluate
9514 their operands more than once, so operands must not have side effects.
9516 @defmac REAL_VALUE_TYPE
9517 The C data type to be used to hold a floating point value in the target
9518 machine's format. Typically this is a @code{struct} containing an
9519 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9523 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9524 Compares for equality the two values, @var{x} and @var{y}. If the target
9525 floating point format supports negative zeroes and/or NaNs,
9526 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
9527 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
9530 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9531 Tests whether @var{x} is less than @var{y}.
9534 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9535 Truncates @var{x} to a signed integer, rounding toward zero.
9538 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9539 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9540 @var{x} is negative, returns zero.
9543 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
9544 Converts @var{string} into a floating point number in the target machine's
9545 representation for mode @var{mode}. This routine can handle both
9546 decimal and hexadecimal floating point constants, using the syntax
9547 defined by the C language for both.
9550 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9551 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9554 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9555 Determines whether @var{x} represents infinity (positive or negative).
9558 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9559 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9562 @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})
9563 Calculates an arithmetic operation on the two floating point values
9564 @var{x} and @var{y}, storing the result in @var{output} (which must be a
9567 The operation to be performed is specified by @var{code}. Only the
9568 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9569 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9571 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9572 target's floating point format cannot represent infinity, it will call
9573 @code{abort}. Callers should check for this situation first, using
9574 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9577 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9578 Returns the negative of the floating point value @var{x}.
9581 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9582 Returns the absolute value of @var{x}.
9585 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
9586 Truncates the floating point value @var{x} to fit in @var{mode}. The
9587 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
9588 appropriate bit pattern to be output as a floating constant whose
9589 precision accords with mode @var{mode}.
9592 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9593 Converts a floating point value @var{x} into a double-precision integer
9594 which is then stored into @var{low} and @var{high}. If the value is not
9595 integral, it is truncated.
9598 @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})
9599 Converts a double-precision integer found in @var{low} and @var{high},
9600 into a floating point value which is then stored into @var{x}. The
9601 value is truncated to fit in mode @var{mode}.
9604 @node Mode Switching
9605 @section Mode Switching Instructions
9606 @cindex mode switching
9607 The following macros control mode switching optimizations:
9609 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9610 Define this macro if the port needs extra instructions inserted for mode
9611 switching in an optimizing compilation.
9613 For an example, the SH4 can perform both single and double precision
9614 floating point operations, but to perform a single precision operation,
9615 the FPSCR PR bit has to be cleared, while for a double precision
9616 operation, this bit has to be set. Changing the PR bit requires a general
9617 purpose register as a scratch register, hence these FPSCR sets have to
9618 be inserted before reload, i.e.@: you can't put this into instruction emitting
9619 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9621 You can have multiple entities that are mode-switched, and select at run time
9622 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9623 return nonzero for any @var{entity} that needs mode-switching.
9624 If you define this macro, you also have to define
9625 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9626 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9627 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9631 @defmac NUM_MODES_FOR_MODE_SWITCHING
9632 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9633 initializer for an array of integers. Each initializer element
9634 N refers to an entity that needs mode switching, and specifies the number
9635 of different modes that might need to be set for this entity.
9636 The position of the initializer in the initializer---starting counting at
9637 zero---determines the integer that is used to refer to the mode-switched
9639 In macros that take mode arguments / yield a mode result, modes are
9640 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9641 switch is needed / supplied.
9644 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9645 @var{entity} is an integer specifying a mode-switched entity. If
9646 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9647 return an integer value not larger than the corresponding element in
9648 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9649 be switched into prior to the execution of @var{insn}.
9652 @defmac MODE_AFTER (@var{mode}, @var{insn})
9653 If this macro is defined, it is evaluated for every @var{insn} during
9654 mode switching. It determines the mode that an insn results in (if
9655 different from the incoming mode).
9658 @defmac MODE_ENTRY (@var{entity})
9659 If this macro is defined, it is evaluated for every @var{entity} that needs
9660 mode switching. It should evaluate to an integer, which is a mode that
9661 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9662 is defined then @code{MODE_EXIT} must be defined.
9665 @defmac MODE_EXIT (@var{entity})
9666 If this macro is defined, it is evaluated for every @var{entity} that needs
9667 mode switching. It should evaluate to an integer, which is a mode that
9668 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9669 is defined then @code{MODE_ENTRY} must be defined.
9672 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9673 This macro specifies the order in which modes for @var{entity} are processed.
9674 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9675 lowest. The value of the macro should be an integer designating a mode
9676 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9677 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9678 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9681 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9682 Generate one or more insns to set @var{entity} to @var{mode}.
9683 @var{hard_reg_live} is the set of hard registers live at the point where
9684 the insn(s) are to be inserted.
9687 @node Target Attributes
9688 @section Defining target-specific uses of @code{__attribute__}
9689 @cindex target attributes
9690 @cindex machine attributes
9691 @cindex attributes, target-specific
9693 Target-specific attributes may be defined for functions, data and types.
9694 These are described using the following target hooks; they also need to
9695 be documented in @file{extend.texi}.
9697 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9698 If defined, this target hook points to an array of @samp{struct
9699 attribute_spec} (defined in @file{tree.h}) specifying the machine
9700 specific attributes for this target and some of the restrictions on the
9701 entities to which these attributes are applied and the arguments they
9705 @deftypefn {Target Hook} bool TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P (const_tree @var{name})
9706 If defined, this target hook is a function which returns true if the
9707 machine-specific attribute named @var{name} expects an identifier
9708 given as its first argument to be passed on as a plain identifier, not
9709 subjected to name lookup. If this is not defined, the default is
9710 false for all machine-specific attributes.
9713 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (const_tree @var{type1}, const_tree @var{type2})
9714 If defined, this target hook is a function which returns zero if the attributes on
9715 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9716 and two if they are nearly compatible (which causes a warning to be
9717 generated). If this is not defined, machine-specific attributes are
9718 supposed always to be compatible.
9721 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9722 If defined, this target hook is a function which assigns default attributes to
9723 the newly defined @var{type}.
9726 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9727 Define this target hook if the merging of type attributes needs special
9728 handling. If defined, the result is a list of the combined
9729 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9730 that @code{comptypes} has already been called and returned 1. This
9731 function may call @code{merge_attributes} to handle machine-independent
9735 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9736 Define this target hook if the merging of decl attributes needs special
9737 handling. If defined, the result is a list of the combined
9738 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9739 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9740 when this is needed are when one attribute overrides another, or when an
9741 attribute is nullified by a subsequent definition. This function may
9742 call @code{merge_attributes} to handle machine-independent merging.
9744 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9745 If the only target-specific handling you require is @samp{dllimport}
9746 for Microsoft Windows targets, you should define the macro
9747 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9748 will then define a function called
9749 @code{merge_dllimport_decl_attributes} which can then be defined as
9750 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9751 add @code{handle_dll_attribute} in the attribute table for your port
9752 to perform initial processing of the @samp{dllimport} and
9753 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9754 @file{i386/i386.c}, for example.
9757 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (const_tree @var{decl})
9758 @var{decl} is a variable or function with @code{__attribute__((dllimport))} specified. Use this hook if the target needs to add extra validation checks to @code{handle_dll_attribute}.
9761 @defmac TARGET_DECLSPEC
9762 Define this macro to a nonzero value if you want to treat
9763 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9764 default, this behavior is enabled only for targets that define
9765 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9766 of @code{__declspec} is via a built-in macro, but you should not rely
9767 on this implementation detail.
9770 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9771 Define this target hook if you want to be able to add attributes to a decl
9772 when it is being created. This is normally useful for back ends which
9773 wish to implement a pragma by using the attributes which correspond to
9774 the pragma's effect. The @var{node} argument is the decl which is being
9775 created. The @var{attr_ptr} argument is a pointer to the attribute list
9776 for this decl. The list itself should not be modified, since it may be
9777 shared with other decls, but attributes may be chained on the head of
9778 the list and @code{*@var{attr_ptr}} modified to point to the new
9779 attributes, or a copy of the list may be made if further changes are
9783 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (const_tree @var{fndecl})
9785 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9786 into the current function, despite its having target-specific
9787 attributes, @code{false} otherwise. By default, if a function has a
9788 target specific attribute attached to it, it will not be inlined.
9791 @deftypefn {Target Hook} bool TARGET_OPTION_VALID_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
9792 This hook is called to parse the @code{attribute(option("..."))}, and
9793 it allows the function to set different target machine compile time
9794 options for the current function that might be different than the
9795 options specified on the command line. The hook should return
9796 @code{true} if the options are valid.
9798 The hook should set the @var{DECL_FUNCTION_SPECIFIC_TARGET} field in
9799 the function declaration to hold a pointer to a target specific
9800 @var{struct cl_target_option} structure.
9803 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr})
9804 This hook is called to save any additional target specific information
9805 in the @var{struct cl_target_option} structure for function specific
9807 @xref{Option file format}.
9810 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct cl_target_option *@var{ptr})
9811 This hook is called to restore any additional target specific
9812 information in the @var{struct cl_target_option} structure for
9813 function specific options.
9816 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (FILE *@var{file}, int @var{indent}, struct cl_target_option *@var{ptr})
9817 This hook is called to print any additional target specific
9818 information in the @var{struct cl_target_option} structure for
9819 function specific options.
9822 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (tree @var{args}, tree @var{pop_target})
9823 This target hook parses the options for @code{#pragma GCC option} to
9824 set the machine specific options for functions that occur later in the
9825 input stream. The options should be the same as handled by the
9826 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
9829 @deftypefn {Target Hook} void TARGET_OPTION_OVERRIDE (void)
9830 Sometimes certain combinations of command options do not make sense on
9831 a particular target machine. You can override the hook
9832 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
9833 once just after all the command options have been parsed.
9835 Don't use this hook to turn on various extra optimizations for
9836 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
9838 If you need to do something whenever the optimization level is
9839 changed via the optimize attribute or pragma, see
9840 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
9843 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
9844 This target hook returns @code{false} if the @var{caller} function
9845 cannot inline @var{callee}, based on target specific information. By
9846 default, inlining is not allowed if the callee function has function
9847 specific target options and the caller does not use the same options.
9851 @section Emulating TLS
9852 @cindex Emulated TLS
9854 For targets whose psABI does not provide Thread Local Storage via
9855 specific relocations and instruction sequences, an emulation layer is
9856 used. A set of target hooks allows this emulation layer to be
9857 configured for the requirements of a particular target. For instance
9858 the psABI may in fact specify TLS support in terms of an emulation
9861 The emulation layer works by creating a control object for every TLS
9862 object. To access the TLS object, a lookup function is provided
9863 which, when given the address of the control object, will return the
9864 address of the current thread's instance of the TLS object.
9866 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
9867 Contains the name of the helper function that uses a TLS control
9868 object to locate a TLS instance. The default causes libgcc's
9869 emulated TLS helper function to be used.
9872 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
9873 Contains the name of the helper function that should be used at
9874 program startup to register TLS objects that are implicitly
9875 initialized to zero. If this is @code{NULL}, all TLS objects will
9876 have explicit initializers. The default causes libgcc's emulated TLS
9877 registration function to be used.
9880 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
9881 Contains the name of the section in which TLS control variables should
9882 be placed. The default of @code{NULL} allows these to be placed in
9886 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
9887 Contains the name of the section in which TLS initializers should be
9888 placed. The default of @code{NULL} allows these to be placed in any
9892 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
9893 Contains the prefix to be prepended to TLS control variable names.
9894 The default of @code{NULL} uses a target-specific prefix.
9897 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
9898 Contains the prefix to be prepended to TLS initializer objects. The
9899 default of @code{NULL} uses a target-specific prefix.
9902 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
9903 Specifies a function that generates the FIELD_DECLs for a TLS control
9904 object type. @var{type} is the RECORD_TYPE the fields are for and
9905 @var{name} should be filled with the structure tag, if the default of
9906 @code{__emutls_object} is unsuitable. The default creates a type suitable
9907 for libgcc's emulated TLS function.
9910 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
9911 Specifies a function that generates the CONSTRUCTOR to initialize a
9912 TLS control object. @var{var} is the TLS control object, @var{decl}
9913 is the TLS object and @var{tmpl_addr} is the address of the
9914 initializer. The default initializes libgcc's emulated TLS control object.
9917 @deftypevr {Target Hook} bool TARGET_EMUTLS_VAR_ALIGN_FIXED
9918 Specifies whether the alignment of TLS control variable objects is
9919 fixed and should not be increased as some backends may do to optimize
9920 single objects. The default is false.
9923 @deftypevr {Target Hook} bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
9924 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
9925 may be used to describe emulated TLS control objects.
9928 @node MIPS Coprocessors
9929 @section Defining coprocessor specifics for MIPS targets.
9930 @cindex MIPS coprocessor-definition macros
9932 The MIPS specification allows MIPS implementations to have as many as 4
9933 coprocessors, each with as many as 32 private registers. GCC supports
9934 accessing these registers and transferring values between the registers
9935 and memory using asm-ized variables. For example:
9938 register unsigned int cp0count asm ("c0r1");
9944 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9945 names may be added as described below, or the default names may be
9946 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9948 Coprocessor registers are assumed to be epilogue-used; sets to them will
9949 be preserved even if it does not appear that the register is used again
9950 later in the function.
9952 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9953 the FPU@. One accesses COP1 registers through standard mips
9954 floating-point support; they are not included in this mechanism.
9956 There is one macro used in defining the MIPS coprocessor interface which
9957 you may want to override in subtargets; it is described below.
9959 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
9960 A comma-separated list (with leading comma) of pairs describing the
9961 alternate names of coprocessor registers. The format of each entry should be
9963 @{ @var{alternatename}, @var{register_number}@}
9969 @section Parameters for Precompiled Header Validity Checking
9970 @cindex parameters, precompiled headers
9972 @deftypefn {Target Hook} {void *} TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
9973 This hook returns a pointer to the data needed by
9974 @code{TARGET_PCH_VALID_P} and sets
9975 @samp{*@var{sz}} to the size of the data in bytes.
9978 @deftypefn {Target Hook} {const char *} TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
9979 This hook checks whether the options used to create a PCH file are
9980 compatible with the current settings. It returns @code{NULL}
9981 if so and a suitable error message if not. Error messages will
9982 be presented to the user and must be localized using @samp{_(@var{msg})}.
9984 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
9985 when the PCH file was created and @var{sz} is the size of that data in bytes.
9986 It's safe to assume that the data was created by the same version of the
9987 compiler, so no format checking is needed.
9989 The default definition of @code{default_pch_valid_p} should be
9990 suitable for most targets.
9993 @deftypefn {Target Hook} {const char *} TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
9994 If this hook is nonnull, the default implementation of
9995 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
9996 of @code{target_flags}. @var{pch_flags} specifies the value that
9997 @code{target_flags} had when the PCH file was created. The return
9998 value is the same as for @code{TARGET_PCH_VALID_P}.
10002 @section C++ ABI parameters
10003 @cindex parameters, c++ abi
10005 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
10006 Define this hook to override the integer type used for guard variables.
10007 These are used to implement one-time construction of static objects. The
10008 default is long_long_integer_type_node.
10011 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
10012 This hook determines how guard variables are used. It should return
10013 @code{false} (the default) if the first byte should be used. A return value of
10014 @code{true} indicates that only the least significant bit should be used.
10017 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
10018 This hook returns the size of the cookie to use when allocating an array
10019 whose elements have the indicated @var{type}. Assumes that it is already
10020 known that a cookie is needed. The default is
10021 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
10022 IA64/Generic C++ ABI@.
10025 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
10026 This hook should return @code{true} if the element size should be stored in
10027 array cookies. The default is to return @code{false}.
10030 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
10031 If defined by a backend this hook allows the decision made to export
10032 class @var{type} to be overruled. Upon entry @var{import_export}
10033 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10034 to be imported and 0 otherwise. This function should return the
10035 modified value and perform any other actions necessary to support the
10036 backend's targeted operating system.
10039 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
10040 This hook should return @code{true} if constructors and destructors return
10041 the address of the object created/destroyed. The default is to return
10045 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
10046 This hook returns true if the key method for a class (i.e., the method
10047 which, if defined in the current translation unit, causes the virtual
10048 table to be emitted) may be an inline function. Under the standard
10049 Itanium C++ ABI the key method may be an inline function so long as
10050 the function is not declared inline in the class definition. Under
10051 some variants of the ABI, an inline function can never be the key
10052 method. The default is to return @code{true}.
10055 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
10056 @var{decl} is a virtual table, virtual table table, typeinfo object, or other similar implicit class data object that will be emitted with external linkage in this translation unit. No ELF visibility has been explicitly specified. If the target needs to specify a visibility other than that of the containing class, use this hook to set @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
10059 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
10060 This hook returns true (the default) if virtual tables and other
10061 similar implicit class data objects are always COMDAT if they have
10062 external linkage. If this hook returns false, then class data for
10063 classes whose virtual table will be emitted in only one translation
10064 unit will not be COMDAT.
10067 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
10068 This hook returns true (the default) if the RTTI information for
10069 the basic types which is defined in the C++ runtime should always
10070 be COMDAT, false if it should not be COMDAT.
10073 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
10074 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10075 should be used to register static destructors when @option{-fuse-cxa-atexit}
10076 is in effect. The default is to return false to use @code{__cxa_atexit}.
10079 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
10080 This hook returns true if the target @code{atexit} function can be used
10081 in the same manner as @code{__cxa_atexit} to register C++ static
10082 destructors. This requires that @code{atexit}-registered functions in
10083 shared libraries are run in the correct order when the libraries are
10084 unloaded. The default is to return false.
10087 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
10088 @var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been defined. Use this hook to make adjustments to the class (eg, tweak visibility or perform any other required target modifications).
10091 @node Named Address Spaces
10092 @section Adding support for named address spaces
10093 @cindex named address spaces
10095 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10096 standards committee, @cite{Programming Languages - C - Extensions to
10097 support embedded processors}, specifies a syntax for embedded
10098 processors to specify alternate address spaces. You can configure a
10099 GCC port to support section 5.1 of the draft report to add support for
10100 address spaces other than the default address space. These address
10101 spaces are new keywords that are similar to the @code{volatile} and
10102 @code{const} type attributes.
10104 Pointers to named address spaces can have a different size than
10105 pointers to the generic address space.
10107 For example, the SPU port uses the @code{__ea} address space to refer
10108 to memory in the host processor, rather than memory local to the SPU
10109 processor. Access to memory in the @code{__ea} address space involves
10110 issuing DMA operations to move data between the host processor and the
10111 local processor memory address space. Pointers in the @code{__ea}
10112 address space are either 32 bits or 64 bits based on the
10113 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10116 Internally, address spaces are represented as a small integer in the
10117 range 0 to 15 with address space 0 being reserved for the generic
10120 To register a named address space qualifier keyword with the C front end,
10121 the target may call the @code{c_register_addr_space} routine. For example,
10122 the SPU port uses the following to declare @code{__ea} as the keyword for
10123 named address space #1:
10125 #define ADDR_SPACE_EA 1
10126 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10129 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_POINTER_MODE (addr_space_t @var{address_space})
10130 Define this to return the machine mode to use for pointers to
10131 @var{address_space} if the target supports named address spaces.
10132 The default version of this hook returns @code{ptr_mode} for the
10133 generic address space only.
10136 @deftypefn {Target Hook} {enum machine_mode} TARGET_ADDR_SPACE_ADDRESS_MODE (addr_space_t @var{address_space})
10137 Define this to return the machine mode to use for addresses in
10138 @var{address_space} if the target supports named address spaces.
10139 The default version of this hook returns @code{Pmode} for the
10140 generic address space only.
10143 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_VALID_POINTER_MODE (enum machine_mode @var{mode}, addr_space_t @var{as})
10144 Define this to return nonzero if the port can handle pointers
10145 with machine mode @var{mode} to address space @var{as}. This target
10146 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10147 except that it includes explicit named address space support. The default
10148 version of this hook returns true for the modes returned by either the
10149 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10150 target hooks for the given address space.
10153 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P (enum machine_mode @var{mode}, rtx @var{exp}, bool @var{strict}, addr_space_t @var{as})
10154 Define this to return true if @var{exp} is a valid address for mode
10155 @var{mode} in the named address space @var{as}. The @var{strict}
10156 parameter says whether strict addressing is in effect after reload has
10157 finished. This target hook is the same as the
10158 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10159 explicit named address space support.
10162 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, enum machine_mode @var{mode}, addr_space_t @var{as})
10163 Define this to modify an invalid address @var{x} to be a valid address
10164 with mode @var{mode} in the named address space @var{as}. This target
10165 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10166 except that it includes explicit named address space support.
10169 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_SUBSET_P (addr_space_t @var{superset}, addr_space_t @var{subset})
10170 Define this to return whether the @var{subset} named address space is
10171 contained within the @var{superset} named address space. Pointers to
10172 a named address space that is a subset of another named address space
10173 will be converted automatically without a cast if used together in
10174 arithmetic operations. Pointers to a superset address space can be
10175 converted to pointers to a subset address space via explicit casts.
10178 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_CONVERT (rtx @var{op}, tree @var{from_type}, tree @var{to_type})
10179 Define this to convert the pointer expression represented by the RTL
10180 @var{op} with type @var{from_type} that points to a named address
10181 space to a new pointer expression with type @var{to_type} that points
10182 to a different named address space. When this hook it called, it is
10183 guaranteed that one of the two address spaces is a subset of the other,
10184 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10188 @section Miscellaneous Parameters
10189 @cindex parameters, miscellaneous
10191 @c prevent bad page break with this line
10192 Here are several miscellaneous parameters.
10194 @defmac HAS_LONG_COND_BRANCH
10195 Define this boolean macro to indicate whether or not your architecture
10196 has conditional branches that can span all of memory. It is used in
10197 conjunction with an optimization that partitions hot and cold basic
10198 blocks into separate sections of the executable. If this macro is
10199 set to false, gcc will convert any conditional branches that attempt
10200 to cross between sections into unconditional branches or indirect jumps.
10203 @defmac HAS_LONG_UNCOND_BRANCH
10204 Define this boolean macro to indicate whether or not your architecture
10205 has unconditional branches that can span all of memory. It is used in
10206 conjunction with an optimization that partitions hot and cold basic
10207 blocks into separate sections of the executable. If this macro is
10208 set to false, gcc will convert any unconditional branches that attempt
10209 to cross between sections into indirect jumps.
10212 @defmac CASE_VECTOR_MODE
10213 An alias for a machine mode name. This is the machine mode that
10214 elements of a jump-table should have.
10217 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10218 Optional: return the preferred mode for an @code{addr_diff_vec}
10219 when the minimum and maximum offset are known. If you define this,
10220 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10221 To make this work, you also have to define @code{INSN_ALIGN} and
10222 make the alignment for @code{addr_diff_vec} explicit.
10223 The @var{body} argument is provided so that the offset_unsigned and scale
10224 flags can be updated.
10227 @defmac CASE_VECTOR_PC_RELATIVE
10228 Define this macro to be a C expression to indicate when jump-tables
10229 should contain relative addresses. You need not define this macro if
10230 jump-tables never contain relative addresses, or jump-tables should
10231 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10235 @deftypefn {Target Hook} {unsigned int} TARGET_CASE_VALUES_THRESHOLD (void)
10236 This function return the smallest number of different values for which it
10237 is best to use a jump-table instead of a tree of conditional branches.
10238 The default is four for machines with a @code{casesi} instruction and
10239 five otherwise. This is best for most machines.
10242 @defmac CASE_USE_BIT_TESTS
10243 Define this macro to be a C expression to indicate whether C switch
10244 statements may be implemented by a sequence of bit tests. This is
10245 advantageous on processors that can efficiently implement left shift
10246 of 1 by the number of bits held in a register, but inappropriate on
10247 targets that would require a loop. By default, this macro returns
10248 @code{true} if the target defines an @code{ashlsi3} pattern, and
10249 @code{false} otherwise.
10252 @defmac WORD_REGISTER_OPERATIONS
10253 Define this macro if operations between registers with integral mode
10254 smaller than a word are always performed on the entire register.
10255 Most RISC machines have this property and most CISC machines do not.
10258 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10259 Define this macro to be a C expression indicating when insns that read
10260 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10261 bits outside of @var{mem_mode} to be either the sign-extension or the
10262 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10263 of @var{mem_mode} for which the
10264 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10265 @code{UNKNOWN} for other modes.
10267 This macro is not called with @var{mem_mode} non-integral or with a width
10268 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10269 value in this case. Do not define this macro if it would always return
10270 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10271 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10273 You may return a non-@code{UNKNOWN} value even if for some hard registers
10274 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10275 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
10276 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10277 integral mode larger than this but not larger than @code{word_mode}.
10279 You must return @code{UNKNOWN} if for some hard registers that allow this
10280 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
10281 @code{word_mode}, but that they can change to another integral mode that
10282 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10285 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10286 Define this macro if loading short immediate values into registers sign
10290 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
10291 Define this macro if the same instructions that convert a floating
10292 point number to a signed fixed point number also convert validly to an
10296 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
10297 When @option{-ffast-math} is in effect, GCC tries to optimize
10298 divisions by the same divisor, by turning them into multiplications by
10299 the reciprocal. This target hook specifies the minimum number of divisions
10300 that should be there for GCC to perform the optimization for a variable
10301 of mode @var{mode}. The default implementation returns 3 if the machine
10302 has an instruction for the division, and 2 if it does not.
10306 The maximum number of bytes that a single instruction can move quickly
10307 between memory and registers or between two memory locations.
10310 @defmac MAX_MOVE_MAX
10311 The maximum number of bytes that a single instruction can move quickly
10312 between memory and registers or between two memory locations. If this
10313 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10314 constant value that is the largest value that @code{MOVE_MAX} can have
10318 @defmac SHIFT_COUNT_TRUNCATED
10319 A C expression that is nonzero if on this machine the number of bits
10320 actually used for the count of a shift operation is equal to the number
10321 of bits needed to represent the size of the object being shifted. When
10322 this macro is nonzero, the compiler will assume that it is safe to omit
10323 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10324 truncates the count of a shift operation. On machines that have
10325 instructions that act on bit-fields at variable positions, which may
10326 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
10327 also enables deletion of truncations of the values that serve as
10328 arguments to bit-field instructions.
10330 If both types of instructions truncate the count (for shifts) and
10331 position (for bit-field operations), or if no variable-position bit-field
10332 instructions exist, you should define this macro.
10334 However, on some machines, such as the 80386 and the 680x0, truncation
10335 only applies to shift operations and not the (real or pretended)
10336 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
10337 such machines. Instead, add patterns to the @file{md} file that include
10338 the implied truncation of the shift instructions.
10340 You need not define this macro if it would always have the value of zero.
10343 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
10344 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
10345 This function describes how the standard shift patterns for @var{mode}
10346 deal with shifts by negative amounts or by more than the width of the mode.
10347 @xref{shift patterns}.
10349 On many machines, the shift patterns will apply a mask @var{m} to the
10350 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
10351 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
10352 this is true for mode @var{mode}, the function should return @var{m},
10353 otherwise it should return 0. A return value of 0 indicates that no
10354 particular behavior is guaranteed.
10356 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
10357 @emph{not} apply to general shift rtxes; it applies only to instructions
10358 that are generated by the named shift patterns.
10360 The default implementation of this function returns
10361 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
10362 and 0 otherwise. This definition is always safe, but if
10363 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
10364 nevertheless truncate the shift count, you may get better code
10368 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
10369 A C expression which is nonzero if on this machine it is safe to
10370 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
10371 bits (where @var{outprec} is smaller than @var{inprec}) by merely
10372 operating on it as if it had only @var{outprec} bits.
10374 On many machines, this expression can be 1.
10376 @c rearranged this, removed the phrase "it is reported that". this was
10377 @c to fix an overfull hbox. --mew 10feb93
10378 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
10379 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
10380 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
10381 such cases may improve things.
10384 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
10385 The representation of an integral mode can be such that the values
10386 are always extended to a wider integral mode. Return
10387 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10388 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10389 otherwise. (Currently, none of the targets use zero-extended
10390 representation this way so unlike @code{LOAD_EXTEND_OP},
10391 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10392 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10393 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
10394 widest integral mode and currently we take advantage of this fact.)
10396 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10397 value even if the extension is not performed on certain hard registers
10398 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10399 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
10401 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10402 describe two related properties. If you define
10403 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10404 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10407 In order to enforce the representation of @code{mode},
10408 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
10412 @defmac STORE_FLAG_VALUE
10413 A C expression describing the value returned by a comparison operator
10414 with an integral mode and stored by a store-flag instruction
10415 (@samp{cstore@var{mode}4}) when the condition is true. This description must
10416 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
10417 comparison operators whose results have a @code{MODE_INT} mode.
10419 A value of 1 or @minus{}1 means that the instruction implementing the
10420 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10421 and 0 when the comparison is false. Otherwise, the value indicates
10422 which bits of the result are guaranteed to be 1 when the comparison is
10423 true. This value is interpreted in the mode of the comparison
10424 operation, which is given by the mode of the first operand in the
10425 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
10426 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10429 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10430 generate code that depends only on the specified bits. It can also
10431 replace comparison operators with equivalent operations if they cause
10432 the required bits to be set, even if the remaining bits are undefined.
10433 For example, on a machine whose comparison operators return an
10434 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10435 @samp{0x80000000}, saying that just the sign bit is relevant, the
10439 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10443 can be converted to
10446 (ashift:SI @var{x} (const_int @var{n}))
10450 where @var{n} is the appropriate shift count to move the bit being
10451 tested into the sign bit.
10453 There is no way to describe a machine that always sets the low-order bit
10454 for a true value, but does not guarantee the value of any other bits,
10455 but we do not know of any machine that has such an instruction. If you
10456 are trying to port GCC to such a machine, include an instruction to
10457 perform a logical-and of the result with 1 in the pattern for the
10458 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10460 Often, a machine will have multiple instructions that obtain a value
10461 from a comparison (or the condition codes). Here are rules to guide the
10462 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10467 Use the shortest sequence that yields a valid definition for
10468 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10469 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10470 comparison operators to do so because there may be opportunities to
10471 combine the normalization with other operations.
10474 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10475 slightly preferred on machines with expensive jumps and 1 preferred on
10479 As a second choice, choose a value of @samp{0x80000001} if instructions
10480 exist that set both the sign and low-order bits but do not define the
10484 Otherwise, use a value of @samp{0x80000000}.
10487 Many machines can produce both the value chosen for
10488 @code{STORE_FLAG_VALUE} and its negation in the same number of
10489 instructions. On those machines, you should also define a pattern for
10490 those cases, e.g., one matching
10493 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10496 Some machines can also perform @code{and} or @code{plus} operations on
10497 condition code values with less instructions than the corresponding
10498 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
10499 machines, define the appropriate patterns. Use the names @code{incscc}
10500 and @code{decscc}, respectively, for the patterns which perform
10501 @code{plus} or @code{minus} operations on condition code values. See
10502 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
10503 find such instruction sequences on other machines.
10505 If this macro is not defined, the default value, 1, is used. You need
10506 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10507 instructions, or if the value generated by these instructions is 1.
10510 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10511 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10512 returned when comparison operators with floating-point results are true.
10513 Define this macro on machines that have comparison operations that return
10514 floating-point values. If there are no such operations, do not define
10518 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10519 A C expression that gives a rtx representing the nonzero true element
10520 for vector comparisons. The returned rtx should be valid for the inner
10521 mode of @var{mode} which is guaranteed to be a vector mode. Define
10522 this macro on machines that have vector comparison operations that
10523 return a vector result. If there are no such operations, do not define
10524 this macro. Typically, this macro is defined as @code{const1_rtx} or
10525 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10526 the compiler optimizing such vector comparison operations for the
10530 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10531 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10532 A C expression that indicates whether the architecture defines a value
10533 for @code{clz} or @code{ctz} with a zero operand.
10534 A result of @code{0} indicates the value is undefined.
10535 If the value is defined for only the RTL expression, the macro should
10536 evaluate to @code{1}; if the value applies also to the corresponding optab
10537 entry (which is normally the case if it expands directly into
10538 the corresponding RTL), then the macro should evaluate to @code{2}.
10539 In the cases where the value is defined, @var{value} should be set to
10542 If this macro is not defined, the value of @code{clz} or
10543 @code{ctz} at zero is assumed to be undefined.
10545 This macro must be defined if the target's expansion for @code{ffs}
10546 relies on a particular value to get correct results. Otherwise it
10547 is not necessary, though it may be used to optimize some corner cases, and
10548 to provide a default expansion for the @code{ffs} optab.
10550 Note that regardless of this macro the ``definedness'' of @code{clz}
10551 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10552 visible to the user. Thus one may be free to adjust the value at will
10553 to match the target expansion of these operations without fear of
10558 An alias for the machine mode for pointers. On most machines, define
10559 this to be the integer mode corresponding to the width of a hardware
10560 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10561 On some machines you must define this to be one of the partial integer
10562 modes, such as @code{PSImode}.
10564 The width of @code{Pmode} must be at least as large as the value of
10565 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10566 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10570 @defmac FUNCTION_MODE
10571 An alias for the machine mode used for memory references to functions
10572 being called, in @code{call} RTL expressions. On most CISC machines,
10573 where an instruction can begin at any byte address, this should be
10574 @code{QImode}. On most RISC machines, where all instructions have fixed
10575 size and alignment, this should be a mode with the same size and alignment
10576 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10579 @defmac STDC_0_IN_SYSTEM_HEADERS
10580 In normal operation, the preprocessor expands @code{__STDC__} to the
10581 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10582 hosts, like Solaris, the system compiler uses a different convention,
10583 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10584 strict conformance to the C Standard.
10586 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
10587 convention when processing system header files, but when processing user
10588 files @code{__STDC__} will always expand to 1.
10591 @defmac NO_IMPLICIT_EXTERN_C
10592 Define this macro if the system header files support C++ as well as C@.
10593 This macro inhibits the usual method of using system header files in
10594 C++, which is to pretend that the file's contents are enclosed in
10595 @samp{extern "C" @{@dots{}@}}.
10600 @defmac REGISTER_TARGET_PRAGMAS ()
10601 Define this macro if you want to implement any target-specific pragmas.
10602 If defined, it is a C expression which makes a series of calls to
10603 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
10604 for each pragma. The macro may also do any
10605 setup required for the pragmas.
10607 The primary reason to define this macro is to provide compatibility with
10608 other compilers for the same target. In general, we discourage
10609 definition of target-specific pragmas for GCC@.
10611 If the pragma can be implemented by attributes then you should consider
10612 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
10614 Preprocessor macros that appear on pragma lines are not expanded. All
10615 @samp{#pragma} directives that do not match any registered pragma are
10616 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
10619 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10620 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10622 Each call to @code{c_register_pragma} or
10623 @code{c_register_pragma_with_expansion} establishes one pragma. The
10624 @var{callback} routine will be called when the preprocessor encounters a
10628 #pragma [@var{space}] @var{name} @dots{}
10631 @var{space} is the case-sensitive namespace of the pragma, or
10632 @code{NULL} to put the pragma in the global namespace. The callback
10633 routine receives @var{pfile} as its first argument, which can be passed
10634 on to cpplib's functions if necessary. You can lex tokens after the
10635 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
10636 callback will be silently ignored. The end of the line is indicated by
10637 a token of type @code{CPP_EOF}. Macro expansion occurs on the
10638 arguments of pragmas registered with
10639 @code{c_register_pragma_with_expansion} but not on the arguments of
10640 pragmas registered with @code{c_register_pragma}.
10642 Note that the use of @code{pragma_lex} is specific to the C and C++
10643 compilers. It will not work in the Java or Fortran compilers, or any
10644 other language compilers for that matter. Thus if @code{pragma_lex} is going
10645 to be called from target-specific code, it must only be done so when
10646 building the C and C++ compilers. This can be done by defining the
10647 variables @code{c_target_objs} and @code{cxx_target_objs} in the
10648 target entry in the @file{config.gcc} file. These variables should name
10649 the target-specific, language-specific object file which contains the
10650 code that uses @code{pragma_lex}. Note it will also be necessary to add a
10651 rule to the makefile fragment pointed to by @code{tmake_file} that shows
10652 how to build this object file.
10657 @defmac HANDLE_SYSV_PRAGMA
10658 Define this macro (to a value of 1) if you want the System V style
10659 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
10660 [=<value>]} to be supported by gcc.
10662 The pack pragma specifies the maximum alignment (in bytes) of fields
10663 within a structure, in much the same way as the @samp{__aligned__} and
10664 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
10665 the behavior to the default.
10667 A subtlety for Microsoft Visual C/C++ style bit-field packing
10668 (e.g.@: -mms-bitfields) for targets that support it:
10669 When a bit-field is inserted into a packed record, the whole size
10670 of the underlying type is used by one or more same-size adjacent
10671 bit-fields (that is, if its long:3, 32 bits is used in the record,
10672 and any additional adjacent long bit-fields are packed into the same
10673 chunk of 32 bits. However, if the size changes, a new field of that
10674 size is allocated).
10676 If both MS bit-fields and @samp{__attribute__((packed))} are used,
10677 the latter will take precedence. If @samp{__attribute__((packed))} is
10678 used on a single field when MS bit-fields are in use, it will take
10679 precedence for that field, but the alignment of the rest of the structure
10680 may affect its placement.
10682 The weak pragma only works if @code{SUPPORTS_WEAK} and
10683 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
10684 of specifically named weak labels, optionally with a value.
10689 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
10690 Define this macro (to a value of 1) if you want to support the Win32
10691 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
10692 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
10693 alignment (in bytes) of fields within a structure, in much the same way as
10694 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
10695 pack value of zero resets the behavior to the default. Successive
10696 invocations of this pragma cause the previous values to be stacked, so
10697 that invocations of @samp{#pragma pack(pop)} will return to the previous
10701 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10702 Define this macro, as well as
10703 @code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the
10704 arguments of @samp{#pragma pack}.
10707 @deftypevr {Target Hook} bool TARGET_HANDLE_PRAGMA_EXTERN_PREFIX
10708 True if @code{#pragma extern_prefix} is to be supported.
10711 @defmac TARGET_DEFAULT_PACK_STRUCT
10712 If your target requires a structure packing default other than 0 (meaning
10713 the machine default), define this macro to the necessary value (in bytes).
10714 This must be a value that would also be valid to use with
10715 @samp{#pragma pack()} (that is, a small power of two).
10718 @defmac DOLLARS_IN_IDENTIFIERS
10719 Define this macro to control use of the character @samp{$} in
10720 identifier names for the C family of languages. 0 means @samp{$} is
10721 not allowed by default; 1 means it is allowed. 1 is the default;
10722 there is no need to define this macro in that case.
10725 @defmac NO_DOLLAR_IN_LABEL
10726 Define this macro if the assembler does not accept the character
10727 @samp{$} in label names. By default constructors and destructors in
10728 G++ have @samp{$} in the identifiers. If this macro is defined,
10729 @samp{.} is used instead.
10732 @defmac NO_DOT_IN_LABEL
10733 Define this macro if the assembler does not accept the character
10734 @samp{.} in label names. By default constructors and destructors in G++
10735 have names that use @samp{.}. If this macro is defined, these names
10736 are rewritten to avoid @samp{.}.
10739 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10740 Define this macro as a C expression that is nonzero if it is safe for the
10741 delay slot scheduler to place instructions in the delay slot of @var{insn},
10742 even if they appear to use a resource set or clobbered in @var{insn}.
10743 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10744 every @code{call_insn} has this behavior. On machines where some @code{insn}
10745 or @code{jump_insn} is really a function call and hence has this behavior,
10746 you should define this macro.
10748 You need not define this macro if it would always return zero.
10751 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10752 Define this macro as a C expression that is nonzero if it is safe for the
10753 delay slot scheduler to place instructions in the delay slot of @var{insn},
10754 even if they appear to set or clobber a resource referenced in @var{insn}.
10755 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10756 some @code{insn} or @code{jump_insn} is really a function call and its operands
10757 are registers whose use is actually in the subroutine it calls, you should
10758 define this macro. Doing so allows the delay slot scheduler to move
10759 instructions which copy arguments into the argument registers into the delay
10760 slot of @var{insn}.
10762 You need not define this macro if it would always return zero.
10765 @defmac MULTIPLE_SYMBOL_SPACES
10766 Define this macro as a C expression that is nonzero if, in some cases,
10767 global symbols from one translation unit may not be bound to undefined
10768 symbols in another translation unit without user intervention. For
10769 instance, under Microsoft Windows symbols must be explicitly imported
10770 from shared libraries (DLLs).
10772 You need not define this macro if it would always evaluate to zero.
10775 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
10776 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10777 any hard regs the port wishes to automatically clobber for an asm.
10778 It should return the result of the last @code{tree_cons} used to add a
10779 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10780 corresponding parameters to the asm and may be inspected to avoid
10781 clobbering a register that is an input or output of the asm. You can use
10782 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10783 for overlap with regards to asm-declared registers.
10786 @defmac MATH_LIBRARY
10787 Define this macro as a C string constant for the linker argument to link
10788 in the system math library, minus the initial @samp{"-l"}, or
10789 @samp{""} if the target does not have a
10790 separate math library.
10792 You need only define this macro if the default of @samp{"m"} is wrong.
10795 @defmac LIBRARY_PATH_ENV
10796 Define this macro as a C string constant for the environment variable that
10797 specifies where the linker should look for libraries.
10799 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10803 @defmac TARGET_POSIX_IO
10804 Define this macro if the target supports the following POSIX@ file
10805 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10806 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10807 to use file locking when exiting a program, which avoids race conditions
10808 if the program has forked. It will also create directories at run-time
10809 for cross-profiling.
10812 @defmac MAX_CONDITIONAL_EXECUTE
10814 A C expression for the maximum number of instructions to execute via
10815 conditional execution instructions instead of a branch. A value of
10816 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10817 1 if it does use cc0.
10820 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10821 Used if the target needs to perform machine-dependent modifications on the
10822 conditionals used for turning basic blocks into conditionally executed code.
10823 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10824 contains information about the currently processed blocks. @var{true_expr}
10825 and @var{false_expr} are the tests that are used for converting the
10826 then-block and the else-block, respectively. Set either @var{true_expr} or
10827 @var{false_expr} to a null pointer if the tests cannot be converted.
10830 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10831 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10832 if-statements into conditions combined by @code{and} and @code{or} operations.
10833 @var{bb} contains the basic block that contains the test that is currently
10834 being processed and about to be turned into a condition.
10837 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10838 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10839 be converted to conditional execution format. @var{ce_info} points to
10840 a data structure, @code{struct ce_if_block}, which contains information
10841 about the currently processed blocks.
10844 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10845 A C expression to perform any final machine dependent modifications in
10846 converting code to conditional execution. The involved basic blocks
10847 can be found in the @code{struct ce_if_block} structure that is pointed
10848 to by @var{ce_info}.
10851 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10852 A C expression to cancel any machine dependent modifications in
10853 converting code to conditional execution. The involved basic blocks
10854 can be found in the @code{struct ce_if_block} structure that is pointed
10855 to by @var{ce_info}.
10858 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10859 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10860 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10863 @defmac IFCVT_EXTRA_FIELDS
10864 If defined, it should expand to a set of field declarations that will be
10865 added to the @code{struct ce_if_block} structure. These should be initialized
10866 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10869 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void)
10870 If non-null, this hook performs a target-specific pass over the
10871 instruction stream. The compiler will run it at all optimization levels,
10872 just before the point at which it normally does delayed-branch scheduling.
10874 The exact purpose of the hook varies from target to target. Some use
10875 it to do transformations that are necessary for correctness, such as
10876 laying out in-function constant pools or avoiding hardware hazards.
10877 Others use it as an opportunity to do some machine-dependent optimizations.
10879 You need not implement the hook if it has nothing to do. The default
10880 definition is null.
10883 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS (void)
10884 Define this hook if you have any machine-specific built-in functions
10885 that need to be defined. It should be a function that performs the
10888 Machine specific built-in functions can be useful to expand special machine
10889 instructions that would otherwise not normally be generated because
10890 they have no equivalent in the source language (for example, SIMD vector
10891 instructions or prefetch instructions).
10893 To create a built-in function, call the function
10894 @code{lang_hooks.builtin_function}
10895 which is defined by the language front end. You can use any type nodes set
10896 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
10897 only language front ends that use those two functions will call
10898 @samp{TARGET_INIT_BUILTINS}.
10901 @deftypefn {Target Hook} tree TARGET_BUILTIN_DECL (unsigned @var{code}, bool @var{initialize_p})
10902 Define this hook if you have any machine-specific built-in functions
10903 that need to be defined. It should be a function that returns the
10904 builtin function declaration for the builtin function code @var{code}.
10905 If there is no such builtin and it cannot be initialized at this time
10906 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
10907 If @var{code} is out of range the function should return
10908 @code{error_mark_node}.
10911 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
10913 Expand a call to a machine specific built-in function that was set up by
10914 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10915 function call; the result should go to @var{target} if that is
10916 convenient, and have mode @var{mode} if that is convenient.
10917 @var{subtarget} may be used as the target for computing one of
10918 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10919 ignored. This function should return the result of the call to the
10923 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (unsigned int @var{loc}, tree @var{fndecl}, void *@var{arglist})
10924 Select a replacement for a machine specific built-in function that
10925 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10926 @emph{before} regular type checking, and so allows the target to
10927 implement a crude form of function overloading. @var{fndecl} is the
10928 declaration of the built-in function. @var{arglist} is the list of
10929 arguments passed to the built-in function. The result is a
10930 complete expression that implements the operation, usually
10931 another @code{CALL_EXPR}.
10932 @var{arglist} really has type @samp{VEC(tree,gc)*}
10935 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, int @var{n_args}, tree *@var{argp}, bool @var{ignore})
10936 Fold a call to a machine specific built-in function that was set up by
10937 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10938 built-in function. @var{n_args} is the number of arguments passed to
10939 the function; the arguments themselves are pointed to by @var{argp}.
10940 The result is another tree containing a simplified expression for the
10941 call's result. If @var{ignore} is true the value will be ignored.
10944 @deftypefn {Target Hook} {const char *} TARGET_INVALID_WITHIN_DOLOOP (const_rtx @var{insn})
10946 Take an instruction in @var{insn} and return NULL if it is valid within a
10947 low-overhead loop, otherwise return a string explaining why doloop
10948 could not be applied.
10950 Many targets use special registers for low-overhead looping. For any
10951 instruction that clobbers these this function should return a string indicating
10952 the reason why the doloop could not be applied.
10953 By default, the RTL loop optimizer does not use a present doloop pattern for
10954 loops containing function calls or branch on table instructions.
10957 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
10959 Take a branch insn in @var{branch1} and another in @var{branch2}.
10960 Return true if redirecting @var{branch1} to the destination of
10961 @var{branch2} is possible.
10963 On some targets, branches may have a limited range. Optimizing the
10964 filling of delay slots can result in branches being redirected, and this
10965 may in turn cause a branch offset to overflow.
10968 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (const_rtx @var{x}, int @var{outer_code})
10969 This target hook returns @code{true} if @var{x} is considered to be commutative.
10970 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
10971 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
10972 of the enclosing rtl, if known, otherwise it is UNKNOWN.
10975 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
10977 When the initial value of a hard register has been copied in a pseudo
10978 register, it is often not necessary to actually allocate another register
10979 to this pseudo register, because the original hard register or a stack slot
10980 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
10981 is called at the start of register allocation once for each hard register
10982 that had its initial value copied by using
10983 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
10984 Possible values are @code{NULL_RTX}, if you don't want
10985 to do any special allocation, a @code{REG} rtx---that would typically be
10986 the hard register itself, if it is known not to be clobbered---or a
10988 If you are returning a @code{MEM}, this is only a hint for the allocator;
10989 it might decide to use another register anyways.
10990 You may use @code{current_function_leaf_function} in the hook, functions
10991 that use @code{REG_N_SETS}, to determine if the hard
10992 register in question will not be clobbered.
10993 The default value of this hook is @code{NULL}, which disables any special
10997 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
10998 This target hook returns nonzero if @var{x}, an @code{unspec} or
10999 @code{unspec_volatile} operation, might cause a trap. Targets can use
11000 this hook to enhance precision of analysis for @code{unspec} and
11001 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
11002 to analyze inner elements of @var{x} in which case @var{flags} should be
11006 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
11007 The compiler invokes this hook whenever it changes its current function
11008 context (@code{cfun}). You can define this function if
11009 the back end needs to perform any initialization or reset actions on a
11010 per-function basis. For example, it may be used to implement function
11011 attributes that affect register usage or code generation patterns.
11012 The argument @var{decl} is the declaration for the new function context,
11013 and may be null to indicate that the compiler has left a function context
11014 and is returning to processing at the top level.
11015 The default hook function does nothing.
11017 GCC sets @code{cfun} to a dummy function context during initialization of
11018 some parts of the back end. The hook function is not invoked in this
11019 situation; you need not worry about the hook being invoked recursively,
11020 or when the back end is in a partially-initialized state.
11021 @code{cfun} might be @code{NULL} to indicate processing at top level,
11022 outside of any function scope.
11025 @defmac TARGET_OBJECT_SUFFIX
11026 Define this macro to be a C string representing the suffix for object
11027 files on your target machine. If you do not define this macro, GCC will
11028 use @samp{.o} as the suffix for object files.
11031 @defmac TARGET_EXECUTABLE_SUFFIX
11032 Define this macro to be a C string representing the suffix to be
11033 automatically added to executable files on your target machine. If you
11034 do not define this macro, GCC will use the null string as the suffix for
11038 @defmac COLLECT_EXPORT_LIST
11039 If defined, @code{collect2} will scan the individual object files
11040 specified on its command line and create an export list for the linker.
11041 Define this macro for systems like AIX, where the linker discards
11042 object files that are not referenced from @code{main} and uses export
11046 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
11047 Define this macro to a C expression representing a variant of the
11048 method call @var{mdecl}, if Java Native Interface (JNI) methods
11049 must be invoked differently from other methods on your target.
11050 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
11051 the @code{stdcall} calling convention and this macro is then
11052 defined as this expression:
11055 build_type_attribute_variant (@var{mdecl},
11057 (get_identifier ("stdcall"),
11062 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
11063 This target hook returns @code{true} past the point in which new jump
11064 instructions could be created. On machines that require a register for
11065 every jump such as the SHmedia ISA of SH5, this point would typically be
11066 reload, so this target hook should be defined to a function such as:
11070 cannot_modify_jumps_past_reload_p ()
11072 return (reload_completed || reload_in_progress);
11077 @deftypefn {Target Hook} reg_class_t TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
11078 This target hook returns a register class for which branch target register
11079 optimizations should be applied. All registers in this class should be
11080 usable interchangeably. After reload, registers in this class will be
11081 re-allocated and loads will be hoisted out of loops and be subjected
11082 to inter-block scheduling.
11085 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
11086 Branch target register optimization will by default exclude callee-saved
11088 that are not already live during the current function; if this target hook
11089 returns true, they will be included. The target code must than make sure
11090 that all target registers in the class returned by
11091 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
11092 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
11093 epilogues have already been generated. Note, even if you only return
11094 true when @var{after_prologue_epilogue_gen} is false, you still are likely
11095 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
11096 to reserve space for caller-saved target registers.
11099 @deftypefn {Target Hook} bool TARGET_HAVE_CONDITIONAL_EXECUTION (void)
11100 This target hook returns true if the target supports conditional execution.
11101 This target hook is required only when the target has several different
11102 modes and they have different conditional execution capability, such as ARM.
11105 @deftypefn {Target Hook} unsigned TARGET_LOOP_UNROLL_ADJUST (unsigned @var{nunroll}, struct loop *@var{loop})
11106 This target hook returns a new value for the number of times @var{loop}
11107 should be unrolled. The parameter @var{nunroll} is the number of times
11108 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11109 the loop, which is going to be checked for unrolling. This target hook
11110 is required only when the target has special constraints like maximum
11111 number of memory accesses.
11114 @defmac POWI_MAX_MULTS
11115 If defined, this macro is interpreted as a signed integer C expression
11116 that specifies the maximum number of floating point multiplications
11117 that should be emitted when expanding exponentiation by an integer
11118 constant inline. When this value is defined, exponentiation requiring
11119 more than this number of multiplications is implemented by calling the
11120 system library's @code{pow}, @code{powf} or @code{powl} routines.
11121 The default value places no upper bound on the multiplication count.
11124 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11125 This target hook should register any extra include files for the
11126 target. The parameter @var{stdinc} indicates if normal include files
11127 are present. The parameter @var{sysroot} is the system root directory.
11128 The parameter @var{iprefix} is the prefix for the gcc directory.
11131 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11132 This target hook should register any extra include files for the
11133 target before any standard headers. The parameter @var{stdinc}
11134 indicates if normal include files are present. The parameter
11135 @var{sysroot} is the system root directory. The parameter
11136 @var{iprefix} is the prefix for the gcc directory.
11139 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11140 This target hook should register special include paths for the target.
11141 The parameter @var{path} is the include to register. On Darwin
11142 systems, this is used for Framework includes, which have semantics
11143 that are different from @option{-I}.
11146 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11147 This target macro returns @code{true} if it is safe to use a local alias
11148 for a virtual function @var{fndecl} when constructing thunks,
11149 @code{false} otherwise. By default, the macro returns @code{true} for all
11150 functions, if a target supports aliases (i.e.@: defines
11151 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11154 @defmac TARGET_FORMAT_TYPES
11155 If defined, this macro is the name of a global variable containing
11156 target-specific format checking information for the @option{-Wformat}
11157 option. The default is to have no target-specific format checks.
11160 @defmac TARGET_N_FORMAT_TYPES
11161 If defined, this macro is the number of entries in
11162 @code{TARGET_FORMAT_TYPES}.
11165 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11166 If defined, this macro is the name of a global variable containing
11167 target-specific format overrides for the @option{-Wformat} option. The
11168 default is to have no target-specific format overrides. If defined,
11169 @code{TARGET_FORMAT_TYPES} must be defined, too.
11172 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11173 If defined, this macro specifies the number of entries in
11174 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11177 @defmac TARGET_OVERRIDES_FORMAT_INIT
11178 If defined, this macro specifies the optional initialization
11179 routine for target specific customizations of the system printf
11180 and scanf formatter settings.
11183 @deftypevr {Target Hook} bool TARGET_RELAXED_ORDERING
11184 If set to @code{true}, means that the target's memory model does not
11185 guarantee that loads which do not depend on one another will access
11186 main memory in the order of the instruction stream; if ordering is
11187 important, an explicit memory barrier must be used. This is true of
11188 many recent processors which implement a policy of ``relaxed,''
11189 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
11190 and ia64. The default is @code{false}.
11193 @deftypefn {Target Hook} {const char *} TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (const_tree @var{typelist}, const_tree @var{funcdecl}, const_tree @var{val})
11194 If defined, this macro returns the diagnostic message when it is
11195 illegal to pass argument @var{val} to function @var{funcdecl}
11196 with prototype @var{typelist}.
11199 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (const_tree @var{fromtype}, const_tree @var{totype})
11200 If defined, this macro returns the diagnostic message when it is
11201 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11202 if validity should be determined by the front end.
11205 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, const_tree @var{type})
11206 If defined, this macro returns the diagnostic message when it is
11207 invalid to apply operation @var{op} (where unary plus is denoted by
11208 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11209 if validity should be determined by the front end.
11212 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, const_tree @var{type1}, const_tree @var{type2})
11213 If defined, this macro returns the diagnostic message when it is
11214 invalid to apply operation @var{op} to operands of types @var{type1}
11215 and @var{type2}, or @code{NULL} if validity should be determined by
11219 @deftypefn {Target Hook} {const char *} TARGET_INVALID_PARAMETER_TYPE (const_tree @var{type})
11220 If defined, this macro returns the diagnostic message when it is
11221 invalid for functions to include parameters of type @var{type},
11222 or @code{NULL} if validity should be determined by
11223 the front end. This is currently used only by the C and C++ front ends.
11226 @deftypefn {Target Hook} {const char *} TARGET_INVALID_RETURN_TYPE (const_tree @var{type})
11227 If defined, this macro returns the diagnostic message when it is
11228 invalid for functions to have return type @var{type},
11229 or @code{NULL} if validity should be determined by
11230 the front end. This is currently used only by the C and C++ front ends.
11233 @deftypefn {Target Hook} tree TARGET_PROMOTED_TYPE (const_tree @var{type})
11234 If defined, this target hook returns the type to which values of
11235 @var{type} should be promoted when they appear in expressions,
11236 analogous to the integer promotions, or @code{NULL_TREE} to use the
11237 front end's normal promotion rules. This hook is useful when there are
11238 target-specific types with special promotion rules.
11239 This is currently used only by the C and C++ front ends.
11242 @deftypefn {Target Hook} tree TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr})
11243 If defined, this hook returns the result of converting @var{expr} to
11244 @var{type}. It should return the converted expression,
11245 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11246 This hook is useful when there are target-specific types with special
11248 This is currently used only by the C and C++ front ends.
11251 @defmac TARGET_USE_JCR_SECTION
11252 This macro determines whether to use the JCR section to register Java
11253 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
11254 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
11258 This macro determines the size of the objective C jump buffer for the
11259 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11262 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11263 Define this macro if any target-specific attributes need to be attached
11264 to the functions in @file{libgcc} that provide low-level support for
11265 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11266 and the associated definitions of those functions.
11269 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
11270 Define this macro to update the current function stack boundary if
11274 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
11275 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11276 different argument pointer register is needed to access the function's
11277 argument list due to stack realignment. Return @code{NULL} if no DRAP
11281 @deftypefn {Target Hook} bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
11282 When optimization is disabled, this hook indicates whether or not
11283 arguments should be allocated to stack slots. Normally, GCC allocates
11284 stacks slots for arguments when not optimizing in order to make
11285 debugging easier. However, when a function is declared with
11286 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11287 cannot safely move arguments from the registers in which they are passed
11288 to the stack. Therefore, this hook should return true in general, but
11289 false for naked functions. The default implementation always returns true.
11292 @deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR
11293 On some architectures it can take multiple instructions to synthesize
11294 a constant. If there is another constant already in a register that
11295 is close enough in value then it is preferable that the new constant
11296 is computed from this register using immediate addition or
11297 subtraction. We accomplish this through CSE. Besides the value of
11298 the constant we also add a lower and an upper constant anchor to the
11299 available expressions. These are then queried when encountering new
11300 constants. The anchors are computed by rounding the constant up and
11301 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11302 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11303 accepted by immediate-add plus one. We currently assume that the
11304 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11305 MIPS, where add-immediate takes a 16-bit signed value,
11306 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11307 is zero, which disables this optimization. @end deftypevr